poly_int: vect_nunits_for_cost
[official-gcc.git] / gcc / ChangeLog
blob52a10ad2305a10629d0f0d25fb37cd5c389cd527
1 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
2             Alan Hayward  <alan.hayward@arm.com>
3             David Sherwood  <david.sherwood@arm.com>
5         * tree-vectorizer.h (vect_nunits_for_cost): New function.
6         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8         (vect_analyze_slp_cost): Likewise.
9         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10         (vect_model_load_cost): Likewise.
12 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13             Alan Hayward  <alan.hayward@arm.com>
14             David Sherwood  <david.sherwood@arm.com>
16         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18         from an unsigned int * to a poly_uint64_pod *.
19         (calculate_unrolling_factor): New function.
20         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
22 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
23             Alan Hayward  <alan.hayward@arm.com>
24             David Sherwood  <david.sherwood@arm.com>
26         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27         from an unsigned int to a poly_uint64.
28         (_loop_vec_info::slp_unrolling_factor): Likewise.
29         (_loop_vec_info::vectorization_factor): Change from an int
30         to a poly_uint64.
31         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32         (vect_get_num_vectors): New function.
33         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
34         (vect_get_num_copies): Use vect_get_num_vectors.
35         (vect_analyze_data_ref_dependences): Change max_vf from an int *
36         to an unsigned int *.
37         (vect_analyze_data_refs): Change min_vf from an int * to a
38         poly_uint64 *.
39         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
40         than an unsigned HOST_WIDE_INT.
41         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
42         (vect_analyze_data_ref_dependence): Change max_vf from an int *
43         to an unsigned int *.
44         (vect_analyze_data_ref_dependences): Likewise.
45         (vect_compute_data_ref_alignment): Handle polynomial vf.
46         (vect_enhance_data_refs_alignment): Likewise.
47         (vect_prune_runtime_alias_test_list): Likewise.
48         (vect_shift_permute_load_chain): Likewise.
49         (vect_supportable_dr_alignment): Likewise.
50         (dependence_distance_ge_vf): Take the vectorization factor as a
51         poly_uint64 rather than an unsigned HOST_WIDE_INT.
52         (vect_analyze_data_refs): Change min_vf from an int * to a
53         poly_uint64 *.
54         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
55         vfm1 as a poly_uint64 rather than an int.  Make the same change
56         for the returned bound_scalar.
57         (vect_gen_vector_loop_niters): Handle polynomial vf.
58         (vect_do_peeling): Likewise.  Update call to
59         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
60         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
61         be constant.
62         * tree-vect-loop.c (vect_determine_vectorization_factor)
63         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
64         (vect_get_known_peeling_cost): Likewise.
65         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
66         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
67         (vect_transform_loop): Likewise.  Use the lowest possible VF when
68         updating the upper bounds of the loop.
69         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
70         rather than an int.
71         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
72         polynomial unroll factors.
73         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
74         (vect_make_slp_decision): Likewise.
75         (vect_supported_load_permutation_p): Likewise, and polynomial
76         vf too.
77         (vect_analyze_slp_cost): Handle polynomial vf.
78         (vect_slp_analyze_node_operations): Likewise.
79         (vect_slp_analyze_bb_1): Likewise.
80         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
81         than an unsigned HOST_WIDE_INT.
82         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
83         (vectorizable_load): Handle polynomial vf.
84         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
85         a poly_uint64.
86         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
88 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
89             Alan Hayward  <alan.hayward@arm.com>
90             David Sherwood  <david.sherwood@arm.com>
92         * match.pd: Handle bit operations involving three constants
93         and try to fold one pair.
95 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
97         * tree-vect-loop-manip.c: Include gimple-fold.h.
98         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
99         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
100         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
101         Add a path that uses a step of VF instead of 1, but disable it
102         for now.
103         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
104         and niters_no_overflow parameters.  Update calls to
105         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
106         Create a new SSA name if the latter choses to use a ste other
107         than zero, and return it via niters_vector_mult_vf_var.
108         * tree-vect-loop.c (vect_transform_loop): Update calls to
109         vect_do_peeling, vect_gen_vector_loop_niters and
110         slpeel_make_loop_iterate_ntimes.
111         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
112         (vect_gen_vector_loop_niters): Update declarations after above changes.
114 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
116         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
117         128-bit round to integer instructions.
118         (ceil<mode>2): Likewise.
119         (btrunc<mode>2): Likewise.
120         (round<mode>2): Likewise.
122 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
124         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
125         unaligned VSX load/store on P8/P9.
126         (expand_block_clear): Allow the use of unaligned VSX
127         load/store on P8/P9.
129 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
131         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
132         New function.
133         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
134         swap associated with both a load and a store.
136 2018-01-02  Andrew Waterman  <andrew@sifive.com>
138         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
139         * config/riscv/riscv.md (clear_cache): Use it.
141 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
143         * web.c: Remove out-of-date comment.
145 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
147         * expr.c (fixup_args_size_notes): Check that any existing
148         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
149         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
150         (emit_single_push_insn): ...here.
152 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
154         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
155         (const_vector_encoded_nelts): New function.
156         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
157         (const_vector_int_elt, const_vector_elt): Declare.
158         * emit-rtl.c (const_vector_int_elt_1): New function.
159         (const_vector_elt): Likewise.
160         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
161         of CONST_VECTOR_ELT.
163 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
165         * expr.c: Include rtx-vector-builder.h.
166         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
167         directly on the tree encoding.
168         (const_vector_from_tree): Likewise.
169         * optabs.c: Include rtx-vector-builder.h.
170         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
171         sequence of "u" values.
172         * vec-perm-indices.c: Include rtx-vector-builder.h.
173         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
174         directly on the vec_perm_indices encoding.
176 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
178         * doc/rtl.texi (const_vector): Describe new encoding scheme.
179         * Makefile.in (OBJS): Add rtx-vector-builder.o.
180         * rtx-vector-builder.h: New file.
181         * rtx-vector-builder.c: Likewise.
182         * rtl.h (rtx_def::u2): Add a const_vector field.
183         (CONST_VECTOR_NPATTERNS): New macro.
184         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
185         (CONST_VECTOR_DUPLICATE_P): Likewise.
186         (CONST_VECTOR_STEPPED_P): Likewise.
187         (CONST_VECTOR_ENCODED_ELT): Likewise.
188         (const_vec_duplicate_p): Check for a duplicated vector encoding.
189         (unwrap_const_vec_duplicate): Likewise.
190         (const_vec_series_p): Check for a non-duplicated vector encoding.
191         Say that the function only returns true for integer vectors.
192         * emit-rtl.c: Include rtx-vector-builder.h.
193         (gen_const_vec_duplicate_1): Delete.
194         (gen_const_vector): Call gen_const_vec_duplicate instead of
195         gen_const_vec_duplicate_1.
196         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
197         (gen_const_vec_duplicate): Use rtx_vector_builder.
198         (gen_const_vec_series): Likewise.
199         (gen_rtx_CONST_VECTOR): Likewise.
200         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
201         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
202         Build a new vector rather than modifying a CONST_VECTOR in-place.
203         (handle_special_swappables): Update call accordingly.
204         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
205         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
206         Build a new vector rather than modifying a CONST_VECTOR in-place.
207         (handle_special_swappables): Update call accordingly.
209 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
211         * simplify-rtx.c (simplify_const_binary_operation): Use
212         CONST_VECTOR_ELT instead of XVECEXP.
214 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
216         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
217         the selector elements to be different from the data elements
218         if the selector is a VECTOR_CST.
219         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
220         ssizetype for the selector.
222 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
224         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
225         before testing each element individually.
226         * tree-vect-generic.c (lower_vec_perm): Likewise.
228 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
230         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
231         * selftest-run-tests.c (selftest::run_tests): Call it.
232         * vector-builder.h (vector_builder::operator ==): New function.
233         (vector_builder::operator !=): Likewise.
234         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
235         (vec_perm_indices::all_from_input_p): New function.
236         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
237         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
238         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
239         instead of reading the VECTOR_CST directly.  Detect whether both
240         vector inputs are the same before constructing the vec_perm_indices,
241         and update the number of inputs argument accordingly.  Use the
242         utility functions added above.  Only construct sel2 if we need to.
244 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
246         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
247         the broadcast of the low byte.
248         (expand_mult_highpart): Use an explicit encoding for the permutes.
249         * optabs-query.c (can_mult_highpart_p): Likewise.
250         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
251         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
252         (vectorizable_bswap): Likewise.
253         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
254         explicit encoding for the power-of-2 permutes.
255         (vect_permute_store_chain): Likewise.
256         (vect_grouped_load_supported): Likewise.
257         (vect_permute_load_chain): Likewise.
259 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
261         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
262         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
263         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
264         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
265         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
266         (vect_gen_perm_mask_any): Likewise.
268 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
270         * int-vector-builder.h: New file.
271         * vec-perm-indices.h: Include int-vector-builder.h.
272         (vec_perm_indices): Redefine as an int_vector_builder.
273         (auto_vec_perm_indices): Delete.
274         (vec_perm_builder): Redefine as a stand-alone class.
275         (vec_perm_indices::vec_perm_indices): New function.
276         (vec_perm_indices::clamp): Likewise.
277         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
278         (vec_perm_indices::new_vector): New function.
279         (vec_perm_indices::new_expanded_vector): Update for new
280         vec_perm_indices class.
281         (vec_perm_indices::rotate_inputs): New function.
282         (vec_perm_indices::all_in_range_p): Operate directly on the
283         encoded form, without computing elided elements.
284         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
285         encoding.  Update for new vec_perm_indices class.
286         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
287         the given vec_perm_builder.
288         (expand_vec_perm_var): Update vec_perm_builder constructor.
289         (expand_mult_highpart): Use vec_perm_builder instead of
290         auto_vec_perm_indices.
291         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
292         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
293         or double series encoding as appropriate.
294         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
295         vec_perm_indices instead of auto_vec_perm_indices.
296         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
297         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
298         (vect_permute_store_chain): Likewise.
299         (vect_grouped_load_supported): Likewise.
300         (vect_permute_load_chain): Likewise.
301         (vect_shift_permute_load_chain): Likewise.
302         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
303         (vect_transform_slp_perm_load): Likewise.
304         (vect_schedule_slp_instance): Likewise.
305         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
306         (vectorizable_mask_load_store): Likewise.
307         (vectorizable_bswap): Likewise.
308         (vectorizable_store): Likewise.
309         (vectorizable_load): Likewise.
310         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
311         vec_perm_indices instead of auto_vec_perm_indices.  Use
312         tree_to_vec_perm_builder to read the vector from a tree.
313         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
314         vec_perm_builder instead of a vec_perm_indices.
315         (have_whole_vector_shift): Use vec_perm_builder and
316         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
317         truncation to calc_vec_perm_mask_for_shift.
318         (vect_create_epilog_for_reduction): Likewise.
319         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
320         from auto_vec_perm_indices to vec_perm_indices.
321         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
322         instead of changing individual elements.
323         (aarch64_vectorize_vec_perm_const): Use new_vector to install
324         the vector in d.perm.
325         * config/arm/arm.c (expand_vec_perm_d::perm): Change
326         from auto_vec_perm_indices to vec_perm_indices.
327         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
328         instead of changing individual elements.
329         (arm_vectorize_vec_perm_const): Use new_vector to install
330         the vector in d.perm.
331         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
332         Update vec_perm_builder constructor.
333         (rs6000_expand_interleave): Likewise.
334         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
335         (rs6000_expand_interleave): Likewise.
337 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
339         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
340         to qimode could truncate the indices.
341         * optabs.c (expand_vec_perm_var): Likewise.
343 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
345         * Makefile.in (OBJS): Add vec-perm-indices.o.
346         * vec-perm-indices.h: New file.
347         * vec-perm-indices.c: Likewise.
348         * target.h (vec_perm_indices): Replace with a forward class
349         declaration.
350         (auto_vec_perm_indices): Move to vec-perm-indices.h.
351         * optabs.h: Include vec-perm-indices.h.
352         (expand_vec_perm): Delete.
353         (selector_fits_mode_p, expand_vec_perm_var): Declare.
354         (expand_vec_perm_const): Declare.
355         * target.def (vec_perm_const_ok): Replace with...
356         (vec_perm_const): ...this new hook.
357         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
358         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
359         * doc/tm.texi: Regenerate.
360         * optabs.def (vec_perm_const): Delete.
361         * doc/md.texi (vec_perm_const): Likewise.
362         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
363         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
364         expand_vec_perm for constant permutation vectors.  Assert that
365         the mode of variable permutation vectors is the integer equivalent
366         of the mode that is being permuted.
367         * optabs-query.h (selector_fits_mode_p): Declare.
368         * optabs-query.c: Include vec-perm-indices.h.
369         (selector_fits_mode_p): New function.
370         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
371         is defined, instead of checking whether the vec_perm_const_optab
372         exists.  Use targetm.vectorize.vec_perm_const instead of
373         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
374         fit in the vector mode before using a variable permute.
375         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
376         vec_perm_indices instead of an rtx.
377         (expand_vec_perm): Replace with...
378         (expand_vec_perm_const): ...this new function.  Take the selector
379         as a vec_perm_indices rather than an rtx.  Also take the mode of
380         the selector.  Update call to shift_amt_for_vec_perm_mask.
381         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
382         Use vec_perm_indices::new_expanded_vector to expand the original
383         selector into bytes.  Check whether the indices fit in the vector
384         mode before using a variable permute.
385         (expand_vec_perm_var): Make global.
386         (expand_mult_highpart): Use expand_vec_perm_const.
387         * fold-const.c: Includes vec-perm-indices.h.
388         * tree-ssa-forwprop.c: Likewise.
389         * tree-vect-data-refs.c: Likewise.
390         * tree-vect-generic.c: Likewise.
391         * tree-vect-loop.c: Likewise.
392         * tree-vect-slp.c: Likewise.
393         * tree-vect-stmts.c: Likewise.
394         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
395         Delete.
396         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
397         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
398         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
399         (aarch64_vectorize_vec_perm_const): ...this new function.
400         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
401         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
402         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
403         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
404         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
405         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
406         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
407         into...
408         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
409         check for NEON modes.
410         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
411         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
412         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
413         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
414         into...
415         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
416         the old VEC_PERM_CONST conditions.
417         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
418         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
419         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
420         (ia64_vectorize_vec_perm_const_ok): Merge into...
421         (ia64_vectorize_vec_perm_const): ...this new function.
422         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
423         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
424         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
425         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
426         * config/mips/mips.c (mips_expand_vec_perm_const)
427         (mips_vectorize_vec_perm_const_ok): Merge into...
428         (mips_vectorize_vec_perm_const): ...this new function.
429         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
430         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
431         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
432         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
433         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
434         (rs6000_expand_vec_perm_const): Delete.
435         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
436         Delete.
437         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
438         (altivec_expand_vec_perm_const_le): Take each operand individually.
439         Operate on constant selectors rather than rtxes.
440         (altivec_expand_vec_perm_const): Likewise.  Update call to
441         altivec_expand_vec_perm_const_le.
442         (rs6000_expand_vec_perm_const): Delete.
443         (rs6000_vectorize_vec_perm_const_ok): Delete.
444         (rs6000_vectorize_vec_perm_const): New function.
445         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
446         an element count and rtx array.
447         (rs6000_expand_extract_even): Update call accordingly.
448         (rs6000_expand_interleave): Likewise.
449         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
450         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
451         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
452         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
453         (rs6000_expand_vec_perm_const): Delete.
454         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
455         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
456         (altivec_expand_vec_perm_const_le): Take each operand individually.
457         Operate on constant selectors rather than rtxes.
458         (altivec_expand_vec_perm_const): Likewise.  Update call to
459         altivec_expand_vec_perm_const_le.
460         (rs6000_expand_vec_perm_const): Delete.
461         (rs6000_vectorize_vec_perm_const_ok): Delete.
462         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
463         reference to the SPE evmerge intructions.
464         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
465         an element count and rtx array.
466         (rs6000_expand_extract_even): Update call accordingly.
467         (rs6000_expand_interleave): Likewise.
468         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
469         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
470         new function.
471         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
473 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
475         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
476         vector mode and that that mode matches the mode of the data
477         being permuted.
478         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
479         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
480         directly using expand_vec_perm_1 when forcing selectors into
481         registers.
482         (expand_vec_perm_var): New function, split out from expand_vec_perm.
484 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
486         * optabs-query.h (can_vec_perm_p): Delete.
487         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
488         * optabs-query.c (can_vec_perm_p): Split into...
489         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
490         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
491         particular selector is valid.
492         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
493         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
494         (vect_grouped_load_supported): Likewise.
495         (vect_shift_permute_load_chain): Likewise.
496         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
497         (vect_transform_slp_perm_load): Likewise.
498         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
499         (vectorizable_bswap): Likewise.
500         (vect_gen_perm_mask_checked): Likewise.
501         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
502         implementations of variable permutation vectors into account
503         when deciding which selector to use.
504         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
505         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
506         with a false third argument.
507         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
508         to test whether the constant selector is valid and can_vec_perm_var_p
509         to test whether a variable selector is valid.
511 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
513         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
514         * optabs-query.c (can_vec_perm_p): Likewise.
515         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
516         instead of vec_perm_indices.
517         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
518         (vect_gen_perm_mask_checked): Likewise,
519         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
520         (vect_gen_perm_mask_checked): Likewise,
522 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
524         * optabs-query.h (qimode_for_vec_perm): Declare.
525         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
526         (qimode_for_vec_perm): ...this new function.
527         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
529 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
531         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
532         does not have a conditional at the top.
534 2018-01-02  Richard Biener  <rguenther@suse.de>
536         * ipa-inline.c (big_speedup_p): Fix expression.
538 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
540         PR target/81616
541         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
542         for generic 4->6.
544 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
546         PR target/81616
547         Generic tuning.
548         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
549         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
550         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
551         cond_taken_branch_cost 3->4.
553 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
555         PR tree-optimization/83581
556         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
557         TODO_cleanup_cfg if any changes have been made.
559         PR middle-end/83608
560         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
561         convert_modes if target mode has the right side, but different mode
562         class.
564         PR middle-end/83609
565         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
566         last argument when extracting from CONCAT.  If either from_real or
567         from_imag is NULL, use expansion through memory.  If result is not
568         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
569         the parts directly to inner mode, if even that fails, use expansion
570         through memory.
572         PR middle-end/83623
573         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
574         check for bswap in mode rather than HImode and use that in expand_unop
575         too.
577 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
579         PR target/83536
580         * config/i386/i386.c (ix86_attribute_table): Remove excess
581         initializer for "shared" attribute.
583 2017-12-30  Tom de Vries  <tom@codesourcery.com>
585         PR libgomp/83046
586         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
587         DECL_PRESERVE_P.
588         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
589         offload_funcs entries that no longer have a corresponding cgraph_node.
590         Mark the remaining ones as DECL_PRESERVE_P.
591         (output_lto): Call prune_offload_funcs.
593 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
595         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
596         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
597         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
598         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
599         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
600         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
601         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
602         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
603         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
604         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
605         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
606         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
607         vaesenc_<mode>, vpclmulqdq_<mode>,
608         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
610 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
612         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
613         _Float<N> and _Float<N>X variants for rounding built-in
614         functions.
615         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
616         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
617         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
618         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
619         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
620         * builtins.c (mathfn_built_in_2): Likewise.
621         * internal-fn.def (CEIL): Likewise.
622         (FLOOR): Likewise.
623         (NEARBYINT): Likewise.
624         (RINT): Likewise.
625         (ROUND): Likewise.
626         (TRUNC): Likewise.
627         * convert.c (convert_to_integer_1): Likewise.
628         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
629         (integer_valued_real_call_p): Likewise.
630         * fold-const-call.c (fold_const_call_ss): Likewise.
631         * gencfn-macros.c (print_case_cfn): Change CFN and operator
632         printers to take a const char * suffix instead of a bool.
633         (print_define_operator_list): Likewise.
634         (fltall_suffixes): New list of suffixes, that include the
635         traditional suffixes as well as all of the _Float<N> and
636         _Float<N>X suffixes.
637         (main): For _Float<N> and _Float<N>X functions, emit both
638         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
639         has the _Float<N> and _Float<N>X case names or operators.  The
640         <name>_ALL variant has both the traditional and the
641         _Float<N>/_Float<N>X case names or operators.
642         * match.pd (COPYSIGN optimizations): Provide optimizations for
643         _Float<N> and _Float<N>X types where possible.
644         (MIN/MAX optimizations): Likewise.
645         (sqrt optimizations): Likewise.
646         (rounding optimizations): Likewise.
648 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
650         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
651         gen_int_mode rather than GEN_INT.
652         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
654 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
656         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
657         instead of CONSTANT_P.
658         (gen_vec_series): Likewise.
659         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
661 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
663         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
664         CC_STATUS_INIT.
666 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
668         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
669         (aarch64_ld1x2<VDC:mode>): Likewise.
670         (aarch64_simd_ld1<mode>_x2): Likewise.
671         (aarch64_simd_ld1<mode>_x2): Likewise.
672         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
673         (vld1_s8_x2): Likewise.
674         (vld1_u16_x2): Likewise.
675         (vld1_s16_x2): Likewise.
676         (vld1_u32_x2): Likewise.
677         (vld1_s32_x2): Likewise.
678         (vld1_u64_x2): Likewise.
679         (vld1_s64_x2): Likewise.
680         (vld1_f16_x2): Likewise.
681         (vld1_f32_x2): Likewise.
682         (vld1_f64_x2): Likewise.
683         (vld1_p8_x2): Likewise.
684         (vld1_p16_x2): Likewise.
685         (vld1_p64_x2): Likewise.
686         (vld1q_u8_x2): Likewise.
687         (vld1q_s8_x2): Likewise.
688         (vld1q_u16_x2): Likewise.
689         (vld1q_s16_x2): Likewise.
690         (vld1q_u32_x2): Likewise.
691         (vld1q_s32_x2): Likewise.
692         (vld1q_u64_x2): Likewise.
693         (vld1q_s64_x2): Likewise.
694         (vld1q_f16_x2): Likewise.
695         (vld1q_f32_x2): Likewise.
696         (vld1q_f64_x2): Likewise.
697         (vld1q_p8_x2): Likewise.
698         (vld1q_p16_x2): Likewise.
699         (vld1q_p64_x2): Likewise.
701 2017-12-27  Martin Liska  <mliska@suse.cz>
703         PR tree-optimization/83552
704         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
705         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
707 2017-12-27  Tom de Vries  <tom@codesourcery.com>
709         * config/nvptx/nvptx.c (nvptx_option_override): Disable
710         -gstatement-frontiers.
712 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
714         PR rtl-optimization/83513
715         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
716         before priority comparison.
718 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
720         PR target/83488
721         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
722         ix86_isa_flags2 to ix86_isa_flags.
723         * config/i386/i386-c.c (ix86_target_macros_internal): Test
724         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
725         isa_flags rather than isa_flags2.
726         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
727         and -mavx512bitalg from isa2_opts to isa_opts.
728         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
729         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
730         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
731         Formatting fixes.
732         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
733         OPTION_MASK_ISA_AVX512F ored with another option similarly to
734         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
735         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
736         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
737         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
738         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
739         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
740         builtins from bdesc_args2 to bdesc_args section.
741         (__builtin_ia32_compressstoreuqi512_mask,
742         __builtin_ia32_compressstoreuhi512_mask,
743         __builtin_ia32_compressstoreuqi256_mask,
744         __builtin_ia32_expandloadqi512_mask,
745         __builtin_ia32_expandloadqi512_maskz,
746         __builtin_ia32_expandloadhi512_mask,
747         __builtin_ia32_expandloadhi512_maskz,
748         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
749         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
750         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
751         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
752         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
753         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
754         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
755         __builtin_ia32_vpshldv_v32hi_maskz,
756         __builtin_ia32_vpopcountb_v64qi_mask,
757         __builtin_ia32_vpopcountw_v32hi_mask,
758         __builtin_ia32_vpshufbitqmb512_mask,
759         __builtin_ia32_vpshufbitqmb256_mask): Add
760         " | OPTION_MASK_ISA_AVX512BW".
761         (__builtin_ia32_expandloadqi256_mask,
762         __builtin_ia32_expandloadqi256_maskz,
763         __builtin_ia32_vpopcountb_v32qi_mask): Add
764         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
765         (__builtin_ia32_expandloadhi256_mask,
766         __builtin_ia32_expandloadhi256_maskz,
767         __builtin_ia32_expandloadqi128_mask,
768         __builtin_ia32_expandloadqi128_maskz,
769         __builtin_ia32_expandloadhi128_mask,
770         __builtin_ia32_expandloadhi128_maskz,
771         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
772         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
773         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
774         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
775         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
776         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
777         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
778         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
779         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
780         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
781         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
782         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
783         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
784         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
785         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
786         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
787         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
788         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
789         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
790         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
791         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
792         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
793         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
794         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
795         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
796         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
797         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
798         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
799         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
800         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
801         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
802         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
803         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
804         __builtin_ia32_vpopcountw_v8hi_mask): Add
805         " | OPTION_MASK_ISA_AVX512VL".
806         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
807         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
808         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
809         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
810         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
811         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
812         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
813         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
814         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
815         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
816         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
817         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
818         _mm_mask_bitshuffle_epi64_mask): Likewise.
819         * common/config/i386/i386-common.c
820         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
821         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
822         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
823         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
824         OPTION_MASK_ISA_AVX512BITALG_UNSET.
825         (OPTION_MASK_ISA2_AVX512F_UNSET,
826         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
827         (ix86_handle_option): For -mno-general-regs-only, clear from
828         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
829         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
830         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
831         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
832         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
833         for moving from ix86_isa_flags2 to ix86_isa_flags.
835 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
837         PR c++/83553
838         * fold-const.c (struct contains_label_data): New type.
839         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
840         inside of a SWITCH_BODY seen during the walk.
841         (contains_label_p): Use walk_tree instead of
842         walk_tree_without_duplicates, prepare data for contains_label_1 and
843         provide own pset.
845 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
847         PR debug/83547
848         * tree-iterator.c (alloc_stmt_list): Start with cleared
849         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
850         or old one reused.
852         PR target/83488
853         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
854         enable avx512f explicitly in #pragma GCC target.
855         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
856         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
857         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
858         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
859         __builtin_ia32_vpdpbusds_v8si_mask,
860         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
861         __builtin_ia32_vpdpbusds_v4si_mask,
862         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
863         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
864         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
865         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
866         __builtin_ia32_vpdpwssds_v8si_mask,
867         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
868         __builtin_ia32_vpdpwssds_v4si_mask,
869         __builtin_ia32_vpdpwssds_v4si_maskz): Use
870         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
871         just OPTION_MASK_ISA_AVX512VNNI.
873 2017-12-22  Martin Jambor  <mjambor@suse.cz>
875         PR lto/82027
876         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
877         clones.
879 2017-12-22  Julia Koval  <julia.koval@intel.com>
880             Sebastian Peryt  <sebastian.peryt@intel.com>
882         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
883         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
884         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
885         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
886         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
887         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
888         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
889         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
890         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
891         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
892         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
893         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
894         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
895         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
896         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
897         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
898         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
899         _mm256_maskz_popcnt_epi64): New intrinsics.
900         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
901         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
902         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
903         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
904         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
905         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
906         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
907         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
908         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
909         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
910         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
911         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
912         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
913         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
914         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
915         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
916         __builtin_ia32_vpshufbitqmb256_mask,
917         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
918         * config/i386/i386-c.c (__AVX512BITALG__): New.
919         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
920         (ix86_valid_target_attribute_inner_p): Ditto.
921         (ix86_expand_args_builtin): Handle new types.
922         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
923         * config/i386/i386.opt: Add -mavx512bitalg.
924         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
925         avx512bitalgintrin.h.
926         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
927         (vpopcount<mode><mask_name>): Add more types.
928         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
929         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
931 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
933         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
934         Or in OPTION_MASK_ISA_AVX512F_SET.
935         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
936         OPTION_MASK_ISA_AVX512VNNI_UNSET.
937         (ix86_handle_option): Adjust for
938         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
939         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
940         section to ARGS.
941         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
942         isa_flag instead of isa_flag2.
943         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
944         isa_opts2 to isa_opts.
945         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
946         to ix86_isa_flags.
948 2017-12-22  Mike Stump  <mikestump@comcast.net>
949             Eric Botcazou  <ebotcazou@adacore.com>
951         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
953 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
955         PR middle-end/83487
956         * config/i386/i386.c (ix86_function_arg_boundary): Return
957         PARM_BOUNDARY for TYPE_EMPTY_P types.
959         PR c/83448
960         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
961         if navail is >= dir.len.
963 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
965         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
966         triplet for ilp32.
968 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
970         PR rtl-optimization/80747
971         PR rtl-optimization/83512
972         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
973         succ edge from ENTRY, copy partition from e->dest to the newly
974         created bb.
975         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
976         ENTRY, use BB_PARTITION of its successor block as current_partition.
977         Don't copy partition when splitting succ edge from ENTRY.
979         PR tree-optimization/83523
980         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
981         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
982         (convert_mult_to_fma): Likewise.
984         PR tree-optimization/83521
985         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
986         gimple_build_assign without code on result of
987         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
988         a VIEW_CONVERT_EXPR.
990 2017-12-21  Andrew Pinski  <apinski@cavium.com>
991             Steve Ellcey  <sellcey@cavium.com>
993         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
994         multi-arch for ilp32.
996 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
998         PR target/83467
999         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1000         constraints to operand 2.
1001         (*ashl<mode>3_mask_1): Ditto.
1002         (*<shift_insn><mode>3_mask): Ditto.
1003         (*<shift_insn><mode>3_mask_1): Ditto.
1004         (*<rotate_insn><mode>3_mask): Ditto.
1005         (*<rotate_insn><mode>3_mask_1): Ditto.
1007 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1009         * reorg.c (make_return_insns): Reemit each insn with its own location.
1011 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1013         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1014         generation for cases where splatting a value is not useful.
1015         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1016         across a vec_duplicate and a paradoxical subreg forming a vector
1017         mode to a vec_concat.
1019 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1021         * combine.c (simplify_set): Do not transform subregs to zero_extends
1022         if the destination is not a scalar int mode.
1024 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1026         PR c++/82872
1027         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1028         shared zero if the input has overflowed.
1030 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1032         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1033         support for Cortex-A55 and Cortex-A75.
1035 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1037         * common/config/arm/arm-common.c (compare_opt_names): Add function
1038         comment.  Use strcmp instead of manual loop.
1040 2017-12-21  Martin Liska  <mliska@suse.cz>
1042         PR gcov-profile/83509
1043         * gcov-dump.c (dump_gcov_file): Do not read info about
1044         support_unexecuted_blocks for gcda files.
1046 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1048         PR rtl-optimization/82973
1049         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1050         (valid_for_const_vector_p): ... this.
1051         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1052         (valid_for_const_vector_p): ... this.  Adjust function comment.
1053         (gen_vec_duplicate): Adjust caller.
1054         * optabs.c (expand_vector_broadcast): Likewise.
1055         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1056         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1057         constant.
1058         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1059         instead of GET_CODE == CONST_FIXED.
1060         (simplify_subreg): Use CONST_FIXED_P macro instead of
1061         GET_CODE == CONST_FIXED.
1063         PR target/83488
1064         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1065         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1066         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1067         (ix86_option_override_internal): Adjust for
1068         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1069         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1070         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1071         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1072         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1073         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1074         def_builtin2 for CET builtins.
1075         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1076         up formatting in IX86_BUILTIN_RDPID code.
1077         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1078         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1079         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1080         ix86_isa_flags2 to ix86_isa_flags.
1081         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1082         ix86_isa_flags2.
1083         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1084         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1085         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1086         of isa_flag2.
1087         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1088         Or in OPTION_MASK_ISA_AVX512F_SET.
1089         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1090         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1091         (ix86_handle_option): Adjust for
1092         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1093         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1095 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1096             Alan Hayward  <alan.hayward@arm.com>
1097             David Sherwood  <david.sherwood@arm.com>
1099         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1100         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1101         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1102         Track polynomial offsets.
1104 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1105             Alan Hayward  <alan.hayward@arm.com>
1106             David Sherwood  <david.sherwood@arm.com>
1108         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1109         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1110         vector_alignment isn't known.
1112 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1113             Alan Hayward  <alan.hayward@arm.com>
1114             David Sherwood  <david.sherwood@arm.com>
1116         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1117         field.
1118         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1119         (vect_loop_versioning): Take the loop versioning threshold as a
1120         separate parameter.
1121         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1122         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1123         versioning_threshold.
1124         (vect_analyze_loop_2): Compute the loop versioning threshold
1125         whenever loop versioning is needed, and store it in the new
1126         field rather than combining it with the cost model threshold.
1127         (vect_transform_loop): Update call to vect_loop_versioning.
1128         Try to combine the loop versioning and cost thresholds here.
1130 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1131             Alan Hayward  <alan.hayward@arm.com>
1132             David Sherwood  <david.sherwood@arm.com>
1134         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1135         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1136         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1137         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1138         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1139         (group_compare_offset): Update accordingly.
1140         (split_small_address_groups_p): Likewise.
1141         (record_use): Take addr_offset as a poly_uint64 rather than
1142         an unsigned HOST_WIDE_INT.
1143         (strip_offset): Return the offset as a poly_uint64 rather than
1144         an unsigned HOST_WIDE_INT.
1145         (record_group_use, split_address_groups): Track polynomial offsets.
1146         (add_iv_candidate_for_use): Likewise.
1147         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1148         than a HOST_WIDE_INT.
1149         (strip_offset_1): Return the offset as a poly_int64 rather than
1150         a HOST_WIDE_INT.
1152 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1153             Alan Hayward  <alan.hayward@arm.com>
1154             David Sherwood  <david.sherwood@arm.com>
1156         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1157         rather than a HOST_WIDE_INT.
1158         * tree.c (get_binfo_at_offset): Likewise.
1160 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1161             Alan Hayward  <alan.hayward@arm.com>
1162             David Sherwood  <david.sherwood@arm.com>
1164         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1165         rather than a HOST_WIDE_INT.
1166         * tree-sra.c (build_ref_for_offset): Likewise.
1168 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1169             Alan Hayward  <alan.hayward@arm.com>
1170             David Sherwood  <david.sherwood@arm.com>
1172         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1173         than an offset_int.
1174         * tree.c (mem_ref_offset): Likewise.
1175         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1176         * builtins.c (get_object_alignment_2): Likewise.
1177         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1178         * gimple-fold.c (get_base_constructor): Likewise.
1179         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1180         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1181         Likewise.
1182         * ipa-polymorphic-call.c
1183         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1184         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1185         (get_ancestor_addr_info): Likewise.
1186         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1187         * match.pd: Likewise.
1188         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1189         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1190         * tree-eh.c (tree_could_trap_p): Likewise.
1191         * tree-object-size.c (addr_object_size): Likewise.
1192         * tree-ssa-address.c (copy_ref_info): Likewise.
1193         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1194         (indirect_refs_may_alias_p): Likewise.
1195         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1196         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1197         (non_rewritable_mem_ref_base): Likewise.
1198         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1199         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1200         * varasm.c (decode_addr_const): Likewise.
1202 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1203             Alan Hayward  <alan.hayward@arm.com>
1204             David Sherwood  <david.sherwood@arm.com>
1206         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1207         offsets for MEM_REFs.
1209 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1210             Alan Hayward  <alan.hayward@arm.com>
1211             David Sherwood  <david.sherwood@arm.com>
1213         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1214         as a poly_uint64 rather than an unsigned int.
1215         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1217 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1218             Alan Hayward  <alan.hayward@arm.com>
1219             David Sherwood  <david.sherwood@arm.com>
1221         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1222         to poly_int64.
1223         (decode_addr_const): Update accordingly.
1225 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1226             Alan Hayward  <alan.hayward@arm.com>
1227             David Sherwood  <david.sherwood@arm.com>
1229         * tree.h (bit_field_size, bit_field_offset): New functions.
1230         * hsa-gen.c (gen_hsa_addr): Use them.
1231         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1232         (simplify_vector_constructor): Likewise.
1233         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1234         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1235         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1236         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1237         tree_fits_uhwi_p.
1239 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1240             Alan Hayward  <alan.hayward@arm.com>
1241             David Sherwood  <david.sherwood@arm.com>
1243         * expr.h (emit_group_load, emit_group_load_into_temps)
1244         (emit_group_store): Take the size as a poly_int64 rather than an int.
1245         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1246         (emit_group_load_into_temp, emit_group_store): Likewise.
1248 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1249             Alan Hayward  <alan.hayward@arm.com>
1250             David Sherwood  <david.sherwood@arm.com>
1252         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1253         an unsigned int to a poly_uint64.
1254         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1255         sizes as poly_uint64s rather than unsigned ints.
1256         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1257         Likewise.
1259 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1260             Alan Hayward  <alan.hayward@arm.com>
1261             David Sherwood  <david.sherwood@arm.com>
1263         * lra-constraints.c (emit_inc): Change inc_amount from an int
1264         to a poly_int64.
1266 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1267             Alan Hayward  <alan.hayward@arm.com>
1268             David Sherwood  <david.sherwood@arm.com>
1270         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1271         to a poly_uint64.
1272         (add_stack_var, stack_var_cmp, partition_stack_vars)
1273         (dump_stack_var_partition): Update accordingly.
1274         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1275         than a HOST_WIDE_INT.
1276         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1277         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1278         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1279         rather than a HOST_WIDE_INT.
1281 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1282             Alan Hayward  <alan.hayward@arm.com>
1283             David Sherwood  <david.sherwood@arm.com>
1285         * target.def (return_pops_args): Treat both the input and output
1286         sizes as poly_int64s rather than HOST_WIDE_INTS.
1287         * targhooks.h (default_return_pops_args): Update accordingly.
1288         * targhooks.c (default_return_pops_args): Likewise.
1289         * doc/tm.texi: Regenerate.
1290         * emit-rtl.h (incoming_args): Change pops_args, size and
1291         outgoing_args_size from int to poly_int64_pod.
1292         * function.h (expr_status): Change x_pending_stack_adjust and
1293         x_stack_pointer_delta from int to poly_int64.
1294         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1295         (ARGS_SIZE_RTX): Update accordingly.
1296         * calls.c (highest_outgoing_arg_in_use): Change from int to
1297         unsigned int.
1298         (stack_usage_watermark, stored_args_watermark): New variables.
1299         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1300         (emit_call_1): Change the stack_size and rounded_stack_size
1301         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1302         as a poly_int64.
1303         (save_fixed_argument_area): Check stack_usage_watermark.
1304         (initialize_argument_information): Change old_pending_adj from
1305         a HOST_WIDE_INT * to a poly_int64_pod *.
1306         (compute_argument_block_size): Return the size as a poly_int64
1307         rather than an int.
1308         (finalize_must_preallocate): Track polynomial argument sizes.
1309         (compute_argument_addresses): Likewise.
1310         (internal_arg_pointer_based_exp): Track polynomial offsets.
1311         (mem_overlaps_already_clobbered_arg_p): Rename to...
1312         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1313         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1314         Check stored_args_used_watermark.
1315         (load_register_parameters): Update accordingly.
1316         (check_sibcall_argument_overlap_1): Likewise.
1317         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1318         args size as a poly_int64 rather than an int.  Return a bool
1319         indicating whether the optimization was possible and return
1320         the new adjustment by reference.
1321         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1322         Update stored_args_watermark.
1323         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1324         (expand_call): Likewise.  Maintain stack_usage_watermark and
1325         stored_args_watermark.  Update calls to
1326         combine_pending_stack_adjustment_and_call.
1327         (emit_library_call_value_1): Handle polynomial argument sizes.
1328         Call stack_region_maybe_used_p and mark_stack_region_used.
1329         Maintain stack_usage_watermark.
1330         (store_one_arg): Likewise.  Update call to
1331         mem_overlaps_already_clobbered_arg_p.
1332         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1333         HOST_WIDE_INT.
1334         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1335         * config/microblaze/microblaze.c (microblaze_function_prologue):
1336         Likewise.
1337         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1338         TARGET_RETURN_POPS_ARGS interface.
1339         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1340         to HOST_WIDE_INT.
1341         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1342         * config/i386/i386.c (ix86_return_pops_args): Update for new
1343         TARGET_RETURN_POPS_ARGS interface.
1344         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1345         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1346         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1347         TARGET_RETURN_POPS_ARGS interface.
1348         * config/vax/vax.c (vax_return_pops_args): Likewise.
1349         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1350         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1351         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1352         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1353         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1354         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1355         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1356         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1357         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1358         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1359         and x_stack_pointer_delta from int to poly_int64.
1360         * dojump.c (do_pending_stack_adjust): Update accordingly.
1361         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1362         stack_pointer_deltas.
1363         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1364         (pad_to_arg_alignment): Track polynomial offsets.
1365         (assign_parm_find_stack_rtl): Likewise.
1366         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1367         * toplev.c (output_stack_usage): Update reference to
1368         current_function_pushed_stack_size.
1370 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1371             Alan Hayward  <alan.hayward@arm.com>
1372             David Sherwood  <david.sherwood@arm.com>
1374         * function.c (in_arg_offset, var_offset, dynamic_offset)
1375         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1376         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1377         rather than a HOST_WIDE_INT.
1378         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1379         (instantiate_virtual_regs_in_insn): Likewise.
1381 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1382             Alan Hayward  <alan.hayward@arm.com>
1383             David Sherwood  <david.sherwood@arm.com>
1385         * rtl.h (get_args_size, add_args_size_note): New functions.
1386         (find_args_size_adjust): Return a poly_int64 rather than a
1387         HOST_WIDE_INT.
1388         (fixup_args_size_notes): Likewise.  Make the same change to the
1389         end_args_size parameter.
1390         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1391         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1392         * calls.c (emit_call_1): Likewise.
1393         * explow.c (adjust_stack_1): Likewise.
1394         * cfgcleanup.c (old_insns_match_p): Update use of
1395         find_args_size_adjust.
1396         * combine.c (distribute_notes): Track polynomial arg sizes.
1397         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1398         end_true_args_size, beg_delay_args_size and end_delay_args_size
1399         from HOST_WIDE_INT to poly_int64.
1400         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1401         than a HOST_WIDE_INT.
1402         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1403         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1404         polynomial arg sizes.
1405         * emit-rtl.c (try_split): Use get_args_size.
1406         * recog.c (peep2_attempt): Likewise.
1407         * reload1.c (reload_as_needed): Likewise.
1408         * expr.c (find_args_size_adjust): Return the adjustment as a
1409         poly_int64 rather than a HOST_WIDE_INT.
1410         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1411         to a poly_int64 and change the return type in the same way.
1412         (emit_single_push_insn): Track polynomial arg sizes.
1414 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1415             Alan Hayward  <alan.hayward@arm.com>
1416             David Sherwood  <david.sherwood@arm.com>
1418         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1419         from HOST_WIDE_INT to poly_int64.
1420         * expr.c (push_block, emit_push_insn): Likewise.
1422 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1423             Alan Hayward  <alan.hayward@arm.com>
1424             David Sherwood  <david.sherwood@arm.com>
1426         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1427         to poly_int64.
1428         (get_frame_size): Return the size as a poly_int64 rather than a
1429         HOST_WIDE_INT.
1430         (frame_offset_overflow): Take the offset as a poly_int64 rather
1431         than a HOST_WIDE_INT.
1432         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1433         (assign_stack_temp): Likewise for the size.
1434         * function.c (get_frame_size): Return a poly_int64 rather than
1435         a HOST_WIDE_INT.
1436         (frame_offset_overflow): Take the offset as a poly_int64 rather
1437         than a HOST_WIDE_INT.
1438         (try_fit_stack_local): Take the start, length and size as poly_int64s
1439         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1440         rather than a HOST_WIDE_INT.
1441         (add_frame_space): Take the start and end as poly_int64s rather than
1442         HOST_WIDE_INTs.
1443         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1444         (assign_stack_temp): Likewise for the size.
1445         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1446         to poly_int64.
1447         (find_temp_slot_from_address): Handle polynomial offsets.
1448         (combine_temp_slots): Likewise.
1449         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1450         to poly_int64.
1451         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1452         poly_int64 rather than a HOST_WIDE_INT.
1453         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1454         than a HOST_WIDE_INT.
1455         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1456         polynomial frame offsets.
1457         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1458         as a poly_int64 rather than an int.
1459         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1460         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1461         * config/v850/v850.c (compute_frame_size): Likewise.
1462         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1463         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1464         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1465         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1466         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1467         rather than a HOST_WIDE_INT.
1468         * explow.c (get_dynamic_stack_base): Likewise.
1469         * final.c (final_start_function): Use the constant lower bound
1470         of the frame size for -Wframe-larger-than.
1471         * ira.c (do_reload): Adjust for new get_frame_size return type.
1472         * lra.c (lra): Likewise.
1473         * reload1.c (reload): Likewise.
1474         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1475         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1476         * rtlanal.c (get_initial_register_offset): Return the offset as
1477         a poly_int64 rather than a HOST_WIDE_INT.
1479 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1480             Alan Hayward  <alan.hayward@arm.com>
1481             David Sherwood  <david.sherwood@arm.com>
1483         * reload1.c (elim_table): Change initial_offset, offset and
1484         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1485         (offsets_at): Change the target array's element type from
1486         HOST_WIDE_INT to poly_int64_pod.
1487         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1488         (elimination_costs_in_insn, update_eliminable_offsets)
1489         (verify_initial_elim_offsets, set_offsets_for_label)
1490         (init_eliminable_invariants): Update after above changes.
1492 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1493             Alan Hayward  <alan.hayward@arm.com>
1494             David Sherwood  <david.sherwood@arm.com>
1496         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1497         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1498         (decomposition): Change start and end from HOST_WIDE_INT
1499         to poly_int64_pod.
1500         (decompose, immune_p): Update accordingly.
1501         (find_inc_amount): Return a poly_int64 rather than an int.
1502         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1503         rather than an int.
1505 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1506             Alan Hayward  <alan.hayward@arm.com>
1507             David Sherwood  <david.sherwood@arm.com>
1509         * tree.h (get_inner_reference): Return the bitsize and bitpos
1510         as poly_int64_pods rather than HOST_WIDE_INT.
1511         * fold-const.h (ptr_difference_const): Return the pointer difference
1512         as a poly_int64_pod rather than a HOST_WIDE_INT.
1513         * expr.c (get_inner_reference): Return the bitsize and bitpos
1514         as poly_int64_pods rather than HOST_WIDE_INT.
1515         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1516         offsets and sizes.
1517         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1518         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1519         (optimize_bit_field_compare): Update call to get_inner_reference.
1520         (decode_field_reference): Likewise.
1521         (fold_unary_loc): Track polynomial offsets and sizes.
1522         (split_address_to_core_and_offset): Return the bitpos as a
1523         poly_int64_pod rather than a HOST_WIDE_INT.
1524         (ptr_difference_const): Likewise for the pointer difference.
1525         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1526         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1527         * dbxout.c (dbxout_expand_expr): Likewise.
1528         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1529         (loc_list_from_tree_1, fortran_common): Likewise.
1530         * gimple-laddress.c (pass_laddress::execute): Likewise.
1531         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1532         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1533         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1534         * tree-affine.c (tree_to_aff_combination): Likewise.
1535         (get_inner_reference_aff): Likewise.
1536         * tree-data-ref.c (split_constant_offset_1): Likewise.
1537         (dr_analyze_innermost): Likewise.
1538         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1539         * tree-sra.c (ipa_sra_check_caller): Likewise.
1540         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1541         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1542         (instrument_bool_enum_load, instrument_object_size): Likewise.
1543         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1544         to get_inner_reference.
1545         * hsa-gen.c (gen_hsa_addr): Likewise.
1546         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1547         * tsan.c (instrument_expr): Likewise.
1548         * match.pd: Update call to ptr_difference_const.
1550 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1551             Alan Hayward  <alan.hayward@arm.com>
1552             David Sherwood  <david.sherwood@arm.com>
1554         * fold-const.c (fold_comparison): Track sizes and offsets as
1555         poly_int64s rather than HOST_WIDE_INTs when folding address
1556         comparisons.
1558 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1559             Alan Hayward  <alan.hayward@arm.com>
1560             David Sherwood  <david.sherwood@arm.com>
1562         * expr.h (get_bit_range): Return the bitstart and bitend as
1563         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1564         as a poly_int64 rather than a HOST_WIDE_INT.
1565         * expr.c (get_bit_range): Likewise.
1566         (expand_assignment): Update call accordingly.
1567         * fold-const.c (optimize_bit_field_compare): Likewise.
1569 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1570             Alan Hayward  <alan.hayward@arm.com>
1571             David Sherwood  <david.sherwood@arm.com>
1573         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1574         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1575         as an optional final parameter.
1576         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1577         (aarch64_legitimate_address_p): Likewise.
1578         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1579         instead of an rtx code.
1580         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1581         (aarch64_legitimate_address_hook_p): Likewise.
1582         (aarch64_print_ldpstp_address): Likewise.
1583         (aarch64_print_operand_address): Likewise.
1584         (aarch64_address_cost): Likewise.
1585         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1586         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1587         (aarch64_mem_pair_lanes_operand): Likewise.
1589 2017-12-20  Richard Biener  <rguenther@suse.de>
1591         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1592         update_stmt_if_modified.
1594 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1596         PR tree-optimization/83491
1597         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1598         before walking uses.  Improve coding style and comments.
1600 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1602         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1604 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1606         PR ipa/83506
1607         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1608         todo_flags_finish.
1609         (pass_ipa_free_fn_summary): Add small_p private data member,
1610         initialize to false in the ctor.
1611         (pass_ipa_free_fn_summary::clone,
1612         pass_ipa_free_fn_summary::set_pass_param,
1613         pass_ipa_free_fn_summary::gate): New methods.
1614         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1615         | TODO_dump_symtab if small_p.
1616         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1617         entry and add another instance of the pass with false parm after
1618         ipa-pure-const.
1619         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1620         ipa_free_fn_summary here.
1622 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1624         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1626 2017-12-20  Martin Sebor  <msebor@redhat.com>
1628         PR testsuite/83131
1629         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1630         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1631         (expand_builtin_strncmp): Same.
1633 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1635         PR bootstrap/83396
1636         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1637         allowed debug stmts before labels.
1638         (expand_gimple_basic_block): Likewise.
1639         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1640         * gimple-iterator.h (gsi_after_labels): Likewise.
1641         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1642         rename reused variable, and simplify using gsi_move_before.
1643         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1644         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1645         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1646         (gimple_verify_flow_info, gimple_block_label): Likewise.
1647         (make_blocks): Move debug markers after adjacent labels.
1648         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1649         allowed debug insns outside blocks.
1650         * df-scan.c (df_insn_delete): Likewise.
1651         * lra-constraints.c (update_ebb_live_info): Likewise.
1652         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1653         (vt_initialize, delete_vta_debug_insns): Likewise.
1654         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1656 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1657             Alan Hayward  <alan.hayward@arm.com>
1658             David Sherwood  <david.sherwood@arm.com>
1660         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1661         (round_up_to_byte_boundary): Likewise.
1662         * expr.h (get_bit_range): Add temporary shim.
1663         * gimple-ssa-store-merging.c (store_operand_info): Change the
1664         bitsize, bitpos, bitregion_start and bitregion_end fields from
1665         unsigned HOST_WIDE_INT to poly_uint64.
1666         (merged_store_group): Likewise load_align_base.
1667         (compatible_load_p, compatible_load_p): Update accordingly.
1668         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1669         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1670         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1671         bitregion_start and bitregion_end as poly_uint64s rather than
1672         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1673         (handled_load): Take the bitsize, bitpos,
1674         bitregion_start and bitregion_end as poly_uint64s rather than
1675         unsigned HOST_WIDE_INTs.
1676         (pass_store_merging::process_store): Update call to
1677         mem_valid_for_store_merging.
1679 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1680             Alan Hayward  <alan.hayward@arm.com>
1681             David Sherwood  <david.sherwood@arm.com>
1683         * builtins.c (get_object_alignment_2): Track polynomial offsets
1684         and sizes.  Update the alignment handling.
1686 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1687             Alan Hayward  <alan.hayward@arm.com>
1688             David Sherwood  <david.sherwood@arm.com>
1690         * tree.h (get_inner_reference): Add a version that returns the
1691         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1692         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1693         the case in which bitpos is not associated with the first byte.
1695 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1696             Alan Hayward  <alan.hayward@arm.com>
1697             David Sherwood  <david.sherwood@arm.com>
1699         * tree-affine.h (get_inner_reference_aff): Return the size as a
1700         poly_widest_int.
1701         * tree-affine.c (get_inner_reference_aff): Likewise.
1702         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1703         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1705 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1706             Alan Hayward  <alan.hayward@arm.com>
1707             David Sherwood  <david.sherwood@arm.com>
1709         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1710         HOST_WIDE_INT rather than a poly_int64.
1712 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1713             Alan Hayward  <alan.hayward@arm.com>
1714             David Sherwood  <david.sherwood@arm.com>
1716         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1717         HOST_WIDE_INT to poly_int64_pod.
1718         (perform_symbolic_merge): Update accordingly.
1719         (bswap_replace): Likewise.
1721 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1722             Alan Hayward  <alan.hayward@arm.com>
1723             David Sherwood  <david.sherwood@arm.com>
1725         * tree-affine.h (aff_tree::offset): Change from widest_int
1726         to poly_widest_int.
1727         (wide_int_ext_for_comb): Delete.
1728         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1729         constants as poly_widest_int rather than widest_int.
1730         (aff_combination_constant_multiple_p): Return the multiplier
1731         as a poly_widest_int.
1732         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1733         polynomial offsets.
1734         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1735         version static and add an overload for poly_widest_int.
1736         (aff_combination_const, aff_combination_add_cst)
1737         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1738         the constants as poly_widest_int rather than widest_int.
1739         (tree_to_aff_combination): Generalize INTEGER_CST case to
1740         poly_int_tree_p.
1741         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1742         (aff_combination_add_product, aff_combination_mult): Handle
1743         polynomial offsets.
1744         (aff_combination_constant_multiple_p): Return the multiplier
1745         as a poly_widest_int.
1746         * tree-predcom.c (determine_offset): Return the offset as a
1747         poly_widest_int.
1748         (split_data_refs_to_components, suitable_component_p): Update
1749         accordingly.
1750         (valid_initializer_p): Update call to
1751         aff_combination_constant_multiple_p.
1752         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1753         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1754         as a poly_int64 rather than a HOST_WIDE_INT.
1755         (get_address_cost): Handle polynomial offsets.
1756         (iv_elimination_compare_lt): Likewise.
1757         (rewrite_use_nonlinear_expr): Likewise.
1759 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1760             Alan Hayward  <alan.hayward@arm.com>
1761             David Sherwood  <david.sherwood@arm.com>
1763         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1764         as a poly_int64_pod rather than a HOST_WIDE_INT.
1765         (get_addr_base_and_unit_offset): Likewise.
1766         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1767         (get_addr_base_and_unit_offset): Likewise.
1768         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1769         to poly_int64 in example.
1770         * fold-const.c (fold_binary_loc): Update call to
1771         get_addr_base_and_unit_offset.
1772         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1773         (maybe_canonicalize_mem_ref_addr): Likewise.
1774         (gimple_fold_stmt_to_constant_1): Likewise.
1775         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1776         Likewise.
1777         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1778         * match.pd: Likewise.
1779         * omp-low.c (lower_omp_target): Likewise.
1780         * tree-sra.c (build_ref_for_offset): Likewise.
1781         (build_debug_ref_for_model): Likewise.
1782         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1783         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1784         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1785         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1786         (constant_pointer_difference): Likewise.
1787         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1788         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1789         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1790         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1791         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1792         (set_ssa_val_to): Likewise.
1793         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1794         (maybe_diag_stxncpy_trunc): Likewise.
1795         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1796         * tree.c (build_simple_mem_ref_loc): Likewise.
1797         (array_at_struct_end_p): Likewise.
1799 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1800             Alan Hayward  <alan.hayward@arm.com>
1801             David Sherwood  <david.sherwood@arm.com>
1803         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1804         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1805         (get_ref_base_and_extent_hwi): Declare.
1806         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1807         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1808         (get_ref_base_and_extent_hwi): New function.
1809         * cfgexpand.c (expand_debug_expr): Update call to
1810         get_ref_base_and_extent.
1811         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1812         * gimple-fold.c (get_base_constructor): Return the offset as a
1813         poly_int64_pod rather than a HOST_WIDE_INT.
1814         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1815         * ipa-polymorphic-call.c
1816         (ipa_polymorphic_call_context::set_by_invariant)
1817         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1818         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1819         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1820         rather than get_ref_base_and_extent.
1821         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1822         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1823         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1824         Likewise.
1825         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1826         call to get_ref_base_and_extent.
1827         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1828         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1829         (stmt_kills_ref_p): Likewise.
1830         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1831         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1832         Likewise.
1833         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1834         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1835         when calling native_encode_expr.
1836         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1837         call to get_ref_base_and_extent.
1838         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1839         get_ref_base_and_extent.
1840         * var-tracking.c (track_expr_p): Likewise.
1842 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1843             Alan Hayward  <alan.hayward@arm.com>
1844             David Sherwood  <david.sherwood@arm.com>
1846         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1847         HOST_WIDE_INT to poly_int64_pod.
1848         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1849         polynomail parameter offsets.
1851 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1852             Alan Hayward  <alan.hayward@arm.com>
1853             David Sherwood  <david.sherwood@arm.com>
1855         * gengtype.c (main): Handle poly_int64_pod.
1856         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1857         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1858         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1859         from HOST_WIDE_INT to poly_int64_pod.
1860         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1861         (copy_cfa): New function.
1862         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1863         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1864         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
1865         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
1866         (initial_return_save): Treat offsets as poly_ints.
1867         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
1868         offset is nonconstant.
1869         (reg_save): Take the offset as a poly_int64.  Fall back to
1870         DW_CFA_expression for nonconstant offsets.
1871         (queue_reg_save): Take the offset as a poly_int64.
1872         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
1874 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1875             Alan Hayward  <alan.hayward@arm.com>
1876             David Sherwood  <david.sherwood@arm.com>
1878         * rtl.h (operand_subword, operand_subword_force): Take the offset
1879         as a poly_uint64 an unsigned int.
1880         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
1882 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1883             Alan Hayward  <alan.hayward@arm.com>
1884             David Sherwood  <david.sherwood@arm.com>
1886         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
1887         'p' format code.  Use INT_LIST rather than SUBREG as the example of
1888         a code with an XINT and an XEXP.  Remove the implication that
1889         accessing an rtx field using XINT is expected to work.
1890         * rtl.def (SUBREG): Change format from "ei" to "ep".
1891         * rtl.h (rtunion::rt_subreg): New field.
1892         (XCSUBREG): New macro.
1893         (SUBREG_BYTE): Use it.
1894         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
1895         Update constructor accordingly.
1896         (subreg_shape::operator ==): Update accordingly.
1897         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
1898         than an unsigned int.
1899         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
1900         a poly_uint64 rather than an unsigned int.
1901         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
1902         than an unsigned int.
1903         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
1904         (subreg_size_highpart_offset): Return a poly_uint64 rather than
1905         an unsigned int.  Take the sizes as poly_uint64s.
1906         (subreg_offset_from_lsb): Return a poly_uint64 rather than
1907         an unsigned int.  Take the shift as a poly_uint64 rather than
1908         an unsigned int.
1909         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
1910         as a poly_uint64 rather than an unsigned int.
1911         (simplify_subreg_regno): Likewise.
1912         (byte_lowpart_offset): Return the memory offset as a poly_int64
1913         rather than an int.
1914         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1915         poly_uint64 rather than an unsigned int.
1916         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
1917         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
1918         poly_uint64 rather than an unsigned int.
1919         * rtl.c (rtx_format): Describe 'p' in comment.
1920         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
1921         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
1922         offset as a poly_uint64 rather than an unsigned int.
1923         (byte_lowpart_offset): Return the memory offset as a poly_int64
1924         rather than an int.
1925         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1926         poly_uint64 rather than an unsigned int.
1927         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
1928         mode sizes as poly_uint64s rather than unsigned ints.  Return a
1929         poly_uint64 rather than an unsigned int.
1930         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
1931         (copy_insn_1): Handle 'p'.
1932         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
1933         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
1934         an unsigned int.  Return the shift in the same way.
1935         (subreg_lsb): Return the shift as a poly_uint64 rather than an
1936         unsigned int.
1937         (subreg_size_offset_from_lsb): Take the sizes and shift as
1938         poly_uint64s rather than unsigned ints.  Return the offset as
1939         a poly_uint64.
1940         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
1941         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
1942         an unsigned int.
1943         * rtlhash.c (add_rtx): Handle 'p'.
1944         * genemit.c (gen_exp): Likewise.
1945         * gengenrtl.c (type_from_format, gendef): Likewise.
1946         * gensupport.c (subst_pattern_match, get_alternatives_number)
1947         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
1948         (subst_dup): Likewise.
1949         * gengtype.c (adjust_field_rtx_def): Likewise.
1950         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
1951         (match_pattern_2): Likewise.
1952         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
1953         (rtx_test::subreg_field): New function.
1954         (operator ==, safe_to_hoist_p, transition_parameter_type)
1955         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
1956         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
1957         * genpeep.c (match_rtx): Likewise.
1958         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
1959         (rtx_writer::print_rtx_operand): Handle 'p'.
1960         (print_value): Handle SUBREG.
1961         * read-rtl.c (apply_int_iterator): Likewise.
1962         (rtx_reader::read_rtx_operand): Handle 'p'.
1963         * alias.c (rtx_equal_for_memref_p): Likewise.
1964         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
1965         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
1966         as poly_ints.
1967         * calls.c (expand_call): Likewise.
1968         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
1969         (make_extraction, gen_lowpart_for_combine): Likewise.
1970         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
1971         Likewise.
1972         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
1973         rather than an unsigned int.  Treat subreg offsets as poly_ints.
1974         (exp_equiv_p): Handle 'p'.
1975         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
1976         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
1977         * dse.c (find_shift_sequence): Likewise.
1978         * dwarf2out.c (rtl_for_decl_location): Likewise.
1979         * expmed.c (extract_low_bits): Likewise.
1980         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
1981         (expand_expr_real_2): Likewise.
1982         * final.c (alter_subreg): Likewise.
1983         (leaf_renumber_regs_insn): Handle 'p'.
1984         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
1985         Treat subreg offsets as poly_ints.
1986         * fwprop.c (forward_propagate_and_simplify): Likewise.
1987         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
1988         * ira.c (get_subreg_tracking_sizes): Likewise.
1989         * ira-conflicts.c (go_through_subreg): Likewise.
1990         * ira-lives.c (process_single_reg_class_operands): Likewise.
1991         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
1992         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
1993         as a poly_uint64 rather than an unsigned int.
1994         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
1995         subreg offsets as poly_ints.
1996         * lra-constraints.c (operands_match_p): Handle 'p'.
1997         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
1998         * lra-spills.c (assign_mem_slot): Likewise.
1999         * postreload.c (move2add_valid_value_p): Likewise.
2000         * recog.c (general_operand, indirect_operand): Likewise.
2001         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2002         (copyprop_hardreg_forward_1): Likewise.
2003         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2004         (record_subregs_of_mode): Likewise.
2005         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2006         * reload.c (operands_match_p): Handle 'p'.
2007         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2008         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2009         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2010         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2011         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2012         (simplify_const_poly_int_tests<N>::run): Likewise.
2013         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2014         a poly_uint64 rather than an unsigned int.
2015         * valtrack.c (debug_lowpart_subreg): Likewise.
2016         * var-tracking.c (var_lowpart): Likewise.
2017         (loc_cmp): Handle 'p'.
2019 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2020             Alan Hayward  <alan.hayward@arm.com>
2021             David Sherwood  <david.sherwood@arm.com>
2023         * ira.c (get_subreg_tracking_sizes): New function.
2024         (init_live_subregs): Take an integer size rather than a register.
2025         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2026         to init_live_subregs.
2028 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2029             Alan Hayward  <alan.hayward@arm.com>
2030             David Sherwood  <david.sherwood@arm.com>
2032         * expr.c (store_constructor_field): Change bitsize from a
2033         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2034         HOST_WIDE_INT to a poly_int64.
2035         (store_constructor): Change size from a HOST_WIDE_INT to
2036         a poly_int64.
2037         (store_field): Likewise bitsize and bitpos.
2039 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2040             Alan Hayward  <alan.hayward@arm.com>
2041             David Sherwood  <david.sherwood@arm.com>
2043         * expmed.h (store_bit_field): Change bitregion_start and
2044         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2045         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2046         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2047         (store_fixed_bit_field, store_split_bit_field): Likewise.
2048         * expr.c (store_constructor_field, store_field): Likewise.
2049         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2050         to bitsize and bitpos.
2051         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2052         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2053         same change in the constructor arguments.
2054         (get_best_mode): Change bitregion_start and bitregion_end from
2055         unsigned HOST_WIDE_INT to poly_uint64.
2056         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2057         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2058         poly_int64.
2059         (bit_field_mode_iterator::next_mode): Update for new types
2060         of m_bitregion_start and m_bitregion_end.
2061         (get_best_mode): Change bitregion_start and bitregion_end from
2062         unsigned HOST_WIDE_INT to poly_uint64.
2064 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2065             Alan Hayward  <alan.hayward@arm.com>
2066             David Sherwood  <david.sherwood@arm.com>
2068         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2069         accepts poly_uint64 offsets.
2070         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2071         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2072         * expmed.c (lowpart_bit_field_p): Likewise.
2073         (extract_bit_field_as_subreg): New function, split out from...
2074         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2075         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2076         extractions, check that BITSIZE matches the size of the extracted
2077         value and that BITNUM is an exact multiple of that size.
2078         If all else fails, try forcing the value into memory if
2079         BITNUM is variable, and adjusting the address so that the
2080         offset is constant.  Split the part that can only handle constant
2081         bitsize and bitnum out into...
2082         (extract_integral_bit_field): ...this new function.
2083         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2084         rather than unsigned HOST_WIDE_INTs.
2086 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2087             Alan Hayward  <alan.hayward@arm.com>
2088             David Sherwood  <david.sherwood@arm.com>
2090         * expmed.h (store_bit_field): Take bitsize and bitnum as
2091         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2092         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2093         that returns the byte size.
2094         (store_bit_field_1): Take bitsize and bitnum as
2095         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2096         to simple_mem_bitfield_p.  Split the part that can only handle
2097         constant bitsize and bitnum out into...
2098         (store_integral_bit_field): ...this new function.
2099         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2100         than unsigned HOST_WIDE_INTs.
2101         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2103 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2104             Alan Hayward  <alan.hayward@arm.com>
2105             David Sherwood  <david.sherwood@arm.com>
2107         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2108         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2109         to poly_int64.
2110         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2111         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2112         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2113         offset as a poly_int64 rather than an int.
2114         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2115         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2116         * lra-constraints.c (equiv_address_substitution): Track offsets
2117         as poly_int64s.
2118         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2119         (curr_insn_transform): Handle the new form of sp_offset.
2120         * lra-eliminations.c (lra_elim_table): Change previous_offset
2121         and offset from HOST_WIDE_INT to poly_int64.
2122         (print_elim_table, update_reg_eliminate): Update accordingly.
2123         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2124         (get_elimination): Update accordingly.
2125         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2126         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2127         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2128         poly_int64 offsets generally.
2129         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2130         (mark_not_eliminable, init_elimination): Update accordingly.
2131         (remove_reg_equal_offset_note): Return a bool and pass the new
2132         offset back by pointer as a poly_int64.
2133         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2134         rather than a HOST_WIDE_INT.
2135         (do_remat): Track offsets poly_int64s.
2136         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2138 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2139             Alan Hayward  <alan.hayward@arm.com>
2140             David Sherwood  <david.sherwood@arm.com>
2142         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2143         offset from HOST_WIDE_INT to poly_int64.
2144         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2145         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2146         (widen_memory_access): Take the sizes and offsets as poly_int64s
2147         rather than HOST_WIDE_INTs.
2148         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2149         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2150         and ints.
2151         (adjust_offset_for_component_ref): Change the offset from a
2152         HOST_WIDE_INT to a poly_int64.
2153         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2154         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2155         * dce.c (find_call_stack_args): Likewise.
2156         * dse.c (record_store): Likewise.
2157         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2158         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2159         * read-rtl-function.c (test_loading_mem): Likewise.
2160         * rtlanal.c (may_trap_p_1): Likewise.
2161         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2162         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2163         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2164         (mem_attrs::mem_attrs): New function.
2165         (set_mem_attributes_minus_bitpos): Change bitpos from a
2166         HOST_WIDE_INT to poly_int64.
2167         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2168         (clear_mem_offset, clear_mem_size, change_address)
2169         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2170         initialize mem_attrs.
2171         (set_mem_offset, set_mem_size, adjust_address_1)
2172         (adjust_automodify_address_1, offset_address, widen_memory_access):
2173         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2175 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2176             Alan Hayward  <alan.hayward@arm.com>
2177             David Sherwood  <david.sherwood@arm.com>
2179         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2180         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2181         rather than 0 to represent an unknown size.  Assert that the size
2182         is known when the mode isn't BLKmode.
2183         (may_trap_p_1): Use -1 for unknown sizes.
2184         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2186 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2187             Alan Hayward  <alan.hayward@arm.com>
2188             David Sherwood  <david.sherwood@arm.com>
2190         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2191         to poly_int64.  Update commentary for positions_needed.large.
2192         (read_info_type): Change offset and width from HOST_WIDE_INT
2193         to poly_int64.
2194         (set_usage_bits): Likewise.
2195         (canon_address): Return the offset as a poly_int64 rather than
2196         a HOST_WIDE_INT.  Use strip_offset_and_add.
2197         (set_all_positions_unneeded, any_positions_needed_p): Use
2198         positions_needed.large to track stores with non-constant widths.
2199         (all_positions_needed_p): Likewise.  Take the offset and width
2200         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2201         (record_store): Cope with non-constant offsets and widths.
2202         Nullify the rhs of an earlier store if we can't tell which bytes
2203         of it are needed.
2204         (find_shift_sequence): Take the access_size and shift as poly_int64s
2205         rather than ints.
2206         (get_stored_val): Take the read_offset and read_width as poly_int64s
2207         rather than HOST_WIDE_INTs.
2208         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2209         non-constant offsets and widths.
2211 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2212             Alan Hayward  <alan.hayward@arm.com>
2213             David Sherwood  <david.sherwood@arm.com>
2215         * inchash.h (inchash::hash::add_poly_int): New function.
2216         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2217         Use poly_int64 rather than HOST_WIDE_INT.
2218         (ao_ref::max_size_known_p): New function.
2219         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2220         rather than HOST_WIDE_INT.
2221         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2222         to temporaries until its interface is adjusted to match.
2223         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2224         (aliasing_component_refs_p, decl_refs_may_alias_p)
2225         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2226         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2227         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2228         ao_ref fields.
2229         * alias.c (ao_ref_from_mem): Likewise.
2230         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2231         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2232         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2233         (maybe_trim_complex_store, maybe_trim_constructor_store)
2234         (live_bytes_read, dse_classify_store): Likewise.
2235         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2236         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2237         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2238         (vn_reference_lookup_3): Likewise.
2239         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2241 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2242             Alan Hayward  <alan.hayward@arm.com>
2243             David Sherwood  <david.sherwood@arm.com>
2245         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2246         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2247         instead of ranges_overlap_p.
2249 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2250             Alan Hayward  <alan.hayward@arm.com>
2251             David Sherwood  <david.sherwood@arm.com>
2253         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2254         sizes as poly_int64s rather than HOST_WIDE_INTs.
2256 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2257             Alan Hayward  <alan.hayward@arm.com>
2258             David Sherwood  <david.sherwood@arm.com>
2260         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2261         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2262         * gimple-fold.c (fold_ctor_reference): Likewise.
2264 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2265             Alan Hayward  <alan.hayward@arm.com>
2266             David Sherwood  <david.sherwood@arm.com>
2268         * target.def (dwarf_poly_indeterminate_value): New hook.
2269         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2270         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2271         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2272         * doc/tm.texi: Regenerate.
2273         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2274         offset as a poly_int64.
2275         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2276         offset as a poly_int64.
2277         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2278         Take the offset as a poly_int64.
2279         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2280         (frame_pointer_fb_offset): Change to a poly_int64.
2281         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2282         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2283         (based_loc_descr): Take the offset as a poly_int64.
2284         Use strip_offset_and_add to handle (plus X (const)).
2285         Use new_reg_loc_descr instead of an open-coded version of the
2286         previous implementation.
2287         (mem_loc_descriptor): Handle CONST_POLY_INT.
2288         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2289         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2291 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2292             Alan Hayward  <alan.hayward@arm.com>
2293             David Sherwood  <david.sherwood@arm.com>
2295         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2296         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2297         * inchash.h (inchash::hash::add_poly_hwi): New function.
2298         * gengtype.c (main): Register poly_int64.
2299         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2300         offset as a poly_int.
2301         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2302         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2303         offset as a poly_int64.
2304         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2305         * print-rtl.c (print_poly_int): New function.
2306         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2307         a poly_int.
2308         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2309         functions.
2310         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2311         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2312         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2313         Enforce track_offset_p for parts of a PARALLEL.
2314         (vt_add_function_parameter): Use const_offset for the final
2315         offset to track.  Use get_tracked_reg_offset for the parts
2316         of a PARALLEL.
2318 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2319             Alan Hayward  <alan.hayward@arm.com>
2320             David Sherwood  <david.sherwood@arm.com>
2322         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2323         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2324         * doc/tm.texi: Regenerate.
2325         * hooks.h (hook_bool_uint_uint_true): Delete.
2326         (hook_bool_puint64_puint64_true): Declare.
2327         * hooks.c (hook_bool_uint_uint_true): Delete.
2328         (hook_bool_puint64_puint64_true): New function.
2329         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2330         instead of unsigned ints.
2331         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2332         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2334 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2335             Alan Hayward  <alan.hayward@arm.com>
2336             David Sherwood  <david.sherwood@arm.com>
2338         * optabs.h (expand_operand): Add an int_value field.
2339         (create_expand_operand): Add an int_value parameter and use it
2340         to initialize the new expand_operand field.
2341         (create_integer_operand): Replace with a declaration of a function
2342         that accepts poly_int64s.  Move the implementation to...
2343         * optabs.c (create_integer_operand): ...here.
2344         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2345         the mode preserves the value of int_value, instead of calling
2346         const_int_operand on the rtx.  Use gen_int_mode to generate
2347         the new rtx.
2349 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2350             Alan Hayward  <alan.hayward@arm.com>
2351             David Sherwood  <david.sherwood@arm.com>
2353         * dumpfile.h (dump_dec): Declare.
2354         * dumpfile.c (dump_dec): New function.
2355         * pretty-print.h (pp_wide_integer): Turn into a function and
2356         declare a poly_int version.
2357         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2359 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2360             Alan Hayward  <alan.hayward@arm.com>
2361             David Sherwood  <david.sherwood@arm.com>
2363         * doc/generic.texi (POLY_INT_CST): Document.
2364         * tree.def (POLY_INT_CST): New tree code.
2365         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2366         * tree-core.h (tree_poly_int_cst): New struct.
2367         (tree_node): Add a poly_int_cst field.
2368         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2369         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2370         instead of a wide_int_ref.
2371         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2372         of a HOST_WIDE_INT.
2373         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2374         instead of an unsigned HOST_WIDE_INT.
2375         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2376         (ptrdiff_tree_p): Declare.
2377         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2378         extern inline implementations if the target doesn't use POLY_INT_CST.
2379         (poly_int_tree_p): New function.
2380         (wi::unextended_tree): New class.
2381         (wi::int_traits <unextended_tree>): New override.
2382         (wi::extended_tree): Add a default constructor.
2383         (wi::extended_tree::get_tree): New function.
2384         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2385         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2386         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2387         (wi::tree_to_poly_wide_ref): New typedefs.
2388         (wi::ints_for): Provide overloads for extended_tree and
2389         unextended_tree.
2390         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2391         (wi::to_wide): New functions.
2392         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2393         * tree.c (poly_int_cst_hasher): New struct.
2394         (poly_int_cst_hash_table): New variable.
2395         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2396         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2397         POLY_INT_CST.
2398         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2399         (init_ttree): Initialize poly_int_cst_hash_table.
2400         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2401         a poly_int64 instead of a HOST_WIDE_INT.
2402         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2403         instead of an unsigned HOST_WIDE_INT.
2404         (wide_int_to_tree): Rename to...
2405         (wide_int_to_tree_1): ...this.
2406         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2407         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2408         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2409         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2410         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2411         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2412         TS_POLY_INT_CST.
2413         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2414         (streamer_read_tree_body): Likewise.
2415         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2416         (streamer_write_tree_body): Likewise.
2417         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2418         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2419         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2420         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2421         * gimple-expr.h (is_gimple_constant): Likewise.
2422         * gimplify.c (maybe_with_size_expr): Likewise.
2423         * print-tree.c (print_node): Likewise.
2424         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2425         * tree-pretty-print.c (dump_generic_node): Likewise.
2426         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2427         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2428         * tree-vrp.c (compare_values_warnv): Likewise.
2429         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2430         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2431         (force_expr_to_var_cost): Likewise.
2432         * tree-ssa-loop.c (for_each_index): Likewise.
2433         * fold-const.h (build_invariant_address, size_int_kind): Take a
2434         poly_int64 instead of a HOST_WIDE_INT.
2435         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2436         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2437         POLY_INT_CST.
2438         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2439         (int_const_binop_2): New function, split out from...
2440         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2441         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2442         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2443         * gimple-ssa-strength-reduction.c (slsr_process_add)
2444         (slsr_process_mul): Check for INTEGER_CSTs before using them
2445         as candidates.
2446         * stor-layout.c (bits_from_bytes): New function.
2447         (bit_from_pos): Use it.
2448         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2449         by BITS_PER_UNIT to get the TYPE_SIZE.
2450         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2451         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2453 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2454             Alan Hayward  <alan.hayward@arm.com>
2455             David Sherwood  <david.sherwood@arm.com>
2457         * doc/rtl.texi (const_poly_int): Document.  Also document the
2458         rtl sharing behavior.
2459         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2460         * rtl.h (const_poly_int_def): New struct.
2461         (rtx_def::u): Add a cpi field.
2462         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2463         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2464         (wi::rtx_to_poly_wide_ref): New typedef
2465         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2466         (poly_int_rtx_p): New functions.
2467         (trunc_int_for_mode): Declare a poly_int64 version.
2468         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2469         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2470         a wide_int_ref.
2471         (strip_offset): Declare.
2472         (strip_offset_and_add): New function.
2473         * rtl.def (CONST_POLY_INT): New rtx code.
2474         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2475         (shared_const_p): Use poly_int_rtx_p.
2476         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2477         HOST_WIDE_INT.
2478         (gen_int_shift_amount): Likewise.
2479         * emit-rtl.c (const_poly_int_hasher): New class.
2480         (const_poly_int_htab): New variable.
2481         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2482         (const_poly_int_hasher::hash): New function.
2483         (const_poly_int_hasher::equal): Likewise.
2484         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2485         (immed_wide_int_const): Rename to...
2486         (immed_wide_int_const_1): ...this and make static.
2487         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2488         instead of a wide_int_ref.
2489         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2490         (gen_lowpart_common): Handle CONST_POLY_INT.
2491         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2492         * cselib.c (cselib_hash_rtx): Likewise.
2493         * dwarf2out.c (const_ok_for_output_1): Likewise.
2494         * expr.c (convert_modes): Likewise.
2495         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2496         * rtlhash.c (add_rtx): Likewise.
2497         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2498         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2499         Handle existing CONST_POLY_INT rtxes.
2500         * expmed.h (expand_shift): Take a poly_int64 instead of a
2501         HOST_WIDE_INT.
2502         * expmed.c (expand_shift): Likewise.
2503         * rtlanal.c (strip_offset): New function.
2504         (commutative_operand_precedence): Give CONST_POLY_INT the same
2505         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2506         and CONST_INT.
2507         * rtl-tests.c (const_poly_int_tests): New struct.
2508         (rtl_tests_c_tests): Use it.
2509         * simplify-rtx.c (simplify_const_unary_operation): Handle
2510         CONST_POLY_INT.
2511         (simplify_const_binary_operation): Likewise.
2512         (simplify_binary_operation_1): Fold additions of symbolic constants
2513         and CONST_POLY_INTs.
2514         (simplify_subreg): Handle extensions and truncations of
2515         CONST_POLY_INTs.
2516         (simplify_const_poly_int_tests): New struct.
2517         (simplify_rtx_c_tests): Use it.
2518         * wide-int.h (storage_ref): Add default constructor.
2519         (wide_int_ref_storage): Likewise.
2520         (trailing_wide_ints): Use GTY((user)).
2521         (trailing_wide_ints::operator[]): Add a const version.
2522         (trailing_wide_ints::get_precision): New function.
2523         (trailing_wide_ints::extra_size): Likewise.
2525 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2526             Alan Hayward  <alan.hayward@arm.com>
2527             David Sherwood  <david.sherwood@arm.com>
2529         * emit-rtl.h (gen_int_shift_amount): Declare.
2530         * emit-rtl.c (gen_int_shift_amount): New function.
2531         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2532         instead of GEN_INT.
2533         * calls.c (shift_return_value): Likewise.
2534         * cse.c (fold_rtx): Likewise.
2535         * dse.c (find_shift_sequence): Likewise.
2536         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2537         (expand_shift, expand_smod_pow2): Likewise.
2538         * lower-subreg.c (shift_cost): Likewise.
2539         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2540         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2541         (expand_vec_perm_var): Likewise.
2542         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2543         (simplify_binary_operation_1): Likewise.
2544         * combine.c (try_combine, find_split_point, force_int_to_mode)
2545         (simplify_shift_const_1, simplify_shift_const): Likewise.
2546         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2548 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2550         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2552 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2554         * doc/invoke.texi (ARM Options): Document accepted extension options
2555         for -march=armv8.3-a.
2557 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2559         PR target/83105
2560         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2561         or --with-float=softfp, set the default CPU to arm10e.
2563 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2565         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2566         * config/visium/predicates.md (const_shift_operand): Likewise.
2567         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2568         (visium_legitimize_reload_address): Likewise.
2570 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2572         PR target/82975
2573         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2575 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2577         PR c++/83490
2578         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2580 2017-12-20  Julia Koval  <julia.koval@intel.com>
2582         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2583         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2584         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2585         * config.gcc: Include vpclmulqdqintrin.h.
2586         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2587         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2588         -mvpclmulqdq.
2589         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2590         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2591         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2592         * config/i386/i386.c (isa2_opts): Add -mcx16.
2593         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2594         (ix86_option_override_internal): Move mcx16 to flags2.
2595         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2596         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2597         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2598         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2599         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2600         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2601         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2602         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2603         * doc/invoke.texi: Add -mvpclmulqdq.
2605 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2607         PR middle-end/83423
2608         * config/i386/i386.c (ix86_static_chain): Move
2609         DECL_STATIC_CHAIN test ...
2610         * calls.c (rtx_for_static_chain): ... here.  New function.
2611         * calls.h (rtx_for_static_chain): Declare.
2612         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2613         instead of targetm.calls.static_chain.
2614         * df-scan.c (df_get_entry_block_def_set): Same.
2616 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2618         PR tree-optimization/83493
2619         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2620         and lb.
2622 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2624         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2625         inform with hardcoded english plural handling.
2627 2017-12-18  Jeff Law  <law@redhat.com>
2629         PR tree-optimization/83477
2630         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2631         a non-virtual PHI, always push a new range.
2633 2017-12-19  Martin Sebor  <msebor@redhat.com>
2635         PR middle-end/77608
2636         * builtins.c (compute_objsize): Handle non-constant offsets.
2638 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2640         PR tree-optimization/83444
2641         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2642         character load case, if get_stridx on MEM_REF's operand doesn't
2643         look usable, retry with get_addr_stridx.
2645 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2647         PR debug/83422
2648         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2649         (variable_tracking_main_1): Keep markers even when VTA fails.
2651         PR bootstrap/83396
2652         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2653         even if there are markers before it.
2654         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2656 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2658         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2659         typical order conditions.
2660         * sel-sched.c (extract_new_fences_from): Likewise.
2661         * config/visium/constraints.md (J, K, L): Likewise.
2662         * config/visium/predicates.md (const_shift_operand): Likewise.
2663         * config/visium/visium.c (visium_legitimize_address,
2664         visium_legitimize_reload_address): Likewise.
2665         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2666         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2667         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2668         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2669         SET_DUMP_DETAIL): Likewise.
2670         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2671         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2672         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2673         avr_set_core_architecture, avr_set_current_function,
2674         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2675         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2676         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2677         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2678         avr_map_decompose, avr_fold_builtin): Likewise.
2679         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2680         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2681         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2682         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2683         * config/m32c/m32c.c (m32c_conditional_register_usage,
2684         m32c_address_cost): Likewise.
2685         * config/m32c/predicates.md (shiftcount_operand,
2686         longshiftcount_operand): Likewise.
2687         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2688         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2689         can_use_cdx_ldstw): Likewise.
2690         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2691         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2692         Likewise.
2693         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2694         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2695         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2696         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2697         * config/vax/vax.c (adjacent_operands_p): Likewise.
2698         * config/ft32/constraints.md (L, b, KA): Likewise.
2699         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2700         Likewise.
2701         * cfgexpand.c (expand_stack_alignment): Likewise.
2702         * gcse.c (insert_expr_in_table): Likewise.
2703         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2704         * cgraphunit.c (cgraph_node::expand): Likewise.
2705         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2706         * emit-rtl.c (add_insn): Likewise.
2707         * input.c (dump_location_info): Likewise.
2708         * passes.c (NEXT_PASS): Likewise.
2709         * read-rtl-function.c (parse_note_insn_name,
2710         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2711         Likewise.
2712         * sched-rgn.c (sched_rgn_init): Likewise.
2713         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2714         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2715         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2716         Likewise.
2717         * explow.c (eliminate_constant_term): Likewise.
2718         * final.c (leaf_renumber_regs_insn): Likewise.
2719         * cfgrtl.c (print_rtl_with_bb): Likewise.
2720         * genhooks.c (emit_init_macros): Likewise.
2721         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2722         * tree-data-ref.c (conflict_fn): Likewise.
2723         * selftest.c (assert_streq): Likewise.
2724         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2725         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2726         fold_binary_loc, multiple_of_p): Likewise.
2727         * reload.c (push_reload, find_equiv_reg): Likewise.
2728         * et-forest.c (et_nca, et_below): Likewise.
2729         * dbxout.c (dbxout_symbol_location): Likewise.
2730         * reorg.c (relax_delay_slots): Likewise.
2731         * dojump.c (do_compare_rtx_and_jump): Likewise.
2732         * gengtype-parse.c (type): Likewise.
2733         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2734         simplify_const_relational_operation): Likewise.
2735         * reload1.c (do_output_reload): Likewise.
2736         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2737         * gengtype.c (type_for_name): Likewise.
2738         * gimple-ssa-sprintf.c (format_directive): Likewise.
2740 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2742         PR target/82975
2743         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2744         accessing it.  Adjust comment.
2746 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2748         PR middle-end/81914
2749         * predict.c (zero_one_minusone): New function.
2750         (apply_return_prediction): Avoid return prediction for functions
2751         returning only -1, 0 and 1 values, unless they only return -1 and 0
2752         or 0 and 1.
2754 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2756         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2757         fall-through warning.
2758         (arc_compute_frame_size): Remove unused variables.
2759         (arc_print_operand): Fix fprintif format.
2760         (arc_can_follow_jump): Clean fall-through warning.
2762 2017-12-19  Marek Polacek  <polacek@redhat.com>
2764         PR c++/83489
2765         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2766         on an error node.
2768 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2770         * config/arc/arc.c (overriderregs): New variable.
2771         (arc_override_options): Track fixed/call saved/call options.
2772         (arc_conditional_register_usage): Check against overrideregs
2773         variable whenever we change register properties.
2775 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2777         * opts.c (finish_options): Don't prefix dump_base_name if it
2778         already contains directories.
2780 2017-12-19  Martin Liska  <mliska@suse.cz>
2782         PR rtl-optimization/82675
2783         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2784         more element in sbitmap.
2786 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2788         * gimplify.c (gimplify_expr): Use error_operand_p.
2790 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2792         PR target/83387
2793         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2794         multilib.
2796 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2798         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2799         the jump is to a label.
2801 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2803         PR tree-optimization/83444
2804         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2805         character loads.
2807         PR ipa/82801
2808         PR ipa/83346
2809         * ipa-inline.c (flatten_remove_node_hook): New function.
2810         (ipa_inline): Keep only nodes with flatten attribute at the end of
2811         the array in the order from ipa_reverse_postorder, only walk that
2812         portion of array for flattening, if there is more than one such
2813         node, temporarily register a removal hook and ignore removed nodes.
2815         PR tree-optimization/80631
2816         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2817         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2818         IFN_REDUC_MAX or IFN_REDUC_MIN.
2820 2017-12-18  Jeff Law  <law@redhat.com>
2822         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2823         record symbolic equivalences from backedges in the CFG.
2825         Revert
2826         2017-11-19  Jeff Law  <law@redhat.com>
2828         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2829         of degenerates resulting from ignoring an edge.
2831 2017-12-18  Martin Sebor  <msebor@redhat.com>
2833         PR middle-end/83373
2834         PR tree-optimization/78450
2835         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2836         (handle_builtin_strlen): Call it.
2838 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2840         PR rtl-optimization/83424
2841         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2843 2017-12-18  Marek Polacek  <polacek@redhat.com>
2845         PR middle-end/83463
2846         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2847         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2848         values.
2850 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2852         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2853         (umaddsidi4, umaddsidi_split): Likewise.
2855 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2857         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2858         constants.
2860 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2862         PR target/83420
2863         * config/s390/s390.c (s390_option_override): Avoid strncpy.
2865 2017-12-18  Richard Biener  <rguenther@suse.de>
2867         PR tree-optimization/81877
2868         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
2869         (outermost_indep_loop): Adjust.
2870         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
2871         (can_sm_ref_p): Adjust.
2873 2017-12-18  Richard Biener  <rguenther@suse.de>
2875         PR middle-end/77291
2876         * tree.c (array_at_struct_end_p): Return true if the underlying
2877         object has space for at least one element in excess of what
2878         the array domain specifies.
2880 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2882         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
2883         example to avoid overfull hbox.
2884         * doc/invoke.texi (Option Summary): Add missing @gol.
2885         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
2886         overfull hbox.
2888 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2889             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2891         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
2892         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
2893         (AArch64 Options): Document -mverbose-cost-dump.
2894         (ARM Options): Likewise, plus -mflip-thumb.
2896 2017-12-17  Martin Sebor  <msebor@redhat.com>
2898         PR bootstrap/83446
2899         * gimple-ssa-warn-restrict.c
2900         (builtin_memref::offset_out_of_bounds): Correct the handling of
2901         anti-ranges.
2903 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2905         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
2907 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
2909         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
2911 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
2913         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
2914         latencies.
2916 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
2918         * doc/invoke.texi: Fix some typos.
2920 2017-12-16  Martin Sebor  <msebor@redhat.com>
2922         PR tree-optimization/78918
2923         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
2924         * builtins.c (check_sizes): Rename...
2925         (check_access): ...to this.  Rename function arguments for clarity.
2926         (check_memop_sizes): Adjust names.
2927         (expand_builtin_memchr, expand_builtin_memcpy): Same.
2928         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
2929         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
2930         (check_strncat_sizes, expand_builtin_strncat): Same.
2931         (expand_builtin_strncpy, expand_builtin_memset): Same.
2932         (expand_builtin_bzero, expand_builtin_memcmp): Same.
2933         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
2934         (maybe_emit_sprintf_chk_warning): Same.
2935         (expand_builtin_strcpy): Adjust.
2936         (expand_builtin_stpcpy): Same.
2937         (expand_builtin_with_bounds): Detect out-of-bounds accesses
2938         in pointer-checking forms of memcpy, memmove, and mempcpy.
2939         (gcall_to_tree_minimal, max_object_size): Define new functions.
2940         * builtins.h (max_object_size): Declare.
2941         * calls.c (alloc_max_size): Call max_object_size instead of
2942         hardcoding ssizetype limit.
2943         (get_size_range): Handle new argument.
2944         * calls.h (get_size_range): Add a new argument.
2945         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
2946         * doc/invoke.texi (-Wrestrict): Adjust, add example.
2947         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
2948         operations.
2949         (gimple_fold_builtin_memory_chk): Same.
2950         (gimple_fold_builtin_stxcpy_chk): New function.
2951         * gimple-ssa-warn-restrict.c: New source.
2952         * gimple-ssa-warn-restrict.h: New header.
2953         * gimple.c (gimple_build_call_from_tree): Propagate location.
2954         * passes.def (pass_warn_restrict): Add new pass.
2955         * tree-pass.h (make_pass_warn_restrict): Declare.
2956         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
2957         operations.
2958         (handle_builtin_strcat): Same.
2959         (strlen_optimize_stmt): Rename...
2960         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
2961         stpncpy, strncpy, and their checking forms.
2963 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
2965         PR rtl-optimization/82849
2966         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
2967         and get_max_loop_iterations_int.
2969 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
2970             Alan Hayward  <alan.hayward@arm.com>
2971             David Sherwood  <david.sherwood@arm.com>
2973         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
2974         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
2975         size as a poly_uint64.
2976         (mode_for_vector, mode_for_int_vector): Take the number of vector
2977         elements as a poly_uint64.
2978         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
2979         size as a poly_uint64.
2980         (mode_for_vector, mode_for_int_vector): Take the number of vector
2981         elements as a poly_uint64.
2983 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
2984             Alan Hayward  <alan.hayward@arm.com>
2985             David Sherwood  <david.sherwood@arm.com>
2987         * machmode.h (MACRO_MODE): New macro.
2988         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
2989         * targhooks.c (default_libcall_value, default_secondary_reload)
2990         (default_memory_move_cost, default_register_move_cost)
2991         (default_class_max_nregs): Likewise.
2993 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
2994             Alan Hayward  <alan.hayward@arm.com>
2995             David Sherwood  <david.sherwood@arm.com>
2997         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
2998         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
2999         target C file.
3000         * genautomata.c (main): Likewise.
3001         * genconditions.c (write_header): Likewise.
3002         * genemit.c (main): Likewise.
3003         * genextract.c (print_header): Likewise.
3004         * genopinit.c (main): Likewise.
3005         * genoutput.c (output_prologue): Likewise.
3006         * genpeep.c (main): Likewise.
3007         * genpreds.c (write_insn_preds_c): Likewise.
3008         * genrecog.c (writer_header): Likewise.
3009         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3010         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3011         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3012         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3013         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3014         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3015         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3016         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3017         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3018         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3019         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3020         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3021         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3022         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3023         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3024         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3025         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3026         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3027         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3028         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3029         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3030         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3031         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3032         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3033         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3034         * config/darwin.c (IN_TARGET_CODE): Likewise.
3035         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3036         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3037         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3038         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3039         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3040         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3041         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3042         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3043         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3044         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3045         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3046         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3047         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3048         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3049         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3050         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3051         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3052         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3053         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3054         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3055         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3056         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3057         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3058         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3059         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3060         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3061         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3062         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3063         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3064         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3065         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3066         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3067         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3068         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3069         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3070         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3071         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3072         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3073         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3074         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3075         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3076         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3077         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3078         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3079         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3080         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3081         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3082         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3083         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3084         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3085         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3086         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3087         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3088         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3089         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3090         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3091         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3092         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3093         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3094         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3095         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3096         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3097         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3098         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3099         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3100         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3101         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3102         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3103         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3104         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3105         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3106         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3107         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3108         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3109         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3110         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3111         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3112         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3113         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3114         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3115         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3116         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3117         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3118         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3119         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3120         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3121         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3122         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3123         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3124         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3125         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3126         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3127         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3128         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3129         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3130         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3131         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3132         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3133         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3134         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3135         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3136         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3137         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3138         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3139         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3140         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3142 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3143             Alan Hayward  <alan.hayward@arm.com>
3144             David Sherwood  <david.sherwood@arm.com>
3146         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3147         checks for MEM_REF.
3149 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3150             Alan Hayward  <alan.hayward@arm.com>
3151             David Sherwood  <david.sherwood@arm.com>
3153         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3154         * doc/md.texi (vec_series@var{m}): Document.
3155         * tree.def (VEC_SERIES_EXPR): New tree code.
3156         * tree.h (build_vec_series): Declare.
3157         * tree.c (build_vec_series): New function.
3158         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3159         * tree-pretty-print.c (dump_generic_node): Likewise.
3160         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3161         * tree-inline.c (estimate_operator_cost): Likewise.
3162         * expr.c (expand_expr_real_2): Likewise.
3163         * optabs-tree.c (optab_for_tree_code): Likewise.
3164         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3165         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3166         * expmed.c (make_tree): Handle VEC_SERIES.
3167         * optabs.def (vec_series_optab): New optab.
3168         * optabs.h (expand_vec_series_expr): Declare.
3169         * optabs.c (expand_vec_series_expr): New function.
3170         * tree-vect-generic.c (expand_vector_operations_1): Check that
3171         the operands also have vector type.
3173 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3174             Alan Hayward  <alan.hawyard@arm.com>
3175             David Sherwood  <david.sherwood@arm.com>
3177         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3178         (VEC_COND_EXPR): Add missing @tindex.
3179         * doc/md.texi (vec_duplicate@var{m}): Document.
3180         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3181         * tree.c (build_vector_from_val): Add stubbed-out handling of
3182         variable-length vectors, using VEC_DUPLICATE_EXPR.
3183         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3184         * cfgexpand.c (expand_debug_expr): Likewise.
3185         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3186         * tree-inline.c (estimate_operator_cost): Likewise.
3187         * tree-pretty-print.c (dump_generic_node): Likewise.
3188         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3189         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3190         (test_vec_duplicate_folding): New function.
3191         (fold_const_c_tests): Call it.
3192         * optabs.def (vec_duplicate_optab): New optab.
3193         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3194         * optabs.h (expand_vector_broadcast): Declare.
3195         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3196         vec_duplicate_optab.
3197         * expr.c (store_constructor): Try using vec_duplicate_optab for
3198         uniform vectors.
3199         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3201 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3203         PR target/83358
3204         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3205         div/mod latencies a bit.
3207 2017-12-15  Jeff Law  <law@redhat.com>
3209         PR tree-optimization/36550
3210         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3211         (mark_threaded_blocks): Rewrite code to avoid block copying when
3212         optimizing for size.  Don't pessimize blocks which will be
3213         copied, but all the statements will be dead.
3215 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3217         PR tree-optimization/81165
3218         * tree-ssa-threadupdate.c (uses_in_bb): New.
3219         (estimate_threading_killed_stmts): New.
3220         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3221         * tree-ssa-threadedge.c
3222         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3223         when its hit.
3225 2017-12-15  Jeff Law  <law@redhat.com>
3227         PR tree-optimization/83410
3228         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3229         threads when parallelizing loops.
3231 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3233         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3234         handler fields.
3235         * config/alpha/alpha.c (vms_attribute_table): Swap
3236         affects_type_identity and handler fields, adjust comments.
3237         * config/mips/mips.c (mips_attribute_table): Likewise.
3238         * config/visium/visium.c (visium_attribute_table): Likewise.
3239         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3240         * config/microblaze/microblaze.c (microblaze_attribute_table):
3241         Likewise.
3242         * config/spu/spu.c (spu_attribute_table): Likewise.
3243         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3244         * config/arc/arc.c (arc_attribute_table): Likewise.
3245         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3246         * config/v850/v850.c (v850_handle_interrupt_attribute,
3247         v850_handle_data_area_attribute): Formatting fixes.
3248         (v850_attribute_table): Swap affects_type_identity and handler
3249         fields, adjust comments.
3250         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3251         * config/arm/arm.c (arm_attribute_table): Likewise.
3252         * config/avr/avr.c (avr_attribute_table): Likewise.
3253         * config/s390/s390.c (s390_attribute_table): Likewise.
3254         * config/sh/sh.c (sh_attribute_table): Likewise.
3255         * config/i386/i386.c (ix86_handle_cconv_attribute,
3256         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3257         (ix86_attribute_table): Swap affects_type_identity and handler
3258         fields, adjust comments.
3259         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3260         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3261         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3262         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3263         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3264         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3265         * config/rx/rx.c (rx_attribute_table): Likewise.
3266         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3267         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3268         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3269         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3270         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3271         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3272         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3273         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3274         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3275         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3276         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3277         exclude.
3278         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3279         handler fields.
3280         (register_scoped_attributes, decl_attributes): Formatting fixes.
3282         PR tree-optimization/83269
3283         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3284         subtraction in arg0's type if type is signed and arg0 is unsigned.
3285         Formatting fix.
3287         PR sanitizer/81281
3288         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3289         convert? on @0 instead of convert.  Check type of @1, not @0.
3290         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3291         convert? on @0 instead of convert.  Check type of @1, not @0.
3292         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3293         only optimize if either both @1 and @2 types are narrower
3294         precision, or both are wider or equal precision, and in the former
3295         case only if both have undefined overflow.
3297 2017-12-15  Richard Biener  <rguenther@suse.de>
3299         PR lto/83388
3300         * internal-fn.def (IFN_NOP): Add.
3301         * internal-fn.c (expand_NOP): Do nothing.
3302         * lto-streamer-in.c (input_function): Instead of removing
3303         sanitizer calls replace them with IFN_NOP calls.
3305 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3306             Alan Hayward  <alan.hayward@arm.com>
3307             David Sherwood  <david.sherwood@arm.com>
3309         * dse.c (store_info, read_info_type): Replace begin and end with
3310         offset and width.
3311         (print_range): New function.
3312         (set_all_positions_unneeded, any_positions_needed_p)
3313         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3314         accordingly.
3315         (record_store): Likewise.  Optimize the case in which all positions
3316         are unneeded.
3317         (get_stored_val): Replace read_begin and read_end with read_offset
3318         and read_width.
3319         (replace_read): Update call accordingly.
3321 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3323         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3324         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3325         (loop_cand::loop_cand): Initialize above members.
3326         (loop_cand::supported_operations): Delete.
3327         (loop_cand::can_interchange_p): Inline above function.
3328         (loop_cand::classify_simple_reduction): Record number of constant
3329         initialized simple reductions.
3330         (should_interchange_loops): New parameters.  Check stmt cost of loops
3331         to be interchange.
3332         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3333         Update call to should_interchange_loops.
3334         (should_interchange_loop_nest): Update call to
3335         should_interchange_loops.
3337 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3339         PR target/66488
3340         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3341         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3342         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3344 2017-12-15  Julia Koval  <julia.koval@intel.com>
3346         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3347         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3348         New.
3349         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3350         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3351         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3353 2017-12-15  Julia Koval  <julia.koval@intel.com>
3355         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3356         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3357         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3358         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3359         _mm_aesenc_epi128): New intrinsics.
3361 2017-12-15  Julia Koval  <julia.koval@intel.com>
3363         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3364         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3365         New.
3366         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3367         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3368         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3370 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3372         * gimple-ssa-strength-reduction.c (analyze_increments):
3373         Distinguish replacement costs for constant strides from those for
3374         unknown strides.
3376 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3378         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3380 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3382         * doc/invoke.texi: Document -Wcast-function-type.
3383         * recog.h (stored_funcptr): Change signature.
3384         * tree-dump.c (dump_node): Avoid warning.
3385         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3387 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3389         PR middle_end/79538
3390         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3391         array.
3393 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3395         PR tree-optimization/83312
3396         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3397         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3398         "val" param, and to cope with arbitrary basic blocks.
3399         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3400         handle NULL_TREE for "val", dropping "bb" param.
3401         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3402         drop "bb" param.  Handle NULL_TREE for "val".
3403         (find_case_label_for_value): Make "switch_stmt" param const.
3404         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3405         of dom_walker.
3406         (vrp_prop::check_all_array_refs): Reimplement as...
3407         (check_array_bounds_dom_walker::before_dom_children): ...this new
3408         vfunc.  Replace linear search through BB block list, excluding
3409         those with non-executable in-edges via dominator walk.
3411 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3413         * config/arm/arm.opt (mverbose-cost-dump): New option.
3414         * config/arm/arm.c (arm_rtx_costs): Use it.
3416 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3418         PR bootstrap/83396
3419         * reload1.c (emit_input_reload_insns): Skip debug markers.
3421 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3423         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3424         for its nop_endbr.
3426         PR bootstrap/83396
3427         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3428         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3429         * config/sh/sh.c: Skip debug insns besides notes.
3430         * config/sh/sh.md: Likewise.
3431         * config/sh/sh_treg_combine.cc: Likewise.
3432         * config/sh/sync.md: Likewise.
3434 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3436         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3437         Add item for weak_undefined.
3439 2017-12-14  Richard Biener  <rguenther@suse.de>
3441         PR tree-optimization/67842
3442         * sese.h (bb_in_region): Remove #if 0'ed code.
3444 2017-12-14  Richard Biener  <rguenther@suse.de>
3446         PR tree-optimization/83326
3447         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3448         may_be_zero parameter and handle it by not marking the first
3449         peeled copy as not exiting the loop.
3450         (try_peel_loop): Likewise.
3451         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3452         to handle the case of constant or zero iterations and perform
3453         loop header copying on-the-fly.
3455 2017-12-14  Richard Biener  <rguenther@suse.de>
3457         PR tree-optimization/83418
3458         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3459         Instead of asserting we don't get unfolded comparisons deal with them.
3461 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3463         PR bootstrap/83396
3464         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3465         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3466         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3467         bb.  Simplify.
3469         PR tree-optimization/83198
3470         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3471         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3472         value if arg is a REAL_CST with incompatible type.
3474 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3475             Bin Cheng  <bin.cheng@arm.com>
3477         PR target/81228
3478         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3479         to CCFPEmode.
3480         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3481         LTGT.
3483 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3485         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3486         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3488 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3489             Alan Hayward  <alan.hayward@arm.com>
3490             David Sherwood  <david.sherwood@arm.com>
3492         * poly-int.h: New file.
3493         * poly-int-types.h: Likewise.
3494         * coretypes.h: Include them.
3495         (POLY_INT_CONVERSION): Define.
3496         * target.def (estimated_poly_value): New hook.
3497         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3498         * doc/tm.texi: Regenerate.
3499         * doc/poly-int.texi: New file.
3500         * doc/gccint.texi: Include it.
3501         * doc/rtl.texi: Describe restrictions on subreg modes.
3502         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3503         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3504         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3505         * targhooks.h (default_estimated_poly_value): Declare.
3506         * targhooks.c (default_estimated_poly_value): New function.
3507         * target.h (estimated_poly_value): Likewise.
3508         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3509         (wi::unary_traits): Delete.
3510         (wi::binary_traits::signed_shift_result_type): Define for
3511         offset_int << HOST_WIDE_INT, etc.
3512         (generic_wide_int::operator <<=): Define for all types and use
3513         wi::lshift instead of <<.
3514         (wi::hwi_with_prec): Add a default constructor.
3515         (wi::ints_for): New class.
3516         (operator <<): Define for all wide-int types.
3517         (operator /): New function.
3518         (operator %): Likewise.
3519         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3520         (ASSERT_MAYBE_NE_AT): New macros.
3522 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3523             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3525         PR middle-end/78468
3526         * emit-rtl.c (init_emit): Remove ??? comment.
3527         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3528         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3529         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3530         alignment of 3 virtual registers to BITS_PER_WORD.
3532         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3534 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3536         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3537         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3539 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3541         PR bootstrap/83396
3542         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3544 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3546         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3547         with the exception of debug begin stmt markers appear before
3548         labels.
3550         PR bootstrap/83396
3551         * final.c (rest_of_handle_final): Call variable_tracking_main only
3552         if !flag_var_tracking.
3554 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3555             Jakub Jelinek  <jakub@redhat.com>
3557         PR bootstrap/83396
3558         PR debug/83391
3559         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3560         labels debug stmts that can only appear after labels.
3562 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3564         PR rtl-optimization/82398
3565         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3566         EXPR_USEFULNESS in priority comparison.
3568 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3570         PR rtl-optimization/83393
3571         * combine.c (move_deaths): If reg_stat points to a too new insn in
3572         last_death, do not use it: find the proper insn instead.
3574 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3576         PR target/82190
3577         * config/rs6000/rs6000-string.c (expand_block_compare,
3578         expand_strn_compare): Fix set_mem_size() calls.
3580 2017-12-12  Jeff Law  <law@redhat.com>
3582         PR tree-optimization/83298
3583         PR tree-optimization/83362
3584         PR tree-optimization/83383
3585         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3586         push_value_range a public interface.  Add new argument to
3587         record_ranges_from_stmt.
3588         * gimple-ssa-evrp-analyze.c
3589         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3590         Update comments.  Handle recording temporary equivalences.
3591         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3592         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3593         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3594         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3595         gimple-ssa-evrp-analyze.h.
3596         (record_temporary_equivalences_from_phis): Add new argument.  When
3597         the PHI arg is an SSA_NAME, set the result's range to the range
3598         of the PHI arg.
3599         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3600         from statements too.
3601         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3602         Pass it down to children as needed.
3603         (thread_outgoing_edges): Likewise.
3604         (thread_across_edge): Likewise.   Push/pop range state as needed.
3605         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3607 2017-12-12  Julia Koval  <julia.koval@intel.com>
3609         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3610         (PTA_CANNONLAKE): Remove PTA_CLWB.
3612 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3614         PR target/83332
3615         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3616         (vcondv2div2df): Likewise.
3617         (vconduv2dfv2di): Likewise.
3619 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3621         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3622         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3623         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3624         ATTR_MATHFN_FPROUNDING_ERRNO.
3626 2017-12-12  Richard Biener  <rguenther@suse.de>
3628         PR tree-optimization/83385
3629         * graphite-scop-detection.c (get_order, order): Remove.
3630         (bb_to_rpo): New global.
3631         (cmp_pbbs): Adjust.
3632         (build_scops): Sort pbbs in RPO order.
3634 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3636         * combine.c (simplify_set): Do not transform subregs to zero_extends
3637         if the destination mode is a vector mode.
3639 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3641         PR tree-optimization/83359
3642         * tree-cfg.h (fold_loop_internal_call): Declare.
3643         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3644         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3645         (find_loop_dist_alias): New function.
3646         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3647         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3648         calls.
3650         PR tree-optimization/80631
3651         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3652         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3653         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3654         hardcoding zero as the value if COND_EXPR is never true.  For
3655         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3656         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3657         hardcoding MAX_EXPR as the reduction operation.
3658         (is_nonwrapping_integer_induction): Allow negative step.
3659         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3660         vect_create_epilog_for_reduction, if no value is suitable, don't
3661         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3663 2017-12-12  Richard Biener  <rguenther@suse.de>
3665         PR tree-optimization/81889
3666         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3667         range info from the non-wrapping IV instead of just the range
3668         of the type.
3670 2017-12-12  Julia Koval  <julia.koval@intel.com>
3672         * config.gcc: Add vaesintrin.h.
3673         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3674         New type.
3675         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3676         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3677         New builtins.
3678         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3679         * config/i386/immintrin.h: Include vaesintrin.h.
3680         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3681         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3682         _mm_aesdec_epi128): New intrinsics.
3684 2017-12-12  Julia Koval  <julia.koval@intel.com>
3686         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3687         OPTION_MASK_ISA_VAES_UNSET): New.
3688         (ix86_handle_option): Handle -mvaes.
3689         * config/i386/cpuid.h: Define bit_VAES.
3690         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3691         * config/i386/i386-c.c (__VAES__): New.
3692         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3693         (ix86_valid_target_attribute_inner_p): Ditto.
3694         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3695         * config/i386/i386.opt: Add -mvaes.
3696         * doc/invoke.texi: Ditto.
3698 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3700         * debug.h (gcc_debug_hooks): Add inline_entry.
3701         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3702         * debug.c (do_nothing_debug_hooks): Likewise.
3703         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3704         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3705         (dwarf2_lineno_debug_hooks): Likewise.
3707         * common.opt (gstatement-frontiers): New, setting
3708         debug_nonbind_markers_p.
3709         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3710         * toplev.c (process_options): Autodetect value for debug statement
3711         frontiers option.
3712         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3713         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3715         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3716         markers.  Integrate source bind into debug stmt expand loop.
3717         (pass_expand::execute): Check debug marker limit.  Avoid deep
3718         TER and expand debug locations for debug bind insns only.
3719         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3720         followed by them.
3721         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3722         * final.c (reemit_insn_block_notes): Take current block from
3723         nonbind markers.  Declare note where it's first set.
3724         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3725         begin stmt markers if enabled.
3726         (notice_source_line): Handle nonbind markers.  Fail if their
3727         location is unknown or that of builtins.
3728         (rest_of_handle_final): Convert begin stmt markers to notes if
3729         var-tracking didn't run.
3730         (rest_of_clean_state): Skip begin stmt markers.
3731         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3732         markers.
3733         * function.c (allocate_struct_function): Set begin_stmt_markers.
3734         * function.h (struct function): Add debug_marker_count counter
3735         and debug_nonbind_markers flag.
3736         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3737         * gimple-low.c (lower_function_body): Adjust
3738         debug_nonbind_markers.
3739         (lower_stmt): Drop or skip gimple debug stmts.
3740         (lower_try_catch): Skip debug stmts.
3741         * gimple.c (gimple_build_debug_begin_stmt): New.
3742         (gimple_copy): Increment debug_marker_count if copying one.
3743         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3744         * gimplify.c (rexpr_location): New.
3745         (rexpr_has_location): New.
3746         (warn_switch_unreachable_r): Handle gimple debug stmts.
3747         (shortcut_cond_r): Call expr_location.
3748         (find_goto): New.
3749         (find_goto_label): New.
3750         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3751         find_goto_label.
3752         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3753         expr_location.
3754         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3755         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3756         (LANG_HOOKS_INITIALIZER): ... this.
3757         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3758         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3759         debug insns.
3760         (update_ebb_live_info): Skip debug insn markers.
3761         * lra.c (debug_insn_static_data): Rename to...
3762         (debug_bind_static_data): ... this.
3763         (debug_marker_static_data): New.
3764         (lra_set_insn_recog_data): Select one of the above depending
3765         on debug insn kind.
3766         (lra_update_isn_regno_info): Don't assume debug insns have
3767         freqs.
3768         (push_insns): Skip debug insns.
3769         * lto-streamer-in.c (input_function): Drop debug stmts
3770         depending on active options.  Adjust debug_nonbind_markers.
3771         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3772         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3773         begin stmt marker notes.
3774         (print_insn): Likewise.
3775         * recog.c (extract_insn): Recognize rtl for debug markers.
3776         * rtl.def (DEBUG_MARKER): New.
3777         * tree-inline.c: Include params.h.
3778         (remap_gimple_stmt): Handle nonbind markers.
3779         (maybe_move_debug_stmts_to_successors): Likewise.
3780         (copy_debug_stmt): Likewise.
3781         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3782         markers regardless of no side effects.
3783         (tsi_link_before): Don't update container's side effects when adding
3784         a begin stmt marker.
3785         (tsi_link_after): Likewise.
3786         (expr_first): Skip begin stmt markers.
3787         (expr_last): Likewise.
3788         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3789         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3790         Disregard nonbind markers.
3791         * tree.c (make_node_stat): Don't set side effects for begin stmt
3792         markers.
3793         (build1_stat): Likewise.
3794         * tree.def (DEBUG_BEGIN_STMT): New.
3795         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3796         * var-tracking.c (delete_debug_insns): Renamed to...
3797         (delete_vta_debug_insns): ... this.
3798         (reemit_marker_as_note): New.
3799         (vt_initialize): Reemit markers.
3800         (delete_vta_debug_insns): Likewise.
3801         (vt_debug_insns_local): Reemit or delete markers.
3802         (variable_tracking_main_1): Likewise.
3803         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3804         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3805         (gimple_debug_nonbind_marker_p): New.
3806         (gimple_build_debug_bind): Adjust.
3807         (gimple_build_debug_begin_stmt): New.
3808         * doc/invoke.texi (max-debug-marker-count): New param.
3809         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3810         (debug_parameter_ref, debug_marker): New.
3811         (NOTE_INSN_BEGIN_STMT): New.
3812         (DEBUG_INSN): Describe begin stmt markers.
3814         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3815         without debug insns, we wouldn't, but clean up debug insns
3816         after a control flow insn nevertheless.
3818         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3819         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3820         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3821         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3822         (rtl_tidy_fallthru_edge): Likewise.
3823         (rtl_verify_fallthru): Likewise.
3824         (rtl_verify_bb_layout): Likewise.
3825         (skip_insns_after_block): Likewise.
3826         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3827         * dwarf2out.c: Include print-rtl.h.
3828         (dwarf2out_next_real_insn): New.
3829         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3830         Dump debug binds in asm comments.
3831         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3832         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3833         callers to use gsi_start_nondebug_bb instead.
3834         (gsi_after_labels): Skip gimple debug stmts.
3835         (gsi_start_nondebug): New.
3836         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3837         (proper_loop_form_for_interchange): Adjust.
3838         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3839         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3840         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3841         (collect_fallthrough_labels): Likewise.
3842         (should_warn_for_implicit_fallthrough): Likewise.
3843         (warn_implicit_fallthrough_r): Likewise.
3844         (expand_FALLTHROUGH_r): Likewise.
3845         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3846         (graphite_copy_stmts_from_block): Skip nonbind markers.
3847         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3848         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3849         * jump.c (clean_barriers): Skip debug insns.
3850         * omp-expand.c (expand_parallel_call): Skip debug insns.
3851         (expand_task_call): Likewise.
3852         (remove_exit_barrier): Likewise.
3853         (expand_omp_taskreg): Likewise.
3854         (expand_omp_for_init_counts): Likewise.
3855         (expand_omp_for_generic): Likewise.
3856         (expand_omp_for_static_nochunk): Likewise.
3857         (expand_omp_for_static_chunk): Likewise.
3858         (expand_omp_simd): Likewise.
3859         (expand_omp_taskloop_for_outer): Likewise.
3860         (expand_omp_taskloop_for_inner): Likewise.
3861         (expand_oacc_for): Likewise.
3862         (expand_omp_sections): Likewise.
3863         (expand_omp_single): Likewise.
3864         (expand_omp_synch): Likewise.
3865         (expand_omp_atomic_load): Likewise.
3866         (expand_omp_atomic_store): Likewise.
3867         (expand_omp_atomic_fetch_op): Likewise.
3868         (expand_omp_atomic_pipeline): Likewise.
3869         (expand_omp_atomic_mutex): Likewise.
3870         (expand_omp_target): Likewise.
3871         (grid_expand_omp_for_loop): Likewise.
3872         (grid_expand_target_grid_body): Likewise.
3873         (build_omp_regions_1): Likewise.
3874         * omp-low.c (check_combined_parallel): Skip debug stmts.
3875         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
3876         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
3877         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
3878         test.
3879         * tree-cfg.c (make_blobs_1): Skip debug stmts.
3880         (make_edges): Likewise.
3881         (cleanup_dead_labels): Likewise.
3882         (gimple_can_merge_blocks_p): Likewise.
3883         (stmt_starts_bb_p): Likewise.
3884         (gimple_block_label): Likewise.
3885         (gimple_redirect_edge_and_branch): Likewise.
3886         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
3887         of debug stmts.
3888         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
3889         TDF_SLIM.
3890         * tree-pretty-print (print_declaration): Omit initializer in slim
3891         dumps.
3892         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
3893         markers.
3894         (eliminate_unnecessary_stmts): Stabilize block removal order.
3895         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
3896         * var-tracking.c (get_first_insn): New.
3897         (vt_emit_notes): Call it.
3898         (vt_initialize): Walk any insns before the first BB.
3899         (delete_debug_insns): Likewise.
3901         * gimple.h (enum gimple_debug_subcode): Add
3902         GIMPLE_DEBUG_BEGIN_STMT.
3903         (gimple_debug_begin_stmt_p): New.
3904         (gimple_debug_nonbind_marker_p): New.
3905         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
3906         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
3907         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
3908         * insn-notes.def (BEGIN_STMT): New.
3909         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
3910         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
3911         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
3912         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
3913         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
3914         (INSN_DEBUG_MARKER_KIND): New.
3915         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
3916         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
3917         (INSN_VAR_LOCATION_PTR): New.
3918         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
3919         only.
3920         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
3921         deps only if debug bind insns are enabled.
3922         (pass_expand::execute): Avoid deep TER and expand
3923         debug locations for debug bind insns only.
3924         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
3925         debug stmts special handling down to debug bind stmts.
3926         * combine.c (try_combine): Narrow debug insns special handling
3927         down to debug bind insns.
3928         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
3929         Narrow debug insns preexisting special handling down to debug
3930         bind insns.
3931         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
3932         handling down to debug bind insns.
3933         * function.c (instantiate_virtual_regs): Skip debug markers,
3934         adjust handling of debug binds.
3935         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
3936         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
3937         * haifa-sched.c (schedule_insn): Narrow special handling of debug
3938         insns to debug bind insns.
3939         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
3940         special handling of debug stmts to debug bind stmts.
3941         * ipa-split.c (split_function): Likewise.
3942         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
3943         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
3944         debug insns.
3945         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
3946         * regrename.c (build_def_use): Likewise.
3947         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
3948         (pass_cprop_hardreg): Narrow special casing of debug insns to
3949         debug bind insns.
3950         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
3951         * reload1.c (reload): Likewise.
3952         * sese.c (sese_insert_phis_for_liveouts): Narrow special
3953         casing of debug stmts to debug bind stmts.
3954         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
3955         * ssa-iterators.h (num_imm_uses): Likewise.
3956         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
3957         debug stmts to debug bind stmts.
3958         * tree-inline.c (tree_function_versioning): Narrow special casing
3959         of debug stmts to debug bind stmts.
3960         * tree-loop-distribution.c (generate_loops_for_partition):
3961         Narrow special casing of debug stmts to debug bind stmts.
3962         * tree-sra.c (analyze_access_subtree): Narrow special casing
3963         of debug stmts to debug bind stmts.
3964         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
3965         stmts to debug bind stmts.
3966         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
3967         casing of debug stmts to debug bind stmts.
3968         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3969         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
3970         casing of debug stmts to debug bind stmts.
3971         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3972         Likewise.
3973         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
3974         debug stmts to debug bind stmts.
3975         (gimple_replace_ssa_lhs): Likewise.
3976         (insert_debug_temp_for_var_def): Likewise.
3977         (insert_debug_temps_for_defs): Likewise.
3978         (reset_debug_uses): Likewise.
3979         * tree-ssanames.c (release_ssa_name_fn): Likewise.
3980         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
3981         (adjust_debug_stmts): Likewise.
3982         (adjust_phi_and_debug_stmts): Likewise.
3983         (vect_do_peeling): Likewise.
3984         * tree-vect-loop.c (vect_transform_loop): Likewise.
3985         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
3986         * var-tracking.c (adjust_mems): Narrow special casing of debug
3987         insns to debug bind insns.
3988         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
3989         (compute_bb_dataflow, vt_find_locations): Likewise.
3990         (vt_expand_loc, emit_notes_for_changes): Likewise.
3991         (vt_init_cfa_base): Likewise.
3992         (vt_emit_notes): Likewise.
3993         (vt_initialize): Likewise.
3994         (vt_finalize): Likewise.
3996         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
3997         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
3998         (next_nonnote_nondebug_insn_bb): New.
3999         (prev_nonnote_nondebug_insn_bb): New.
4000         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4001         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4002         (prev_nonnote_nondebug_insn_bb): Declare.
4003         (next_nonnote_nondebug_insn_bb): Declare.
4004         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4005         * cfgrtl.c (get_last_bb_insn): Likewise.
4006         * lra.c (push_insns): Likewise.
4008 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4010         PR c/82050
4011         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4012         to...
4013         * selftest.c (selftest::test_runner::test_runner): New ctor.
4014         (selftest::test_runner::~test_runner): New dtor.
4015         * selftest.h (class selftest::test_runner): New class.
4017 2017-12-11  Carl Love  <cel@us.ibm.com>
4019         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4020         vec_extract_fp32_from_shortl]): Add #defines.
4021         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4022         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4023         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4024         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4025         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4026         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4027         * doc/extend.texi: Add documentation for the added builtins.
4029 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4031         PR rtl-optimization/80693
4032         PR rtl-optimization/81019
4033         PR rtl-optimization/81020
4034         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4035         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4036         possibly modified to REG_DEAD, if it did not originate in i3.
4038 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4040         * recog.c (store_data_bypass_p_1): New function.
4041         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4042         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4044 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4046         PR rtl-optimization/83361
4047         * ifcvt.c (if_convert): Call fixup_partitions.
4049 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4051         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4052         early folding of splat_u{8,16,32}.
4054 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4056         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4057         output_operand_lossage first argument with capital letter.
4058         (aarch64_override_options): Don't start error and sorry first argument
4059         with capital letter.
4061 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4063         PR gcov-profile/83355
4064         * auto-profile.c (string_table::get_index_by_decl): Don't
4065         recurse when abstract origin points to itself.
4067 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4069         PR tree-optimization/83320
4070         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4071         (prune_datarefs_not_in_loop): Ditto.
4073 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4075         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4077 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4079         PR tree-optimization/83337
4080         * gimple-loop-interchange.cc (compute_access_stride): Handle
4081         bitfield DRs properly.
4083 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4085         PR tree-optimization/83338
4086         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4087         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4088         vector of unsigned integers to vector of signed integers.
4090 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4092         PR rtl-optimization/83317
4093         * lra-constraints.c (process_address_1): Add insn code check.
4095 2017-12-08  Michael Matz  <matz@suse.de>
4097         Fix PR tree-optimization/83323
4098         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4099         head-controlled loops and loop BBs.
4100         * common.opt (funroll-and-jam): Remove, instead ...
4101         (floop-unroll-and-jam): ... reuse this option.
4102         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4103         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4104         (-floop-unroll-and-jam): ... this option.
4106 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4108         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4109         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4110         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4112 2017-12-08  Richard Biener  <rguenther@suse.de>
4114         PR middle-end/81782
4115         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4116         handle accesses outside of zero-sized vars.
4118 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4120         PR tree-optimization/83141
4121         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4122         test for MEM_REFs implicitely changing types with padding.  Remove
4123         inline keyword.
4124         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4126 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4128         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4129         the comment.
4130         * config/avr/avr.c (avr_attribute_table): Likewise.
4131         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4132         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4133         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4134         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4135         of struct attribute_spec.
4136         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4138 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4140         PR target/82960
4141         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4143 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4145         PR middle-end/83609
4146         * profile-count.c (profile_count::from_gcov_type): Move from
4147         profile-count.h; handle overflow.
4148         * profile-count.h (profile_count::from_gcov_type): Move offline.
4150 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4152         PR rtl-optimization/83304
4153         * combine.c (move_deaths): If we do not know where a register died,
4154         search for it.
4156 2017-12-08  Richard Biener  <rguenther@suse.de>
4158         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4159         Provide -fopt-info-loop feedback when we interchange in a nest.
4161 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4163         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4164         for armv6 ARM CPU IDs.
4166 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4168         * common/config/arm/arm-common.c: Include <algorithm>.
4169         (INCLUDE_VECTOR): Define.
4170         (compare_opt_names): New function.
4171         (arm_rewrite_selected_arch): Only strip out extensions that can be
4172         expressed through -mfpu.  Sort the remaining extensions
4173         alphabetically.
4175 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4177         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4178         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4179         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4180         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4182 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4184         PR ada/81470
4185         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4186         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4188 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4190         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4191         of struct attribute_spec.
4193 2017-12-08  Julia Koval  <julia.koval@intel.com>
4195         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4196         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4197         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4198         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4199         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4200         _mm_maskz_dpwssds_epi32): Ditto.
4202 2017-12-08  Richard Biener  <rguenther@suse.de>
4204         PR tree-optimization/81303
4205         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4206         conditions try to create a comparison vector type matching
4207         the data vector type.
4208         (vectorizable_condition): Adjust.
4209         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4210         Leave invariant conditions alone in case we can vectorize those.
4212 2017-12-08  Julia Koval  <julia.koval@intel.com>
4214         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4215         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4216         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4217         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4218         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4220 2017-12-08  Julia Koval  <julia.koval@intel.com>
4222         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4223         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4224         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4225         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4226         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4227         _mm_maskz_dpbusds_epi32): New intrinsics.
4229 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4231         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4232         operand.
4233         (ld<bh>io_signed): Likewise.
4234         (st<bhw>io): Likewise.
4235         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4236         SMALL_INT12 constant integer operand.
4238 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4240         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4241         Rename to...
4242         (nios2_large_constant_allowed): ...this.  Adjust uses.
4243         (nios2_plus_symbolic_constant_p): Rename to...
4244         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4245         (nios2_legitimate_address_p): Correct CONST_INT handling.
4246         (nios2_symbolic_memory_operand_p): Rename to...
4247         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4248         (nios2_large_constant_p): Check for large constant integers too.
4249         (nios2_split_large_constant): Handle constant integers.
4250         (nios2_split_symbolic_memory_operand): Rename to...
4251         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4252         (nios2_legitimize_constant_address): Handle constant integers.
4253         (r0rel_constant_p): Handle small constant integers.
4254         (nios2_print_operand_address): Handle r0-relative integer addresses.
4255         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4256         * config/nios2/nios2.md: Adjust for renamed functions.
4258 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4260         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4261         (TARGET_HAVE_SRODATA_SECTION): New define.
4262         (riscv_select_section): New function.
4264 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4265             Alexander Monakov  <amonakov@ispras.ru>
4266             Jakub Jelinek  <jakub@redhat.com>
4268         PR target/81906
4269         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4271 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4273         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4274         a bool success value.  Don't call output_operand_lossage here.
4275         (aarch64_print_ldpstp_address): Return a bool success value.
4276         (aarch64_print_operand_address): Call output_addr_const if
4277         aarch64_print_address_internal fails.
4278         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4279         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4280         if aarch64_print_ldpstp_address fails.
4282 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4284         * tree-vector-builder.h
4285         (tree_vector_builder::binary_encoded_nelts): Declare.
4286         * tree-vector-builder.c
4287         (tree_vector_builder::binary_encoded_nelts): New function.
4288         * fold-const.c (negate_expr_p): Likewise.
4289         (operand_equal_p, fold_checksum_tree): Likewise.
4290         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4291         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4292         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4293         (uniform_vector_p): Likewise.
4294         * varasm.c (const_hash_1, compare_constant): Likewise.
4295         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4296         (valid_lattice_transition): Operate directly on the VECTOR_CST
4297         encoding.
4298         * ipa-icf.c: Include tree-vector-builder.h.
4299         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4300         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4302 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4304         * tree.c (build_vector): Delete.
4305         * tree.h (build_vector): Make static and move into the self-testing
4306         block.
4308 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4310         * vector-builder.h (vector_builder::derived): New const overload.
4311         (vector_builder::elt): New function.
4312         * tree-vector-builder.h (tree_vector_builder::type): New function.
4313         (tree_vector_builder::apply_step): Declare.
4314         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4315         function.
4316         * gimple-fold.h (tree_vector_builder): Declare.
4317         (gimple_build_vector): Take a tree_vector_builder instead of a
4318         type and vector of elements.
4319         * gimple-fold.c (gimple_build_vector): Likewise.
4320         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4321         accordingly.
4322         (get_initial_defs_for_reduction): Likewise.
4323         (vectorizable_induction): Likewise.
4325 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4327         * tree-vector-builder.h
4328         (tree_vector_builder::new_binary_operation): Declare.
4329         * tree-vector-builder.c
4330         (tree_vector_builder::new_binary_operation): New function.
4331         * fold-const.c (fold_relational_const): Use it.
4332         (const_binop): Likewise.  Check that both input vectors have
4333         the same number of elements, thus excluding things like WIDEN_SUM.
4334         Check whether it is possible to operate directly on the encodings
4335         of stepped inputs.
4337 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4339         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4340         new_unary_operation, operating only on the encoded elements.
4341         (const_unop): Likewise.
4342         (exact_inverse): Likewise.
4343         (distributes_over_addition_p): New function.
4344         (const_binop): Use tree_vector_builder and new_unary_operation
4345         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4346         on the encoded elements unless the encoding is strided and the
4347         operation does not distribute over addition.
4348         (fold_convert_const):  Use tree_vector_builder and
4349         new_unary_operation.  Operate only on the encoded elements
4350         for truncating integer conversions, or for non-stepped encodings.
4352 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4354         * config/sparc/sparc.c: Include tree-vector-builder.h.
4355         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4356         * expmed.c: Include tree-vector-builder.h.
4357         (make_tree): Use tree_vector_builder instead of build_vector.
4358         * fold-const.c: Include tree-vector-builder.h.
4359         (const_binop): Use tree_vector_builder instead of build_vector.
4360         (const_unop): Likewise.
4361         (native_interpret_vector): Likewise.
4362         (fold_vec_perm): Likewise.
4363         (fold_ternary_loc): Likewise.
4364         * gimple-fold.c: Include tree-vector-builder.h.
4365         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4366         of build_vector.
4367         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4368         (simplify_vector_constructor): Use tree_vector_builder instead
4369         of build_vector.
4370         * tree-vect-generic.c: Include tree-vector-builder.h.
4371         (add_rshift): Use tree_vector_builder instead of build_vector.
4372         (expand_vector_divmod): Likewise.
4373         (optimize_vector_constructor): Likewise.
4374         * tree-vect-loop.c: Include tree-vector-builder.h.
4375         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4376         of build_vector.  Explicitly use a stepped encoding for
4377         { 1, 2, 3, ... }.
4378         * tree-vect-slp.c: Include tree-vector-builder.h.
4379         (vect_get_constant_vectors): Use tree_vector_builder instead
4380         of build_vector.
4381         (vect_transform_slp_perm_load): Likewise.
4382         (vect_schedule_slp_instance): Likewise.
4383         * tree-vect-stmts.c: Include tree-vector-builder.h.
4384         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4385         (vect_gen_perm_mask_any): Likewise.
4386         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4387         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4388         of build_vector.
4389         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4390         encoding.
4392 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4394         * doc/generic.texi (VECTOR_CST): Describe new representation of
4395         vector constants.
4396         * vector-builder.h: New file.
4397         * tree-vector-builder.h: Likewise.
4398         * tree-vector-builder.c: Likewise.
4399         * Makefile.in (OBJS): Add tree-vector-builder.o.
4400         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4401         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4402         (tree_vector): Change the number of elements to
4403         vector_cst_encoded_nelts.
4404         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4405         (VECTOR_CST_ELTS): Delete.
4406         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4407         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4408         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4409         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4410         (VECTOR_CST_ENCODED_ELT): Likewise.
4411         (vector_cst_encoded_nelts): New function.
4412         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4413         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4414         (vector_cst_int_elt, vector_cst_elt): Declare.
4415         * tree.c: Include tree-vector-builder.h.
4416         (tree_code_size): Abort if passed VECTOR_CST.
4417         (tree_size): Update for new VECTOR_CST layout.
4418         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4419         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4420         (build_vector): Use tree_vector_builder.
4421         (vector_cst_int_elt, vector_cst_elt): New functions.
4422         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4423         encoded elements and then create the vector in the canonical form.
4424         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4425         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4426         (tree_c_tests): Call test_vector_cst_patterns.
4427         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4428         VECTOR_CST fields.
4429         (hash_tree): Likewise.
4430         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4431         (streamer_write_tree_header): Likewise.
4432         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4433         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4434         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4436 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4438         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4439         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4440         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4442 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4443             Richard Biener  <rguenther@suse.de>
4445         PR tree-optimization/81303
4446         * Makefile.in (gimple-loop-interchange.o): New object file.
4447         * common.opt (floop-interchange): Reuse the option from graphite.
4448         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4449         -floop-interchange and mention it for -O3.
4450         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4451         * gimple-loop-interchange.cc: New file.
4452         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4453         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4454         * passes.def (pass_linterchange): New pass.
4455         * timevar.def (TV_LINTERCHANGE): New time var.
4456         * tree-pass.h (make_pass_linterchange): New declaration.
4457         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4458         interchange.  Record IV before/after increment in new parameters.
4459         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4460         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4461         path check into...
4462         (check_reduction_path): ...New function here.
4463         * tree-vectorizer.h (check_reduction_path): New declaration.
4465 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4467         PR target/83252
4468         PR rtl-optimization/80818
4469         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4470         always early clobbered.
4471         * lra-lives.c (process_bb_lives): Check input hard regs for early
4472         clobbered non-operand hard reg.
4474 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4476         PR middle-end/83164
4477         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4478         types_compatible_p, just that TYPE_MODE is the same.
4480 2017-12-07  Martin Sebor  <msebor@redhat.com>
4482         PR c/81544
4483         * attribs.c (empty_attribute_table): Initialize new member of
4484         struct attribute_spec.
4485         (decl_attributes): Add argument.  Handle mutually exclusive
4486         combinations of attributes.
4487         (selftests::test_attribute_exclusions): New function.
4488         (selftests::attribute_c_tests): Ditto.
4489         * attribs.h (decl_attributes): Add default argument.
4490         * selftest.h (attribute_c_tests): Declare.
4491         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4492         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4493         member.
4494         * doc/extend.texi (Common Function Attributes): Update const and pure.
4495         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4496         of struct attribute_spec.
4497         * config/arc/arc.c (arc_attribute_table): Same.
4498         * config/arm/arm.c (arm_attribute_table): Same.
4499         * config/avr/avr.c ( avr_attribute_table): Same.
4500         * config/bfin/bfin.c (bfin_attribute_table): Same.
4501         * config/cr16/cr16.c (cr16_attribute_table): Same.
4502         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4503         * config/h8300/h8300.c (h8300_attribute_table): Same.
4504         * config/i386/i386.c (ix86_attribute_table): Same.
4505         * config/ia64/ia64.c (ia64_attribute_table): Same.
4506         * config/m32c/m32c.c (m32c_attribute_table): Same.
4507         * config/m32r/m32r.c (m32r_attribute_table): Same.
4508         * config/m68k/m68k.c (m68k_attribute_table): Same.
4509         * config/mcore/mcore.c (mcore_attribute_table): Same.
4510         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4511         * config/mips/mips.c (mips_attribute_table): Same.
4512         * config/msp430/msp430.c (msp430_attribute_table): Same.
4513         * config/nds32/nds32.c (nds32_attribute_table): Same.
4514         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4515         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4516         * config/rl78/rl78.c (rl78__attribute_table): Same.
4517         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4518         * onfig/rx/rx.c (rx_attribute_table): Same.
4519         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4520         * config/sh/sh.c (sh_attribute_table): Same.
4521         * config/sparc/sparc.c (sparc_attribute_table): Same.
4522         * config/spu/spu.c (spu_attribute_table): Same.
4523         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4524         * config/v850/v850.c (v850_attribute_table): Same.
4525         * config/visium/visium.c (visium_attribute_table): Same.
4527 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4529         PR target/82641
4530         * config/arm/arm.c (INCLUDE_STRING): Define.
4531         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4532         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4533         and .fpu.
4535 2017-12-07  Michael Matz  <matz@suse.de>
4537         Add unroll and jam pass
4539         * gimple-loop-jam.c: New file.
4540         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4541         * common.opt (funroll-and-jam): New option.
4542         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4543         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4544         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4545         * passes.def: Add pass_loop_jam.
4546         * timevar.def (TV_LOOP_JAM): Add.
4547         * tree-pass.h (make_pass_loop_jam): Declare.
4548         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4549         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4550         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4551         to flow_loop_tree_node_add.
4552         (duplicate_subloops, copy_loops_to): Append to sibling list.
4553         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4554         * doc/invoke.texi (-funroll-and-jam): Document new option.
4555         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4557 2017-12-07  Richard Biener  <rguenther@suse.de>
4559         PR tree-optimization/83296
4560         PR tree-optimization/67769
4561         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4562         flow sensitive info in an unrelated BB.
4563         (value_replacement): Use reset_flow_sensitive_info.
4564         (minmax_replacement): Reset flow sensitive info on the def
4565         we move.  Do not reset flow sensitive info in the whole BB
4566         we move the stmt to.
4567         (abs_replacement): Likewise.
4569 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4571         PR target/43871
4572         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4573         rs6000_cpu to the given -mcpu=, or to the default processor.
4575 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4577         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4578         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4579         instead of rs6000_cpu_attr.
4580         (rs6000_adjust_cost): Ditto.
4581         (is_microcoded_insn): Ditto.
4582         (rs6000_adjust_priority): Ditto.
4583         (rs6000_issue_rate): Ditto.
4584         (rs6000_use_sched_lookahead): Ditto.
4585         (rs6000_use_sched_lookahead_guard): Ditto.
4586         (rs6000_sched_reorder): Ditto.
4587         (force_new_group): Ditto.
4588         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4589         (group_ending_nop): Ditto.
4591 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4593         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4594         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4595         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4596         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4597         appropriate.
4598         (rs6000_reassociation_width): Ditto.
4599         (rs6000_emit_epilogue): Ditto.
4600         (rs6000_adjust_cost): Ditto.
4601         (is_microcoded_insn): Ditto.
4602         (is_cracked_insn): Ditto.
4603         (rs6000_adjust_priority): Ditto.
4604         (rs6000_sched_reorder): Ditto.
4605         (rs6000_sched_reorder2): Ditto.
4606         (insn_must_be_first_in_group): Ditto.
4607         (insn_must_be_last_in_group): Ditto.
4608         (rs6000_register_move_cost): Ditto.
4609         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4610         rs6000_cpu.
4612 2017-12-07  Julia Koval  <julia.koval@intel.com>
4614         * config.gcc: Add vaesintrin.h.
4615         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4616         New type.
4617         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4618         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4619         New builtins.
4620         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4621         * config/i386/immintrin.h: Include vaesintrin.h.
4622         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4623         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4624         _mm_aesdec_epi128): New intrinsics.
4626 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4628         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4629         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4630         c-family/c-spellcheck.cc.
4631         (best_macro_match::best_macro_match): Likewise.
4632         * spellcheck-tree.h
4633         (struct edit_distance_traits<cpp_hashnode *>): Move to
4634         c-family/c-spellcheck.h.
4635         (class best_macro_match): Likewise.
4637 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4639         PR tree-optimization/83293
4640         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4641         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4642         might insert into empty bb.
4644         PR sanitizer/81281
4645         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4646         simplify for plus with :c added, and pointer_plus without that.
4647         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4648         with undefined overflow and the conversion is not widening,
4649         perform negation in utype and only convert to type afterwards.
4650         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4651         simplify for plus with :c added, and pointer_plus without that.
4652         If type is integral with undefined overflow and the conversion is
4653         not widening, perform minus in utype and only convert to type
4654         afterwards.  Move the last pointer_diff_expr simplify into the
4655         two outermost ifs.
4657 2017-12-06  Martin Sebor  <msebor@redhat.com>
4659         PR tree-optimization/82646
4660         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4661         strncpy, not maxlen.
4663 2017-12-06  Martin Sebor  <msebor@redhat.com>
4665         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4666         nonstring.
4668         PR tree-optimization/83075
4669         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4670         strncat/strncpy don't change length of source string.
4672 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4674         Revert
4675         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4677         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4678         to prevent b2bst errata sequence.
4679         (sqrtdf2_fix): Likewise.
4681 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4683         PR tree-optimization/81945
4684         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4685         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4686         to dest_cfun has orig_loop_num set, either remap it to the new
4687         loop number if the loop got moved too, or clear it.
4689 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4691         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4692         to reservation.
4693         (thunderx2t99_nothing): New insn reservation.
4694         (thunderx2t99_mrs): New insn reservation.
4695         (thunderx2t99_multiple): New insn reservation.
4696         (thunderx2t99_alu_basi): Add bfx to reservation.
4697         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4699 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4701         PR target/82248
4702         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4704 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4706         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4707         tree-ssa-pre.c::remove_dead_inserted_code.
4708         * tree-ssa-dce.h: New file.
4709         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4710         (remove_dead_inserted_code): Move and rename to function
4711         tree-ssa-dce.c::simple_dce_from_worklist.
4712         (pass_pre::execute): Update use.
4714 2017-12-05  Richard Biener  <rguenther@suse.de>
4716         PR tree-optimization/83277
4717         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4718         to code-gen liveout vars.
4720 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4722         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4723         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4724         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4725         split condition.
4727 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4729         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4730         function.
4731         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4732         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4733         ASAN is enabled.
4735 2017-12-05  Richard Biener  <rguenther@suse.de>
4737         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4738         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4739         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4740         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4742 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4744         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4746 2017-12-05  Martin Liska  <mliska@suse.cz>
4747             Jakub Jelinek  <jakub@redhat.com>
4749         * doc/invoke.texi: Document the options.
4750         * flag-types.h (enum sanitize_code): Add
4751         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4752         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4753         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4754         * opts.c: Define new sanitizer options.
4755         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4756         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4758 2017-12-05  Julia Koval  <julia.koval@intel.com>
4760         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4761         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4762         (ix86_handle_option): Handle -mavx512vnni.
4763         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4764         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4765         * config/i386/i386-c (__AVX512VNNI__): New.
4766         * config/i386/i386.c (ix86_target_string): Handle new option.
4767         (ix86_valid_target_attribute_inner_p): Handle new option.
4768         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4769         * config/i386/i386.opt (mavx512vnni): New option.
4771 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4773         PR target/81616
4774         * athlon.md: Disable for generic.
4775         * haswell.md: Enable for generic.
4776         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4777         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4778         to 4.
4779         (ix86_adjust_cost): Move generic to haswell path.
4781 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4783         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4784         instead of MEM_P in a couple more places.  Fix formatting issues.
4786 2017-12-04  Jim Wilson  <jimw@sifive.com>
4788         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4789         instead of GP_REG_LAST-1.
4790         (riscv_adjust_libcall_cfi_prologue): Likewise.
4791         (riscv_adjust_libcall_cri_epilogue): Likewise.
4792         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4793         comment.
4795 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4797         * ipa-pure-const.c (check_decl): Add missing newline.
4798         (state_from_flags): Likewise.
4800 2017-12-04  Jeff Law  <law@redhat.com>
4802         PR tree-optimizatin/78496
4803         * gimple-ssa-evrp-analyze.h
4804         (evrp_range_analyzer::get_vr_values): Simplify.
4805         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4806         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4807         and gimple-ssa-evrp-analyze.h.
4808         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4809         (simplify_stmt_for_jump_threading): Copy a blob of code from
4810         tree-vrp.c to use ranges to simplify statements.
4811         (dom_opt_dom_walker::before_dom_children): Call
4812         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4813         (dom_opt_dom_walker::after_dom_children): Similarly for
4814         evrp_range_analyzer::leave.
4815         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4816         conditionals.
4818         * gimple-ssa-evrp-analyze.c
4819         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4820         vr_values::update_value_range so preexisting range info is
4821         medged with new range info, even if the new range is VR_VARYING.
4823 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4825         * combine.c: Adjust comment.
4826         (use_crosses_set_p): Delete.
4827         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4828         (try_combine): Ditto.
4830 2017-12-04  Richard Biener  <rguenther@suse.de>
4832         PR tree-optimization/83255
4833         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4834         Re-add zero-iteration check.
4836 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4838         PR rtl-optimization/83245
4839         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4840         hard registers as earlyclobber, also if not in an asm.
4842 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4844         PR bootstrap/83265
4845         Revert
4846         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4848         PR target/43871
4849         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4850         rs6000_cpu based on cpu_index, not tune_index.
4852 2017-12-04  Richard Biener  <rguenther@suse.de>
4854         PR tree-optimization/83238
4855         * graphite-scop-detection.c (scop_detection::merge_sese): Make
4856         code match comment, rejecting invalid SESE regions.
4858 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
4860         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4861         require base operand is a REG_POINTER prior to reload on targets
4862         with non-equivalent space registers.
4864 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4866         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
4867         (update_specialized_profile): Fix updating of counts.
4868         (perhaps_add_new_callers): Likewise.
4870 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4872         PR target/81616
4873         * x86-tune.def: Remove obsolette FIXMEs.
4874         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
4875         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
4876         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
4877         Enable for generic.
4878         (X86_TUNE_PAD_RETURNS): Disable for generic.
4880 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
4882         PR tree-optimization/83170
4883         PR tree-optimization/83241
4884         * gimple-ssa-store-merging.c
4885         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
4886         gimple_vuse (ins_stmt) in case it has changed.
4887         (imm_store_chain_info::output_merged_store): Likewise.
4889         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
4890         POINTER_DIFF_EXPR.
4892         PR c++/81212
4893         * tree-cfg.c (pass_warn_function_return::execute): Handle
4894         __builtin_ubsan_handle_missing_return like __builtin_unreachable
4895         with BUILTINS_LOCATION.
4897         PR target/78643
4898         PR target/80583
4899         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
4900         is BLKmode for vector field with vector raw mode, use TYPE_MODE
4901         instead of DECL_MODE.
4903         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
4904         last argument to rtx pointer.
4905         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
4906         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
4907         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
4908         depending on the chosen ISAs.
4909         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
4910         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
4911         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
4912         callers.
4913         * config/i386/sse.md (mov<mode>_internal): Likewise.
4914         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
4916 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4918         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
4920 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4922         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
4923         parameters from prototype.
4924         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
4925         parameters.  Don't print a comment.
4926         (emit_fusion_gpr_load): Adjust.
4927         (emit_fusion_load_store): Adjust.
4928         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
4929         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
4930         comment on the second line.
4932 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4934         PR target/43871
4935         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4936         rs6000_cpu based on cpu_index, not tune_index.
4938 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4940         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
4941         which_alternative instead of which_alternative + 1.
4942         (output_asm_insn): Print an extra tab if the template is short.
4944 2017-12-01  Jim Wilson  <jimw@sifive.com>
4946         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
4947         comment.
4948         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
4949         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
4950         reference.
4951         * doc/tm.texi: Regenerate.
4953 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
4955         PR target/81959
4956         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
4957         whether we can allocate pseudos before trying to fix an address.
4958         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
4959         memory address is indexed or indirect.
4960         (floatuns_<mode>si2_hw2): Likewise.
4962 2017-12-01  Jason Merrill  <jason@redhat.com>
4964         * Makefile.in (TAGS): Add c-family/*.cc.
4966 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
4968         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
4969         (call_value_insn): Likewise.
4970         (sibcall_insn): Likewise.
4971         (sibcall_value_insn): Likewise.
4972         (movsi_aarch64): Likewise.
4973         (movdi_aarch64): Likewise.
4974         (add_losym_): Likewise.
4975         (ldr_got_small_): Likewise.
4976         (ldr_got_small_sidi): Likewise.
4977         (ldr_got_small_28k_): Likewise.
4978         (ldr_got_small_28k_sidi): Likewise.
4979         * config/aarch64/aarch64.c (aarch64_print_address_internal):
4980         Move output_addr_const to symbolic case. Add error check.
4982 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4984         * config/s390/predicates.md (plus16_Q_operand): New predicate.
4985         * config/s390/s390.md: Disable MVC merging peephole if it would
4986         disable operand forwarding.
4987         (new peephole2): Split MVCs if it would turn them into up to 2
4988         forwardable MVCs.
4990 2017-12-01  Richard Biener  <rguenther@suse.de>
4992         PR tree-optimization/83232
4993         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
4994         detection of same access. Instead of breaking the group here
4995         do not consider the duplicate.  Add comment explaining real fix.
4997 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
4999         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5001 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5003         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5005 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5007         * function.h (struct function): Remove cilk_frame_decl,
5008         is_cilk_function and calls_cilk_spawn fields.
5009         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5010         field.
5011         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5012         cilk_elemental field.
5013         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5014         * target.def: Adjust comment.
5015         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5016         Don't test cilk_elemental.
5018         PR tree-optimization/83233
5019         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5020         bswap_stat name for the struct.
5022         PR c/79153
5023         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5024         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5025         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5026         SWITCH_BREAK_LABEL_P set on the label.
5027         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5028         added for default case if it was missing and not all cases covered.
5029         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5030         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5031         set on the label.
5032         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5033         to avoid -Wimplicit-fallthrough warning.
5034         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5035         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5037         PR tree-optimization/83221
5038         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5039         down by 16.
5040         (init_reassoc): Formatting fix.
5042         PR sanitizer/81275
5043         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5044         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5045         is diagnosed.
5047         PR sanitizer/83219
5048         * tree-cfg.c: Include asan.h.
5049         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5051 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5053         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5055 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5057         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5058         vpcmpeqd instruction.
5060 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5062         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5063         registers type for 128bit mode.
5065 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5067         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5068         of ;;.
5069         * gengtype-state.c (read_state_pair): Likewise.
5070         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5071         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5072         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5073         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5074         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5075         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5076         * builtins.c (fold_builtin_3): Likewise.
5077         * graphite-scop-detection.c
5078         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5079         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5080         Likewise.
5082 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5084         PR sanitizer/81697
5085         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5086         parameter. Return true if ignore_decl_rtl_set_p is true and other
5087         conditions are satisfied.
5088         * asan.h (asan_protect_global): Add new parameter.
5089         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5090         to asan_protect_global calls.
5092 2017-11-30  Jim Wilson  <jimw@sifive.com>
5094         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5095         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5096         -mstrict-align, add info on default value.  Delete redundant lines for
5097         -mabi.  Add missing -mexplicit-relocs docs.
5099 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5101         * config/arc/arc.md (trap): New pattern.
5103 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5105         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5106         instruction to end into a delay slot.
5107         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5108         can be placed into a delay slot against reg_note.
5110 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5112         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5113         labels number of usages.
5115 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5117         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5118         function.
5119         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5121 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5123         PR target/83210
5124         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5125         multiplication by power of 2 constant into two shifts + comparison.
5127 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5129         PR target/81616
5130         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5132 2017-11-30  Richard Biener  <rguenther@suse.de>
5134         PR tree-optimization/83202
5135         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5136         allow_peel argument and guard peeling.
5137         (canonicalize_loop_induction_variables): Likewise.
5138         (canonicalize_induction_variables): Pass false.
5139         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5140         peeling from cunrolli.
5142 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5144         * combine.c (try_combine): Print a message to dump file whenever
5145         I0, I1, or I2 cannot be combined into I3.
5147 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5149         PR rtl-optimization/83156
5150         PR rtl-optimization/82621
5151         * combine.c (try_combine): Don't split an I2 if one of the dests is
5152         set again before I3.  Allow unused dests.
5154 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5156         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5158 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5160         PR rtl-optimization/80818
5161         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5162         recursively.  Use insn code for clobber.
5163         (lra_set_insn_recog_data): Pass the new arg to
5164         collect_non_operand_hard_regs.
5165         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5166         code for clobber.
5167         (lra_update_insn_regno_info): Pass insn to
5168         add_regs_to_insn_regno_info.
5170 2017-11-29  Jim Wilson  <jimw@sifive.com>
5171             Andrew Waterman  <andrew@sifive.com>
5173         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5174         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5175         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5176         use SINGLE_SHIFT_COST.
5177         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5178         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5180 2017-11-29  Julia Koval  <julia.koval@intel.com>
5182         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5183         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5184         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5185         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5186         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5187         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5188         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5189         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5190         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5191         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5192         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5193         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5194         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5195         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5196         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5197         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5198         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5199         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5200         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5201         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5202         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5203         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5204         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5205         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5206         __builtin_ia32_vpshldv_v2di_mask,
5207         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5208         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5209         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5211 2017-11-29  Julia Koval  <julia.koval@intel.com>
5213         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5214         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5215         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5216         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5217         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5218         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5219         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5220         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5221         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5222         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5223         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5224         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5225         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5226         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5227         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5228         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5229         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5230         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5231         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5232         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5233         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5235 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5237         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5238         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5239         workaround.
5241 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5243         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5244         to prevent b2bst errata sequence.
5245         (sqrtdf2_fix): Likewise.
5247 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5249         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5250         (div_sqrt_insn_p): New function.
5251         (sparc_do_work_around_errata): Insert NOP instructions to
5252         prevent sequences that could trigger the TN-0013 errata for
5253         certain LEON3 processors.
5254         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5255         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5256         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5257         (in_branch_delay): Prevent div and sqrt in delay slot if
5258         fix_lost_divsqrt.
5259         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5261 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5263         * config/sparc/sparc.c (atomic_insn_p): New function.
5264         (sparc_do_work_around_errata): Insert NOP instructions to
5265         prevent sequences that could trigger the TN-0010 errata for
5266         UT700.
5267         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5268         instruction referable in atomic_insns_p.
5270 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5272         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5273         (atomic_compare_and_swap_leon3_1): Likewise.
5274         (ldstub): Likewise.
5276 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5278         * config/sparc/sparc.c (fpop_insn_p): New function.
5279         (sparc_do_work_around_errata): Insert NOP instructions to
5280         prevent sequences that could trigger the TN-0012 errata for
5281         GR712RC.
5282         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5283         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5284         (in_branch_annul_delay): Prevent floating-point instructions
5285         in delay slot of annulled integer branch.
5287 2017-11-29  Richard Biener  <rguenther@suse.de>
5289         PR tree-optimization/83202
5290         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5291         (bst_fail): Use it.
5292         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5293         nodes vectorized to the same stmts multiple times.
5294         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5295         (vect_analyze_slp_instance): Adjust.
5296         (scalar_stmts_to_slp_tree_map_t): New typedef.
5297         (vect_schedule_slp_instance): Add a map recording the SLP node
5298         representing the vectorized stmts for a set of scalar stmts.
5299         Avoid code-generating redundancies.
5300         (vect_schedule_slp): Allocate map and pass it down.
5302 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5304         PR c++/83187
5305         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5306         type if it is new.
5308 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5310         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5311         cases for printing LDP/STP memory addresses.
5312         (aarch64_print_address_internal): Renamed from
5313         aarch64_print_operand_address, added parameter, add Pmode check.
5314         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5315         (aarch64_print_operand_address): Indirect to
5316         aarch64_print_address_internal.
5317         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5318         'y' operand output specifier.
5320 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5322         PR middle-end/83185
5323         * tree.c (build_simple_mem_ref_loc): Handle
5324         get_addr_base_and_unit_offset returning a MEM_REF.
5326         PR middle-end/80929
5327         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5329         PR target/80819
5330         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5331         alternative.
5333 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5335         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5336         m_x_offset.
5337         (layout::move_to_column): Likewise.
5339 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5341         PR sanitizer/81275
5342         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5343         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5345 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5346             Martin Jambor  <mjambor@suse.cz>
5348         PR ipa/82808
5349         * tree.h (expr_type_first_operand_type_p): Declare
5350         * tree.c (expr_type_first_operand_type_p): New function.
5351         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5352         (ipa_value_from_jfunc): Adjust declaration.
5353         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5354         Use it as result type for arithmetics, unless it is NULL in which case
5355         be more conservative.
5356         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5357         ipa_get_jf_pass_through_result.
5358         (propagate_vals_across_pass_through): Likewise.
5359         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5360         is to propagate_vals_across_pass_through.
5361         (propagate_constants_across_call): Pass PARM_TYPE to
5362         propagate_scalar_across_jump_function.
5363         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5364         ipa_value_from_jfunc.
5365         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5366         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5367         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5368         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5369         target_type, pass it to ipa_value_from_jfunc.
5370         (update_indirect_edges_after_inlining): Pass parameter type to
5371         try_make_edge_direct_simple_call.
5373 2017-11-28  Jeff Law  <law@redhat.com>
5375         * gimple-ssa-evrp-analyze.c
5376         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5377         refine ranges if scev_initialized_p returns true.
5378         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5380 2017-11-28  Julia Koval  <julia.koval@intel.com>
5382         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5383         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5384         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5385         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5386         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5387         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5388         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5389         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5390         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5391         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5392         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5393         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5394         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5395         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5396         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5397         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5398         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5399         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5400         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5401         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5402         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5403         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5405 2017-11-28  Julia Koval  <julia.koval@intel.com>
5407         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5408         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5409         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5410         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5411         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5412         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5413         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5414         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5415         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5416         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5417         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5418         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5419         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5420         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5421         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5422         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5423         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5424         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5425         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5426         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5427         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5428         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5429         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5430         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5431         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5432         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5433         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5434         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5435         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5436         __builtin_ia32_vpshld_v2di_mask): New builtins.
5437         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5438         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5440 2017-11-28  Richard Biener  <rguenther@suse.de>
5442         PR tree-optimization/80776
5443         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5444         Declare.
5445         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5446         New function.
5447         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5448         If the incoming edge is an effective fallthru because the other
5449         edge only reaches a __builtin_unreachable () then record ranges
5450         derived from the controlling condition in SSA info.
5451         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5452         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5454 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5456         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5457         on mingw build hosts.
5459 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5461         PR debug/81307
5462         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5464 2017-11-28  Richard Biener  <rguenther@suse.de>
5466         PR middle-end/83141
5467         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5468         copies generated from memcpy use a character array as reference
5469         type.
5471 2017-11-28  Julia Koval  <julia.koval@intel.com>
5472             Sebastian Peryt  <sebastian.peryt@intel.com>
5474         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5475         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5476         cilk-common.o, cilk.h, cilk-common.c): Remove.
5477         * builtin-types.def
5478         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5479         * builtins.c (is_builtin_name): Remove cilkplus condition.
5480         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5481         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5482         cilk-builtins.def, cilkplus.def): Remove.
5483         * cif-code.def (CILK_SPAWN): Remove.
5484         * cilk-builtins.def: Delete.
5485         * cilk-common.c: Ditto.
5486         * cilk.h: Ditto.
5487         * cilkplus.def: Ditto.
5488         * config/darwin.h (fcilkplus): Delete.
5489         * cppbuiltin.c: Ditto.
5490         * doc/extend.texi: Remove cilkplus doc.
5491         * doc/generic.texi: Ditto.
5492         * doc/invoke.texi: Ditto.
5493         * doc/passes.texi: Ditto.
5494         * gcc.c (fcilkplus): Remove.
5495         * gengtype.c (cilk.h): Remove.
5496         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5497         support.
5498         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5499         Remove.
5500         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5501         gimplify_call_expr,
5502         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5503         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5504         cilkplus conditions.
5505         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5506         inline_read_section): Ditto.
5507         * ipa-inline-analysis.c (cilk.h): Remove.
5508         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5509         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5510         append_linker_options): Remove condition for fcilkplus.
5511         * lto/lto-lang.c (cilk.h): Remove.
5512         (lto_init): Remove condition for fcilkplus.
5513         * omp-expand.c (expand_cilk_for_call): Delete.
5514         (expand_omp_taskreg, expand_omp_for_static_chunk,
5515         expand_omp_for): Remove cilkplus
5516         conditions.
5517         (expand_cilk_for): Delete.
5518         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5519         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5520         execute_lower_omp, diagnose_sb_0): Ditto.
5521         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5522         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5523         * tree-nested.c: Ditto.
5524         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5525         (dump_generic_node): Ditto.
5526         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5527         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5528         Delete.
5529         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5531 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5533         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5534         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5535         for vector mode and !TARGET_SIMD.
5537 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5539         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5540         Adjust comment.
5541         * tree.h (SWITCH_LABELS): Remove.
5542         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5543         assert SWITCH_BODY is non-NULL.
5544         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5545         handling.
5546         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5548         PR tree-optimization/80788
5549         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5550         has TREE_OVERFLOW set, call drop_tree_overflow.
5552 2017-11-28  Richard Biener  <rguenther@suse.de>
5554         PR tree-optimization/83158
5555         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5557 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5559         PR 81288/target
5560         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5561         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5563 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5565         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5566         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5568 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5570         PR middle_end/82333
5571         * varasm.c (compare_constant): Take the mode of the constants into
5572         account when comparing floating point constants.
5574 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5576         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5577         from explicit instantiation of debug_helper.
5578         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5580 2017-11-27  Richard Biener  <rguenther@suse.de>
5582         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5583         refactor a bit.
5585 2017-11-27  Richard Biener  <rguenther@suse.de>
5587         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5588         (type_hash_canon): Also clear int_cst_hash_table entry for
5589         TYPE_MIN/MAX_VALUE.
5590         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5592 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5594         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5595         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5596         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5598 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5600         * hash-map.h (gt_cleare_cache): Avoid UB.
5602 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5604         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5605         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5606         (decide_unroll_runtime_iterations): Likewise.
5607         (decide_unroll_stupid): Likewise.
5609 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5611         PR target/83109
5612         * config/i386/i386.md: Add a loop with incssp.
5614 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5616         PR tree-optimization/81248
5617         * tree-sra.c (splice_param_accesses): Remove size check.
5618         (decide_one_param_reduction): Fix size check.
5619         * gimple-pretty-print.c (dump_profile): Silence warning.
5620         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5622 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5624         PR debug/81307
5625         * dbxout.c (lastlineno): New variable.
5626         (dbx_debug_hooks): Use dbxout_switch_text_section as
5627         switch_text_section debug hook.
5628         (dbxout_function_end): Switch to current_function_section
5629         rather than function_section.  If crtl->has_bb_partition,
5630         output just one N_FUN, depending on in_cold_section_p.
5631         (dbxout_source_line): Remember last lineno in lastlineno.
5632         (dbxout_switch_text_section): New function.
5633         (dbxout_function_decl): Adjust dbxout_block caller.
5634         (dbx_block_with_cold_children): New function.
5635         (dbxout_block): Return true if any LBRAC/RBRAC have been
5636         emitted.  Use dbx_block_with_cold_children at depth == 0
5637         in second partition.  Add PARENT_BLOCKNUM argument, pass
5638         it optionally adjusted to children.  Output LBRAC/RBRAC
5639         around recursive call only if the block is in the current
5640         partition, if not and anything was output, emit empty
5641         range LBRAC/RBRAC.
5642         * final.c (final_scan_insn): Compute cold_function_name
5643         before calling switch_text_section debug hook.  Call
5644         that hook even if dwarf2out_do_frame if not emitting
5645         dwarf debug info.
5647         PR target/83100
5648         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5649         TREE_READONLY decls.
5651 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5653         PR rtl-optimization/82488
5654         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5656 2017-11-26  Julia Koval  <julia.koval@intel.com>
5658         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5659         skylake-avx512.
5660         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5661         skylake_cost): New.
5663 2017-11-26  Julia Koval  <julia.koval@intel.com>
5665         * config/i386/driver-i386.c (host_detect_local_cpu):
5666         Detect skylake-avx512.
5668 2017-11-26  Julia Koval  <julia.koval@intel.com>
5670         * config.gcc: Add -march=cannonlake.
5671         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5672         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5673         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5674         (PTA_CANNONLAKE): New.
5675         (processor_target_table): Add cannonlake.
5676         (ix86_option_override_internal): Ditto.
5677         (fold_builtin_cpu): Ditto.
5678         (get_builtin_code_for_version): Handle cannonlake.
5679         (M_INTEL_COREI7_CANNONLAKE): New.
5680         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5681         * doc/invoke.texi: Add -march=cannonlake.
5683 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5685         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5686         (try_init_one_plugin): Alternative implementation for MinGW.
5687         * Makefile.in (plugin_implib): New.
5688         (gengtype-lex.c): Fix broken AIX workaround.
5689         * configure: Regenerate.
5690         * doc/plugins.texi: Document support for MinGW.
5692 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5694         PR rtl-optimization/81553
5695         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5696         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5697         is a shift where C1 has different mode than the whole shift, use C1's
5698         mode for MULT rather than the shift's mode.
5700         PR target/82848
5701         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5702         builtins not enabled in the currently selected ISA.
5704 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5706         PR tree-optimization/71026
5707         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5708         (insert_reciprocals): Change to insert reciprocals before a division
5709         by a square and to insert the square of a reciprocal.
5710         (execute_cse_reciprocals_1): Change to consider division by a square.
5711         (register_division_in): Add importance parameter.
5713 2017-11-24  Richard Biener  <rguenther@suse.de>
5715         PR tree-optimization/82402
5716         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5717         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5719 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5721         * match.pd (0-ptr): New transformation.
5723 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5725         PR bootstrap/83015
5726         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5727         when skipping checking.
5729 2017-11-24  Richard Biener  <rguenther@suse.de>
5731         PR tree-optimization/83128
5732         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5733         (vn_reference_lookup_3): Likewise.
5735 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5737         PR sanitizer/83014
5738         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5739         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5740         tree_to_uhwi twice.
5742         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5743         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5744         check for the latter with gimple_call_builtin_p.  Do not handle
5745         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5747 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5749         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5750         vdot_u32.
5752 2017-11-23  Julia Koval  <julia.koval@intel.com>
5754         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5755         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5756         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5757         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5758         _mm512_maskz_expandloadu_epi16): New intrinsics.
5759         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5760         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5761         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5762         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5763         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5764         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5765         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5766         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5767         _mm256_maskz_expandloadu_epi8): New intrinsics.
5768         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5769         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5770         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5771         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5772         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5773         * config/i386/sse.md (VI248_VLBW): New iterator.
5774         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5776 2017-11-23  Julia Koval  <julia.koval@intel.com>
5778         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5779         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5780         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5781         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5782         _mm512_mask_compressstoreu_epi16): New.
5783         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5784         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5785         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5786         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5787         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5788         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5789         _mm256_mask_compressstoreu_epi8): New.
5790         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5791         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5792         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5793         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5794         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5795         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5796         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5797         __builtin_ia32_compresshi128_mask,
5798         __builtin_ia32_compressstoreuqi512_mask,
5799         __builtin_ia32_compressstoreuhi512_mask,
5800         __builtin_ia32_compressstoreuqi256_mask,
5801         __builtin_ia32_compressstoreuqi128_mask,
5802         __builtin_ia32_compressstoreuhi256_mask,
5803         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5804         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5805         array for flags2.
5806         (ix86_expand_special_args_builtin): Handle new types.
5807         (s4fma_expand): Handle new builtin array.
5808         * config/i386/immintrin.h: Include new headers.
5809         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5810         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5812 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5814         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5815         fixes.  Declare temp and g variables at the top in order to avoid
5816         {} in most of the cases.
5818 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5820         * match.pd (ptr-0): New transformation.
5822 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5824         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5825         (struct addr_mode_cost_table): New.
5826         (struct tune_params): Add field addr_mode_costs.
5827         * config/arm/arm.c (generic_addr_mode_costs): New.
5828         (arm_slowmul_tune): Initialise addr_mode_costs field.
5829         (arm_fastmul_tune): Likewise.
5830         (arm_strongarm_tune): Likewise.
5831         (arm_xscale_tune): Likewise.
5832         (arm_9e_tune): Likewise.
5833         (arm_marvell_pj4_tune): Likewise.
5834         (arm_v6t2_tune): Likewise.
5835         (arm_cortex_tune): Likewise.
5836         (arm_cortex_a8_tune): Likewise.
5837         (arm_cortex_a7_tune): Likewise.
5838         (arm_cortex_a15_tune): Likewise.
5839         (arm_cortex_a35_tune): Likewise.
5840         (arm_cortex_a53_tune): Likewise.
5841         (arm_cortex_a57_tune): Likewise.
5842         (arm_exynosm1_tune): Likewise.
5843         (arm_xgene1_tune): Likewise.
5844         (arm_cortex_a5_tune): Likewise.
5845         (arm_cortex_a9_tune): Likewise.
5846         (arm_cortex_a12_tune): Likewise.
5847         (arm_cortex_a73_tune): Likewise.
5848         (arm_v7m_tune): Likewise.
5849         (arm_cortex_m7_tune): Likewise.
5850         (arm_v6m_tune): Likewise.
5851         (arm_fa726te_tune): Likewise.
5852         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5854 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5856         * config/arm/arm.c (arm_mem_costs): New function.
5857         (arm_rtx_costs_internal): Use arm_mem_costs.
5859 2017-11-23  Mark Wielaard  <mark@klomp.org>
5861         * dwarf2out.c (init_sections_and_labels): Use generation to create
5862         unique ranges_section_label and ranges_base_label. Return generation.
5863         (output_rnglists): Add generation argument. Use generation to create
5864         unique ranges labels.
5865         (dwarf2out_finish): Get generation from init_sections_and_labels
5866         and pass generation to output_rnglists.
5868 2017-11-23  Mike Stump  <mikestump@comcast.net>
5869             Eric Botcazou  <ebotcazou@adacore.com>
5871         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
5872         * cfgloop.h (struct loop): Add unroll field.
5873         * function.h (struct function): Add has_unroll bitfield.
5874         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
5875         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
5876         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
5877         (pass_rtl_unroll_loops::gate): Likewise.
5878         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
5879         for which loop->unroll==1.
5880         (decide_unroll_constant_iterations): Use note for consistency and
5881         take loop->unroll into account.  Return early if loop->unroll is set.
5882         Fix thinko in existing test.
5883         (decide_unroll_runtime_iterations): Use note for consistency and
5884         take loop->unroll into account.
5885         (decide_unroll_stupid): Likewise.
5886         * lto-streamer-in.c (input_cfg): Read loop->unroll.
5887         * lto-streamer-out.c (output_cfg): Write loop->unroll.
5888         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
5889         New case.
5890         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
5891         (print_loop): Print loop->unroll if set.
5892         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
5893         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
5894         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
5895         New case.
5896         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
5897         loop->unroll is set and smaller than the trip count.  Otherwise bypass
5898         entirely the heuristics if loop->unroll is set.  Remove dead note.
5899         Fix off-by-one bug in other note.
5900         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
5901         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
5902         is greater than 1.
5903         (tree_unroll_loops_completely): Make static.
5904         (pass_complete_unroll::execute): Use correct type for variable.
5905         (pass_complete_unrolli::execute): Fix formatting.
5906         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
5908 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5910         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
5911         enable when TARGET_PREFER_AVX128 is set.
5913 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5915         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
5916         frequency.
5917         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
5918         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
5920 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5922         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
5923         frequencies. Fix estimation of aggregate parameters.
5925 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5927         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
5928         when inlining.
5930 2017-11-23  Tom de Vries  <tom@codesourcery.com>
5932         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
5934 2017-11-23  Tom de Vries  <tom@codesourcery.com>
5936         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
5937         macro.
5939 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
5941         PR target/83111
5942         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
5943         sibcall_value_pcrel_fdpic): Use local variable instead of
5944         operands[3].
5945         (calli_tbr_rel): Add missing operand 2.
5946         (call_valuei_tbr_rel): Add missing operand 3.
5948 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5950         PR middle-end/82253
5951         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
5952         bitpos/bitsize covering the whole destination, use store_expr only if
5953         the complex mode is the same.  Otherwise, use expand_normal and if
5954         it returns CONCAT, subreg each part separately instead of trying to
5955         subreg the whole result.
5957 2017-11-23  Richard Biener  <rguenther@suse.de>
5959         PR tree-optimization/23094
5960         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
5961         come here from walking over backedges in the first iteration.
5962         (vn_reference_lookup_3): Skip clobbers that store the same value.
5964 2017-11-23  Richard Biener  <rguenther@suse.de>
5966         PR tree-optimization/81403
5967         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
5968         a block we need a leader relative to.
5969         (phi_translate_1): For nary processing require a leader from
5970         get_representative_for given we run expression simplification
5971         using match-and-simplify.  Remove previous fix.
5973 2017-11-22  Jeff Law  <law@redhat.com>
5975         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
5976         Use new method allocate_value_range rather than accessing the
5977         vrp_value_range_pool data member directly.
5978         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
5979         to use extract_range_from_stmt method to avoid need for
5980         extract_range_from_assignment method.
5981         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
5982         method rather than setting values_propgated data member directly.
5983         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
5984         and values propagated data members and extract_range_from_assignment
5985         method.  Reorder private data members to conform to standards.
5986         Add new methods set_lattice_propagation_complete and
5987         allocate_value_range.
5989 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
5991         PR rtl-optimization/83030
5992         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
5993         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
5994         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
5995         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
5996         CROSSING_JUMP_P flag is already set before setting it.
5997         * cfgrtl.c (fixup_partition_crossing): Likewise.
5998         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
5999         insn as useless.
6001 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6003         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6004         Handle the case where both arguments are using gen_const_vec_series.
6006 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6008         PR c++/62170
6009         * pretty-print.c (pp_format): Move quoting implementation to
6010         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6011         to pass address of "quote" local.
6012         (pp_begin_quote): New function.
6013         (pp_end_quote): New function.
6014         * pretty-print.h (printer_fn): Convert penultimate param from bool
6015         to bool *.
6016         (pp_begin_quote): New decl.
6017         (pp_end_quote): New decl.
6018         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6019         param from bool to bool *.
6020         * tree-diagnostic.h (default_tree_printer): Likewise.
6022 2017-11-22  Jeff Law  <law@redhat.com>
6024         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6025         Initialize vr_values.
6026         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6027         vr_values via vr_values class instance rather than delegators.
6028         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6029         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6030         (evrp_range_analyzer::push_value_range): Likewise.
6031         (evrp_range_analyzer::pop_value_range): Likewise.
6032         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6033         most delegators.  Those remaining are exposed as public interfaces.
6034         Make vr_values a pointer and private.
6035         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6036         vr_values.
6037         (evrp_range_analyzer::get_vr_value): New method.
6038         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6039         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6040         (class evrp_dom_walker): Attach evrp_folder class, initialize
6041         it in the ctor.  Remove temporary delegators.
6042         (evrp_dom_walker::before_dom_children): Call methods in attached
6043         evrp_range_analyzer class via class instance pointer.  Use
6044         free value_range_constant_singleton to remove need for
6045         op_with_constant_singleton_value delegator method.  Do not
6046         create a vrp_prop class instance for every call!  Narrow
6047         scope of a couple variables.
6048         (evrp_dom_walker::cleanup): Call methods in attached
6049         evrp_range_analyzer class via class instance pointer.
6050         * vr-values.h (class vr_values): Privatize many methods and
6051         data members.
6053 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6055         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6056         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6057         * expr.c (expand_expr_real_2): Likewise.
6058         * fold-const.c (const_unop): Likewise.
6059         * optabs-tree.c (optab_for_tree_code): Likewise.
6060         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6061         * tree-inline.c (estimate_operator_cost): Likewise.
6062         * tree-pretty-print.c (dump_generic_node): Likewise.
6063         (op_code_prio): Likewise.
6064         (op_symbol_code): Likewise.
6065         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6066         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6067         * internal-fn.c (direct_internal_fn_optab): New function.
6068         (direct_internal_fn_array, direct_internal_fn_supported_p
6069         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6070         * fold-const-call.c (fold_const_reduction): New function.
6071         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6072         CFN_REDUC_MIN.
6073         * tree-vect-loop.c: Include internal-fn.h.
6074         (reduction_code_for_scalar_code): Rename to...
6075         (reduction_fn_for_scalar_code): ...this and return an internal
6076         function.
6077         (vect_model_reduction_cost): Take an internal_fn rather than
6078         a tree_code.
6079         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6080         than assignments.
6081         (vectorizable_reduction): Use internal functions rather than tree
6082         codes for the reduction operation.  Update calls to the functions
6083         above.
6084         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6085         Use calls to internal functions rather than REDUC tree codes.
6086         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6088 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6090         * config/vxworks.c (vxworks_override_options): Pick default
6091         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6092         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6093         DWARF_GNAT_ENCODINGS_DEFAULT.
6094         * config/vxworksae.h: Likewise.
6096 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6098         PR tree-optimization/83104
6099         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6100         not result.
6102 2017-11-22  Marek Polacek  <polacek@redhat.com>
6103             H.J. Lu  <hongjiu.lu@intel.com>
6104             Jason Merrill  <jason@redhat.com>
6106         PR c++/60336
6107         PR middle-end/67239
6108         PR target/68355
6109         * calls.c (initialize_argument_information): Call
6110         warn_parameter_passing_abi target hook.
6111         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6112         argument onto stack.
6113         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6114         * common.opt: Update -fabi-version description.
6115         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6116         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6117         int_size_in_bytes.
6118         (ix86_is_empty_record): New function.
6119         (ix86_warn_parameter_passing_abi): New function.
6120         (TARGET_EMPTY_RECORD_P): Redefine.
6121         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6122         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6123         * doc/tm.texi: Regenerated.
6124         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6125         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6126         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6127         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6128         instead of int_size_in_bytes.
6129         * expr.c (copy_blkmode_to_reg): Likewise.
6130         * function.c (aggregate_value_p): Return 0 for empty types.
6131         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6132         target hook.
6133         (locate_and_pad_parm): Call arg size_in_bytes instead
6134         size_in_bytes.
6135         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6136         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6137         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6138         hooks.
6139         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6140         (std_gimplify_va_arg_expr): Skip empty records.  Call
6141         arg_size_in_bytes instead size_in_bytes.
6142         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6143         * tree-core.h (tree_type_common): Add empty_flag.
6144         (tree_decl_common): Update comments.
6145         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6146         DECL_PADDING_P.
6147         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6148         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6149         DECL_PADDING_P.
6150         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6151         * tree.c (default_is_empty_type): New function.
6152         (default_is_empty_record): New function.
6153         (arg_int_size_in_bytes): New function.
6154         (arg_size_in_bytes): New function.
6155         (get_ultimate_context): New function.
6156         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6157         TRANSLATION_UNIT_WARN_EMPTY_P.
6158         (default_is_empty_record, arg_int_size_in_bytes,
6159         arg_size_in_bytes, get_ultimate_context): Declare.
6161 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6163         * config/arm/arm.c (cmse_clear_registers): New function.
6164         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6165         code by call to cmse_clear_registers.
6166         (cmse_nonsecure_entry_clear_before_return): Likewise.
6168 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6170         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6171         (vdot_s32, vdotq_s32): New.
6172         (vdot_lane_u32, vdotq_lane_u32): New.
6173         (vdot_lane_s32, vdotq_lane_s32): New.
6176 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6178         PR middle-end/82547
6179         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6180         for unsigned values with fewer HWIs than the precision.
6181         (test_overflow): New function.
6182         (wide_int_cc_tests): Call it.
6184 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6186         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6187         has the same mode class as Pmode.
6189 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6191         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6192         padding_bits_to_clear_ptr.
6193         (cmse_nonsecure_entry_clear_before_return): Likewise.
6195 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6197         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6198         auto_sbitap instead of integer bitfield to control register needing
6199         clearing.
6201 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6203         PR tree-optimization/83044
6204         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6205         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6206         that need the upper bound.  Subtract offset from
6207         get_addr_base_and_unit_offset only if positive and subtract it
6208         before division by eltsize rather than after it.
6210         PR debug/83084
6211         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6212         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6213         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6214         asm.
6216         PR middle-end/82875
6217         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6218         expand_binop with *mul_widen_optab, make sure at least one of the
6219         operands doesn't have VOIDmode.
6221         PR debug/83034
6222         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6224         PR rtl-optimization/82044
6225         PR tree-optimization/82042
6226         * dse.c (record_store): Check for overflow.
6227         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6228         add_wild_read instead of clear_rhs_from_active_local_stores on
6229         overflow and log it into dump_file.
6231 2017-11-22  Richard Biener  <rguenther@suse.de>
6233         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6234         fake edges to exit when looking for a place to insert.
6235         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6236         and remove.
6237         (insert_into_preds_of_block): Commit edge insertion immediately,
6238         assert that doesn't require new BBs.
6239         (fini_pre): Release expressions.
6240         (pass_pre::execute): Shuffle things around a bit, if the fn
6241         is too large do not compute AVAIL either as this is really the
6242         quadratic bit.
6244 2017-11-22  Richard Biener  <rguenther@suse.de>
6246         PR tree-optimization/83089
6247         * tree-if-conv.c (pass_if_conversion::execute): If anything
6248         changed reset SCEV and free the number of iteration estimates.
6250 2017-11-21  Martin Sebor  <msebor@redhat.com>
6252         PR tree-optimization/82945
6253         * calls.h (warn_nonstring_bound): Remove unused function.
6255 2017-11-21  Martin Sebor  <msebor@redhat.com>
6257         PR tree-optimization/82945
6258         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6259         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6260         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6261         functions.
6262         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6263         * calls.h (get_attr_nonstring_decl): Declare new function.
6264         * doc/extend.texi (attribute nonstring): Update.
6265         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6266         get_attr_nonstring_decl and handle it.
6267         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6268         detection of nul-termination.
6269         (strlen_to_stridx): Change to a pointer.
6270         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6271         (pass_strlen::execute): Same.
6273 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6275         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6276         for the new option -mprefer-vector-width=[none|128|256|512].
6277         * config/i386/i386.c (ix86_target_string): remove old style options
6278         -mprefer-avx256 and make -mprefer-avx128 as alias.
6279         (ix86_option_override_internal):  Apply defaults for the
6280         -mprefer-vector-width=[128|256] option.
6281         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6282         Implement macros to work with -mprefer-vector-width=.
6283         * config/i386/i386.opt: Implemented option
6284         -mprefer-vector-width=[none|128|256|512].
6285         * doc/invoke.texi: Documentation for
6286         -mprefer-vector-width=[none|128|256|512].
6288 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6290         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6292 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6294         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6296 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6298         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6299         POINTER_PLUS_EXPR.
6300         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6301         * expr.c (expand_expr_real_2): Likewise.
6302         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6303         fold_binary_loc): Likewise.
6304         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6305         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6306         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6307         MINUS_EXPR transformations.
6308         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6309         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6310         * tree-inline.c (estimate_operator_cost): Likewise.
6311         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6312         op_symbol_code): Likewise.
6313         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6314         * vr-values.c (extract_range_from_binary_expr): Likewise.
6315         * varasm.c (initializer_constant_valid_p_1): Likewise.
6316         * tree.def: New tree code POINTER_DIFF_EXPR.
6318 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6320         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6321         integer suffix to movbe mnemonic.
6322         (*bswaphi2_movbe): Ditto.
6323         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6325 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6327         PR c++/83045
6328         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6329         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6330         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6332 2017-11-21  Martin Liska  <mliska@suse.cz>
6334         * tree-inline.c (expand_call_inline): Remove not needed
6335         xstrdup_for_dump.
6337 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6338             Jakub Jelinek  <jakub@redhat.com>
6340         PR target/82880
6341         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6342         Remove static keyword from f variable.
6344 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6346         PR tree-optimization/83086
6347         * gimple-ssa-store-merging.c
6348         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6349         rather than n.base_addr.
6351 2017-11-21  Martin Liska  <mliska@suse.cz>
6353         PR rtl-optimization/82044
6354         PR tree-optimization/82042
6355         * dse.c (check_mem_read_rtx): Check for overflow.
6357 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6359         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6360         typo in comment.
6362 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6364         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6365         and its corresponding call arg location note.
6367 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6368             Andrew Burgess  <andrew.burgess@embecosm.com>
6370         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6371         declaration.
6372         (arc_return_slot_offset): Likewise.
6373         (arc_eh_return_address_location): New declaration.
6374         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6375         (MUST_SAVE_REGISTER): Add exception handler case.
6376         (MUST_SAVE_RETURN_ADDR): Likewise.
6377         (arc_frame_pointer_required): Likewise.
6378         (arc_frame_pointer_needed): New function.
6379         (arc_compute_frame_size): Changed.
6380         (arc_expand_prologue): Likewise.
6381         (arc_expand_epilogue): Likewise.
6382         (arc_initial_elimination_offset): Likewise.
6383         (arc_return_slot_offset): Delete.
6384         (arc_eh_return_address_location): New function.
6385         (arc_builtin_setjmp_frame_value): Likewise.
6386         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6387         (EH_RETURN_STACKADJ_RTX): Define.
6388         (EH_RETURN_HANDLER_RTX): Likewise.
6389         * config/arc/arc.md (eh_return): Delete.
6391 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6393         * print-tree.h (debug_vec_tree): Remove prototype.
6394         * gdbinit.in (pvt): Remove macro.
6396 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6398         PR tree-optimization/83047
6399         * gimple-ssa-store-merging.c
6400         (imm_store_chain_info::output_merged_store): If the loads with the
6401         same vuse are in different basic blocks, for load_gsi pick a load
6402         location that is dominated by the other loads.
6404         PR c++/83059
6405         * config/i386/i386.c (ix86_memmodel_check): Start
6406         -Winvalid-memory-model diagnostics with lowercase letter.
6408         PR debug/82718
6409         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6410         set in_cold_section_p to the partition containing loc_list->first.
6411         When seeing loc_list->last_before_switch node, update secname and
6412         perform range_across_switch second partition handling only after that.
6414         PR debug/82933
6415         * run-rtl-passes.c: Include debug.h.
6416         (run_rtl_passes): Call debug_hooks->assembly_start.
6417         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6418         multiple times.
6420         PR target/82981
6421         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6422         OPTAB_DIRECT in calls to expand_simple_binop.
6424 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6426         PR c/81404
6427         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6429 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6431         PR c++/72786
6432         * spellcheck.h (best_match::blithely_get_best_candidate): New
6433         accessor.
6435 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6437         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6438         with lower case letter.
6440 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6442         * config/i386/i386.md (bswaphi2): New expander.
6443         (*bswaphi2_movbe): New insn pattern.
6444         (bswaphi -> rorhi pepehole2): New peephole pattern.
6446 2017-11-20  Jeff Law  <law@redhat.com>
6448         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6449         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6450         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6451         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6452         gimple-ssa-evrp-analyze.h.
6454         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6455         set BB_VISITED here.
6456         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6458 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6460         * tree-predcom.c: Add general comment on Store-Store chains.
6461         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6462         flag in component.
6463         (get_chain_last_ref_at): Rename into...
6464         (get_chain_last_write_at): ...this.
6465         (get_chain_last_write_before_load): New function.
6466         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6467         CT_STORE_STORE when write reference is added.
6468         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6469         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6470         (initialize_root_vars_store_elim_1): Ditto.
6471         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6472         definition is created.
6473         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6474         chain by replacing it with dominant stored value.
6476 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6478         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6480 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6482         * vr-values.c (extract_range_from_binary_expr): Use a full range
6483         for VR_VARYING.
6485 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6487         * config/arm/arm.md (R4_REGNUM): Define constant.
6488         (nonsecure_call_internal): Remove r4 clobber.
6489         (nonsecure_call_value_internal): Likewise.
6490         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6491         clobber and resequence match_operands.
6492         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6493         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6494         (nonsecure_call_value_reg_thumb2): Likewise.
6496 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6498         PR tree-optimization/78821
6499         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6500         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6501         to the address of the base rather than the base itself.
6502         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6503         (find_bswap_or_nop_finalize): New function.
6504         (find_bswap_or_nop): Use it.
6505         (bswap_replace): Return a tree rather than bool, change first
6506         argument from gimple * to gimple_stmt_iterator, allow inserting
6507         into an empty sequence, allow ins_stmt to be NULL - then emit
6508         all stmts into gsi.  Fix up MEM_REF address gimplification.
6509         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6510         (struct store_immediate_info): Add N and INS_STMT non-static
6511         data members.
6512         (store_immediate_info::store_immediate_info): Initialize them
6513         from newly added ctor args.
6514         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6515         bitpos at the end.
6516         (stmts_may_clobber_ref_p): For stores call also
6517         refs_anti_dependent_p.
6518         (gather_bswap_load_refs): New function.
6519         (imm_store_chain_info::try_coalesce_bswap): New method.
6520         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6521         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6522         (imm_store_chain_info::output_merged_store): Fail if number of
6523         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6524         and NOP_EXPR rhs_code.
6525         (pass_store_merging::process_store): Compute n and ins_stmt, if
6526         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6527         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6528         ctor.
6529         (pass_store_merging::execute): Calculate dominators.
6531         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6532         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6533         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6534         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6535         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6536         class pass_optimize_bswap, bswap_replace,
6537         pass_optimize_bswap::execute): Moved to ...
6538         * gimple-ssa-store-merging.c: ... this file.
6539         Include optabs-tree.h.
6540         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6541         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6542         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6543         anonymous namespace, remove static keywords.
6544         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6545         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6547 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6549         PR bootstrap/83062
6550         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6552 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6554         * vec.h (debug_helper): New function.
6555         (DEFINE_DEBUG_VEC): New macro.
6556         * hash-set.h (debug_helper): New function.
6557         (DEFINE_DEBUG_HASH_SET): New macro.
6558         * cfg.c (debug_slim (edge)): New function.
6559         Call DEFINE_DEBUG_VEC for edges.
6560         Call DEFINE_DEBUG_HASH_SET for edges.
6561         * cfghooks.c (debug_slim (basic_block)): New function.
6562         Call DEFINE_DEBUG_VEC for basic blocks.
6563         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6564         * print-tree.c (debug_slim): New function to handle trees.
6565         Call DEFINE_DEBUG_VEC for trees.
6566         Call DEFINE_DEBUG_HASH_SET for trees.
6567         (debug (vec<tree, va_gc>) &): Remove.
6568         (debug (<vec<tree, va_gc>) *): Remove.
6569         * print-rtl.c (debug_slim): New function to handle const_rtx.
6570         Call DEFINE_DEBUG_VEC for rtx_def.
6571         Call DEFINE_DEBUG_VEC for rtx_insn.
6572         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6573         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6574         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6575         (debug (vec<rtx_insn *> *ptr): Remove.
6576         (debug_insn_vector): Remove.
6577         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6579 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6581         PR rtl-optimization/82020
6582         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6583         IF_THEN_ELSE condition.
6585 2017-11-19  Jeff Law  <law@redhat.com>
6587         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6588         of degenerates resulting from ignoring an edge.
6590 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6592         PR ipa/81360
6593         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6595 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6597         PR ipa/83001
6598         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6599         for uninitialied counts.
6601 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6603         PR ipa/60243
6604         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6606 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6608         PR target/82713
6609         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6611 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6613         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6614         "do while (0)".
6616 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6618         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6619         "do {} while (0)".
6621 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6623         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6624         macro body.
6626 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6628         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6629         "do {} while (0)".
6630         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6632 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6634         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6635         "do {} while (0)".
6636         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6637         semicolon after MCORE_EXPORT_NAME call.
6639 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6641         PR target/82961
6642         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6643         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6645 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6647         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6648         (cgraph_edge::clone): Cleanup updating of profile.
6649         * ipa-cp.c (update_profiling_info): Likewise.
6650         * ipa-inline-transform.c (inline_transform): Likewise.
6651         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6652         * ipa-split.c (execute_split_functions): Do not split when function
6653         is cold.
6654         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6655         * profile-count.c (profile_count::dump): Add global0.
6656         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6657         undefined.
6658         (profile_count::to_sreal_scale): Likewise.
6659         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6660         (profile_count::combine_with_ipa_count): New function.
6661         * profile-count.h (profile_guessed_global0adjusted): New.
6662         (profile_count::adjusted_zero): New.
6663         (profile_count::global0adjusted): New.
6664         (profile_count::combine_with_ipa_count): New.
6665         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6666         correct profile of return block of split functions.
6667         (copy_cfg_body): Remove unused profile_count.
6668         (copy_body): Likewise.
6669         (expand_call_inline): Update.
6670         (tree_function_versioning): Update.
6672 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6674         * hash-set.h (hash_set::empty): New.
6675         * tree-ssa-threadbackward.h: Delete.
6676         * tree-ssa-threadbackward.c (class thread_jumps): New.
6677         Move max_threaded_paths into class.
6678         (fsm_find_thread_path): Remove arguments that are now in class.
6679         (profitable_jump_thread_path): Rename to...
6680         (thread_jumps::profitable_jump_thread_path): ...this.
6681         (convert_and_register_jump_thread_path): Rename to...
6682         (thread_jumps::convert_and_register_current_path): ...this.
6683         (check_subpath_and_update_thread_path): Rename to...
6684         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6685         (register_jump_thread_path_if_profitable): Rename to...
6686         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6687         (handle_phi): Rename to...
6688         (thread_jumps::handle_phi): ...this.
6689         (handle_assignment): Rename to...
6690         (thread_jumps::handle_assignment): ...this.
6691         (fsm_find_control_statement_thread_paths): Rename to...
6692         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6693         (find_jump_threads_backwards): Rename to...
6694         (thread_jumps::find_jump_threads_backwards): ...this.
6695         Initialize path local data.
6696         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6697         from within thread_jumps class.
6698         (pass_early_thread_jumps::execute): Same.
6700 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6702         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6704 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6706         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6707         to canonical location.
6709 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6711         PR target/81356
6712         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6713         Remove.
6714         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6716 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6718         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6719         rcrt1.o%s/grcrt1.o%s for -static-pie.
6721 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6723         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6724         ix86_shift_rotate_cost): Break out from ...
6725         (ix86_rtx_costs): ... here.
6726         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6727         vector operations.
6729 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6731         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6732         when entry block was promoted unlikely.
6733         (estimate_bb_frequencies): Increase frequency scale.
6734         * profile-count.h (profile_count): Export precision info.
6736 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6738         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6739         disturbin profile of entry block.
6741 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6743         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6744         roundoff errors.
6746 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6748         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6749         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6751 2017-11-17  Jeff Law  <law@redhat.com>
6753         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6754         from evrp_dom_walker class.  Various methods moved into new class.
6755         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6756         (evrp_range_analyzer::enter): New method.
6757         (evrp_range_analyzer::leave): New method.
6758         (evrp_dom_walker): Remove delegators no longer needed by this class.
6759         Replace vr_values data member with evrp_range_analyzer
6761         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6762         method extracted from evrp_dom_walker::before_dom_children.
6763         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6764         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6766         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6767         Add private copy constructor and move assignment operators.
6768         Privatize methods and class data where trivially possible.
6769         (evrp_dom_walker::cleanup): New function, extracted from
6770         execute_early_vrp.  Simplify access to class data.
6772         * vr-values.h (get_output_for_vrp): Prototype.
6773         * vr-values.c (get_output_for_vrp): New function extracted from
6774         vrp_visit_assignment_or_call and extract_range_from_stmt.
6775         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6777 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6779         * config/aarch64/aarch64.c
6780         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6781         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6782         tune_params::AUTOPREFETCHER_WEAK.
6784 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6786         PR target/82641
6787         * config/arm/arm.c (arm_valid_target_attribute_rec):
6788         Parse "arch=" and "+<ext>".
6789         (arm_valid_target_attribute_tree): Re-init global options.
6790         (arm_option_override): Make non-static.
6791         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6792         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6793         (__ARM_FEATURE_CRC32): Support undef.
6794         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6795         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6797 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6799         * gdbinit.in (break-on-diagnostic): New command.
6801 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6803         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6804         return to indirect jump for EH return if control-flow
6805         protection is enabled. Change explicit 'false' argument in
6806         pro_epilogue_adjust_stack with a value of
6807         flag_cf_protection.
6808         * config/i386/i386.md (simple_return_indirect_internal):
6809         Remove SImode restriction to support 64-bit.
6811 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6813         * combine.c (added_notes_insn): New.
6814         (try_combine): Handle added_notes_insn like added_links_insn.
6815         Rewrite return value code.
6816         (distribute_notes): Set added_notes_insn to the earliest insn we added
6817         a note to.
6819 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6821         PR rtl-optimization/82621
6822         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6823         dest of one of those SETs is unused.
6825 2017-11-17  Richard Biener  <rguenther@suse.de>
6827         PR fortran/83017
6828         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6829         * tree-pretty-print.c (dump_generic_node): Handle
6830         annot_expr_parallel_kind.
6831         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6832         * gimplify.c (gimple_boolify): Likewise.
6834 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6836         * config.gcc (extra_headers): Add cet.h for x86 targets.
6837         * config/i386/cet.h: New file.
6838         * doc/install.texi: Add --enable-cet/--disable-cet.
6840 2017-11-17  Richard Biener  <rguenther@suse.de>
6842         PR tree-optimization/83017
6843         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6844         (gen_parallel_loop): Properly count iterations.
6845         (parallelize_loops): Handle loop->can_be_parallel independent
6846         of flag_loop_parallelize_all.  Make static profitability test match
6847         the runtime one.
6848         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6849         * invoke.texi (parloops-min-per-thread): Document.
6851 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6853         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6854         upstreaming review comments.
6856 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6858         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6859         with fast unaligned access.
6860         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6862 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6864         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
6865         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
6867 2017-11-17  Richard Biener  <rguenther@suse.de>
6869         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
6870         folding of references.
6872 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
6874         PR middle-end/78809
6875         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
6876         of replacing call to strncmp with corresponding call to strcmp when
6877         meeting conditions.
6879 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6881         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
6882         option prefer-avx256 for skylake-avx512 configuration.
6883         * config/i386/i386.c (ix86_option_override_internal): Ditto.
6884         (get_builtin_code_for_version): Ditto.
6886 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6887             Monk Chiang  <sh.chiang04@gmail.com>
6889         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
6890         (FIXED_REGISTERS): Reserve more register numbers.
6891         (CALL_USED_REGISTERS): Likewise.
6892         (REG_ALLOC_ORDER): Likewise.
6893         (REG_CLASS_CONTENTS): Likewise.
6894         (REGISTER_NAMES): Likewise.
6896 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6897             Kito Cheng  <kito.cheng@gmail.com>
6899         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
6900         V2SI.
6901         * config/nds32/iterators.md: Add vector mode iterators and attributes.
6903 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
6905         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
6906         parameter list for vec_splats.
6908 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
6910         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
6911         date of C17.
6912         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
6914 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
6916         PR bootstrap/82856
6917         * doc/install.texi: Document incompatibility of Perl >=5.6.26
6918         with the required version of automake 1.11.6.
6920 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
6922         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
6923         for it.
6924         (DU_C2_3_power9): Correct reservation combinations.
6925         (FP_DIV_power9, VEC_DIV_power9): New.
6926         (power9-alu): Split out rotate/shift...
6927         (power9-rot): ...to here, correct dispatch resource.
6928         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
6929         resource.
6930         (power9-fp): Correct latency.
6931         (power9-sdiv): Add div/sqrt resource.
6932         (power9-ddiv): Correct latency, add div/sqrt resource.
6933         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
6934         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
6935         resource.
6936         (power9-qpdiv, power9-qpmul): Adjust resource usage.
6938 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
6940         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
6941         switch statement mapping KF built-ins to TF built-ins if we don't
6942         have the proper ISA 3.0 assembler support.
6944 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
6946         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
6947         (lower_emutls_function_body): Do not compute it.
6949 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
6951         * ipa-split.c (split_bb_info): Turn time to sreal.
6952         (split_point): Likewise.
6953         (dump_split_point): Likewise.
6954         (fine_split_points): Likewise.
6955         (execute_split_functions): Only zero split_bbs; turn time to sreals.
6957 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
6959         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
6960         in sreal.
6962 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
6964         * predict.c (combine_predictions_for_bb): Preserve zero predicted
6965         edges.
6966         (expensive_function_p): Remove useless assert.
6967         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
6969 2017-11-16  Martin Sebor  <msebor@redhat.com>
6971         PR tree-optimization/82588
6972         PR tree-optimization/82583
6973         * tree-vrp.c (check_array_ref): Handle flexible array members,
6974         string literals, and inner indices.
6975         (search_for_addr_array): Add detail to diagnostics.
6977 2017-11-16  Nathan Sidwell  <nathan@acm.org>
6979         PR c++/82836
6980         PR c++/82737
6981         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
6982         (SET_DECL_ASSEMBLER_NAME): Forward to
6983         overwrite_decl_assembler_name.
6984         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
6985         (overwrite_decl_assembler_name): Declare.
6986         * tree.c (overwrite_decl_assembler_name): New.
6987         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
6988         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
6989         (LANG_HOOKS_INITIALIZER): Add it.
6990         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
6991         * langhooks.c (lhd_set_decl_assembler_name): Use
6992         SET_DECL_ASSEMBLER_NAME.
6993         (lhd_overwrite_decl_assembler_name): Default implementation.
6995 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
6996             Jackson Woodruff  <jackson.woodruff@arm.com>
6998         PR tree-optimization/71026
6999         * match.pd: Canonicalize constant multiplies in division.
7001 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7003         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7004         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7005         Remove OPT_fomit_frame_pointer entry.
7006         * common/config/arc/arc-common.c: Likewise.
7007         * common/config/arm/arm-common.c: Likewise.
7008         * common/config/avr/avr-common.c: Likewise.
7009         * common/config/c6x/c6x-common.c: Likewise.
7010         * common/config/cr16/cr16-common.c: Likewise.
7011         * common/config/cris/cris-common.c: Likewise.
7012         * common/config/epiphany/epiphany-common.c: Likewise.
7013         * common/config/fr30/fr30-common.c: Likewise.
7014         * common/config/frv/frv-common.c: Likewise.
7015         * common/config/ia64/ia64-common.c: Likewise.
7016         * common/config/iq2000/iq2000-common.c: Likewise.
7017         * common/config/lm32/lm32-common.c: Likewise.
7018         * common/config/m32r/m32r-common.c: Likewise.
7019         * common/config/mcore/mcore-common.c: Likewise.
7020         * common/config/microblaze/microblaze-common.c: Likewise.
7021         * common/config/mips/mips-common.c: Likewise.
7022         * common/config/mmix/mmix-common.c: Likewise.
7023         * common/config/mn10300/mn10300-common.c: Likewise.
7024         * common/config/nios2/nios2-common.c: Likewise.
7025         * common/config/pa/pa-common.c: Likewise.
7026         * common/config/pdp11/pdp11-common.c: Likewise.
7027         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7028         * common/config/riscv/riscv-common.c: Likewise.
7029         * common/config/rs6000/rs6000-common.c: Likewise.
7030         * common/config/rx/rx-common.c: Likewise.
7031         * common/config/s390/s390-common.c: Likewise.
7032         * common/config/sh/sh-common.c: Likewise.
7033         * common/config/sparc/sparc-common.c: Likewise.
7034         * common/config/tilegx/tilegx-common.c: Likewise.
7035         * common/config/tilepro/tilepro-common.c: Likewise.
7036         * common/config/v850/v850-common.c: Likewise.
7037         * common/config/visium/visium-common.c: Likewise.
7038         * common/config/xstormy16/xstormy16-common.c: Likewise.
7039         * common/config/xtensa/xtensa-common.c: Likewise.
7040         * invoke.texi (-fomit-frame-pointer): Update documentation.
7042 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7044         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7046 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7048         * cfg.c (scale_bbs_frequencies_int,
7049         cale_bbs_frequencies_gcov_type): Remove.
7050         * cfg.h (scale_bbs_frequencies_int,
7051         cale_bbs_frequencies_gcov_type): Remove.
7053 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7055         * tree-ssa-loop-manip.c
7056         (scale_dominated_blocks_in_loop): Update to profile counts.
7057         (tree_transform_and_unroll_loop): Likewise.
7059 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7061         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7062         scale_bbs_frequencies_int.
7064 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7066         * final.c (compute_alignments): Use counts rather than frequencies.
7068 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7070         * cfgloopanal.c: Include sreal.h
7071         (average_num_loop_insns): Use counts and sreal for accounting.
7073 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7075         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7076         manipulation.
7078 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7079             Kito Cheng  <kito.cheng@gmail.com>
7081         * config/nds32/constraints.md: Provide more constraints.
7082         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7083         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7084         support constraints usage.
7086 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7088         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7090 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7091             Kito Cheng  <kito.cheng@gmail.com>
7093         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7094         * config/nds32/nds32.opt: Refine the layout.
7095         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7096         TARGET_EXT_STRING): Support new options.
7097         * config/nds32/nds32.h: Likewise.
7098         * config/nds32/nds32.md: Likewise.
7099         * config/nds32/nds32-predicates.c: Likewise.
7100         * config/nds32/constraints.md: Likewise.
7101         * common/config/nds32/nds32-common.c: Likewise.
7103 2017-11-16  Julia Koval  <julia.koval@intel.com>
7105         PR target/82983
7106         * config/i386/gfniintrin.h: Add sse check.
7107         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7109 2017-11-16  Julia Koval  <julia.koval@intel.com>
7111         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7112         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7113         (ix86_handle_option): Handle -mavx512vbmi2.
7114         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7115         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7116         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7117         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7118         (ix86_valid_target_attribute_inner_p): Ditto.
7119         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7120         * config/i386/i386.opt (mavx512vbmi2): New option.
7121         * doc/invoke.texi: Add new option.
7123 2017-11-16  Julia Koval  <julia.koval@intel.com>
7125         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7126         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7127         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7128         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7129         _mm512_gf2p8mul_epi8): New intrinsics.
7130         * config/i386/i386-builtin-types.def
7131         (V64QI_FTYPE_V64QI_V64QI): New type.
7132         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7133         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7134         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7135         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7136         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7137         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7139 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7141         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7142         explicitly as a stream of bytes.
7144 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7146         * config/rs6000/altivec.h (vec_xst_be): New #define.
7147         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7148         and externalize from *altivec_vperm_<mode>_internal.
7149         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7150         instantiation.
7151         (XL_BE_V8HI): Likewise.
7152         (XL_BE_V4SI): Likewise.
7153         (XL_BE_V4SI): Likewise.
7154         (XL_BE_V2DI): Likewise.
7155         (XL_BE_V4SF): Likewise.
7156         (XL_BE_V2DF): Likewise.
7157         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7158         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7159         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7160         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7161         VSX_BUILTIN_VEC_XST_BE.
7162         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7163         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7164         built-ins.
7165         (altivec_init_builtins): Replace conditional calls to def_builtin
7166         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7167         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7168         with unconditional calls.  Remove calls to def_builtin for
7169         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7170         __builtin_vec_xst_be.
7171         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7172         to define_expand, and add alternate RTL generation for P8.
7173         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7174         vsx_ld_elemrev_v8hi.
7175         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7176         add alternate RTL generation for P8.
7177         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7178         vsx_ld_elemrev_v16qi.
7179         (vsx_st_elemrev_v8hi): Convert define_insn
7180         to define_expand, and add alternate RTL generation for P8.
7181         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7182         vsx_st_elemrev_v8hi.
7183         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7184         add alternate RTL generation for P8.
7185         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7186         vsx_st_elemrev_v16qi.
7188 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7190         PR target/82990
7191         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7192         TARGET_AVX512ER check.
7193         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7194         neither -mzeroupper nor -mno-zeroupper is used and
7195         TARGET_EMIT_VZEROUPPER is set.
7196         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7197         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7199 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7201         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7202         folding of vector compares.
7203         (fold_build_vec_cmp): New helper function.
7204         (fold_compare_helper): New helper function.
7205         (builtin_function_type): Add compare builtins to the list of functions
7206         having unsigned arguments.  Cosmetic updates to comment indentation.
7207         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7208         the not+eq combination.
7210 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7212         PR tree-optimization/82726
7213         PR tree-optimization/70754
7214         * tree-predcom.c (order_drefs_by_pos): New function.
7215         (combine_chains): Move code setting has_max_use_after to...
7216         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7217         according to position information.
7218         (tree_predictive_commoning_loop): Update call to above function.
7219         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7221 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7223         PR tree-optimization/82726
7224         Revert
7225         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7227         PR tree-optimization/70754
7228         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7229         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7230         combined stmt before it if not NULL.
7231         (combine_chains): Process refs reversely and compute dominance point
7232         for root ref.
7234         Revert
7235         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7237         PR tree-optimization/79663
7238         * tree-predcom.c (combine_chains): Process refs in reverse order
7239         only for ZERO length chains, and add explaining comment.
7241 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7243         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7244         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7245         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7247 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7249         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7251 2017-11-15  Martin Liska  <mliska@suse.cz>
7253         * tree-cfg.c (pass_warn_function_return::execute):
7254         Compare warn_return_type for greater than zero.
7256 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7258         PR target/82941
7259         PR target/82942
7260         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7261         to return true on Xeon and not on Xeon Phi.
7262         (ix86_check_avx256_register): Changed to ...
7263         (ix86_check_avx_upper_register): ... this. Add extra check for
7264         VALID_AVX512F_REG_OR_XI_MODE.
7265         (ix86_avx_u128_mode_needed): Changed
7266         ix86_check_avx256_register to ix86_check_avx_upper_register.
7267         (ix86_check_avx256_stores): Changed to ...
7268         (ix86_check_avx_upper_stores): ... this. Changed
7269         ix86_check_avx256_register to ix86_check_avx_upper_register.
7270         (ix86_avx_u128_mode_after): Changed
7271         avx_reg256_found to avx_upper_reg_found. Changed
7272         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7273         (ix86_avx_u128_mode_entry): Changed
7274         ix86_check_avx256_register to ix86_check_avx_upper_register.
7275         (ix86_avx_u128_mode_exit): Ditto.
7276         * config/i386/i386.h: (host_detect_local_cpu): New define.
7278 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7280         * config/arm/xgene1.md (xgene1): Split into automatons
7281         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7282         (xgene1_f_load): Adjust reservations and/or types.
7283         (xgene1_f_store): Likewise.
7284         (xgene1_load_pair): Likewise.
7285         (xgene1_store_pair): Likewise.
7286         (xgene1_fp_load1): Likewise.
7287         (xgene1_load1): Likewise.
7288         (xgene1_store1): Likewise.
7289         (xgene1_move): Likewise.
7290         (xgene1_alu): Likewise.
7291         (xgene1_simd): Likewise.
7292         (xgene1_bfm): Likewise.
7293         (xgene1_neon_load1): Likewise.
7294         (xgene1_neon_store1): Likewise.
7295         (xgene1_neon_logic): Likewise.
7296         (xgene1_neon_st1): Likewise.
7297         (xgene1_neon_ld1r): Likewise.
7298         (xgene1_alu_cond): Added.
7299         (xgene1_shift_reg): Likwise.
7300         (xgene1_bfx): Likewise.
7301         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7303 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7305         PR target/82981
7306         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7307         ssa-iterators.h.
7308         (can_widen_mult_without_libcall): New function.
7309         (expand_mul_overflow): If only checking unsigned mul overflow,
7310         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7311         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7312         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7313         (expand_DIVMOD): Formatting fix.
7314         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7315         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7316         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7318         PR tree-optimization/82977
7319         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7320         constructed temporary to strlen_to_stridx.put.
7322 2017-11-15  Martin Liska  <mliska@suse.cz>
7324         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7325         * configure: Regenerate.
7327 2017-11-15  Martin Liska  <mliska@suse.cz>
7329         PR target/82927
7330         * config/sh/sh-mem.cc: Use proper probability for
7331         REG_BR_PROB_NOTE.
7333 2017-11-14  Jeff Law  <law@redhat.com>
7335         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7336         the red zone for stack_clash_protection_final_dynamic_probe targets
7337         when the total dynamic stack size is zero bytes.
7339         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7340         blocks is post order.
7342 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7344         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7345         * final.c (rest_of_clean_state): Set it for the
7346         -fcompare-debug dump.
7347         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7348         class when TDF_COMPARE_DEBUG is set.
7350         * dwarf2out.c (gen_producer_string): Discard
7351         OPT_fcompare_debug.
7353 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7355         PR c/81156
7356         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7357         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7358         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7359         (__TGMATH_REAL_2_3): Remove macros.
7360         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7361         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7362         __builtin_tgmath.
7363         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7364         __TGMATH_REAL_2.
7365         (remquo): Define using __TGMATH_REAL_3.
7367 2017-11-14  Jeff Law  <law@redhat.com>
7369         * vr-values.c: New file with contents extracted from tree-vrp.c.
7370         * Makefile.in (OBJS): Add vr-values.o
7371         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7372         (set_value_range, set_and_canonicalize_value_range): Likewise.
7373         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7374         (value_range_constant_singleton, symbolic_range_p): Likewise.
7375         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7376         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7377         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7378         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7379         (find_case_label_range, find_case_label_index): Likewise.
7380         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7381         (range_int_cst_singleton_p, value_inside_range): Likewise.
7382         (get_single_symbol): Likewise.
7383         (switch_update): Move structure definition here.
7384         (to_remove_edges, to_update_switch_stmts): Provide externs.
7385         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7386         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7387         (vrp_val_is_min, set_value_range): Likewise.
7388         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7389         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7390         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7391         (range_is_nonnull, range_int_cst_p): Likewwise.
7392         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7393         (get_single_symbol, operand_less_p): Likewise
7394         (compare_values_warnv, compare_values): Likewise.
7395         (value_inside_range, value_range_constant_singleton): Likewise.
7396         (zero_nonzero_bitgs_from_vr): Likewise.
7397         (extract_range_from_binary_expr_1): Likewise.
7398         (overflow_comparison_p): Likewise.
7399         (to_remove_edges, to_update_switch_stmts): Likewise.
7400         (find_case_label-index, find_case_label_range): Likewise.
7401         (switch_update, set_value_range_to_nonnegative): Remove.
7402         (set_value_range_to_truthvalue): Likewise.
7403         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7404         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7405         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7406         (find_case_label_ranges, test_for_singularity): Likewise.
7407         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7408         (x_vr_values): Move to its remaining use site.
7410 2017-11-10  Jeff Law  <law@redhat.com>
7412         * vr-values.h (VR_INITIALIZER): Move #define here.
7413         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7414         * Makefile.in (OBJS): Add tree-evrp.o
7415         * tree-vrp.h (assert_info): Move structure definition here.
7416         (set_value_range_to_varying): Prototype.
7417         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7418         (infer_value_range, register_edge_assert_for): Likewise.
7419         (stmt_interesting_for_vrp): Likewise.
7420         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7421         (set_value_range_to_varying): No longer static.
7422         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7423         (infer_value_range, register_edge_assert_for): Likewise.
7425 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7427         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7428         generate the XXBRD instruction.
7430         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7431         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7432         as being compatible if -mabi=ieeelongdouble.
7433         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7434         to setup float128 built-ins with hardware support.
7435         (BU_FLOAT128_HW_2): Likewise.
7436         (BU_FLOAT128_HW_3): Likewise.
7437         (BU_FLOAT128_HW_VSX_1): Likewise.
7438         (BU_FLOAT128_HW_VSX_2): Likewise.
7439         (scalar_extract_expq): Change float128 built-in functions to
7440         accommodate having both KFmode and TFmode functions.  Use the
7441         KFmode variant as the default.
7442         (scalar_extract_sigq): Likewise.
7443         (scalar_test_neg_qp): Likewise.
7444         (scalar_insert_exp_q): Likewise.
7445         (scalar_insert_exp_qp): Likewise.
7446         (scalar_test_data_class_qp): Likewise.
7447         (sqrtf128_round_to_odd): Delete processing the round to odd
7448         built-in functions as special built-in functions, and define them
7449         as float128 built-ins.  Use the KFmode variant as the default.
7450         (truncf128_round_to_odd): Likewise.
7451         (addf128_round_to_odd): Likewise.
7452         (subf128_round_to_odd): Likewise.
7453         (mulf128_round_to_odd): Likewise.
7454         (divf128_round_to_odd): Likewise.
7455         (fmaf128_round_to_odd): Likewise.
7456         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7457         support for KFmode and TFmode xststdcqp calls.
7458         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7459         point, switch the built-in handlers for the get/set float128
7460         exponent, get float128 mantissa, float128 test built-ins, and the
7461         float128 round to odd built-in functions.  Eliminate creating the
7462         float128 round to odd built-in functions as special built-ins.
7463         (rs6000_init_builtins): Eliminate special creation of the float128
7464         round to odd built-in functions.
7465         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7466         function insns to support both TFmode and KFmode variants.
7467         (xsxsigqp_<mode>): Likewise.
7468         (xsiexpqpf_<mode>): Likewise.
7469         (xsiexpqp_<mode>): Likewise.
7470         (xststdcqp_<mode>): Likewise.
7471         (xststdcnegqp_<mode>): Likewise.
7472         (xststdcqp_<mode>): Likewise.
7474 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7476         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7477         unused path_in_freq_ptr parameter.
7478         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7480 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7482         * ipa-inline.c (edge_badness): Dump sreal frequency.
7483         (compute_inlined_call_time): Match natural implementaiton ...
7484         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7485         forgotten division by CGRAPH_FREQ_BASE.
7487 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7489         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7490         Solaris 11.  Update comment.
7491         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7492         renaming.
7493         * config/sol2.h (STARTFILE_SPEC): Likewise.
7494         * configure: Regenerate.
7496 2017-11-14  Carl Love  <cel@us.ibm.com>
7498         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7499         le_ and be_ prefixes to swap* variables.  Remove
7500         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7501         statements.
7503 2017-11-14  Jason Merrill  <jason@redhat.com>
7505         Support GTY((cache)) on hash_map.
7506         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7507         (ggc_cache_remove): Override it instead of ggc_mx.
7508         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7509         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7510         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7511         (simple_cache_map_traits): Override maybe_mx.
7512         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7513         (hash_map): Friend gt_cleare_cache.
7514         (gt_cleare_cache): New.
7515         * tree.h (tree_cache_traits): New hash_map traits class.
7516         (tree_cache_map): New typedef.
7518 2017-11-14  Richard Biener  <rguenther@suse.de>
7520         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7521         paramter and handling.
7522         (cleanup_control_flow_bb): Likewise.
7523         (cleanup_control_flow_pre): New helper performing a DFS walk
7524         to call cleanup_control_flow_bb in PRE order.
7525         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7526         via cleanup_control_flow_pre.
7528 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7530         * config/aarch64/aarch64-simd.md
7531         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7532         (*aarch64_simd_bsl<mode>_alt): Likewise.
7533         (aarch64_simd_bsldi_internal): New.
7534         (aarch64_simd_bsldi_alt): Likewise.
7536 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7538         * tracer.c (better_p): Do not compare frequencies.
7539         * reg-stack.c (better_edge): Likewise.
7540         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7541         and back.
7543 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7545         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7546         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7547         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7548         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7549         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7550         * predict.c (maybe_hot_frequency_p): Inline to ...
7551         (maybe_hot_count_p): ... here; rewrite to counts.
7552         (counts_to_freqs): Rename to ...
7553         (update_max_bb_count): ... this one.
7554         (expensive_function_p): Use counts.
7555         (estimate_bb_frequencies): Update.
7556         (rebuild_frequencies): Update.
7557         * predict.h (counts_to_freqs): Rename to ...
7558         (update_max_bb_count): ... this one.
7559         * profile.c (compute_branch_probabilities): Add debug info
7560         * tree-inline.c (expand_call_inline): Update debug info.
7561         (optimize_inline_calls): Use update_max_bb_count..
7562         (tree_function_versioning): Use update_max_bb_count..
7563         * value-prof.c (gimple_value_profile_transformations):
7564         Do not use update_max_bb_count.
7566 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7568         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7569         always use frequencies.
7571 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7573         * bb-reorder.c: Remove frequencies from comments.
7574         (better_edge_p): Use profile counts.
7575         (find_traces): Dump profile counts.
7576         (rotate_loop): Use profile counts.
7577         (find_traces_1_round): Likewise.
7578         (connect_better_edge_p): Use counts instead of probabilities for
7579         reverse walk.
7580         (copy_bb_p): Drop early check for non-0 frequency.
7581         (sanitize_hot_paths): Update comments.
7583 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7585         * ipa-split.c (struct split_point): Add count.
7586         (consider_split): Do not compute incoming frequency; compute incoming
7587         count and store it to split_point.
7588         (split_function): Set count of the call to split part correctly.
7590 2017-11-13  Carl Love  <cel@us.ibm.com>
7592         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7594 2017-11-13  Tom Tromey  <tom@tromey.com>
7596         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7598 2017-11-13  Carl Love  <cel@us.ibm.com>
7600         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7601         Add support for builtins:
7602         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7603         vector {un,}signed {char,int,short},
7604         vector {un,}signed {char,int,short}) arguments.
7605         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7606         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7607         Add BU_P9V_AV_2 expansions for the builtins.
7608         * config/rs6000/altivec.h (vec_first_match_index,
7609         vec_first_mismatch_index, vec_first_match_or_eos_index,
7610         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7611         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7612         new extern declaration.
7613         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7614         * config/rs6000/vsx.md (first_match_index_<mode>,
7615         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7616         first_mismatch_or_eos_index_<mode>): Add define expand.
7617         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7618         * doc/extend.texi: Update the built-in documenation file for the new
7619         built-in functions.
7621 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7623         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7624         into the min/max operations for _Float<N> and _Float<N>X types.
7626 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7628         PR lto/81351
7629         * dwarf2out.c (do_eh_frame): New static variable.
7630         (dwarf2out_begin_prologue): Set it.
7631         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7633 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7635         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7637         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7639 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7641         PR lto/81351
7642         * debug.h (dwarf2out_do_eh_frame): Declare.
7643         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7644         (dwarf2out_do_frame): Use it.
7645         (dwarf2out_do_cfi_asm): Likewise.
7646         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7647         (dwarf2out_assembly_start): Likewise.
7648         (dwarf2out_begin_prologue): Fix comment.
7649         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7650         if the target needs either debug or unwind DWARF2 info.
7651         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7652         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7653         -ftrapping-math, -ftrapv and -fwrapv.
7655 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7657         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7658         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7659         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7660         (estimate_edge_size_and_time): Likewise.
7661         (ipa_merge_fn_summary_after_inlining): Likewise.
7662         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7663         (compute_uninlined_call_time): Use sreal_frequency.
7664         (compute_inlined_call_time): Likewise.
7665         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7666         * profile-count.c: Include sreal.h.
7667         (profile_count::to_sreal_scale): New.
7668         * profile-count.h: Forward declare sreal.
7669         (profile_count::to_sreal_scale): Declare.
7671 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7673         * diagnostic.c (maybe_line_and_column): New.
7674         (diagnostic_get_location_text): Use it.
7675         (diagnostic_report_current_module): Likewise.
7676         (test_diagnostic_get_location_text): Add tests.
7678 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7680         * doc/md.texi (Specifying processor pipeline description): Fix
7681         incorrect latency for the div instruction example.
7683 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7685         PR tree-optimization/78821
7686         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7687         that bit_not_p is the same.
7688         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7689         (split_group): Count precisely bit_not_p bits in each statement.
7690         (invert_op): New function.
7691         (imm_store_chain_info::output_merged_store): Use invert_op to
7692         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7693         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7695 2017-11-13  Martin Liska  <mliska@suse.cz>
7697         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7698         (struct source_info): Likewise.
7699         (add_branch_counts): Likewise.
7700         (add_line_counts): Likewise.
7701         (function_summary): Likewise.
7702         (output_intermediate_line): Likewise.
7703         (generate_results): Likewise.
7705 2017-11-13  Martin Liska  <mliska@suse.cz>
7707         * gcov.c (struct block_info): Remove typedef for block_t.
7708         (struct line_info): Likewise.
7709         (line_info::has_block): Likewise.
7710         (EXIT_BLOCK): Likewise.
7711         (unblock): Likewise.
7712         (circuit): Likewise.
7713         (get_cycles_count): Likewise.
7714         (process_file): Likewise.
7715         (read_graph_file): Likewise.
7716         (solve_flow_graph): Likewise.
7717         (find_exception_blocks): Likewise.
7718         (add_line_counts): Likewise.
7719         (accumulate_line_info): Likewise.
7720         (output_line_details): Likewise.
7722 2017-11-13  Martin Liska  <mliska@suse.cz>
7724         * gcov.c (struct arc_info): Remove typedef for arc_t.
7725         (struct line_info): Likewise.
7726         (add_branch_counts): Likewise.
7727         (output_branch_count): Likewise.
7728         (function_info::~function_info): Likewise.
7729         (circuit): Likewise.
7730         (output_intermediate_line): Likewise.
7731         (read_graph_file): Likewise.
7732         (solve_flow_graph): Likewise.
7733         (find_exception_blocks): Likewise.
7734         (add_line_counts): Likewise.
7735         (accumulate_line_info): Likewise.
7736         (output_line_details): Likewise.
7737         (output_function_details): Likewise.
7739 2017-11-13  Martin Liska  <mliska@suse.cz>
7741         * gcov.c (struct function_info): Remove typedef for function_t.
7742         (struct source_info): Likewise.
7743         (source_info::get_functions_at_location): Likewise.
7744         (solve_flow_graph): Likewise.
7745         (find_exception_blocks): Likewise.
7746         (add_line_counts): Likewise.
7747         (output_intermediate_file): Likewise.
7748         (process_file): Likewise.
7749         (generate_results): Likewise.
7750         (release_structures): Likewise.
7751         (read_graph_file): Likewise.
7752         (read_count_file): Likewise.
7753         (accumulate_line_counts): Likewise.
7754         (output_lines): Likewise.
7756 2017-11-13  Martin Liska  <mliska@suse.cz>
7758         * gcov.c (function_info::function_info): Remove num_counts
7759         and add vector<gcov_type>.
7760         (function_info::~function_info): Use the vector.
7761         (process_file): Likewise.
7762         (read_graph_file): Likewise.
7763         (read_count_file): Likewise.
7764         (solve_flow_graph): Likewise.
7766 2017-11-13  Martin Liska  <mliska@suse.cz>
7768         * gcov.c (function_info::is_artificial): New function.
7769         (process_file): Erase all artificial early.
7770         (generate_results): Skip as all artificial are already
7771         removed.
7773 2017-11-13  Martin Liska  <mliska@suse.cz>
7775         * gcov.c (read_graph_file): Store to global vector of functions.
7776         (read_count_file): Iterate the vector.
7777         (process_file): Likewise.
7778         (generate_results): Likewise.
7779         (release_structures): Likewise.
7781 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7783         PR tree-optimization/82954
7784         * gimple-ssa-store-merging.c
7785         (imm_store_chain_info::coalesce_immediate_stores): If
7786         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7788 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7790         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7791         Upddate call to ENDIAN_LANE_N.
7792         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7793         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7794         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7795         and use aarch64_endian_lane_rtx.
7796         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7798 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7800         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7802 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7804         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7806 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7808         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7809         body.
7810         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7811         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7812         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7814 2017-11-11  Martin Sebor  <msebor@redhat.com>
7816         PR c/81117
7817         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7819         PR bootstrap/82948
7820         * prefic.c (translate_name): Replace strncpy with memcpy to
7821         avoid -Wstringop-truncation.
7823 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7825         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7827 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7829         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7831 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7833         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7834         merging.
7836 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7838         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7839         enable generating XXBRH if the value is in a vector register.
7840         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7841         value is in a vector register.
7842         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7843         register to register bswap64's instead of doing the GPR sequence
7844         used on previous machines.
7845         (bswapdi2_xxbrd): New insn.
7846         (bswapdi2_reg): Disallow on ISA 3.0.
7847         (register to register bswap64 splitter): Do not split the insn on
7848         ISA 3.0 systems that use XXBRD.
7850 2017-11-10  Martin Sebor  <msebor@redhat.com>
7852         PR c/81117
7853         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7854         with memcpy.
7855         (find_subframework_file): Same.
7857 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7859         * auto-profile.c (afdo_indirect_call): Drop frequency.
7860         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7861         (cgraph_node::create_edge): Drop frequency argument.
7862         (cgraph_node::create_indirect_edge): Drop frequency argument.
7863         (cgraph_edge::make_speculative): Drop frequency arguments.
7864         (cgraph_edge::resolve_speculation): Do not update frequencies
7865         (cgraph_edge::dump_edge_flags): Do not dump frequency.
7866         (cgraph_node::dump): Check consistency in IPA mode.
7867         (cgraph_edge::maybe_hot_p): Use IPA counter.
7868         (cgraph_edge::verify_count_and_frequency): Rename to ...
7869         (cgraph_edge::verify_count): ... this one; drop frequency checking.
7870         (cgraph_node::verify_node): Update.
7871         * cgraph.h (struct cgraph_edge): Drop frequency.
7872         (cgraph_edge::frequency): New function.
7873         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
7874         frequencies.
7875         (cgraph_edge::rebuild_edges): Likewise.
7876         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
7877         (duplicate_thunk_for_node): Do not pass frequency.
7878         (cgraph_node::create_clone): Scale only counts.
7879         (cgraph_node::create_virtual_clone): Do not pass frequency.
7880         (cgraph_node::create_edge_including_clones): Do not pass frequency.
7881         (cgraph_node::create_version_clone): Do not pass frequency.
7882         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
7883         (cgraph_node::expand_thunk): Do not pass frequency.
7884         (cgraph_node::create_wrapper): Do not pass frequency.
7885         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
7886         frequency.
7887         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
7888         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
7889         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
7890         (gather_caller_stats): Use frequency function.
7891         (ipcp_cloning_candidate_p): Use frequency function.
7892         (ipcp_propagate_stage): Use frequency function.
7893         (get_info_about_necessary_edges): Use frequency function.
7894         (update_profiling_info): Update only IPA profile.
7895         (update_specialized_profile): Use frequency functoin.
7896         (perhaps_add_new_callers): Update only IPA profile.
7897         * ipa-devirt.c (ipa_devirt): Use IPA profile.
7898         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
7899         (dump_ipa_call_summary): Use frequency function.
7900         (estimate_edge_size_and_time): Use frequency function.
7901         (ipa_merge_fn_summary_after_inlining): Use frequency function.
7902         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
7903         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
7904         (update_noncloned_counts): ... ths one; scale counts only.
7905         (clone_inlined_nodes): Do not scale frequency.
7906         (inline_call): Do not pass frequency.
7907         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
7908         (compute_inlined_call_time): Use IPA profile.
7909         (want_inline_small_function_p): Use IPA profile.
7910         (want_inline_self_recursive_call_p): Use IPA profile.
7911         (edge_badness): Use IPA profile.
7912         (lookup_recursive_calls): Use IPA profile.
7913         (recursive_inlining): Do not pass frequency.
7914         (resolve_noninline_speculation): Do not update frequency.
7915         (inline_small_functions): Collect max of IPA profile.
7916         (dump_overall_stats): Dump IPA porfile.
7917         (dump_inline_stats): Dump IPA porfile.
7918         (ipa_inline): Collect IPA stats.
7919         * ipa-inline.h (clone_inlined_nodes): Update prototype.
7920         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
7921         (ipa_propagate_frequency): Use frequency function.
7922         (ipa_profile): Cleanup.
7923         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
7924         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
7925         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
7926         (input_node): Do not stream frequency.
7927         (input_edge): Do not stream frequency.
7928         (merge_profile_summaries): Scale only IPA profiles.
7929         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
7930         * predict.c (drop_profile): Do not recompute frequency.
7931         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
7932         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
7933         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
7934         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
7935         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
7936         * tree-inline.c (copy_bb): Do not scale frequency.
7937         (expand_call_inline): Do not scale frequency.
7938         (tree_function_versioning): Do not scale frequency.
7939         * ubsan.c (ubsan_create_edge): Do not pass frequency.
7941 2017-11-10  Julia Koval  <julia.koval@intel.com>
7943         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
7944         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
7945         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
7946         (_mm256_maskz_gf2p8affine_epi64_epi8)
7947         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
7948         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
7949         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
7950         (__builtin_ia32_vgf2p8affineqb_v32qi)
7951         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
7952         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
7954 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
7956         PR target/82641
7957         * config/arm/arm.c
7958         (arm_option_override): Refactor.
7959         (arm_option_reconfigure_globals): New.
7960         (arm_options_perform_arch_sanity_checks): New.
7961         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
7962         New prototype.
7963         (arm_options_perform_arch_sanity_checks): Likewise
7965 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
7967         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
7968         (power9-qpmul): New.
7969         * rs6000/rs6000.md ("type" attr): Add qmul.
7970         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
7971         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
7972         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
7974 2017-11-10  Martin Sebor  <msebor@redhat.com>
7976         PR c/81117
7977         * builtins.c (compute_objsize): Handle arrays that
7978         compute_builtin_object_size likes to fail for.  Make extern.
7979         * builtins.h (compute_objsize): Declare.
7980         (check_strncpy_sizes): New function.
7981         (expand_builtin_strncpy): Call check_strncpy_sizes.
7982         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
7983         -Wstringop-truncation.
7984         (gimple_fold_builtin_strncat): Same.
7985         * gimple.c (gimple_build_call_from_tree): Set call location.
7986         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
7987         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
7988         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
7989         (handle_builtin_strlen): Use strlen_to_stridx.
7990         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
7991         stpncpy.
7992         Use strlen_to_stridx.
7993         (pass_strlen::execute): Release strlen_to_stridx.
7994         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
7995         (-Wstringop-truncation): Document new option.
7997 2017-11-10  Martin Liska  <mliska@suse.cz>
7999         PR gcov-profile/82702
8000         * gcov.c (main): Handle intermediate files in a different
8001         way.
8002         (get_gcov_intermediate_filename): New function.
8003         (output_gcov_file): Remove support of intermediate files.
8004         (generate_results): Allocate intermediate file.
8005         (release_structures): Clean-up properly fn_end.
8006         (output_intermediate_file): Start iterating with line 1.
8008 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8010         PR tree-optimization/82929
8011         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8012         ops_swapped_p non-static data member.
8013         (store_immediate_info::store_immediate_info): Clear it.
8014         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8015         ops set ops_swapped_p.
8016         (count_multiple_uses): Handle ops_swapped_p.
8018 2017-11-10  Martin Liska  <mliska@suse.cz>
8020         * coverage.c (coverage_init): Stream information about
8021         support of has_unexecuted_blocks.
8022         * doc/gcov.texi: Document that.
8023         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8024         * gcov.c (read_graph_file): Likewise.
8025         (output_line_beginning): Fix a small issue with
8026         color output.
8028 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8030         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8031         reference of trivial component.
8033 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8035         PR bootstrap/82916
8036         * gimple-ssa-store-merging.c
8037         (pass_store_merging::terminate_all_aliasing_chains): For
8038         gimple_store_p stmts also call refs_output_dependent_p.
8040         PR rtl-optimization/82913
8041         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8042         single set.
8044 2017-11-09  Jeff Law  <law@redhat.com>
8046         * vr-values.h: New file with vr_values class.
8047         * tree-vrp.c: Include vr-values.h
8048         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8049         data objects into the vr_values class.
8050         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8051         (get_value_range): Make it a member function within vr_values class.
8052         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8053         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8054         (op_with_constant_singleton_value_range): Likewise.
8055         (extract_range_for_var_from_comparison_expr): Likewise.
8056         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8057         (extract_range_from_binary_expr): Likewise.
8058         (extract_range_from_unary_expr): Likewise.
8059         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8060         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8061         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8062         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8063         (compare_name_with_value, compare_names): Likewise.
8064         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8065         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8066         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8067         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8068         (extract_range_from_phi_node): Likewise.
8069         (simplify_truth_ops_using_ranges): Likewise.
8070         (simplify_div_or_mod_using_ranges): Likewise.
8071         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8072         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8073         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8074         (simplify_float_conversion_using_ranges): Likewise.
8075         (simplify_internal_call_using_ranges): Likewise.
8076         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8077         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8078         poitner via x_vr_values for calls into gimple folder.
8079         (vrp_initialize_lattice): Make this the vr_values ctor.
8080         (vrp_free_lattice): Make this the vr_values dtor.
8081         (set_vr_value): New function.
8082         (class vrp_prop): Add vr_values data member.  Add various member
8083         functions as well as member functions that delegate to vr_values.
8084         (check_array_ref): Make a member function within vrp_prop class.
8085         (search_for_addr_array, vrp_initialize): Likewise.
8086         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8087         vr_value, values_propagated, etc.
8088         (check_array_bounds): Extract vrp_prop class instance pointer from
8089         walk info structure.  Use it to call member functions.
8090         (check_all_array_refs): Make a member function within vrp_prop class.
8091         Smuggle class instance pointer via walk info structure.
8092         (x_vr_values): New local static.
8093         (vrp_valueize): Use x_vr_values to get class instance.
8094         (vr_valueize_1): Likewise.
8095         (class vrp_folder): Add vr_values data member.  Add various member
8096         functions as well as member functions that delegate to vr_values.
8097         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8098         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8099         class instance from vr_values.  Use it to call member functions.
8100         (vrp_dom_walker): Add vr_values data member.
8101         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8102         instance via x_vr_values.
8103         (identify_jump_threads): Accept vr_values as argument.  Store
8104         it into the walker structure.
8105         (evrp_dom_walker): Add vr_values class data member.  Add various
8106         delegators.
8107         (evrp_dom_walker::try_find_new_range): Use vr_values data
8108         member to access the memory allocator.
8109         (evrp_dom_walker::before_dom_children): Store vr_values class
8110         instance into the vrp_folder class.
8111         (evrp_dom_walker::push_value_range): Rework to avoid direct
8112         access to num_vr_values and vr_value.
8113         (evrp_dom_walker::pop_value_range): Likewise.
8114         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8115         Use vr_values to get to dump_all_value_ranges member function.
8116         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8117         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8118         Pass vr_values class instance down to identify_jump_threads.
8119         Remove call to vrp_free_lattice.
8120         (debug_all_value_ranges): Remove.
8122         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8123         (vrp_folder): Likewise.
8125         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8126         Get it from the existing bitmap instead.
8127         (vrp_intersect_ranges_1): Likewise.
8129 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8131         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8132         bit_not_p field.
8133         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8134         set bit_not_p to it.
8135         (imm_store_chain_info::coalesce_immediate_stores): Break group
8136         if bit_not_p is different.
8137         (count_multiple_uses, split_group,
8138         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8139         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8140         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8141         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8142         to store_immediate_info ctor.
8144 2017-11-09  Jim Wilson  <jimw@sifive.com>
8146         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8147         (scan_prog_file): Likewise.
8149 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8151         * bb-reorder.c (max_entry_frequency): Remove.
8152         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8153         connect_traces, push_to_next_round_p): Remove prototypes.
8154         (find_traces_1_round): Use counts only.
8155         (push_to_next_round_p): Likewise.
8156         (find_traces): Likewise.
8157         (rotate_loop): Likewise.
8158         (find_traces_1_round): Likewise.
8159         (connect_traces): Likewise.
8160         (edge_order): Likewise.
8162 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8164         * config/arm/arm.c (output_return_instruction): Add comments to
8165         indicate requirement for cmse_nonsecure_entry return to account
8166         for the size of clearing instruction output here.
8167         (thumb_exit): Likewise.
8168         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8169         return in hardfloat mode.
8171 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8173         * config/rs6000/rs6000.c (machine_function): Add a bool,
8174         "toc_is_wrapped_separately".
8175         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8176         if it wasn't explicitly set or unset, we are optimizing for speed, and
8177         doing separate shrink-wrapping.
8178         (rs6000_get_separate_components): Enable the TOC component if
8179         saving the TOC register in the prologue.
8180         (rs6000_components_for_bb): Handle the TOC component.
8181         (rs6000_emit_prologue_components): Store the TOC register where needed.
8182         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8183         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8185 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8187         * ipa-param-manipulation.c: New file.
8188         * ipa-param-manipulation.h: Likewise.
8189         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8190         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8191         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8192         (ipa_parm_adjustment): Likewise.
8193         (ipa_parm_adjustment_vec): Likewise.
8194         (ipa_get_vector_of_formal_parms): Moved declaration to
8195         ipa-param-manipulation.h.
8196         (ipa_get_vector_of_formal_parm_types): Likewise.
8197         (ipa_modify_formal_parameters): Likewise.
8198         (ipa_modify_call_arguments): Likewise.
8199         (ipa_combine_adjustments): Likewise.
8200         (ipa_dump_param_adjustments): Likewise.
8201         (ipa_modify_expr): Likewise.
8202         (ipa_get_adjustment_candidate): Likewise.
8203         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8204         ipa-param-manipulation.c.
8205         (ipa_get_vector_of_formal_parm_types): Likewise.
8206         (ipa_modify_formal_parameters): Likewise.
8207         (ipa_modify_call_arguments): Likewise.
8208         (ipa_modify_expr): Likewise.
8209         (get_ssa_base_param): Likewise.
8210         (ipa_get_adjustment_candidate): Likewise.
8211         (index_in_adjustments_multiple_times_p): Likewise.
8212         (ipa_combine_adjustments): Likewise.
8213         (ipa_dump_param_adjustments): Likewise.
8214         * tree-sra.c: Also include ipa-param-manipulation.h
8215         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8216         ipa-param.h.
8218 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8219             Alan Hayward  <alan.hayward@arm.com>
8220             David Sherwood  <david.sherwood@arm.com>
8222         * doc/sourcebuild.texi (vect_masked_store): Document.
8224 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8225             Alan Hayward  <alan.hayward@arm.com>
8226             David Sherwood  <david.sherwood@arm.com>
8228         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8230 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8231             Alan Hayward  <alan.hayward@arm.com>
8232             David Sherwood  <david.sherwood@arm.com>
8234         * doc/sourcebuild.texi (vect_variable_length): Document.
8236 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8237             Alan Hayward  <alan.hayward@arm.com>
8238             David Sherwood  <david.sherwood@arm.com>
8240         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8242 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8243             Alan Hayward  <alan.hayward@arm.com>
8244             David Sherwood  <david.sherwood@arm.com>
8246         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8248 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8249             Alan Hayward  <alan.hayward@arm.com>
8250             David Sherwood  <david.sherwood@arm.com>
8252         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8253         previously undocumented selectors.
8254         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8256 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8258         * doc/rtl.texi (const_vector): Say that elements can be
8259         const_wide_ints too.
8260         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8261         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8262         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8263         * optabs.c (expand_vector_broadcast): Likewise.
8265 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8266             Alan Hayward  <alan.hayward@arm.com>
8267             David Sherwood  <david.sherwood@arm.com>
8269         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8270         scaled index even if the unscaled address was invalid.
8271         Don't increase the complexity of using a scale in that case.
8273 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8274             Alan Hayward  <alan.hayward@arm.com>
8275             David Sherwood  <david.sherwood@arm.com>
8277         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8278         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8279         UNITS_PER_WORD bytes.
8280         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8281         based on REGMODE_NATURAL_SIZE of the inner mode.
8282         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8283         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8284         * expmed.c (lowpart_bit_field_p): Divide the value up into
8285         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8286         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8287         whether something is likely to occupy more than one register.
8289 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8291         PR ipa/82879
8292         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8293         profile_count::adjust_for_ipa_scaling.
8294         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8295         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8296         function.
8297         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8299 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8301         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8302         (split_group): Add total_orig and total_new arguments, estimate the
8303         number of statements related to the store group without store merging
8304         and with store merging.
8305         (imm_store_chain_info::output_merged_store): Adjust split_group
8306         callers, punt if estimated number of statements with store merging
8307         is not smaller than estimated number of statements without it.
8308         Formatting fix.
8309         (handled_load): Remove has_single_use checks.
8310         (pass_store_merging::process_store): Likewise.
8312 2017-11-09  Richard Biener  <rguenther@suse.de>
8314         PR tree-optimization/82902
8315         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8317 2017-11-09  Martin Liska  <mliska@suse.cz>
8319         PR target/82863
8320         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8321         uninitialized.
8323 2017-11-09  Martin Liska  <mliska@suse.cz>
8325         PR tree-optimization/82669
8326         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8328 2017-11-09  Martin Liska  <mliska@suse.cz>
8330         PR gcov-profile/48463
8331         * coverage.c (coverage_begin_function): Output also end locus
8332         of a function and information whether the function is
8333         artificial.
8334         * gcov-dump.c (tag_function): Parse and print the information.
8335         * gcov.c (INCLUDE_MAP): Add include.
8336         (INCLUDE_SET): Likewise.
8337         (struct line_info): Move earlier in the source file because
8338         of vector<line_info> in function_info structure.
8339         (line_info::line_info): Likewise.
8340         (line_info::has_block): Likewise.
8341         (struct source_info): Add new member index.
8342         (source_info::get_functions_at_location): New function.
8343         (function_info::group_line_p): New function.
8344         (output_intermediate_line): New function.
8345         (output_intermediate_file): Use the mentioned function.
8346         (struct function_start): New.
8347         (struct function_start_pair_hash): Likewise.
8348         (process_file): Add code that identifies group functions.
8349         Assign lines either to global or function scope.
8350         (generate_results): Skip artificial functions.
8351         (find_source): Assign index for each source file.
8352         (read_graph_file): Read new flag artificial and end_line.
8353         (add_line_counts): Assign it either to global of function scope.
8354         (accumulate_line_counts): Isolate core of the function to
8355         accumulate_line_info and call it for both function and global
8356         scope lines.
8357         (accumulate_line_info): New function.
8358         (output_line_beginning): Fix GNU coding style.
8359         (print_source_line): New function.
8360         (output_line_details): Likewise.
8361         (output_function_details): Likewise.
8362         (output_lines): Iterate both source (global) scope and function
8363         scope.
8364         (struct function_line_start_cmp): New class.
8365         * doc/gcov.texi: Reflect changes in documentation.
8367 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8369         PR debug/82837
8370         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8371         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8372         and similarly for not instead of neg.
8374 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8376         * config/i386/i386.opt: Add -mforce-indirect-call.
8377         * config/i386/predicates.md: Check for flag_force_indirect_call.
8378         * doc/invoke.texi: Document -mforce-indirect-call
8380 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8382         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8383         New extern.
8384         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8385         config/riscv/riscv.c (predict.h): New include.
8386         (riscv_slow_unaligned_access_p): No longer static.
8387         (riscv_block_move_straight): Add require.
8388         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8390 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8392         PR target/82855
8393         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8394         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8395         nonimmediate_operand predicate for operand 1 instead of
8396         register_operand.
8398 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8400         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8401         New pattern.
8402         * config/aarch64/constraints.md (Uml): New constraint.
8403         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8404         predicate.
8406 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8408         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8409         of two vec_duplicates into a vec_concat.
8411 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8413         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8414         Simplify vec_merge of vec_duplicate and vec_concat.
8415         * config/aarch64/constraints.md (Utq): New constraint.
8416         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8417         define_insn.
8419 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8421         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8422         Simplify vec_merge of vec_duplicate and const_vector.
8423         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8424         New predicate.
8425         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8426         mode iterator.  Update predicate on operand 1 to
8427         handle non-const_vec constants.  Delete constraints.
8428         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8430 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8432         PR tree-optimization/78821
8433         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8434         data member.
8435         (store_operand_info::store_operand_info): Initialize it to false.
8436         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8437         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8438         store in the group, and if chain_info is non-NULL, to ignore altogether
8439         that chain.
8440         (compatible_load_p): Fail if bit_not_p does not match.
8441         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8442         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8443         (pass_store_merging::process_store): Adjust
8444         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8445         call terminate_all_aliasing_chains newly when adding a store into
8446         a chain with non-NULL chain_info.
8448 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8450         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8452 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8454         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8455         Remove.
8456         (aarch64_layout_frame): Initialise emit_frame_chain.
8457         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8458         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8460 2017-11-08  Martin Liska  <mliska@suse.cz>
8462         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8463         of gimple_call_internal_p.
8465 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8467         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8469 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8471         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8472         "do {} while (0)".
8474 2017-11-08  Martin Liska  <mliska@suse.cz>
8476         PR sanitizer/82792
8477         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8479 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8481         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8483 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8485         PR target/82855
8486         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8487         (*cmp<mode>_ccz_1): New insn with $k alternative.
8489         PR target/82855
8490         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8491         RTX_COMM_COMPARE as commutative as well.
8492         (ix86_binary_operator_ok): Formatting fix.
8493         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8494         *<code><mode>3<mask_name><round_saeonly_name>,
8495         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8496         *<s>mul<mode>3_highpart<mask_name>,
8497         *vec_widen_umult_even_v16si<mask_name>,
8498         *vec_widen_umult_even_v8si<mask_name>,
8499         *vec_widen_umult_even_v4si<mask_name>,
8500         *vec_widen_smult_even_v16si<mask_name>,
8501         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8502         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8503         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8504         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8505         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8506         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8507         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8508         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8509         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8510         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8511         ix86_binary_operator_ok.  Formatting fixes.
8512         (*<plusminus_insn><mode>3<mask_name><round_name>,
8513         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8514         fixes.
8516 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8518         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8519         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8520         eq and ne if TARGET_ISEL.
8521         (cmp): New code_iterator.
8522         (UNS, UNSU_, UNSIK): New code_attrs.
8523         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8524         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8525         to...
8526         ("eq<mode>3"): ... this.
8527         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8528         to...
8529         ("ne<mode>3"): ... this.
8531 2017-11-07  Julia Koval  <julia.koval@intel.com>
8533         PR target/82812
8534         * common/config/i386/i386-common.c
8535         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8536         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8537         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8538         * config/i386/i386.opt: Ditto.
8539         * config/i386/i386.c (ix86_target_string): Ditto.
8540         (ix86_option_override_internal): Ditto.
8541         (ix86_init_mpx_builtins): Move MPX to args2.
8542         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8543         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8544         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8545         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8546         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8547         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8548         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8550 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8552         PR target/80425
8553         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8554         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8555         (zero-extendsidi peephole2): Remove peephole.
8557 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8559         PR c/53037
8560         * stor-layout.c: Include attribs.h.
8561         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8562         explicit lookup of "aligned" attribute.
8564 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8566         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8567         (riscv_expand_block_move): Likewise.
8568         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8569         implementation.
8570         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8571         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8572         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8573         (riscv_adjust_block_mem): Likewise.
8574         (riscv_block_move_loop): Likewise.
8575         (riscv_expand_block_move): Likewise.
8576         * config/riscv/riscv.md (movmemsi): New pattern.
8578 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8580         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8581         (MUSL_DYNAMIC_LINKER): Likewise.
8583 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8585         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8586         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8587         nonmmory_operand.
8589 2017-11-07  Richard Biener  <rguenther@suse.de>
8591         * match.pd: Fix build.
8593 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8594             Jackson Woodruff  <jackson.woodruff@arm.com>
8596         PR tree-optimization/71026
8597         * match.pd: Canonicalize negate in division.
8599 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8601         PR middle-end/80131
8602         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8604 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8606         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8607         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8609 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8611         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8612         non-scalar integral types.
8613         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8614         (-(A-B), -(~A)): New transformations.
8616 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8618         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8619         semicolon after "do {} while (0)".
8620         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8621         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8622         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8623         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8624         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8626 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8628         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8629         after "do {} while (0)".
8630         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8631         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8632         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8633         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8634         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8636 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8638         PR other/82784
8639         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8640         "while {} do (0)".
8641         (arm_rtx_costs_internal): Add missing semicolon after
8642         HANDLE_NARROW_SHIFT_ARITH call.
8644 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8646         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8647         disable isel if it was not set explicitly.
8649 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8651         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8652         (add_type_attribute) likewise.
8654 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8656         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8657         of struct ix86_frame.
8658         (ix86_initial_elimination_offset): Likewise.
8659         (ix86_expand_split_stack_prologue): Likewise.
8661 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8663         * tree-vrp.h (enum value_range_type): Update stale comment.
8665 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8666             Alan Hayward  <alan.hayward@arm.com>
8667             David Sherwood  <david.sherwood@arm.com>
8669         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8670         (aarch64_expand_vec_perm_const): Take the number of units too.
8671         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8672         (aarch64_expand_vec_perm_const): Likewise.
8673         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8674         (vec_perm<mode>): Update accordingly.
8676 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8677             Alan Hayward  <alan.hayward@arm.com>
8678             David Sherwood  <david.sherwood@arm.com>
8680         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8681         Take the number of units too.
8682         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8683         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8684         but check for a vector mode before rather than after the call.
8685         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8686         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8687         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8688         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8689         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8690         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8691         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8692         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8693         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8694         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8695         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8696         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8697         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8698         (aarch64_sqdmull2_n<mode>): Update accordingly.
8700 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8701             Alan Hayward  <alan.hayward@arm.com>
8702             David Sherwood  <david.sherwood@arm.com>
8704         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8705         the number of units too.
8706         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8707         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8708         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8709         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8710         (vec_store_lanesxi<mode>): Update accordingly.
8712 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8713             Alan Hayward  <alan.hayward@arm.com>
8714             David Sherwood  <david.sherwood@arm.com>
8716         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8717         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8718         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8719         of units rather than the mode.
8720         * config/aarch64/iterators.md (nunits): New mode attribute.
8721         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8722         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8723         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8724         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8725         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8726         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8727         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8728         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8729         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8730         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8731         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8732         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8733         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8734         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8735         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8736         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8737         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8738         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8739         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8740         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8741         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8742         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8743         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8744         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8745         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8746         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8747         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8748         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8749         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8750         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8751         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8752         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8753         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8754         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8755         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8756         (aarch64_simd_vec_setv2di): Likewise.
8758 2017-11-06  Carl Love  <cel@us.ibm.com>
8760         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8761         definitions.
8762         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8763         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8764         to power 8.
8765         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8766         extern declaration.
8767         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8768         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8769         Add power 8 macro expansions.
8770         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8771         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8772         power 8 instructions.  (VSX_XXBR): Add iterator.
8774 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8776         * config/arm/arm.md (predicable_short_it): Change default to "no",
8777         improve documentation, remove uses that are identical to the default.
8778         (enabled_for_depr_it): Rename to enabled_for_short_it.
8779         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8780         Remove default uses.
8781         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8782         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8783         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8784         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8786 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8788         PR target/82748
8789         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8790         float128 helper macros, which are no longer used after deleting
8791         the old 'q' built-in functions, and moving the round to odd
8792         built-in functions to being special built-in functions.
8793         (BU_FLOAT128_2): Likewise.
8794         (BU_FLOAT128_1_HW): Likewise.
8795         (BU_FLOAT128_2_HW): Likewise.
8796         (BU_FLOAT128_3_HW): Likewise.
8797         (FABSQ): Delete old 'q' built-in functions.
8798         (COPYSIGNQ): Likewise.
8799         (SQRTF128_ODD): Move round to odd built-in functions to be
8800         special built-in functions, so that we can handle
8801         -mabi=ieeelongdouble.
8802         (TRUNCF128_ODD): Likewise.
8803         (ADDF128_ODD): Likewise.
8804         (SUBF128_ODD): Likewise.
8805         (MULF128_ODD): Likewise.
8806         (DIVF128_ODD): Likewise.
8807         (FMAF128_ODD): Likewise.
8808         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8809         built-in names to 'f128'.
8810         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8811         old 'q' built-in functions, as the machine independent code for
8812         'f128' built-in functions handles this.
8813         (rs6000_expand_builtin): Add expansion for float128 round to odd
8814         functions, keying off on -mabi=ieeelongdouble of whether to use
8815         the KFmode or TFmode variant.
8816         (rs6000_init_builtins): Initialize the _Float128 round to odd
8817         built-in functions.
8818         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8819         _Float128 'q' built-in functions are now mapped into the new
8820         'f128' built-in functions.
8822 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8824         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8825         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8827 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8829         PR jit/82826
8830         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8831         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8832         * toplev.c: Include "ipa-fnsummary.h".
8833         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8835 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8837         PR tree-optimization/82838
8838         * gimple-ssa-store-merging.c
8839         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8840         on a separate gimple_seq which is then appended to seq.
8842 2017-11-06  Jeff Law  <law@redhat.com>
8844         PR target/82788
8845         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8846         (get_probe_interval): New functions.
8847         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8848         (ix86_adjust_stack_and_probe): Likewise.
8849         (output_adjust_stack_and_probe): Likewise.
8850         (ix86_emit_probe_stack_range): Likewise.
8851         (ix86_expand_prologue): Likewise.
8853 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8855         PR tree-optimization/82816
8856         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8857         if the modes of the two types are the same.
8858         (convert_plusminus_to_widen): Likewise.
8860 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8862         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8863         p9_vadu<mode>3.
8864         (usadv16qi): New define_expand.
8865         (usadv8hi): New define_expand.
8867 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
8869         PR bootstrap/82832
8870         * ipa-inline-transform.c (update_noncloned_frequencies): Always
8871         scale.
8872         (inline_transform): Likewise.
8873         * predict.c (counts_to_freqs): Remove useless conditional.
8874         * profile-count.h (profile_count::apply_scale): Move sanity check.
8875         * tree-inline.c (copy_bb): Always scale.
8876         (copy_cfg_body): Likewise.
8878 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
8880         PR target/67591
8881         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
8882         attribute.
8883         (*cmp_ite0): Add enabled_for_depr_it attribute.
8884         (*cmp_ite1): Likewise.
8886 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8888         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
8889         TYPE_MFCRF.
8891 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8893         * tree-vrp.c (vrp_int_const_binop): Return true on success and
8894         return the value by pointer.
8895         (extract_range_from_multiplicative_op_1): Update accordingly.
8896         Return as soon as an operation fails.
8898 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8900         PR other/82784
8901         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
8902         (DEF_SANITIZER_BUILTIN): ... here.
8903         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
8904         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
8906 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8908         PR other/82784
8909         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
8910         macro body.
8911         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
8912         ASM_OUTPUT_BEFORE_CASE_LABEL call.
8913         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
8914         after macro body.
8915         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8916         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8917         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8919 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8921         PR other/82784
8922         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
8923         "do {} while (0)".
8925 2017-11-04  Michael Clark  <michaeljclark@mac.com>
8927         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
8928         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
8929         (adddi3): Likewise.
8930         (*addsi3_extended): Likewise.
8931         (*addsi3_extended2): Likewise.
8932         (<optab>si3): Likewise.
8933         (<optab>di3): Likewise.
8934         (<optab><mode>3): Likewise.
8935         (<*optabe>si3_internal): Likewise.
8936         (zero_extendqi<SUPERQI:mode>2): Likewise.
8937         (*add<mode>hi3): Likewise.
8938         (*xor<mode>hi3): Likewise.
8939         (<optab>di3): Likewise.
8940         (*<optab>si3_extend): Likewise.
8941         (*sge<u>_<X:mode><GPR:mode>): Likewise.
8942         (*slt<u>_<X:mode><GPR:mode>): Likewise.
8943         (*sle<u>_<X:mode><GPR:mode>): Likewise.
8945 2017-11-04  Andrew Waterman  <andrew@sifive.com>
8947         * config/riscv/riscv.c (riscv_option_override): Conditionally set
8948         TARGET_STRICT_ALIGN based upon -mtune argument.
8950 2017-11-04  Andrew Waterman  <andrew@sifive.com>
8952         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
8954 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
8956         * config/i386/i386.c (choose_basereg): Use optional scratch
8957         register and add assertion.
8958         (x86_emit_outlined_ms2sysv_save): Use scratch register when
8959         needed, and don't allocate stack.
8960         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
8961         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
8962         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
8964 2017-11-03  Jeff Law  <law@redhat.com>
8966         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
8967         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
8968         to probe at the start of a noreturn function.
8970 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
8972         PR tree-optimization/78821
8973         * gimple-ssa-store-merging.c: Update the file comment.
8974         (MAX_STORE_ALIAS_CHECKS): Define.
8975         (struct store_operand_info): New type.
8976         (store_operand_info::store_operand_info): New constructor.
8977         (struct store_immediate_info): Add rhs_code and ops data members.
8978         (store_immediate_info::store_immediate_info): Add rhscode, op0r
8979         and op1r arguments to the ctor, initialize corresponding data members.
8980         (struct merged_store_group): Add load_align_base and load_align
8981         data members.
8982         (merged_store_group::merged_store_group): Initialize them.
8983         (merged_store_group::do_merge): Update them.
8984         (merged_store_group::apply_stores): Pick the constant for
8985         encode_tree_to_bitpos from one of the two operands, or skip
8986         encode_tree_to_bitpos if neither operand is a constant.
8987         (class pass_store_merging): Add process_store method decl.  Remove
8988         bool argument from terminate_all_aliasing_chains method decl.
8989         (pass_store_merging::terminate_all_aliasing_chains): Remove
8990         var_offset_p argument and corresponding handling.
8991         (stmts_may_clobber_ref_p): New function.
8992         (compatible_load_p): New function.
8993         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
8994         if there is overlap and rhs_code is not INTEGER_CST.  For
8995         non-overlapping stores terminate group if rhs is not mergeable.
8996         (get_alias_type_for_stmts): Change first argument from
8997         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
8998         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
8999         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9000         alias type.
9001         (get_location_for_stmts): Change first argument from
9002         auto_vec<gimple *> & to vec<gimple *> &.
9003         (struct split_store): Remove orig_stmts data member, add orig_stores.
9004         (split_store::split_store): Create orig_stores rather than orig_stmts.
9005         (find_constituent_stmts): Renamed to ...
9006         (find_constituent_stores): ... this.  Change second argument from
9007         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9008         to info structures rather than the statements.
9009         (split_group): Rename ALLOW_UNALIGNED argument to
9010         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9011         it.  Adjust find_constituent_stores caller.
9012         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9013         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9014         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9015         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9016         (mem_valid_for_store_merging): New function.
9017         (handled_load): New function.
9018         (pass_store_merging::process_store): New method.
9019         (pass_store_merging::execute): Use process_store method.  Adjust
9020         terminate_all_aliasing_chains caller.
9022 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9024         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9025         Return true for more constants, symbols and label references.
9026         (aarch64_valid_floating_const): Remove unused function.
9028 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9030         PR target/82786
9031         * config/aarch64/aarch64.c (aarch64_layout_frame):
9032         Undo forcing of LR at bottom of frame.
9034 2017-11-03  Jeff Law  <law@redhat.com>
9036         PR target/82823
9037         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9038         for int_registers_saved.
9040         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9041         extracted from tree-ssa-dom.c.
9042         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9043         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9044         (record_equivalences_from_incoming_edge): Add additional argument
9045         to single_pred_edge_ignoring_loop_edges call.
9046         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9047         (uncprop_dom_walker::before_dom_children): Add additional argument
9048         to single_pred_edge_ignoring_loop_edges call.
9049         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9050         single_pred_edge_ignoring_loop_edges rather than open coding.
9051         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9053 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9055         * match.pd (-(-A)): Rewrite.
9057 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9059         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9060         (rs6000_emit_int_cmove): New declaration.
9061         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9062         (rs6000_emit_sISEL): Delete.
9063         (rs6000_emit_int_cmove): Make non-static.
9064         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9065         instead of rs6000_emit_sISEL.
9067 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9069         * asan.c (create_cond_insert_point): Maintain profile.
9070         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9071         merged.
9072         * basic-block.h (struct basic_block_def): Remove frequency.
9073         (EDGE_FREQUENCY): Use to_frequency
9074         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9075         heuristics.
9076         (find_traces): Update to use to_frequency.
9077         (find_traces_1_round): Likewise; use only IPA counts.
9078         (bb_to_key): Likewise.
9079         (connect_traces): Use IPA counts only.
9080         (copy_bb_p): Update to use to_frequency.
9081         (fix_up_crossing_landing_pad): Likewise.
9082         (sanitize_hot_paths): Likewise.
9083         * bt-load.c (basic_block_freq): Likewise.
9084         * cfg.c (init_flow): Set count_max to uninitialized.
9085         (check_bb_profile): Remove frequencies; check counts.
9086         (dump_bb_info): Do not dump frequencies.
9087         (update_bb_profile_for_threading): Update counts only.
9088         (scale_bbs_frequencies_int): Likewise.
9089         (MAX_SAFE_MULTIPLIER): Remove.
9090         (scale_bbs_frequencies_gcov_type): Update counts only.
9091         (scale_bbs_frequencies_profile_count): Update counts only.
9092         (scale_bbs_frequencies): Update counts only.
9093         * cfg.h (struct control_flow_graph): Add count-max.
9094         (update_bb_profile_for_threading): Update prototype.
9095         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9096         (find_many_sub_basic_blocks): Likewise.
9097         * cfgcleanup.c (try_forward_edges): Likewise.
9098         (try_crossjump_to_edge): Likewise.
9099         * cfgexpand.c (expand_gimple_cond): Likewise.
9100         (expand_gimple_tailcall): Likewise.
9101         (construct_init_block): Likewise.
9102         (construct_exit_block): Likewise.
9103         * cfghooks.c (verify_flow_info): Check consistency of counts.
9104         (dump_bb_for_graph): Do not dump frequencies.
9105         (split_block_1): Do not update frequencies.
9106         (split_edge): Do not update frequencies.
9107         (make_forwarder_block): Do not update frequencies.
9108         (duplicate_block): Do not update frequencies.
9109         (account_profile_record): Do not update frequencies.
9110         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9111         for global heuristics.
9112         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9113         (expected_loop_iterations_unbounded): Use counts only.
9114         * cfgloopmanip.c (scale_loop_profile): Simplify.
9115         (create_empty_loop_on_edge): Simplify
9116         (loopify): Simplify
9117         (duplicate_loop_to_header_edge): Simplify
9118         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9119         (update_br_prob_note): Take care of removing note when profile
9120         becomes undefined.
9121         (relink_block_chain): Do not dump frequency.
9122         (rtl_account_profile_record): Use to_frequency.
9123         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9124         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9125         (cgraph_update_edges_for_call_stmt_node): Likewise.
9126         (cgraph_edge::verify_count_and_frequency): Update.
9127         (cgraph_node::verify_node): Temporarily disable frequency verification.
9128         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9129         to_cgraph_frequency.
9130         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9131         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9132         frequencies.
9133         (cgraph_node::expand_thunk): Update profile.
9134         * except.c (dw2_build_landing_pads): Do not update frequency.
9135         * final.c (compute_alignments): Use to_frequency.
9136         (dump_basic_block_info): Do not dump frequency.
9137         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9138         (dump_gimple_bb_header): Do not dump frequency.
9139         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9140         do update count.
9141         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9142         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9143         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9144         (init_before_recovery): Do not update frequency.
9145         (sched_create_recovery_edges): Do not update frequency.
9146         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9147         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9148         (ipa_cp_c_finalize): Set max_count to uninitialized.
9149         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9150         (param_change_prob): Use counts.
9151         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9152         local profiles.
9153         * ipa-split.c (consider_split): Use to_frequency.
9154         (split_function): Use to_frequency.
9155         * ira-build.c (loop_compare_func): Likewise.
9156         (mark_loops_for_removal): Likewise.
9157         (mark_all_loops_for_removal): Likewise.
9158         * loop-doloop.c (doloop_modify): Do not update frequency.
9159         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9160         frequency.
9161         * lto-streamer-in.c (input_function): Update count_max.
9162         * omp-expand.c (expand_omp_taskreg): Update count_max.
9163         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9164         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9165         (maybe_hot_count_p): Use ipa counts only.
9166         (maybe_hot_bb_p): Simplify.
9167         (maybe_hot_edge_p): Simplify.
9168         (probably_never_executed): Do not take frequency argument.
9169         (probably_never_executed_bb_p): Do not pass frequency.
9170         (probably_never_executed_edge_p): Likewise.
9171         (combine_predictions_for_bb): Check that profile is nonzero.
9172         (propagate_freq): Do not set frequency.
9173         (drop_profile): Simplify.
9174         (counts_to_freqs): Simplify.
9175         (expensive_function_p): Use to_frequency.
9176         (propagate_unlikely_bbs_forward): Simplify.
9177         (determine_unlikely_bbs): Simplify.
9178         (estimate_bb_frequencies): Add hack to silence graphite issues.
9179         (compute_function_frequency): Use ipa counts.
9180         (pass_profile::execute): Update.
9181         (rebuild_frequencies): Use counts only.
9182         (force_edge_cold): Use counts only.
9183         * profile-count.c (profile_count::dump): Dump new count types.
9184         (profile_count::differs_from_p): Check compatiblity.
9185         (profile_count::to_frequency): New function.
9186         (profile_count::to_cgraph_frequency): New function.
9187         * profile-count.h (struct function): Declare.
9188         (enum profile_quality): Add profile_guessed_local and
9189         profile_guessed_global0.
9190         (class profile_proability): Decrease number of bits to 29;
9191         update from_reg_br_prob_note and to_reg_br_prob_note.
9192         (class profile_count: Update comment; decrease number of bits
9193         to 61. Check compatibility.
9194         (profile_count::compatible_p): New private member function.
9195         (profile_count::ipa_p): New member function.
9196         (profile_count::operator<): Handle global zero correctly.
9197         (profile_count::operator>): Handle global zero correctly.
9198         (profile_count::operator<=): Handle global zero correctly.
9199         (profile_count::operator>=): Handle global zero correctly.
9200         (profile_count::nonzero_p): New member function.
9201         (profile_count::force_nonzero): New member function.
9202         (profile_count::max): New member function.
9203         (profile_count::apply_scale): Handle IPA scalling.
9204         (profile_count::guessed_local): New member function.
9205         (profile_count::global0): New member function.
9206         (profile_count::ipa): New member function.
9207         (profile_count::to_frequency): Declare.
9208         (profile_count::to_cgraph_frequency): Declare.
9209         * profile.c (OVERLAP_BASE): Delete.
9210         (compute_frequency_overlap): Delete.
9211         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9212         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9213         * sched-ebb.c (rank): Use counts only.
9214         * shrink-wrap.c (handle_simple_exit): Use counts only.
9215         (try_shrink_wrapping): Use counts only.
9216         (place_prologue_for_one_component): Use counts only.
9217         * tracer.c (find_best_predecessor): Use to_frequency.
9218         (find_trace): Use to_frequency.
9219         (tail_duplicate): Use to_frequency.
9220         * trans-mem.c (expand_transaction): Do not update frequency.
9221         * tree-call-cdce.c: Do not update frequency.
9222         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9223         (gimple_merge_blocks): Likewise.
9224         (gimple_split_edge): Likewise.
9225         (gimple_duplicate_sese_region): Likewise.
9226         (gimple_duplicate_sese_tail): Likewise.
9227         (move_sese_region_to_fn): Likewise.
9228         (gimple_account_profile_record): Likewise.
9229         (insert_cond_bb): Likewise.
9230         * tree-complex.c (expand_complex_div_wide): Likewise.
9231         * tree-eh.c (lower_resx): Update profile.
9232         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9233         frequencies.
9234         (initialize_cfun): Do not initialize frequencies
9235         (freqs_to_counts): Delete.
9236         (copy_cfg_body): Ignore count parameter.
9237         (copy_body): Update.
9238         (expand_call_inline): Update count_max.
9239         (optimize_inline_calls): Update count_max.
9240         (tree_function_versioning): Update count_max.
9241         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9242         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9243         frequency.
9244         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9245         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9246         (try_peel_loop): Likewise.
9247         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9248         to_frequency.
9249         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9250         (tree_transform_and_unroll_loop): Do not use frequencies
9251         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9252         Use reliable prediction only.
9253         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9254         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9255         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9256         probability scaling.
9257         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9258         not update frequency
9259         (any_remaining_duplicated_blocks): Likewise.
9260         (update_profile): Likewise.
9261         (estimated_freqs_path): Delete.
9262         (freqs_to_counts_path): Delete.
9263         (clear_counts_path): Delete.
9264         (ssa_fix_duplicate_block_edges): Likewise.
9265         (duplicate_thread_path): Likewise.
9266         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9267         * tree-tailcall.c (decrease_profile): Do not update frequency.
9268         (eliminate_tail_call): Likewise.
9269         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9270         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9271         (optimize_mask_stores): Likewise.
9272         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9273         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9274         (ubsan_expand_ptr_ifn): Update profile.
9275         * value-prof.c (gimple_ic): Simplify.
9276         * value-prof.h (gimple_ic): Update prototype.
9277         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9278         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9279         counts are nonzero.
9280         (want_inline_self_recursive_call_p): Likewise.
9281         (resolve_noninline_speculation): Only cummulate defined counts.
9282         (inline_small_functions): Use nonzero_p.
9283         (ipa_inline): Do not access freed node.
9285 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9287         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9288         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9290 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9292         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9293         non-legitimate address.
9295 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9297         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9298         (*lt0_<mode>di, *lt0_<mode>si): New.
9300 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9302         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9303         TARGET_PAIRED_FLOAT.
9305 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9306             Jim Wilson  <jim.wilson@linaro.org>
9308         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9309         * config/aarch64/aarch64-tune.md: Regenerated.
9310         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9311         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9313 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9315         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9316         (arc_expand_prologue): Restore blink for millicode.
9317         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9319 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9321         PR target/82809
9322         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9323         gen_vec_duplicate after forcing the scalar into a register.
9325 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9327         * combine (try_combine): Print the insns input to try_combine to the
9328         dump file.
9330 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9332         PR target/79868
9333         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9334         Remove second argument from aarch64_process_target_attr call.
9335         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9336         Ditto.
9337         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9338         field type.
9339         (aarch64_handle_attr_arch): Remove second argument.
9340         (aarch64_handle_attr_cpu): Ditto.
9341         (aarch64_handle_attr_tune): Ditto.
9342         (aarch64_handle_attr_isa_flags): Ditto.
9343         (aarch64_process_one_target_attr): Ditto.
9344         (aarch64_process_target_attr): Ditto.
9345         (aarch64_option_valid_attribute_p): Remove second argument.
9346         on aarch64_process_target_attr call.
9348 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9350         * diagnostic.c: Include "selftest-diagnostic.h".
9351         (selftest::assert_location_text): New function.
9352         (selftest::test_diagnostic_get_location_text): New function.
9353         (selftest::diagnostic_c_tests): Call it.
9355 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9357         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9358         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9359         (class selftest::test_diagnostic_context): Move to...
9360         * selftest-diagnostic.c: New file.
9361         * selftest-diagnostic.h: New file.
9363 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9365         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9366         offset range for FT32B.
9367         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9368         * config/ft32/ft32.md: Add TARGET_NOPM.
9369         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9370         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9372 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9374         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9376 2017-11-02  Jeff Law  <law@redhat.com>
9378         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9379         virtual keyword on FINAL OVERRIDE members.
9381         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9382         virtuals together.  Add virtual destructor.
9383         (substitute_and_fold_engine): Similarly.
9385 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9387         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9389 2017-11-02  Richard Biener  <rguenther@suse.de>
9391         PR tree-optimization/82795
9392         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9394 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9396         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9397         gcc_SUN_LD_VERSION.
9398         (gcc_GAS_CHECK_FEATURE): Remove.
9399         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9400         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9401         * configure: Regenerate.
9403 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9405         * config/arc/arc.c (hwloop_optimize): Account for empty
9406         body loops.
9408 2017-11-02  Richard Biener  <rguenther@suse.de>
9410         PR middle-end/82765
9411         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9412         Truncate ARRAY_REF index and element size.
9414 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9416         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9418 2017-11-01  Jeff Law  <law@redhat.com>
9420         * tree-ssa-ccp.c (ccp_folder): New class derived from
9421         substitute_and_fold_engine.
9422         (ccp_folder::get_value): New member function.
9423         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9424         (ccp_fold_stmt): Remove prototype.
9425         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9426         * tree-ssa-copy.c (copy_folder): New class derived from
9427         substitute_and_fold_engine.
9428         (copy_folder::get_value): Renamed from get_value.
9429         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9430         * tree-vrp.c (vrp_folder): New class derived from
9431         substitute_and_fold_engine.
9432         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9433         (vrp_folder::get_value): New member function.
9434         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9435         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9436         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9437         provide a class interface to folder/substitute routines.
9438         (ssa_prop_fold_stmt_fn): Remove typedef.
9439         (ssa_prop_get_value_fn): Likewise.
9440         (subsitute_and_fold): Remove prototype.
9441         (replace_uses_in): Likewise.
9442         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9443         Renamed from replace_uses_in.  Call the virtual member function
9444         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9445         (substitute_and_fold_dom_walker): Remove initialization of
9446         data member entries for calbacks.  Add substitute_and_fold_engine
9447         member and initialize it.
9448         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9449         member functions for get_value, replace_phi_args_in c
9450         replace_uses_in, and fold_stmt calls.
9451         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9452         substitute_and_fold.  Remove assert.   Update ctor call.
9454         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9455         (ssa_prop_visit_phi_fn): Likewise.
9456         (class ssa_propagation_engine): New class to provide an interface
9457         into ssa_propagate.
9458         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9459         variable.
9460         (ssa_prop_visit_phi): Likewise.
9461         (ssa_propagation_engine::simulate_stmt): Moved into class.
9462         Call visit_phi/visit_stmt from the class rather than via
9463         file scoped static variables.
9464         (ssa_propagation_engine::simulate_block): Moved into class.
9465         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9466         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9467         set file scoped statics for the visit_stmt/visit_phi callbacks.
9468         * tree-complex.c (complex_propagate): New class derived from
9469         ssa_propagation_engine.
9470         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9471         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9472         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9473         class.
9474         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9475         ssa_propagation_engine.
9476         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9477         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9478         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9479         * tree-ssa-copy.c (copy_prop): New class derived from
9480         ssa_propagation_engine.
9481         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9482         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9483         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9484         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9485         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9486         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9487         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9489 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9491         PR rtl-optimization/82778
9492         PR rtl-optimization/82597
9493         * compare-elim.c (struct comparison): Add in_a_setter field.
9494         (find_comparison_dom_walker::before_dom_children): Remove killed
9495         bitmap and df_simulate_find_defs call, instead walk the defs.
9496         Compute last_setter and initialize in_a_setter.  Merge definitions
9497         with first initialization for a few variables.
9498         (try_validate_parallel): Use insn_invalid_p instead of
9499         recog_memoized.  Return insn rather than just the pattern.
9500         (try_merge_compare): Fix up comment.  Don't uselessly test if
9501         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9502         chains.
9503         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9504         call.
9506 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9507             Alan Hayward  <alan.hayward@arm.com>
9508             David Sherwood  <david.sherwood@arm.com>
9510         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9511         aarch64_hard_regno_nregs to get the number of registers
9512         in a mode.
9514 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9515             Alan Hayward  <alan.hayward@arm.com>
9516             David Sherwood  <david.sherwood@arm.com>
9518         * config/aarch64/constraints.md (Upl): Rename to...
9519         (Uaa): ...this.
9520         * config/aarch64/aarch64.md
9521         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9522         Update accordingly.
9524 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9525             Alan Hayward  <alan.hayward@arm.com>
9526             David Sherwood  <david.sherwood@arm.com>
9528         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9529         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9530         earlier in file.
9532 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9533             Alan Hayward  <alan.hayward@arm.com>
9534             David Sherwood  <david.sherwood@arm.com>
9536         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9537         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9538         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9539         named expanders.
9540         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9541         of a single element.
9542         * config/aarch64/iterators.md: Add a comment above the permute
9543         unspecs to say that they are generated directly by
9544         aarch64_expand_vec_perm_const.
9545         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9547 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9549         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9551 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9553         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9554         and medany code models, and describe what they do.
9556 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9558         Revert accidental duplicate:
9560         * combine.c (can_change_dest_mode): Reject changes in
9561         REGMODE_NATURAL_SIZE.
9563 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9565         PR rtl-optimization/64682
9566         PR rtl-optimization/69567
9567         PR rtl-optimization/69737
9568         PR rtl-optimization/82683
9569         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9570         register mentioned in the note, drop the note, unless it came from I3,
9571         in which case it should go to I3 again.
9573 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9575         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9576         and return false if not.
9577         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9579 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9581         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9582         range is known to be empty.
9584 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9585             Alan Hayward  <alan.hayward@arm.com>
9586             David Sherwood  <david.sherwood@arm.com>
9588         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9589         and CONST_VECTOR_NUNITS instead of computing the number of units from
9590         the byte sizes of the vector and element.
9591         (simplify_binary_operation_1): Likewise.
9592         (simplify_const_binary_operation): Likewise.
9593         (simplify_ternary_operation): Likewise.
9595 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9596             Alan Hayward  <alan.hayward@arm.com>
9597             David Sherwood  <david.sherwood@arm.com>
9599         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9600         (int_mem_offset): ...this new function.
9601         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9602         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9603         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9604         Update accordingly.
9606 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9607             Alan Hayward  <alan.hayward@arm.com>
9608             David Sherwood  <david.sherwood@arm.com>
9610         * lower-subreg.c (interesting_mode_p): New function.
9611         (compute_costs, find_decomposable_subregs, decompose_register)
9612         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9613         (resolve_clobber, dump_choices): Use it.
9615 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9616             Alan Hayward  <alan.hayward@arm.com>
9617             David Sherwood  <david.sherwood@arm.com>
9619         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9621 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9622             Alan Hayward  <alan.hayward@arm.com>
9623             David Sherwood  <david.sherwood@arm.com>
9625         * alias.c (find_base_value, find_base_term): Only process integer
9626         truncations.  Check the precision rather than the size.
9628 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9629             Alan Hayward  <alan.hayward@arm.com>
9630             David Sherwood  <david.sherwood@arm.com>
9632         * machmode.h (is_narrower_int_mode): New function
9633         * optabs.c (expand_float, expand_fix): Use it.
9634         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9636 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9637             Alan Hayward  <alan.hayward@arm.com>
9638             David Sherwood  <david.sherwood@arm.com>
9640         * rtl.h (narrower_subreg_mode): New function.
9641         * ira-color.c (update_costs_from_allocno): Use it.
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         * optabs-query.h (convert_optab_p): New function, split out from...
9648         (convert_optab_handler): ...here.
9649         (widening_optab_handler): Delete.
9650         (find_widening_optab_handler): Remove permit_non_widening parameter.
9651         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9652         override that operates on mode class wrappers.
9653         * optabs-query.c (widening_optab_handler): Delete.
9654         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9655         parameter.  Assert that the two modes are the same class and that
9656         the "from" mode is narrower than the "to" mode.  Use
9657         convert_optab_handler instead of widening_optab_handler.
9658         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9659         instead of widening_optab_handler.
9660         * expr.c (expand_expr_real_2): Update calls to
9661         find_widening_optab_handler.
9662         * optabs.c (expand_widen_pattern_expr): Likewise.
9663         (expand_binop_directly): Take the insn_code as a parameter.
9664         (expand_binop): Only call find_widening_optab_handler for
9665         conversion optabs; use optab_handler otherwise.  Update calls
9666         to find_widening_optab_handler and expand_binop_directly.
9667         Use convert_optab_handler instead of widening_optab_handler.
9668         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9669         find_widening_optab_handler and use scalar_mode rather than
9670         machine_mode.
9671         (convert_plusminus_to_widen): Likewise.
9673 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9674             Alan Hayward  <alan.hayward@arm.com>
9675             David Sherwood  <david.sherwood@arm.com>
9677         * machmode.h (fixed_size_mode): New class.
9678         * rtl.h (get_pool_mode): Return fixed_size_mode.
9679         * gengtype.c (main): Add fixed_size_mode.
9680         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9681         (get_raw_arg_mode): Likewise.
9682         * doc/tm.texi: Regenerate.
9683         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9684         * targhooks.c (default_get_reg_raw_mode): Likewise.
9685         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9686         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9687         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9688         (msp430_get_raw_result_mode): Likewise.
9689         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9690         * dbxout.c (dbxout_parms): Require fixed-size modes.
9691         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9692         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9693         * omp-low.c (lower_oacc_reductions): Likewise.
9694         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9695         (simplify_subreg): Update accordingly.
9696         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9697         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9698         that aren't fixed-size.
9699         (get_pool_mode): Return a fixed_size_mode.
9700         (output_constant_pool_2): Take a fixed_size_mode.
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         * doc/rtl.texi (vec_series): Document.
9707         (const): Say that the operand can be a vec_series.
9708         * rtl.def (VEC_SERIES): New rtx code.
9709         * rtl.h (const_vec_series_p_1): Declare.
9710         (const_vec_series_p): New function.
9711         * emit-rtl.h (gen_const_vec_series): Declare.
9712         (gen_vec_series): Likewise.
9713         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9714         (gen_vec_series): Likewise.
9715         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9716         * simplify-rtx.c (simplify_unary_operation): Handle negations
9717         of vector series.
9718         (simplify_binary_operation_series): New function.
9719         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9720         (test_vector_ops_series): New function.
9721         (test_vector_ops): Call it.
9722         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9723         gen_const_vec_series.
9724         (altivec_lvsr): Likewise.
9725         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9727 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9728             Alan Hayward  <alan.hayward@arm.com>
9729             David Sherwood  <david.sherwood@arm.com>
9731         * doc/rtl.texi (const): Update description of address constants.
9732         Say that vector constants are allowed too.
9733         * common.md (E, F): Use CONSTANT_P instead of checking for
9734         CONST_VECTOR.
9735         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9736         checking for CONST_VECTOR.
9737         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9738         VEC_DUPLICATE.
9739         * expr.c (expand_expr_real_2): Check for vector modes instead
9740         of checking for CONST_VECTOR.
9741         * rtl.h (const_vec_p): New function.
9742         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9743         (unwrap_const_vec_duplicate): Handle them here too.
9745 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9746             David Malcolm  <dmalcolm@redhat.com>
9747             Alan Hayward  <alan.hayward@arm.com>
9748             David Sherwood  <david.sherwood@arm.com>
9750         * rtl.h (vec_duplicate_p): New function.
9751         * selftest-rtl.c (assert_rtx_eq_at): New function.
9752         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9753         (assert_rtx_eq_at): Declare.
9754         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9755         * selftest-run-tests.c (selftest::run_tests): Call it.
9756         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9757         (simplify_unary_operation_1): Recursively handle vector duplicates.
9758         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9759         vector duplicates.
9760         (simplify_subreg): Handle subregs of vector duplicates.
9761         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9762         (selftest::simplify_rtx_c_tests): New functions.
9764 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9765             Alan Hayward  <alan.hayward@arm.com>
9766             David Sherwood  <david.sherwood@arm.com>
9768         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9769         (gen_vec_duplicate): Likewise.
9770         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9771         out from...
9772         (gen_const_vector): ...here.
9773         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9774         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9775         whose elements are all equal.
9776         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9777         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9778         (simplify_relational_operation): Likewise.
9779         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9780         Likewise.
9781         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9782         (aarch64_expand_vector_init): Likewise.
9783         * config/arm/arm.c (neon_vdup_constant): Likewise.
9784         (neon_expand_vector_init): Likewise.
9785         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9786         (arm_block_set_unaligned_vect): Likewise.
9787         (arm_block_set_aligned_vect): Likewise.
9788         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9789         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9790         (expand_vec_perm_even_odd_pack): Likewise.
9791         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9792         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9793         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9794         gen_const_vec_duplicate.
9795         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9796         * config/mips/mips.c (mips_gen_const_int_vector): Use
9797         gen_const_vec_duplicate.
9798         (mips_expand_vector_init): Use CONST0_RTX.
9799         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9800         (define_split): Use gen_const_vec_duplicate.
9801         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9802         (define_split): Use gen_const_vec_duplicate.
9803         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9804         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9805         * config/spu/spu.c (spu_const): Likewise.
9807 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9808             Alan Hayward  <alan.hayward@arm.com>
9809             David Sherwood  <david.sherwood@arm.com>
9811         * combine.c (can_change_dest_mode): Reject changes in
9812         REGMODE_NATURAL_SIZE.
9814 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9816         * configure.ac (--enable-libssp): New.
9817         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9818         trying to determine target-specific default.  Adjust indentation.
9819         * configure: Regenerated.
9820         * doc/install.texi (Configuration): Expand --disable-libssp
9821         documentation.
9823 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9825         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9826         calculation.
9828 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9830         PR c++/81702
9831         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9833 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9835         * auto-profile.c (autofdo_source_profile::read): Use
9836         UNKNOWN_LOCATION rather than 0.
9837         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9838         (warning_at): ...this overload.
9839         (warning_at_rich_loc_n): Rename to...
9840         (warning_n): ...this overload.
9841         (error_at_rich_loc): Rename to...
9842         (error_at): ...this overload.
9843         (pedwarn_at_rich_loc): Rename to...
9844         (pedwarn): ...this overload.
9845         (permerror_at_rich_loc): Rename to...
9846         (permerror): ...this overload.
9847         (inform_at_rich_loc): Rename to...
9848         (inform): ...this overload.
9849         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9850         (diagnostic_n_impl_richloc): Rename to...
9851         (diagnostic_n_impl): ...this rich_location *-based decl.
9852         (inform_at_rich_loc): Rename to...
9853         (inform): ...this, and add an assertion.
9854         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9855         (warning_at_rich_loc): Rename to...
9856         (warning_at): ...this, and add an assertion.
9857         (warning_at_rich_loc_n): Rename to...
9858         (warning_n): ...this, and add an assertion.
9859         (warning_n): Update location_t-based implementation for removal of
9860         location_t-based diagnostic_n_impl.
9861         (pedwarn_at_rich_loc): Rename to...
9862         (pedwarn): ...this, and add an assertion.
9863         (permerror_at_rich_loc): Rename to...
9864         (permerror): ...this, and add an assertion.
9865         (error_n): Update for removal of location_t-based diagnostic_n_impl.
9866         (error_at_rich_loc): Rename to...
9867         (error_at): ...this, and add an assertion.
9868         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
9869         (driver::do_spec_on_infiles): Likewise.
9870         * substring-locations.c (format_warning_va): Update for renaming
9871         of inform_at_rich_loc.
9873 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
9875         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
9876         _Float<N>X built-in functions so that the variant without the
9877         "__builtin_" prefix is only enabled for the GNU C and Objective C
9878         languages when they are in non-strict ANSI/ISO mode.
9879         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
9880         * target.def (floatn_builtin_p): Add a target hook to control
9881         whether _Float<N> and _Float<N>X built-in functions without the
9882         "__builtin_" prefix are enabled, and return true for C and
9883         Objective C in the default hook.  Include langhooks.h in
9884         targhooks.c.
9885         * targhooks.h (default_floatn_builtin_p): Likewise.
9886         * targhooks.c (default_floatn_builtin_p): Likewise.
9887         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
9888         floatn_builtin_p target hook.
9889         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
9891 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
9892             Eric Botcazou  <ebotcazou@adacore.com>
9894         PR rtl-optimization/81803
9895         * lra-constraints.c (curr_insn_transform): Also reload the whole
9896         register for a strict subreg no wider than a word if this is for
9897         a WORD_REGISTER_OPERATIONS target.
9899 2017-10-31  Jason Merrill  <jason@redhat.com>
9901         * gdbinit.in: Skip over inlines from timevar.h.
9903 2017-10-31  Martin Liska  <mliska@suse.cz>
9905         * doc/gcov.texi: Document new option.
9906         * gcov.c (print_usage): Likewise print it.
9907         (process_args): Support the argument.
9908         (format_count): New function.
9909         (format_gcov): Use the function.
9911 2017-10-31  Martin Liska  <mliska@suse.cz>
9913         * gcov.c (struct name_map): do not use typedef.
9914         Define operator== and operator<.
9915         (name_search): Remove.
9916         (name_sort): Remove.
9917         (main): Do not allocate names.
9918         (process_file): Add vertical space.
9919         (generate_results): Use std::find.
9920         (release_structures): Do not release memory.
9921         (find_source): Use std::find.
9923 2017-10-31  Martin Liska  <mliska@suse.cz>
9925         * gcov.c (struct line_info): Remove it's typedef.
9926         (line_info::line_info): Add proper ctor.
9927         (line_info::has_block): Do not use a typedef.
9928         (struct source_info): Do not use typedef.
9929         (circuit): Likewise.
9930         (get_cycles_count): Likewise.
9931         (output_intermediate_file): Iterate via vector iterator.
9932         (add_line_counts): Use std::vector methods.
9933         (accumulate_line_counts): Likewise.
9934         (output_lines): Likewise.
9936 2017-10-31  Martin Liska  <mliska@suse.cz>
9938         * gcov.c (struct source_info): Remove typedef.
9939         (source_info::source_info): Add proper ctor.
9940         (accumulate_line_counts): Use struct, not it's typedef.
9941         (output_gcov_file): Likewise.
9942         (output_lines): Likewise.
9943         (main): Do not allocate an array.
9944         (output_intermediate_file): Use size of vector container.
9945         (process_file): Resize the vector.
9946         (generate_results): Do not preallocate, use newly added vector
9947         lines.
9948         (release_structures): Do not release sources.
9949         (find_source): Use vector methods.
9950         (add_line_counts): Do not use typedef.
9952 2017-10-31  Martin Liska  <mliska@suse.cz>
9954         * doc/gcov.texi: Document that.
9955         * gcov.c (add_line_counts): Mark lines with a non-executed
9956         statement.
9957         (output_line_beginning): Handle such lines.
9958         (output_lines): Pass new argument.
9959         (output_intermediate_file): Print it in intermediate format.
9961 2017-10-31  Martin Liska  <mliska@suse.cz>
9963         * color-macros.h: New file.
9964         * diagnostic-color.c: Factor out color related to macros to
9965         color-macros.h.
9966         * doc/gcov.texi: Document -k option.
9967         * gcov.c (INCLUDE_STRING): Include string.h.
9968         (print_usage): Add -k option.
9969         (process_args): Parse it.
9970         (pad_count_string): New function.
9971         (output_line_beginning): Likewise.
9972         (DEFAULT_LINE_START): New macro.
9973         (output_lines): Support color output.
9975 2017-10-31  Martin Liska  <mliska@suse.cz>
9977         PR gcov-profile/82633
9978         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
9979         their interaction with GCOV infrastructure.
9980         * configure.ac: Add -fkeep-{inline,static}-functions to
9981         coverage_flags.
9982         * configure: Regenerate.
9984 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
9986         PR target/82772
9987         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
9989 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
9991         PR target/82674
9992         * config/rs6000/rs6000.md (allocate_stack): Force update interval
9993         into a register if it does not fit into an immediate offset field.
9995 2017-10-31  Olivier Hainque  <hainque@adacore.com>
9997         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
9999 2017-10-31  Julia Koval  <julia.koval@intel.com>
10001         * config.gcc: Add gfniintrin.h.
10002         * config/i386/gfniintrin.h: New.
10003         * config/i386/i386-builtin-types.def
10004         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10005         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10006         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10007         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10008         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10009         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10010         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10011         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10012         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10013         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10014         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10015         * config/i386/immintrin.h: Include gfniintrin.h.
10016         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10018 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10020         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10022 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10024         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10025         (arm_ashldi3_1bit): Remove pattern.
10026         (ashrdi3): Remove shift by 1 expansion.
10027         (arm_ashrdi3_1bit): Remove pattern.
10028         (lshrdi3): Remove shift by 1 expansion.
10029         (arm_lshrdi3_1bit): Remove pattern.
10030         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10031         cost of ashldi3 by 1.
10032         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10033         (<shift>di3_neon): Likewise.
10035 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10037         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10038         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10039         and (*aarch64_simd_mov<VQ:mode>).
10040         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10041         pattern alternative.
10042         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10043         attributes to match pattern alternative.
10045 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10047         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10048         * config/rs6000/emmintrin.h: New file.
10049         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10051 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10053         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10054         * (movdi_vfp_cortexa8): Remove pattern.
10056 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10058         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10059         etc. reference.
10060         (Specific, alpha*-dec-osf5.1): Remove.
10061         (Specific, mips-sgi-irix5): Remove.
10062         (Specific, mips-sgi-irix6): Remove.
10064 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10066         PR middle-end/22141
10067         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10068         arguments to clear_bit_region_be.
10070 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10072         * gimplify.c: Include memmodel.h.
10074 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10076         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10077         condition whether loop should be executed at all.
10079 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10081         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10082         gimple folding of vec_madd() intrinsics.
10083         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10084         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10085         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10087 2017-10-30  Richard Biener  <rguenther@suse.de>
10089         PR tree-optimization/82762
10090         Revert
10091         2017-10-23  Richard Biener  <rguenther@suse.de>
10093         PR tree-optimization/82129
10094         Revert
10095         2017-08-01  Richard Biener  <rguenther@suse.de>
10097         PR tree-optimization/81181
10098         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10099         (compute_antic): ... end of iteration here.
10101 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10103         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10104         -std=iso9899:2017 and -std=gnu17.
10105         * doc/standards.texi (C Language): Document C17 support.
10106         * doc/cpp.texi (Overview): Mention -std=c17.
10107         (Standard Predefined Macros): Document C11 and C17 values of
10108         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10109         * doc/extend.texi (Inline): Do not list individual options for
10110         standards newer than C99.
10111         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10112         "GNU C17".
10113         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10114         language name.
10116 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10118         * asan.c (asan_finish_file): Align asan globals array by shadow
10119         granularity.
10121 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10123         PR middle-end/22141
10124         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10125         (struct store_immediate_info): Add bitregion_start and bitregion_end
10126         fields.
10127         (store_immediate_info::store_immediate_info): Add brs and bre
10128         arguments and initialize bitregion_{start,end} from those.
10129         (struct merged_store_group): Add bitregion_start, bitregion_end,
10130         align_base and mask fields.  Drop unnecessary struct keyword from
10131         struct store_immediate_info.  Add do_merge method.
10132         (clear_bit_region_be): Use memset instead of loop storing zeros.
10133         (merged_store_group::do_merge): New method.
10134         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10135         stores as long as the surrounding bitregions have no gaps.
10136         (merged_store_group::merge_overlapping): Use do_merge.
10137         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10138         is byte aligned, rather than requiring that start and width are
10139         byte aligned.  Drop unnecessary struct keyword from
10140         struct store_immediate_info.  Allocate and populate also mask array.
10141         Make start of the arrays relative to bitregion_start rather than
10142         start and size them according to bitregion_{end,start} difference.
10143         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10144         struct store_immediate_info.
10145         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10146         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10147         struct keyword from struct store_immediate_info.
10148         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10149         between stores as long as the surrounding bitregions have no gaps.
10150         Formatting fixes.
10151         (struct split_store): Add orig non-static data member.
10152         (split_store::split_store): Initialize orig to false.
10153         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10154         if there is exactly a single original stmt.  Change stmts argument
10155         to pointer from reference, if NULL, don't push anything to it.  Add
10156         first argument, use it to optimize skipping over orig stmts that
10157         are known to be before bitpos already.  Simplify.
10158         (split_group): Return unsigned int count how many stores are or
10159         would be needed rather than a bool.  Add allow_unaligned argument.
10160         Change split_stores argument from reference to pointer, if NULL,
10161         only do a dry run computing how many stores would be produced.
10162         Rewritten algorithm to use both alignment and misalign if
10163         !allow_unaligned and handle bitfield stores with gaps.
10164         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10165         from bitregion_start instead of start.  Compute allow_unaligned
10166         here, if true, do 2 split_group dry runs to compute which one
10167         produces fewer stores and prefer aligned if equal.  Punt if
10168         new count is bigger or equal than original before emitting any
10169         statements, rather than during that.  Remove no longer needed
10170         new_ssa_names tracking.  Replace num_stmts with
10171         split_stores.length ().  Use 32-bit stack allocated entries
10172         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10173         if possible.  Handle bitfields with gaps.
10174         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10175         Compute bitregion_{start,end} for the stores and construct
10176         store_immediate_info with that.  Formatting fixes.
10178 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10180         PR target/82725
10181         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10182         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10184 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10186         * gimplify.c: Include tm_p.h.
10188         * common.opt (gcoff): Re-add as ignored option.
10189         (gcoff1, gcoff2, gcoff3): Likewise.
10191         * Makefile.in (OBJS): Delete sdbout.o.
10192         (GTFILES): Delete $(srcdir)/sdbout.c.
10193         * debug.h: Delete sdb_debug_hooks.
10194         * final.c: Delete sdbout.h include.
10195         (final_scan_insn): Delete SDB_DEBUG check.
10196         (rest_of_clean_state): Likewise.
10197         * output.h: Delete sdb_begin_function_line.
10198         * sdbout.c: Delete.
10199         * sdbout.h: Delete.
10200         * toplev.c: Delete sdbout.h include.
10201         (process_options): Delete SDB_DEBUG check.
10202         * tree-core.h (tree_type_common): Delete pointer field of
10203         tree_type_symtab.
10204         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10205         TYPE_SYMTAB_POINTER.
10206         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10207         (TYPE_SYMTAB_IS_POINTER): Delete.
10208         (TYPE_SYMTAB_IS_DIE): Renumber.
10209         * xcoffout.c: Refer to former sdbout.c file.
10210         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10212         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10213         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10214         (Debugging Options): Delete -gcoff.
10215         (-gstabs): Delete SDB reference.
10216         (-gcoff): Delete.
10217         (-gcoff@var{level}): Delete.
10218         * doc/passes.texi (Debugging information output): Delete SDB and
10219         sdbout.c references.
10220         * doc/tm.texi: Regenerate.
10221         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10222         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10223         references.
10224         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10225         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10226         (SDB_DEBUGGING_INFO): Delete.
10227         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10228         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10229         * target.def (output_source_filename): Delete COFF reference.
10231         * common.opt (gcoff): Delete.
10232         (gxcoff+): Update Negative chain.
10233         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10234         SDB_DEBUG.
10235         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10236         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10237         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10238         and SDB references.
10239         (expand_function_start): Change sdb reference to past tense.
10240         (expand_function_end): Change sdb reference to past tense.
10241         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10242         * opts.c (debug_type_names): Delete coff entry.
10243         (common_handle_option): Delete OPT_gcoff case.
10244         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10246         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10247         * config/cris/cris.h: Delete SDB reference in comment.
10248         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10249         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10250         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10251         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10252         to past tense.
10253         (ix86_expand_prologue): Likewise.
10254         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10255         * config/ia64/ia64.h: Likewise.
10256         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10257         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10258         support.
10259         * config/mmix/mmix.h: Likewise.
10260         * config/nds32/nds32.c: Likewise.
10261         * config/stormy/storym16.h: Likewise.
10262         * config/visium/visium.h: Likewise.
10263         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10265 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10267         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10268         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10269         FRAME_POINTER_REGNUM point at high end of local var area.
10271 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10273         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10274         Move comment around.  Do not reset best_edge for a copiable
10275         destination if the copy would cause a partition change.
10276         (better_edge_p): Remove redundant check.
10278 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10280         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10282 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10284         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10285         for math functions that have _Float<N> and _Float<N>X variants.
10286         (mathfn_built_in_2): Add support for math functions that have
10287         _Float<N> and _Float<N>X variants.
10288         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10289         (expand_builtin_mathfn_ternary): Add support for fma with
10290         _Float<N> and _Float<N>X variants.
10291         (expand_builtin): Likewise.
10292         (fold_builtin_3): Likewise.
10293         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10294         create math function _Float<N> and _Float<N>X variants as external
10295         library builtins.
10296         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10297         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10298         the __builtin_ prefix and if not strict ansi, without the prefix.
10299         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10300         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10301         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10302         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10303         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10304         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10305         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10306         function signatures for fma _Float<N> and _Float<N>X variants.
10307         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10308         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10309         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10310         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10311         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10312         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10313         * gencfn-macros.c (print_case_cfn): Add support for math functions
10314         that have _Float<N> and _Float<N>X variants.
10315         (print_define_operator_list): Likewise.
10316         (fltfn_suffixes): Likewise.
10317         (main): Likewise.
10318         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10319         for math functions that have _Float<N> and _Float<N>X variants.
10320         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10321         and _Float<N>X variants.
10322         (COPYSIGN): Likewise.
10323         (FMIN): Likewise.
10324         (FMAX): Likewise.
10325         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10326         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10327         variants.
10328         (integer_valued_read_call_p): Likewise.
10329         * fold-const-call.c (fold_const_call_ss): Likewise.
10330         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10331         _Float<N> and _Float<N>X variants.
10332         (fold_const_call_ssss): Add support for fma _Float<N> and
10333         _Float<N>X variants.
10334         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10335         support for copysign and fma _Float<N> and _Float<N>X variants.
10336         (backprop::process_builtin_call_use): Likewise.
10337         * tree-call-cdce.c (can_test_argument_range); Add support for
10338         sqrt _Float<N> and _Float<N>X variants.
10339         (edom_only_function): Likewise.
10340         (get_no_error_domain): Likewise.
10341         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10342         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10343         copysign _Float<N> and _Float<N>X variants.
10344         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10345         handled by machine independent code.
10346         (FMAF128): Likewise.
10347         * doc/cpp.texi (Common Predefined Macros): Document defining
10348         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10349         fma _Float<N> and _Float<N>X variants.
10351 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10353         PR target/82692
10354         * config/i386/i386-modes.def (CCFPU): Remove definition.
10355         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10356         (ix86_cc_modes_compatible): Ditto.
10357         (ix86_expand_carry_flag_compare): Ditto.
10358         (ix86_expand_int_movcc): Ditto.
10359         (ix86_expand_int_addcc): Ditto.
10360         (ix86_reverse_condition): Ditto.
10361         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10362         Return true/false for unordered/ordered fp comparisons.
10363         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10364         (ix86_prepare_fp_compare_args): Update for rename.
10365         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10366         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10367         (ix86_expand_sse_compare_and_jump): Ditto.
10368         * config/i386/predicates.md (fcmov_comparison_operator):
10369         Remove CCFPU mode handling.
10370         (ix86_comparison_operator): Ditto.
10371         (ix86_carry_flag_operator): Ditto.
10372         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10373         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10374         (*cmpu<mode>_cc_i387): Ditto.
10375         (FPCMP): Remove mode iterator.
10376         (unord): Remove mode attribute.
10377         (unord_subst): New define_subst transformation
10378         (unord): New define_subst attribute.
10379         (unordered): Ditto.
10380         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10381         (*cmpi<unord>xf_i387): Ditto.
10382         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10383         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10384         using unord_subst transformation.
10385         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10386         (round_saeonly): Also handle CCFP mode.
10387         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10388         Remove UNSPEC_SAHF unspec handling.
10390 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10392         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10394 2017-10-27  Jeff Law  <law@redhat.com>
10396         * gimple-ssa-sprintf.c: Include domwalk.h.
10397         (class sprintf_dom_walker): New class, derived from dom_walker.
10398         (sprintf_dom_walker::before_dom_children): New function.
10399         (struct call_info): Moved into sprintf_dom_walker class
10400         (compute_formath_length, handle_gimple_call): Likewise.
10401         (sprintf_length::execute): Call the dominator walker rather
10402         than walking the statements.
10404         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10405         gimple statement locations.
10406         (check_array_bounds): Corresponding changes.  Get the statement's
10407         location directly from wi->stmt.
10409 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10411         PR target/82717
10412         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10414 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10416         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10417         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10419 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10421         PR target/82703
10422         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10423         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10424         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10425         maybe_get_pool_constant.
10426         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10427         Likewise.
10429 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10431         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10432         2.26 caveat.  Update gas and gld versions.
10433         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10434         reference.
10436 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10438         * cgraph.h (set_malloc_flag): Declare.
10439         * cgraph.c (set_malloc_flag_1): New function.
10440         (set_malloc_flag): Likewise.
10441         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10442         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10443         false.
10444         (read_ipa_call_summary): Add support for reading is_return_callee.
10445         (write_ipa_call_summary): Stream is_return_callee.
10446         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10447         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10448         ipa-prop.h, ipa-fnsummary.h.
10449         (pure_const_names): Change to static.
10450         (malloc_state_e): Define.
10451         (malloc_state_names): Define.
10452         (funct_state_d): Add field malloc_state.
10453         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10454         (check_retval_uses): New function.
10455         (malloc_candidate_p): Likewise.
10456         (analyze_function): Add support for malloc attribute.
10457         (pure_const_write_summary): Stream malloc_state.
10458         (pure_const_read_summary): Add support for reading malloc_state.
10459         (dump_malloc_lattice): New function.
10460         (propagate_malloc): New function.
10461         (warn_function_malloc): New function.
10462         (ipa_pure_const::execute): Call propagate_malloc and
10463         ipa_free_fn_summary.
10464         (pass_local_pure_const::execute): Add support for malloc attribute.
10465         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10466         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10468 2017-10-27  Martin Liska  <mliska@suse.cz>
10470         PR gcov-profile/82457
10471         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10472         for fork-like functions to be properly instrumented.
10474 2017-10-27  Richard Biener  <rguenther@suse.de>
10476         PR middle-end/81659
10477         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10478         info when we redirected EH.
10480 2017-10-26  Michael Collison  <michael.collison@arm.com>
10482         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10483         New pattern.
10484         (<optab>_trunchf<GPI:mode>2: New pattern.
10485         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10486         * config/aarch64/iterators.md (wv): New mode attribute.
10487         (vf, VF): New mode attributes.
10488         (vgp, VGP): New mode attributes.
10489         (s): Update attribute with SImode and DImode prefixes.
10491 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10493         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10494         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10495         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10496         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10497         addressing with PIC.
10498         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10499         (nios2_symbolic_constant_p): Likewise.
10500         (nios2_legitimate_address_p): Likewise.
10501         (nios2_r0rel_section_name_p): New.
10502         (nios2_symbol_ref_in_r0rel_data_p): New.
10503         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10504         (r0rel_constant_p): New.
10505         (nios2_print_operand_address): Handle r0rel_constant_p.
10506         (nios2_cdx_narrow_form_p): Likewise.
10507         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10508         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10509         (Nios II Options): Document -mr0rel-sec.
10511 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10513         * config/nios2/nios2.c: Include xregex.h.
10514         (nios2_gprel_sec_regex): New.
10515         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10516         addressing with PIC.
10517         (nios2_small_section_name_p): Check for regex match.
10518         * config/nios2/nios2.opt (mgprel-sec=): New option.
10519         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10520         (Nios II Options): Document -mgprel-sec.
10522 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10524         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10526 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10528         PR tree-optimization/82707
10529         * gimple.c (gimple_copy): Fix unsharing of
10530         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10532 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10534         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10535         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10536         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10537         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10538         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10539         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10540         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10541         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10542         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10543         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10544         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10545         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10546         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10547         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10548         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10549         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10550         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10552 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10554         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10555         default to IBM.
10556         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10557         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10558         warning to rs6000.c.  Remove the Undocumented flag, since it has
10559         been documented.
10560         (-mabi=ibmlongdouble): Likewise.
10561         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10562         already set, set the default format for long double.
10563         (rs6000_debug_reg_global): Print whether long double is IBM or
10564         IEEE.
10565         (rs6000_option_override_internal): Rework setting long double
10566         format.  Only warn if the user is changing the long double default
10567         and they did not use -Wno-psabi.
10568         * doc/invoke.texi (PowerPC options): Update the documentation for
10569         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10571 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10572             Alan Hayward  <alan.hayward@arm.com>
10573             David Sherwood  <david.sherwood@arm.com>
10575         * rtl.h (wider_subreg_mode): New function.
10576         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10577         rather than an unsigned int *.
10578         * ira-color.c (regno_max_ref_width): Replace with...
10579         (regno_max_ref_mode): ...this new variable.
10580         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10581         Use wider_subreg_mode.
10582         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10583         rather than an unsigned int *.
10584         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10585         (process_alt_operands): Likewise.
10586         (invariant_p): Likewise.
10587         * lra-spills.c (assign_mem_slot): Likewise.
10588         (add_pseudo_to_slot): Likewise.
10589         * lra.c (collect_non_operand_hard_regs): Likewise.
10590         (add_regs_to_insn_regno_info): Likewise.
10591         * reload1.c (regno_max_ref_width): Replace with...
10592         (regno_max_ref_mode): ...this new variable.
10593         (reload): Update accordingly.  Update call to
10594         ira_sort_regnos_for_alter_reg.
10595         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10596         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10597         (scan_paradoxical_subregs): Likewise.
10599 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10601         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10602         (aarch64_frame): Add emit_frame_chain boolean.
10603         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10604         Move eh_return case to aarch64_layout_frame.
10605         (aarch64_layout_frame): Initialize emit_frame_chain.
10606         (aarch64_expand_prologue): Use emit_frame_chain.
10608 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10610         * config/aarch64/aarch64.c (aarch64_layout_frame):
10611         Ensure LR is always stored at the bottom of the callee-saves.
10612         Remove rarely used frame layout which saves callee-saves at top of
10613         frame, so the store of LR can be used as a valid probe in all cases.
10615 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10617         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10618         Improve unaligned TImode/TFmode base/offset split.
10620 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10621             Alan Hayward  <alan.hayward@arm.com>
10622             David Sherwood  <david.sherwood@arm.com>
10624         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10625         * combine.c (find_single_use_1): Likewise.
10626         (expand_field_assignment): Likewise.
10627         (move_deaths): Likewise.
10628         * lra-constraints.c (simplify_operand_subreg): Likewise.
10629         (curr_insn_transform): Likewise.
10630         * lra.c (collect_non_operand_hard_regs): Likewise.
10631         (add_regs_to_insn_regno_info): Likewise.
10632         * rtlanal.c (reg_referenced_p): Likewise.
10633         (covers_regno_no_parallel_p): Likewise.
10635 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10637         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10638         Don't print any bits outside the precision of the value.
10639         * wide-int.cc (test_printing): Add some new tests.
10641 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10643         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10644         supports -xbrace_comment option.
10645         * configure: Regenerate.
10646         * config.in: Regenerate.
10647         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10648         (ASM_CPU_SPEC): Use it.
10650 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10652         * target.def (static_rtx_alignment): New hook.
10653         * targhooks.h (default_static_rtx_alignment): Declare.
10654         * targhooks.c (default_static_rtx_alignment): New function.
10655         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10656         * doc/tm.texi: Regenerate.
10657         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10658         instead of targetm.constant_alignment.  Remove call to
10659         set_mem_attributes.
10660         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10661         (cris_preferred_mininum_alignment): New function, split out from...
10662         (cris_constant_alignment): ...here.
10663         (cris_static_rtx_alignment): New function.
10664         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10665         split out from...
10666         (ix86_constant_alignment): ...here.
10667         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10668         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10669         (mmix_static_rtx_alignment): New function.
10670         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10671         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10673 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10675         PR target/81800
10676         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10677         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10679 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10681         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10682         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10683         mark as a sign-extending load.
10684         (local_pic_load_u): Define.
10686 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10688         PR middle-end/82062
10689         * fold-const.c (operand_equal_for_comparison_p): Also return true
10690         if ARG0 is a simple variant of ARG1 with narrower precision.
10691         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10693 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10695         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10696         cost correctly.
10697         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10698         scatter_static, scatter_per_elt.
10699         * x86-tune-costs.h: Add new cost entries.
10701 2017-10-25  Richard Biener  <rguenther@suse.de>
10703         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10704         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10705         class pass_fre): Move to ...
10706         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10707         class pass_fre): ... here and adjust for statistics.
10709 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10711         PR libstdc++/81706
10712         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10713         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10714         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10715         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10716         declarations.
10718 2017-10-25  Richard Biener  <rguenther@suse.de>
10720         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10721         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10722         eliminate_push_avail, eliminate_insert): Move inside...
10723         (class eliminate_dom_walker): ... this class in preparation
10724         of move.
10725         (fini_eliminate): Remove by merging with ...
10726         (eliminate): ... this function.  Adjust for class changes.
10727         (pass_pre::execute): Remove fini_eliminate call.
10728         (pass_fre::execute): Likewise.
10730 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10732         PR target/82460
10733         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10734         (VPERMI2, VPERMI2I): New mode iterators.
10735         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10736         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10737         patterns.
10738         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10739         mode iterator.  Remove 3 old define_insn patterns.
10740         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10741         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10742         VPERMI2 mode iterator, remove the other two expanders.
10743         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10744         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10745         instructions, remove the other two patterns.
10746         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10747         mode iterator, remove the other two patterns.
10748         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10749         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10750         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10751         and adjust argument order accordingly.
10752         (ix86_expand_vec_perm): Adjust caller.
10753         (expand_vec_perm_1): Likewise.
10754         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10755         (expand_vec_perm_vpermt2_vpshub2): ... this.
10756         (ix86_expand_vec_perm_const_1): Adjust caller.
10757         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10759         PR target/82370
10760         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10761         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10762         (vec_shl_<mode>): Remove unused expander.
10763         (avx512bw_<shift_insn><mode>3): New define_insn.
10764         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10765         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10767 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10769         PR c++/82466
10770         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10771         description.
10773 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10775         PR rtl-optimization/82396
10776         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10777         (autopref_multipass_init): Simplify initialization.
10778         (autopref_rank_data): Simplify sort order.
10779         * gcc/sched-int.h (autopref_multipass_data_): Remove
10780         multi_mem_insn_p, min_offset and max_offset.
10782 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10784         PR middle-end/60580
10785         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10786         Check special value of flag_omit_frame_pointer.
10787         (aarch64_can_eliminate): Likewise.
10788         (aarch64_override_options_after_change_1): Simplify handling of
10789         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10791 2017-10-24  Richard Biener  <rguenther@suse.de>
10793         PR tree-optimization/82697
10794         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10795         zero for conditional load and unconditional store.
10797 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10799         * doc/install.texi: Document bootstrap-cet.
10801 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10803         PR target/82659
10804         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10805         ENDBR instruction at function entrance if function is only
10806         called directly.
10808 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10810         PR target/82628
10811         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10812         patterns to better describe from which operation the CF is computed.
10813         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10814         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10815         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10816         is 0, use _0 suffixed expanders instead of emitting a comparison
10817         before it.
10819 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10821         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10822         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10824 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10826         PR middle-end/82569
10827         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10828         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10829         * loop-iv.c (iv_get_reaching_def): Likewise.
10830         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10831         variable is promoted and the partition contains undefined values.
10833 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10835         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10836         reflect reality.
10837         (nios2_address_cost): Define.
10838         (nios2_legitimize_address): Recognize (exp + constant) directly.
10839         (TARGET_ADDRESS_COST): Define.
10841 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10843         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10844         (nios2_symbolic_memory_operand_p): Declare.
10845         (nios2_split_large_constant): Declare.
10846         (nios2_split_symbolic_memory_operand): Declare.
10847         * config/nios2/nios2.c: Adjust includes.
10848         (nios2_symbolic_constant_allowed): New.
10849         (nios2_symbolic_constant_p): New.
10850         (nios2_plus_symbolic_constant_p): New.
10851         (nios2_valid_addr_expr_p): Recognize addresses involving
10852         symbolic constants.
10853         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10854         (nios2_symbolic_memory_operand_p): New.
10855         (nios2_large_constant_p): New.
10856         (nios2_split_large_constant): New.
10857         (nios2_split_plus_large_constant): New.
10858         (nios2_split_symbolic_memory_operand): New.
10859         (nios2_legitimize_address): Code refactoring.  Handle addresses
10860         involving symbolic constants.
10861         (nios2_emit_move_sequence): Likewise.
10862         (nios2_print_operand): Improve error output.
10863         (nios2_print_operand_address): Handle LO_SUM.
10864         (nios2_cdx_narrow_form_p): Likewise.
10865         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
10866         operands involving symbolic constants.
10867         (movhi_internal, movsi_internal): Likewise.
10868         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
10869         (extendhisi2, extendqi<mode>2): Likewise.
10871 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10873         * tree-pass.h (PROP_rtl_split_insns): Define.
10874         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
10876 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10878         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
10880 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
10882         PR debug/82630
10883         * target.def (const_not_ok_for_debug_p): Default to
10884         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
10885         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
10886         * targhooks.c (default_const_not_ok_for_debug_p): New function.
10887         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
10888         which targetm.const_not_ok_for_debug_p returned true.
10889         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
10890         for UNSPECs.
10891         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
10892         Likewise.
10893         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
10894         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
10895         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
10896         if !base_term_p.
10897         (ix86_const_not_ok_for_debug_p): New function.
10898         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
10899         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
10901 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
10903         PR bootstrap/82610
10904         * system.h: Conditionally include "unique-ptr.h" if
10905         INCLUDE_UNIQUE_PTR is defined.
10906         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
10907         of defining INCLUDE_UNIQUE_PTR before including "system.h".
10909 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
10911         * config/rl78/rl78.md: New define_expand "subdi3".
10913 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
10915         PR target/82673
10916         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
10917         DF_REF_INSN if DF_REF_INSN_INFO is false.
10919 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
10921         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
10922         xmm_move instead of sse_move.
10923         (sse_store_index): New function.
10924         (ix86_register_move_cost): Be more sensible about mismatch stall;
10925         model AVX moves correctly; make difference between sse->integer and
10926         integer->sse.
10927         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
10928         moves; make difference between SSE and AVX.
10929         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
10930         and zmm_move. Increase size of sse load and store tables;
10931         add unaligned load and store tables; add ssemmx_to_integer.
10932         * x86-tune-costs.h: Update all entries according to real
10933         move latencies from Agner Fog's manual and chip documentation.
10935 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
10937         PR target/82628
10938         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
10939         * config/i386/constraints.md (Wf): New constraint.
10940         * config/i386/i386.md (UNSPEC_SBB): New unspec.
10941         (cmp<dwi>_doubleword): Removed.
10942         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
10943         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
10944         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
10945         expand with cmp<dwi>_doubleword.  For LTU and GEU use
10946         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
10948         * common.opt (gcolumn-info): Enable by default.
10949         * doc/invoke.texi (gcolumn-info): Document new default.
10951 2017-10-23  Richard Biener  <rguenther@suse.de>
10953         PR tree-optimization/82672
10954         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
10955         Fold the stmt if we propagated into it.
10957 2017-10-23  Richard Biener  <rguenther@suse.de>
10959         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
10960         (bitmap_remove_expr_from_set): ... this.  All callers call this
10961         for non-constant values.
10962         (bitmap_set_subtract): Rename to...
10963         (bitmap_set_subtract_expressions): ... this.  Adjust and
10964         optimize.
10965         (bitmap_set_contains_value): Remove superfluous check.
10966         (bitmap_set_replace_value): Inline into single caller ...
10967         (bitmap_value_replace_in_set): ... here and simplify.
10968         (dependent_clean): Merge into ...
10969         (clean): ... this using an overload.  Adjust.
10970         (prune_clobbered_mems): Adjust.
10971         (compute_antic_aux): Likewise.
10972         (compute_partial_antic_aux): Likewise.
10974 2017-10-23  Richard Biener  <rguenther@suse.de>
10976         PR tree-optimization/82129
10977         Revert
10978         2017-08-01  Richard Biener  <rguenther@suse.de>
10980         PR tree-optimization/81181
10981         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10982         (compute_antic): ... end of iteration here.
10984 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
10986         * target.def (starting_frame_offset): New hook.
10987         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
10988         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
10989         * doc/tm.texi.in: Regenerate.
10990         * hooks.h (hook_hwi_void_0): Declare.
10991         * hooks.c (hook_hwi_void_0): New function.
10992         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
10993         STARTING_FRAME_OFFSET.
10994         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
10995         * reload1.c (reload): Likewise.
10996         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
10997         instead of STARTING_FRAME_OFFSET.
10998         * function.c (try_fit_stack_local): Likewise.
10999         (assign_stack_local_1): Likewise
11000         (instantiate_virtual_regs): Likewise.
11001         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11002         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11003         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11004         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11005         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11006         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11007         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11008         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11009         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11010         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11011         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11012         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11013         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11014         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11015         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11016         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11017         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11018         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11019         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11020         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11021         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11022         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11023         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11024         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11025         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11026         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11027         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11028         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11029         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11030         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11031         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11032         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11033         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11034         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11035         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11036         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11037         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11038         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11039         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11040         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11041         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11042         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11043         return a HOST_WIDE_INT.
11044         (avr_builtin_setjmp_frame_value): Use it instead of
11045         STARTING_FRAME_OFFSET.
11046         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11047         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11048         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11049         New function.
11050         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11051         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11052         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11053         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11054         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11055         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11056         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11057         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11058         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11059         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11060         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11061         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11062         New function.
11063         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11064         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11065         * config/mips/mips.c (mips_compute_frame_info): Refer to
11066         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11067         (mips_starting_frame_offset): New function.
11068         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11069         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11070         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11071         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11072         and return a HOST_WIDE_INT.
11073         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11074         (mmix_initial_elimination_offset): Refer to
11075         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11076         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11077         * config/pa/pa.c (pa_starting_frame_offset): New function.
11078         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11079         (pa_expand_prologue): Likewise.
11080         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11081         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11082         !FRAME_GROWS_DOWNWARD handling to...
11083         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11084         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11085         !FRAME_GROWS_DOWNWARD handling to...
11086         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11087         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11088         !FRAME_GROWS_DOWNWARD handling to...
11089         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11090         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11091         Redefine.
11092         (rs6000_starting_frame_offset): New function.
11093         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11094         !FRAME_GROWS_DOWNWARD handling to...
11095         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11096         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11097         !FRAME_GROWS_DOWNWARD handling to...
11098         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11099         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11100         !FRAME_GROWS_DOWNWARD handling to...
11101         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11102         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11103         (rs6000_starting_frame_offset): New function.
11104         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11105         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11106         * config/vax/vax.c (vax_starting_frame_offset): New function.
11107         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11108         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11109         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11110         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11111         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11112         * system.h (STARTING_FRAME_OFFSET): Poison.
11114 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11116         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11117         SCALAR_TYPE_MODE instead of TYPE_MODE.
11119 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11120             Alan Hayward  <alan.hayward@arm.com>
11121             David Sherwood  <david.sherwood@arm.com>
11123         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11125 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11126             Alan Hayward  <alan.hayward@arm.com>
11127             David Sherwood  <david.sherwood@arm.com>
11129         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11131 2017-10-23  Richard Biener  <rguenther@suse.de>
11133         PR tree-optimization/82129
11134         * tree-ssa-pre.c (bitmap_set_and): Remove.
11135         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11136         canonicalizing expressions in the set to those with lowest
11137         ID rather than taking that from the first edge.
11139 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11141         * combine.c (rtx_equal_for_field_assignment_p): Use
11142         byte_lowpart_offset.
11144 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11145             Alan Hayward  <alan.hayward@arm.com>
11146             David Sherwood  <david.sherwood@arm.com>
11148         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11149         to a SUBREG_PROMOTED_VAR.
11151 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11152             Alan Hayward  <alan.hayward@arm.com>
11153             David Sherwood  <david.sherwood@arm.com>
11155         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11156         (expand_debug_source_expr): Likewise.
11157         * combine.c (combine_simplify_rtx): Likewise.
11158         * cse.c (fold_rtx): Likewise.
11159         * optabs.c (expand_float): Likewise.
11160         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11161         (simplify_binary_operation_1): Likewise.
11163 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11164             Alan Hayward  <alan.hayward@arm.com>
11165             David Sherwood  <david.sherwood@arm.com>
11167         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11168         (record_promoted_value): Likewise.
11169         * expr.c (expand_expr_real_2): Likewise.
11170         * ree.c (update_reg_equal_equiv_notes): Likewise.
11171         (combine_set_extension): Likewise.
11172         * rtlanal.c (low_bitmask_len): Likewise.
11173         * simplify-rtx.c (neg_const_int): Likewise.
11174         (simplify_binary_operation_1): Likewise.
11176 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11177             Alan Hayward  <alan.hayward@arm.com>
11178             David Sherwood  <david.sherwood@arm.com>
11180         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11181         * regcprop.c (maybe_mode_change): Likewise.
11182         * reload1.c (alter_reg): Likewise.
11184 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11186         * inchash.h (inchash::hash::add_wide_int): New function.
11187         * lto-streamer-out.c (hash_tree): Use it.
11189 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11191         * inchash.h (inchash::hash::add_wide_int): Rename to...
11192         (inchash::hash::add_hwi): ...this.
11193         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11194         (polymorphic_call_target_hasher::hash): Likewise.
11195         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11196         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11197         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11198         * lto-streamer-out.c (hash_tree): Likewise.
11199         * optc-save-gen.awk: Likewise.
11200         * tree.c (add_expr): Likewise.
11202 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11204         PR target/52451
11205         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11206         for ordered inequality comparisons even with TARGET_IEEE_FP.
11208 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11210         PR target/82628
11211         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11212         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11213         Expand with cmp<dwi>_doubleword.
11215 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11217         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11218         List CET intrinsics.
11219         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11220         specific to -fcf-protection option.
11222 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11224         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11225         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11226         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11227         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11228         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11229         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11230         (extra_objs): Add cet.o for Linux/x86 targets.
11231         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11232         * config/i386/cet.c: New file.
11233         * config/i386/cetintrin.h: Likewise.
11234         * config/i386/t-cet: Likewise.
11235         * config/i386/cpuid.h (bit_SHSTK): New.
11236         (bit_IBT): Likewise.
11237         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11238         pass IBT and SHSTK bits.
11239         * config/i386/i386-builtin-types.def
11240         (VOID_FTYPE_UNSIGNED_PVOID): New.
11241         (VOID_FTYPE_UINT64_PVOID): Likewise.
11242         * config/i386/i386-builtin.def: Add CET intrinsics.
11243         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11244         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11245         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11246         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11247         prototype.
11248         * config/i386/i386.c (rest_of_insert_endbranch): New.
11249         (pass_data_insert_endbranch): Likewise.
11250         (pass_insert_endbranch): Likewise.
11251         (make_pass_insert_endbranch): Likewise.
11252         (ix86_notrack_prefixed_insn_p): Likewise.
11253         (ix86_target_string): Add -mibt, -mshstk flags.
11254         (ix86_option_override_internal): Add flag_cf_protection
11255         processing.
11256         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11257         (ix86_print_operand): Add 'notrack' prefix output.
11258         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11259         (ix86_expand_builtin): Expand CET intrinsics.
11260         (x86_output_mi_thunk): Add 'endbranch' instruction.
11261         * config/i386/i386.h (TARGET_IBT): New.
11262         (TARGET_IBT_P): Likewise.
11263         (TARGET_SHSTK): Likewise.
11264         (TARGET_SHSTK_P): Likewise.
11265         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11266         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11267         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11268         (builtin_setjmp_setup): New pattern.
11269         (builtin_longjmp): Likewise.
11270         (rdssp<mode>): Likewise.
11271         (incssp<mode>): Likewise.
11272         (saveprevssp): Likewise.
11273         (rstorssp): Likewise.
11274         (wrss<mode>): Likewise.
11275         (wruss<mode>): Likewise.
11276         (setssbsy): Likewise.
11277         (clrssbsy): Likewise.
11278         (nop_endbr): Likewise.
11279         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11280         options.
11281         * config/i386/immintrin.h: Include <cetintrin.h>.
11282         * config/i386/linux-common.h
11283         (file_end_indicate_exec_stack_and_cet): New prototype.
11284         (TARGET_ASM_FILE_END): New.
11286 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11288         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11289         latencies instead of having separate table; make difference between
11290         integer and float costs.
11291         * i386.h (processor_costs): Remove scalar_stmt_cost,
11292         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11293         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11294         vec_store_cost.
11295         * x86-tune-costs.h: Remove entries which has been removed in
11296         procesor_costs from all tables; make cond_taken_branch_cost
11297         and cond_not_taken_branch_cost COST_N_INSNS based.
11299 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11301         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11303 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11305         * config/i386/i386.md (isa): Remove fma_avx512f.
11306         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11307         <avx512>_fmadd_<mode>_mask3<round_name>,
11308         <avx512>_fmsub_<mode>_mask<round_name>,
11309         <avx512>_fmsub_<mode>_mask3<round_name>,
11310         <avx512>_fnmadd_<mode>_mask<round_name>,
11311         <avx512>_fnmadd_<mode>_mask3<round_name>,
11312         <avx512>_fnmsub_<mode>_mask<round_name>,
11313         <avx512>_fnmsub_<mode>_mask3<round_name>,
11314         <avx512>_fmaddsub_<mode>_mask<round_name>,
11315         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11316         <avx512>_fmsubadd_<mode>_mask<round_name>,
11317         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11318         (*vec_widen_umult_even_v16si<mask_name>,
11319         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11320         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11322 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11324         * extend.texi: Add 'nocf_check' documentation.
11325         * gimple.texi: Add second parameter to
11326         gimple_build_call_from_tree.
11327         * invoke.texi: Add -fcf-protection documentation.
11328         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11330 2017-10-20  Richard Biener  <rguenther@suse.de>
11332         PR tree-optimization/82473
11333         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11334         the largest input type.
11336 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11338         * c-attribs.c (handle_nocf_check_attribute): New function.
11339         (c_common_attribute_table): Add 'nocf_check' handling.
11340         * gimple-parser.c: Add second argument NULL to
11341         gimple_build_call_from_tree.
11342         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11343         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11344         call insn.
11345         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11346         * common.opt: Add fcf-protection flag.
11347         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11348         * flag-types.h: Add enum cf_protection_level.
11349         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11350         Add 'nocf_check' attribute propagation to gimple call.
11351         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11352         (gimple_build_call_from_tree): Update prototype.
11353         (gimple_call_nocf_check_p): New function.
11354         (gimple_call_set_nocf_check): Likewise.
11355         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11356         * ipa-icf.c: Add nocf_check attribute in statement hash.
11357         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11358         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11359         * toplev.c (process_options): Add flag_cf_protection handling.
11361 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11363         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11365 2017-10-20  Richard Biener  <rguenther@suse.de>
11367         PR tree-optimization/82603
11368         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11369         remove false predicated stores.
11371 2017-10-20  Richard Biener  <rguenther@suse.de>
11373         * graphite-isl-ast-to-gimple.c
11374         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11375         Remove return value and simplify, dump copied stmt after lhs
11376         adjustment.
11377         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11378         Reduce dump verbosity.
11379         (gsi_insert_earliest): Likewise.
11380         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11381         * graphite.c (print_global_statistics): Adjust dumping.
11382         (print_graphite_scop_statistics): Likewise.
11383         (print_graphite_statistics): Do not dump loops here.
11384         (graphite_transform_loops): But here.
11386 2017-10-20  Nicolas Roche  <roche@adacore.com>
11388         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11389         * configure: Regenerate.
11391 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11393         PR target/82158
11394         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11395         functions when optimizing replace GIMPLE_RETURN stmts with
11396         calls to __builtin_unreachable ().
11398         PR sanitizer/82595
11399         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11400         for -fsanitize=thread link of executables.
11401         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11402         link of executables.
11404         PR target/82370
11405         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11406         New mode iterators.
11407         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11408         define_insns for logical vector shifts to use VI248_AVX512BW
11409         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11410         condition, useless isa and prefix attributes.  Change the first
11411         2 of these define_insns to ...
11412         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11413         define_insn for avx512vl.
11414         (<shift_insn><mode>3): ... and this, new define_insn without
11415         masking for non-avx512vl.
11417         PR target/82370
11418         * config/i386/sse.md (*andnot<mode>3,
11419         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11420         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11421         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11422         not applied use empty suffix even for TARGET_AVX512VL.
11423         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11424         is applied, supply evex,evex or evex,evex,evex instead of just
11425         evex.
11427 2017-10-20  Julia Koval  <julia.koval@intel.com>
11429         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11430         (OPTION_MASK_ISA_GFNI_UNSET): New.
11431         (ix86_handle_option): Handle OPT_mgfni.
11432         * config/i386/cpuid.h (bit_GFNI): New.
11433         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11434         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11435         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11436         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11437         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11438         * config/i386/i386.opt: Add mgfni.
11440 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11442         * config/msp430/msp430.c (msp430_option_override): Disable
11443         -fdelete-null-pointer-checks.
11444         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11446 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11448         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11449         of x87 and SSE instructions.
11451 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11453         * asan.c (create_cond_insert_point): Do not update edge count.
11454         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11455         (afdo_propagate_circuit): Likewise.
11456         (afdo_calculate_branch_prob): Likewise.
11457         (afdo_annotate_cfg): Likewise.
11458         * basic-block.h (struct edge_def): Remove count.
11459         (edge_def::count): New accessor.
11460         * bb-reorder.c (rotate_loop): Update.
11461         (find_traces_1_round): Update.
11462         (connect_traces): Update.
11463         (sanitize_hot_paths): Update.
11464         * cfg.c (unchecked_make_edge): Update.
11465         (make_single_succ_edge): Update.
11466         (check_bb_profile): Update.
11467         (dump_edge_info): Update.
11468         (update_bb_profile_for_threading): Update.
11469         (scale_bbs_frequencies_int): Update.
11470         (scale_bbs_frequencies_gcov_type): Update.
11471         (scale_bbs_frequencies_profile_count): Update.
11472         (scale_bbs_frequencies): Update.
11473         * cfganal.c (connect_infinite_loops_to_exit): Update.
11474         * cfgbuild.c (compute_outgoing_frequencies): Update.
11475         (find_many_sub_basic_blocks): Update.
11476         * cfgcleanup.c (try_forward_edges): Update.
11477         (try_crossjump_to_edge): Update
11478         * cfgexpand.c (expand_gimple_cond): Update
11479         (expand_gimple_tailcall): Update
11480         (construct_exit_block): Update
11481         * cfghooks.c (verify_flow_info): Update
11482         (redirect_edge_succ_nodup): Update
11483         (split_edge): Update
11484         (make_forwarder_block): Update
11485         (duplicate_block): Update
11486         (account_profile_record): Update
11487         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11488         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11489         * cfgloopmanip.c (scale_loop_profile): Update.
11490         (loopify): Update.
11491         (lv_adjust_loop_entry_edge): Update.
11492         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11493         (force_nonfallthru_and_redirect): Update.
11494         (purge_dead_edges): Update.
11495         (rtl_flow_call_edges_add): Update.
11496         * cgraphunit.c (init_lowered_empty_function): Update.
11497         (cgraph_node::expand_thunk): Update.
11498         * gimple-pretty-print.c (dump_probability): Update.
11499         (dump_edge_probability): Update.
11500         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11501         * haifa-sched.c (sched_create_recovery_edges): Update.
11502         * hsa-gen.c (convert_switch_statements): Update.
11503         * ifcvt.c (dead_or_predicable): Update.
11504         * ipa-inline-transform.c (inline_transform): Update.
11505         * ipa-split.c (split_function): Update.
11506         * ipa-utils.c (ipa_merge_profiles): Update.
11507         * loop-doloop.c (add_test): Update.
11508         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11509         * lto-streamer-in.c (input_cfg): Update.
11510         (input_function): Update.
11511         * lto-streamer-out.c (output_cfg): Update.
11512         * modulo-sched.c (sms_schedule): Update.
11513         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11514         * predict.c (maybe_hot_edge_p): Update.
11515         (unlikely_executed_edge_p): Update.
11516         (probably_never_executed_edge_p): Update.
11517         (dump_prediction): Update.
11518         (drop_profile): Update.
11519         (propagate_unlikely_bbs_forward): Update.
11520         (determine_unlikely_bbs): Update.
11521         (force_edge_cold): Update.
11522         * profile.c (compute_branch_probabilities): Update.
11523         * reg-stack.c (better_edge): Update.
11524         * shrink-wrap.c (handle_simple_exit): Update.
11525         * tracer.c (better_p): Update.
11526         * trans-mem.c (expand_transaction): Update.
11527         (split_bb_make_tm_edge): Update.
11528         * tree-call-cdce.c: Update.
11529         * tree-cfg.c (gimple_find_sub_bbs): Update.
11530         (gimple_split_edge): Update.
11531         (gimple_duplicate_sese_region): Update.
11532         (gimple_duplicate_sese_tail): Update.
11533         (gimple_flow_call_edges_add): Update.
11534         (insert_cond_bb): Update.
11535         (execute_fixup_cfg): Update.
11536         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11537         * tree-complex.c (expand_complex_div_wide): Update.
11538         * tree-eh.c (lower_resx): Update.
11539         (unsplit_eh): Update.
11540         (cleanup_empty_eh_move_lp): Update.
11541         * tree-inline.c (copy_edges_for_bb): Update.
11542         (freqs_to_counts): Update.
11543         (copy_cfg_body): Update.
11544         * tree-ssa-dce.c (remove_dead_stmt): Update.
11545         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11546         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11547         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11548         (unloop_loops): Update.
11549         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11550         * tree-ssa-loop-split.c (connect_loops): Update.
11551         (split_loop): Update.
11552         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11553         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11554         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11555         * tree-ssa-reassoc.c (branch_fixup): Update.
11556         * tree-ssa-tail-merge.c (replace_block_by): Update.
11557         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11558         (compute_path_counts): Update.
11559         (update_profile): Update.
11560         (recompute_probabilities): Update.
11561         (update_joiner_offpath_counts): Update.
11562         (estimated_freqs_path): Update.
11563         (freqs_to_counts_path): Update.
11564         (clear_counts_path): Update.
11565         (ssa_fix_duplicate_block_edges): Update.
11566         (duplicate_thread_path): Update.
11567         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11568         (case_bit_test_cmp): Update.
11569         (collect_switch_conv_info): Update.
11570         (gen_inbound_check): Update.
11571         (do_jump_if_equal): Update.
11572         (emit_cmp_and_jump_insns): Update.
11573         * tree-tailcall.c (decrease_profile): Update.
11574         (eliminate_tail_call): Update.
11575         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11576         (vect_do_peeling): Update.
11577         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11578         * ubsan.c (ubsan_expand_null_ifn): Update.
11579         (ubsan_expand_ptr_ifn): Update.
11580         * value-prof.c (gimple_divmod_fixed_value): Update.
11581         (gimple_mod_pow2): Update.
11582         (gimple_mod_subtract): Update.
11583         (gimple_ic): Update.
11584         (gimple_stringop_fixed_value): Update.
11586 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11588         PR target/82618
11589         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11591 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11593         PR rtl-optimization/82395
11594         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11595         based on non_spilled_static_chain_regno_p.
11597 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11599         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11600         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11601         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11603 2017-10-19  Martin Sebor  <msebor@redhat.com>
11605         PR tree-optimization/82596
11606         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11608 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11610         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11611         (get_mem_refs_of_builtin_call): Likewise.
11612         * builtins.c (expand_builtin_apply): Adjust call to
11613         allocate_dynamic_stack_space.
11614         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11615         the third argument to allocate_dynamic_stack_space, otherwise -1.
11616         (expand_builtin): Deal with all alloca variants.
11617         (is_inexpensive_builtin): Likewise.
11618         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11619         * calls.c (special_function_p): Deal with all alloca variants.
11620         (initialize_argument_information): Adjust call to
11621         allocate_dynamic_stack_space.
11622         (expand_call): Likewise.
11623         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11624         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11625         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11626         use it for the stack usage computation.
11627         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11628         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11629         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11630         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11631         (in_loop_p): Remove first argument and useless check.
11632         (pass_walloca::execute): Remove useless test and adjust call to above.
11633         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11634         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11635         (gimplify_call_expr): Deal with all alloca variants.
11636         * hsa-gen.c (gen_hsa_alloca): Likewise.
11637         (gen_hsa_insns_for_call): Likewise.
11638         * ipa-pure-const.c (special_builtin_state): Likewise.
11639         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11640         * tree-object-size.c (alloc_object_size): Likewise.
11641         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11642         (call_may_clobber_ref_p_1): Likewise.
11643         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11644         (ccp_fold_stmt): Likewise.
11645         (optimize_stack_restore): Likewise.
11646         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11647         (mark_all_reaching_defs_necessary_1): Likewise.
11648         (propagate_necessity): Likewise.
11649         (eliminate_unnecessary_stmts): Likewise.
11650         * tree.c (build_common_builtin_nodes): Build
11651         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11652         (build_alloca_call_expr): New function.
11653         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11654         (CASE_BUILT_IN_ALLOCA): Likewise.
11655         (build_alloca_call_expr): Declare.
11656         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11658 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11660         PR debug/82509
11661         * dwarf2out.c (new_die_raw): New static inline function.
11662         (new_die): Use it to create the DIE.
11663         (add_AT_external_die_ref): Likewise.
11664         (clone_die): Likewise.
11665         (clone_as_declaration): Likewise.
11666         (dwarf2out_vms_debug_main_pointer): Likewise.
11667         (base_type_die): Likewise.  Remove early return for corner cases.
11668         Do not call add_pubtype on the DIE here.
11669         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11670         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11671         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11672         native order exists for base types, attach the DIE manually and call
11673         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11675 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11677         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11678         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11679         the mem into it.
11680         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11682 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11684         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11685         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11686         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11687         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11688         * opts.c (sanitizer_opts): Add builtin.
11689         * ubsan.c (instrument_builtin): New function.
11690         (pass_ubsan::execute): Call it.
11691         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11692         * doc/invoke.texi: Document -fsanitize=builtin.
11694         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11695         builtins, store max (log2 (align), 0) into uchar field instead of
11696         align into uptr field.
11697         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11698         store uchar 0 field instead of uptr 0 field.
11699         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11700         instead of passing one address of struct with 2 locations pass
11701         two addresses of structs with 1 location each.
11702         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11703         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11704         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11705         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11706         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11707         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11708         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11709         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11711 2017-10-19  Martin Liska  <mliska@suse.cz>
11713         PR driver/81829
11714         * file-find.c (remove_prefix): Remove.
11715         * file-find.h (remove_prefix): Likewise.
11716         * gcc-ar.c: Remove smartness of lookup.
11718 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11720         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11721         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11722         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11724 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11726         PR target/82580
11727         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11728         (setcc + and to xor + setcc): New peephole2.
11730 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11732         * doc/sourcebuild.texi (Test Directives, Variants of
11733         dg-require-support): Add dg-require-stack-size.
11735 2017-10-19  Martin Liska  <mliska@suse.cz>
11737         PR sanitizer/82517
11738         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11739         that have a large alignment.
11740         (gimplify_target_expr): Likewise.
11742 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11744         PR rtl-optimization/82602
11745         * ira.c (rtx_moveable_p): Return false for volatile asm.
11747 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11749         PR target/82580
11750         * config/i386/i386-modes.def (CCGZ): New CC mode.
11751         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11752         * config/i386/predicates.md (ix86_comparison_operator):
11753         Handle CCGZmode.
11754         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11755         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11756         with double-word subtraction.
11757         (put_condition_code): Handle CCGZmode.
11759 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11761         * wide-int.cc (debug (const wide_int &)): New.
11762         (debug (const wide_int *)): New.
11763         (debug (const widest_int &)): New.
11764         (debug (const widest_int *)): New.
11766 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11768         PR middle-end/82556
11769         * lra-constraints.c (curr_insn_transform): Use non-input operand
11770         instead of output one for matched reload.
11772 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11774         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11775         (tree-ssa-loop-ivopts.h): New header file.
11776         (struct builtin_info): New fields.
11777         (classify_builtin_1): Compute and record base and offset parts for
11778         memset builtin partition by calling strip_offset.
11779         (offset_cmp, fuse_memset_builtins): New functions.
11780         (finalize_partitions): Fuse adjacent memset partitions by calling
11781         above function.
11782         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11783         Expose the interface.
11784         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11786 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11788         PR tree-optimization/82574
11789         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11790         that data reference must be executed exactly once per iteration
11791         against the outermost loop in nest.
11792         (classify_partition): Update call to above function.
11794 2017-10-18  Richard Biener  <rguenther@suse.de>
11796         PR tree-optimization/82591
11797         * graphite.c (graphite_transform_loops): Move code gen message
11798         printing ...
11799         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11800         Here.  Handle scop_to_isl_ast failing.
11801         (scop_to_isl_ast): Limit the number of ISL operations.
11803 2017-10-18  Richard Biener  <rguenther@suse.de>
11805         * graphite-isl-ast-to-gimple.c
11806         (translate_isl_ast_to_gimple::set_rename): Simplify.
11807         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11808         (graphite_copy_stmts_from_block): ... here.
11809         (copy_bb_and_scalar_dependences): Simplify.
11810         (add_parameters_to_ivs_params): Canonicalize.
11811         (generate_entry_out_of_ssa_copies): Simplify.
11812         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11813         by passing in ISL dimension.
11814         (parameter_index_in_region_1): Rename to ...
11815         (parameter_index_in_region): ... this.
11816         (extract_affine): Adjust assert, pass down parameter index.
11817         (add_param_constraints): Use range-info when available.
11818         (build_scop_context): Adjust.
11819         * sese.c (new_sese_info): Adjust.
11820         (free_sese_info): Likewise.
11821         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11822         Remove unused typedefs.
11823         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11825 2017-10-18  Martin Liska  <mliska@suse.cz>
11827         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11829 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11831         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11832         (s390_sched_init): Do not reset s390_sched_state if we entered the
11833         current basic block via a fallthru edge and all others are unlikely.
11835 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11837         * config/s390/s390.c (NUM_SIDES): New variable.
11838         (LONGRUNNING_THRESHOLD): New variable.
11839         (LATENCY_FACTOR): New variable.
11840         (s390_sched_score): Decrease score for long-running instructions on
11841         wrong side.
11842         (s390_sched_variable_issue): Perform bookkeeping for long-running
11843         instructions.
11845 2017-10-18  Richard Biener  <rguenther@suse.de>
11847         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11848         Simplify with removal of the parameter rename map.
11849         (set_rename): Likewise.
11850         (should_copy_to_new_region): Likewise.
11851         (graphite_copy_stmts_from_block): Likewise.
11852         (copy_bb_and_scalar_dependences): Remove initialization of
11853         unused copied_bb_map.
11854         (copy_def): Remove.
11855         (copy_internal_parameters): Likewise.
11856         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11857         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11858         Use INTEGRAL_TYPE_P.
11859         (parameter_index_in_region_1): Rename to ...
11860         (assign_parameter_index_in_region): ... this.  Assert we have
11861         a parameter we handle.
11862         (scan_tree_for_params): Adjust.
11863         * sese.h (parameter_rename_map_t): Remove.
11864         (struct sese_info_t): Remove unused parameter_rename_map and
11865         copied_bb_map members.
11866         * sese.c (new_sese_info): Adjust.
11867         (free_sese_info): Likewise.
11869 2017-10-18  Martin Liska  <mliska@suse.cz>
11871         PR sanitizer/82545
11872         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
11873         on an abnormal edge.
11875 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11877         * doc/invoke.texi (ffunction-sections and fdata-sections):
11878         Update.
11880 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11882         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
11883         the use statement can throw internally.
11885 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11887         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
11888         any RTX present on the RHS of a SET.
11889         * compare-elim.c (try_eliminate_compare): Restore comment.
11891 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
11893         * langhooks.h (struct lang_hooks): Document that tree_size langhook
11894         may be also called on tcc_type nodes.
11895         * langhooks.c (lhd_tree_size): Likewise.
11897 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
11899         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
11900         format_warning_at_substring.
11901         (maybe_warn): Convert source_range * param to a location_t.  Pass
11902         UNKNOWN_LOCATION rather than NULL to fmtwarn.
11903         (format_directive): Remove code to extract source_ranges and
11904         source_range * in favor of just a location_t.
11905         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
11906         fmtwarn.
11907         * substring-locations.c (format_warning_va): Convert
11908         source_range * param to a location_t.
11909         (format_warning_at_substring): Likewise.
11910         * substring-locations.h (format_warning_va): Likewise.
11911         (format_warning_at_substring): Likewise.
11913 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
11915         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
11916         vec_scatter_store
11917         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
11918         and scatter/gather ops.
11920         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
11921         vec_gather_load and vec_scatter_store.
11922         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
11923         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
11924         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
11925         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
11926         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
11927         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
11929 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
11931         * reg-stack.c (compare_for_stack_reg): Add bool argument.
11932         Detect FTST instruction and handle its register pops.  Only pop
11933         second operand if can_pop_second_op is true.
11934         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
11935         set can_pop_second_op to false in the compare_for_stack_reg call.
11937         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
11938         output_fp_compare for stack register operands.
11939         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
11940         instructions here.  Do not emit stack register pops here.  Assert
11941         that FCOMPP pops next to top stack register.  Rewrite function.
11943 2017-10-17  Nathan Sidwell  <nathan@acm.org>
11945         PR middle-end/82577
11946         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
11947         use DECL_ASSEMBLER_NAME_RAW.
11949         PR middle-end/82546
11950         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
11951         TYPE nodes.
11953 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
11954             Wilco Dijkstra <wilco.dijkstra@arm.com>
11956         * builtins.c (expand_builtin_update_setjmp_buf): Add a
11957         converstion to Pmode from the buf_addr.
11959 2017-10-17  Richard Biener  <rguenther@suse.de>
11961         * graphite-dependences.c (scop_get_reads_and_writes): Change
11962         output parameters to references.
11964 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
11966         PR 71026/tree-optimization
11967         * fold-const.c (distribute_real_division): Removed.
11968         (fold_binary_loc): Remove calls to distribute_real_divison.
11970 2017-10-17  Richard Biener  <rguenther@suse.de>
11972         * graphite-scop-detection.c
11973         (scop_detection::stmt_has_simple_data_refs_p): Always use
11974         the full nest as region.
11975         (try_generate_gimple_bb): Likewise.
11976         * sese.c (scalar_evolution_in_region): Simplify now that
11977         SCEV can handle instantiation in regions.
11978         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
11979         in the non-loop part of a function if requested.
11981 2017-10-17  Richard Biener  <rguenther@suse.de>
11983         PR tree-optimization/82563
11984         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
11985         New function.
11986         (graphite_regenerate_ast_isl): Call it.
11987         * graphite-scop-detection.c (build_scops): Remove entry edge split.
11989 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
11991         PR tree-optimization/82549
11992         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
11993         Formatting fixes.  Instead of calling make_bit_field_ref with negative
11994         bitpos return 0.
11996 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
11998         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
11999         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12000         _mm_maskz_reduce_ss): New.
12001         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12002         __builtin_ia32_reducess_mask): Ditto..
12003         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12004         * config/i386/sse.md (reduces<mode>): Renamed to ...
12005         (reduces<mode><mask_scalar_name>): ... this.
12006         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12007         Changed to ...
12008         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12009         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12011 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12013         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12014         * selftest-run-tests.c (selftest::run_tests): Call
12015         selftest::unique_ptr_tests_cc_tests.
12016         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12017         * unique-ptr-tests.cc: New file.
12019 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12021         PR sanitizer/82353
12022         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12023         locations.
12024         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12025         (make_hard_regno_born, make_hard_regno_dead): Update
12026         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12028 2017-10-16  Jeff Law  <law@redhat.com>
12030         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12032 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12034         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12036 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12038         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12039         with_cpu if we were configured for an e500v2 target cpu name.
12041 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12043         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12044         * doc/invoke.texi: Document +nodsp as a valid extension for
12045         -mcpu=cortex-m33.
12047 2017-10-16  Martin Liska  <mliska@suse.cz>
12049         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12050         (test_set_range): Likewise.
12051         (test_range_functions): Rename to ...
12052         (test_bit_in_range): ... this.
12053         (sbitmap_c_tests): Add new test.
12055 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12057         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12058         New.
12059         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12060         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12062 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12064         * config/aarch64/aarch64-builtins.c
12065         (aarch64_types_quadopu_lane_qualifiers): New.
12066         (TYPES_QUADOPU_LANE): New.
12067         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12068         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12069         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12070         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12071         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12072         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12073         (Vdottype, DOTPROD): New.
12074         (sur): Add SDOT and UDOT.
12076 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12078         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12079         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12080         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12081         Add TARGET_DOTPROD.
12082         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12083         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12084         Enable TARGET_DOTPROD.
12085         (cortex-a75.cortex-a55): Likewise.
12086         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12088 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12090         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12091         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12092         New.
12093         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12094         New.
12095         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12096         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12097         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12098         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12099         * config/arm/types.md (neon_dot, neon_dot_q): New.
12100         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12102 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12104         * config/arm/arm.h (TARGET_DOTPROD): New.
12105         * config/arm/arm.c (arm_arch_dotprod): New.
12106         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12107         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12108         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12109         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12110         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12111         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12112         * doc/invoke.texi (armv8.2-a): Document dotprod
12114 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12116         * i386.c (ix86_vec_cost): New function.
12117         (ix86_rtx_costs): Handle vector operations better.
12118         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12119         * x86-tune-costs.h: Add new costs to all tables.
12121 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12123         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12124         operations.
12125         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12126         divsd, sqrtss and sqrtsd
12127         * x86-tune-costs.h: Add new entries to all costs.
12128         (znver1_cost): Fix to match real instruction latencies.
12130 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12131             Michael Collison <michael.collison@arm.com>
12133         * compare-elim.c: Include emit-rtl.h.
12134         (can_merge_compare_into_arith): New function.
12135         (try_validate_parallel): Likewise.
12136         (try_merge_compare): Likewise.
12137         (try_eliminate_compare): Call the above when no previous clobber
12138         is available.
12139         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12140         dataflow problems.
12142 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12144         PR middle-end/62263
12145         PR middle-end/82498
12146         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12147         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12149         PR middle-end/62263
12150         PR middle-end/82498
12151         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12152         to be any operand_equal_p operands.  For & (B - 1) require
12153         B to be power of 2.  Recognize
12154         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12156 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12158         PR bootstrap/82553
12159         * optabs.c (expand_memory_blockage): Fix call of
12160         targetm.have_memory_blockage.
12162 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12164         PR bootstrap/82548
12165         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12166         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12167         objects to extra_objs instead of overwriting it.
12169 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12171         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12172         Use any_fp_register_operand as operand[3] predicate.  Simplify
12173         equality test for operands[2] and operands[4] memory location.
12174         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12175         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12176         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12177         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12178         any_fp_register_operand as operand[1] predicate.  Simplify
12179         equality test for operands[0] and operands[3] memory location.
12180         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12181         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12182         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12184 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12186         * target-insns.def: Add memory_blockage.
12187         * optabs.c (expand_memory_blockage): New function.
12188         (expand_asm_memory_barrier): Rename ...
12189         (expand_asm_memory_blockage): ... to this.
12190         (expand_mem_thread_fence): Call expand_memory_blockage
12191         instead of expand_asm_memory_barrier.
12192         (expand_mem_singnal_fence): Ditto.
12193         (expand_atomic_load): Ditto.
12194         (expand_atomic_store): Ditto.
12195         * doc/md.texi (Standard Pattern Names For Generation):
12196         Document memory_blockage instruction pattern.
12198 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12200         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12201         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12202         * config/rl78/rl78.md: New define_expand "adddi3".
12204 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12206         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12207         are set correctly.
12209 2017-10-13  Jeff Law  <law@redhat.com>
12211         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12213 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12215         PR target/82274
12216         * internal-fn.c (expand_mul_overflow): If both operands have
12217         the same highpart of -1 or 0 and the topmost bit of lowpart
12218         is different, overflow is if res <= 0 rather than res < 0.
12220 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12222         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12223         TARGET_P9_VECTOR code for unaligned_load case.
12225 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12227         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12229 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12231         * tree-core.h (tree_contains_struct): Make bool.
12232         * tree.c (tree_contains_struct): Likewise.
12233         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12234         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12235         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12236         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12237         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12239 2017-10-13  Richard Biener  <rguenther@suse.de>
12241         * graphite-isl-ast-to-gimple.c
12242         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12243         parameters and dominance check.
12244         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12245         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12246         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12247         Do not update SSA form here or do intermediate IL verification.
12248         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12249         (graphite_initialize): Remove check on the number of loops in
12250         the function and inline into graphite_transform_loops.
12251         (graphite_finalize): Inline into graphite_transform_loops.
12252         (graphite_transform_loops): Perform SSA update and IL verification
12253         here.
12254         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12256 2017-10-13  Richard Biener  <rguenther@suse.de>
12258         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12259         graphite_expression_type_precision): Avoid global constructor
12260         by moving ...
12261         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12262         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12263         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12264         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12265         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12266         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12268 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12270         PR target/82499
12271         * config/i386/i386.h (ix86_red_zone_size): New.
12272         * config/i386/i386.md (push peephole2s): Replace
12273         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12275 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12276             Alan Hayward  <alan.hayward@arm.com>
12277             David Sherwood  <david.sherwood@arm.com>
12279         * combine.c (can_change_dest_mode): Reject changes in
12280         REGMODE_NATURAL_SIZE.
12282 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12283             Alan Hayward  <alan.hayward@arm.com>
12284             David Sherwood  <david.sherwood@arm.com>
12286         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12287         (expand_debug_source_expr): Likewise.
12288         * combine.c (combine_simplify_rtx): Likewise.
12289         * cse.c (fold_rtx): Likewise.
12290         * fwprop.c (canonicalize_address): Likewise.
12291         * targhooks.c (default_shift_truncation_mask): Likewise.
12293 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12294             Alan Hayward  <alan.hayward@arm.com>
12295             David Sherwood  <david.sherwood@arm.com>
12297         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12298         (widened_mode): Likewise.
12299         (expand_unop): Likewise.
12300         * ree.c (transform_ifelse): Likewise.
12301         (merge_def_and_ext): Likewise.
12302         (combine_reaching_defs): Likewise.
12303         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12305 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12306             Alan Hayward  <alan.hayward@arm.com>
12307             David Sherwood  <david.sherwood@arm.com>
12309         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12310         * combine.c (gen_lowpart_for_combine): Likewise.
12311         * dwarf2out.c (rtl_for_decl_location): Likewise.
12312         * final.c (alter_subreg): Likewise.
12313         * rtlhooks.c (gen_lowpart_general): Likewise.
12314         (gen_lowpart_if_possible): Likewise.
12316 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12317             Alan Hayward  <alan.hayward@arm.com>
12318             David Sherwood  <david.sherwood@arm.com>
12320         * calls.c (expand_call): Use subreg_lowpart_offset.
12321         * cse.c (cse_insn): Likewise.
12322         * regcprop.c (copy_value): Likewise.
12323         (copyprop_hardreg_forward_1): Likewise.
12325 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12327         PR target/82524
12328         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12329         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12330         =Q constraints to +Q and into insn condition add check
12331         that operands[0] and operands[1] are equal.
12332         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12333         =Q constraints to +Q and into insn condition add check
12334         that operands[0] is equal to either operands[1] or operands[2].
12336         PR target/82498
12337         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12338         instead of handling MINUS_EXPR twice (once for each argument),
12339         canonicalize operand order and handle just once, use rtype where
12340         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12342         PR target/82498
12343         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12344         any values of __C while still being pattern recognizable as a simple
12345         rotate instruction.
12347 2017-10-13  Richard Biener  <rguenther@suse.de>
12349         PR tree-optimization/82451
12350         Revert
12351         2017-10-02  Richard Biener  <rguenther@suse.de>
12353         PR tree-optimization/82355
12354         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12355         a mapping for the enclosing loop but avoid generating one for
12356         the loop tree root.
12357         (copy_bb_and_scalar_dependences): Remove premature codegen
12358         error on PHIs in blocks duplicated into multiple places.
12359         * graphite-scop-detection.c
12360         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12361         in the region use it as loop and nest to analyze the DR in.
12362         (try_generate_gimple_bb): Likewise.
12363         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12364         (add_loop_constraints): For blocks in a loop not in the region
12365         create a dimension with a single iteration.
12366         * sese.h (gbb_loop_at_index): Remove assert.
12368         * cfgloop.c (loop_preheader_edge): For the loop tree root
12369         return the single successor of the entry block.
12370         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12371         Reset the SCEV hashtable and niters.
12372         * graphite-scop-detection.c
12373         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12374         assert that we only have POLYNOMIAL_CHREC that vary in loops
12375         contained in the region.
12376         (scop_detection::graphite_can_represent_expr): Adjust.
12377         (scop_detection::stmt_has_simple_data_refs_p): For loops
12378         not in the region set loop to NULL.  The nest is now the
12379         entry edge to the region.
12380         (try_generate_gimple_bb): Likewise.
12381         * sese.c (scalar_evolution_in_region): Adjust for
12382         instantiate_scev change.
12383         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12384         Make nest parameter the edge into the region.
12385         (create_data_ref): Likewise.
12386         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12387         entry edge into a region and adjust instantiate_scev calls.
12388         (create_data_ref): Likewise.
12389         (graphite_find_data_references_in_stmt): Likewise.
12390         (find_data_references_in_stmt): Pass the loop preheader edge
12391         from the nest argument.
12392         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12393         parameter the edge into the region.
12394         (instantiate_parameters): Use the loop preheader edge as entry.
12395         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12396         NULL loop.
12397         (get_instantiated_value_entry): Make instantiate_below parameter
12398         the edge into the region.
12399         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12400         when we cannot use loop-based instantiation instantiate by
12401         walking use-def chains.
12402         (instantiate_scev_poly): Adjust.
12403         (instantiate_scev_binary): Likewise.
12404         (instantiate_scev_convert): Likewise.
12405         (instantiate_scev_not): Likewise.
12406         (instantiate_array_ref): Remove.
12407         (instantiate_scev_3): Likewise.
12408         (instantiate_scev_2): Likewise.
12409         (instantiate_scev_1): Likewise.
12410         (instantiate_scev_r): Do not blindly handle N-operand trees.
12411         Do not instantiate array-refs.  Handle all constants and invariants.
12412         (instantiate_scev): Make instantiate_below parameter
12413         the edge into the region.
12414         (resolve_mixers): Use the loop preheader edge for the region
12415         parameter to instantiate_scev_r.
12416         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12418 2017-10-13  Richard Biener  <rguenther@suse.de>
12420         PR tree-optimization/82525
12421         * graphite-isl-ast-to-gimple.c
12422         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12423         out from ...
12424         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12425         Fail code generation when we cannot represent the isl integer.
12426         (binary_op_to_tree): Elide modulo operations that are no-ops
12427         in the type we code generate.  Remove now superfluous code
12428         generation errors.
12430 2017-10-13  Richard Biener  <rguenther@suse.de>
12432         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12433         (scop_detection::harmful_loop_in_region): Remove premature
12434         IV type restriction.
12435         (scop_detection::graphite_can_represent_scev): We can handle
12436         pointer IVs just fine.
12438 2017-10-13  Alan Modra  <amodra@gmail.com>
12440         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12441         "Clobbers and Scratch Registers".  Add paragraph on
12442         alternative to clobbers for scratch registers and OpenBLAS
12443         example.
12445 2017-10-13  Alan Modra  <amodra@gmail.com>
12447         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12448         example of a memory input for a string of known length.  Move
12449         commentary out of table.  Add a number of new examples
12450         covering array memory inputs.
12452 2017-10-12  Martin Liska  <mliska@suse.cz>
12454         PR tree-optimization/82493
12455         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12456         (test_range_functions): New function.
12457         (sbitmap_c_tests): Likewise.
12458         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12459         * selftest.h (sbitmap_c_tests): New function.
12461         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12463 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12465         * config/rs6000/amo.h: Fix spacing issue.
12467 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12469         PR target/82498
12470         * config/i386/i386.md (*ashl<mode>3_mask_1,
12471         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12472         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12473         patterns.
12475 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12477         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12478         (profile_probability): Set max_probability
12479         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12480         in temporaries.
12481         * profile-count.c (profile_probability::differs_from_p): Do not
12482         rely on max_probaiblity == 10000
12484 2017-10-12  Jeff Law  <law@redhat.com>
12486         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12487         negative offsets.
12489 2017-10-12  Martin Sebor  <msebor@redhat.com>
12491         PR other/82301
12492         PR c/82435
12493         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12494         (handle_alias_pairs): Call it.
12495         * common.opt (-Wattribute-alias): New option.
12496         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12497         * doc/invoke.texi (-Wattribute-alias): Document.
12499 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12501         Revert
12502         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12503         PR sanitizer/82353
12504         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12505         locations.
12506         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12507         (make_hard_regno_born, make_hard_regno_dead): Update
12508         bb_killed_pseudos and bb_gen_pseudos.
12510 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12512         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12514 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12516         * config/alpha/alpha.c (alpha_split_conditional_move):
12517         Use std::swap instead of manually swapping.
12518         (alpha_stdarg_optimize_hook): Ditto.
12519         (alpha_canonicalize_comparison): Ditto.
12521 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12523         * tree-loop-distribution.c (struct builtin_info): New struct.
12524         (struct partition): Refactor fields into struct builtin_info.
12525         (partition_free): Free struct builtin_info.
12526         (build_size_arg_loc, build_addr_arg_loc): Delete.
12527         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12528         information from struct builtin_info.
12529         (find_single_drs): New function refactored from classify_partition.
12530         Also moved builtin validity checks to this function.
12531         (compute_access_range, alloc_builtin): New functions.
12532         (classify_builtin_st, classify_builtin_ldst): New functions.
12533         (classify_partition): Refactor code into functions find_single_drs,
12534         classify_builtin_st and classify_builtin_ldst.
12535         (distribute_loop): Don't do runtime alias check when distributing
12536         loop nest.
12537         (find_seed_stmts_for_distribution): New function.
12538         (pass_loop_distribution::execute): Refactor code finding seed
12539         stmts into above function.  Support distribution for the innermost
12540         two-level loop nest.  Adjust dump information.
12542 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12544         * tree-loop-distribution.c: Adjust the general comment.
12545         (NUM_PARTITION_THRESHOLD): New macro.
12546         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12547         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12548         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12549         in call to build_partition_graph.
12550         (finalize_partitions): New parameter.  Make loop distribution more
12551         conservative by fusing more partitions.
12552         (distribute_loop): Don't do runtime alias check in case of loop nest
12553         distribution.
12554         (find_seed_stmts_for_distribution): New function.
12555         (prepare_perfect_loop_nest): New function.
12556         (pass_loop_distribution::execute): Refactor code finding seed stmts
12557         and loop nest into above functions.  Support loop nest distribution.
12558         Adjust dump information accordingly.
12560 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12562         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12563         and set PTYPE_SEQUENTIAL for merged partition.
12565 2017-10-12  Richard Biener  <rguenther@suse.de>
12567         PR tree-optimization/69728
12568         Revert
12569         2017-09-19  Richard Biener  <rguenther@suse.de>
12571         PR tree-optimization/69728
12572         * graphite-sese-to-poly.c (schedule_error): New global.
12573         (add_loop_schedule): Handle empty domain by failing the
12574         schedule.
12575         (build_original_schedule): Handle schedule_error.
12577         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12578         domain by returning an unchanged schedule.
12580 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12582         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12583         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12585 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12587         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12588         Handle params.def.
12590 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12592         PR c++/82159
12593         * expr.c (store_field): Don't optimize away bitsize == 0 store
12594         from CALL_EXPR with addressable return type.
12596 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12598         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12599         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12600         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12601         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12602         TARGET_ISEL instead of TARGET_ISEL<sel>.
12604 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12606         * config/rs6000/rs6000.c
12607         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12609 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12611         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12612         Move up in file.
12613         (reg_or_cint_operand): Fix comment.
12614         (reg_or_zero_operand): New predicate.
12615         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12616         * config/rs6000/rs6000.c (output_isel): Delete.
12617         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12618         instead of reg_or_cint_operand.  Output instruction directly (not via
12619         output_isel).
12620         (isel_unsigned_<mode>): Ditto.
12621         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12622         gpc_reg_operand.  Add an instruction alternative for this.  Output
12623         instruction directly.
12624         (*isel_reversed_unsigned_<mode>): Ditto.
12626 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12628         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12629         (TARGET_CANONICALIZE_COMPARISON): Define.
12631 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12633         PR target/81422
12634         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12635         Check whether the dest is REG before adding REG_EQUIV note.
12637 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12639         PR sanitizer/82353
12640         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12641         locations.
12642         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12643         (make_hard_regno_born, make_hard_regno_dead): Update
12644         bb_killed_pseudos and bb_gen_pseudos.
12646 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12648         * incpath.h (enum incpath_kind): Name enum, prefix values.
12649         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12650         * incpath.c (heads, tails): Use INC_MAX.
12651         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12652         (merge_include_chains, split_quote_chain,
12653         register_include_chains): Update incpath_kind names.
12654         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12655         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12656         names.
12657         (add_framework_path, darwin_register_objc_includes): Likewise.
12658         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12660 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12662         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12663         Do not use float_operator operator predicate.
12664         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12665         * config/i386/predicates.md (float_operator): Remove predicate.
12667 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12669         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12670         (*jccxf_i387): Ditto.
12671         (*jcc<mode>_i387): Ditto.
12672         (*jccu<mode>_i387): Ditto.
12673         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12674         (*jcc_*_i387 splitters): Remove.
12675         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12676         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12677         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12678         Remove predicate.
12680 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12682         * profile-count.h (slow_safe_scale_64bit): New function.
12683         (safe_scale_64bit): New inline.
12684         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12685         * profile-count.c: Include wide-int.h
12686         (slow_safe_scale_64bit): New.
12688 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12690         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12691         HAS_DECL_ASSEMBLER_NAME_P.
12692         * gimple-expr.c (gimple_decl_printable_name: Check
12693         HAS_DECL_ASSEMBLER_NAME_P too.
12694         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12695         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12696         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12697         * passes.c (rest_of_decl_compilation): Check
12698         HAS_DECL_ASSEMBLER_NAME_P too.
12699         * recog.c (verify_changes): Likewise.
12700         * tree-pretty-print.c (dump_decl_name): Likewise.
12701         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12703         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12704         (SET_DECL_ASSEMBLER_NAME): Use it.
12705         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12706         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12707         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12709 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12711         * config.gcc (i386, x86_64): Add extra objects.
12712         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12713         (ix86_min_insn_size): Declare.
12714         (ix86_issue_rate): Declare.
12715         (ix86_adjust_cost): Declare.
12716         (ia32_multipass_dfa_lookahead): Declare.
12717         (ix86_macro_fusion_p): Declare.
12718         (ix86_macro_fusion_pair_p): Declare.
12719         (ix86_bd_has_dispatch): Declare.
12720         (ix86_bd_do_dispatch): Declare.
12721         (ix86_core2i7_init_hooks): Declare.
12722         (ix86_atom_sched_reorder): Declare.
12723         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12724         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12725         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12726         (rip_relative_addr_p): Rename to ...
12727         (ix86_rip_relative_addr_p): ... this one; export.
12728         (memory_address_length): Update.
12729         (ix86_issue_rate): Move to x86-tune-sched.c.
12730         (ix86_flags_dependent): Move to x86-tune-sched.c.
12731         (ix86_agi_dependent): Move to x86-tune-sched.c.
12732         (exact_dependency_1): Move to x86-tune-sched.c.
12733         (exact_store_load_dependency): Move to x86-tune-sched.c.
12734         (ix86_adjust_cost): Move to x86-tune-sched.c.
12735         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12736         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12737         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12738         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12739         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12740         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12741         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12742         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12743         (min_insn_size): Rename to ...
12744         (ix86_min_insn_size): ... this one; export.
12745         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12746         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12747         (core2i7_first_cycle_multipass_backtrack): Move to
12748         x86-tune-sched-core.c.
12749         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12750         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12751         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12752         (ix86_avoid_jump_mispredicts): Update.
12753         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12754         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12755         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12756         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12757         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12758         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12759         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12760         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12761         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12762         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12763         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12764         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12765         (BIG): Move to ix86-tune-sched-bd.c.
12766         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12767         (enum insn_path): Move to ix86-tune-sched-bd.c.
12768         (get_mem_group): Move to ix86-tune-sched-bd.c.
12769         (is_cmp): Move to ix86-tune-sched-bd.c.
12770         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12771         (is_branch): Move to ix86-tune-sched-bd.c.
12772         (is_prefetch): Move to ix86-tune-sched-bd.c.
12773         (init_window): Move to ix86-tune-sched-bd.c.
12774         (allocate_window): Move to ix86-tune-sched-bd.c.
12775         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12776         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12777         (process_end_window): Move to ix86-tune-sched-bd.c.
12778         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12779         (find_constant): Move to ix86-tune-sched-bd.c.
12780         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12781         (has_immediate): Move to ix86-tune-sched-bd.c.
12782         (get_insn_path): Move to ix86-tune-sched-bd.c.
12783         (get_insn_group): Move to ix86-tune-sched-bd.c.
12784         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12785         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12786         (add_insn_window): Move to ix86-tune-sched-bd.c.
12787         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12788         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12789         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12790         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12791         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12792         (do_dispatch): Move to ix86-tune-sched-bd.c.
12793         (has_dispatch): Move to ix86-tune-sched-bd.c.
12794         * config/i386/t-i386: Add new object files.
12795         * config/i386/x86-tune-costs.h: New file.
12796         * config/i386/x86-tune-sched-atom.c: New file.
12797         * config/i386/x86-tune-sched-bd.c: New file.
12798         * config/i386/x86-tune-sched-core.c: New file.
12799         * config/i386/x86-tune-sched.c: New file.
12801 2017-10-11  Liu Hao  <lh_mouse@126.com>
12803         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12804         the generic version below instead.
12805         (should_colorize): Recognize Windows consoles as terminals
12806         for MinGW targets.
12807         * pretty-print.c [__MINGW32__] (write_all): New function.
12808         [__MINGW32__] (find_esc_head): Likewise.
12809         [__MINGW32__] (find_esc_terminator): Likewise.
12810         [__MINGW32__] (eat_esc_sequence): Likewise.
12811         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12812         ANSI escape codes.
12813         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12814         for MinGW targets.
12816 2017-10-11  Richard Biener  <rguenther@suse.de>
12818         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12819         Properly call analyze_scalar_evolution with the loop of the stmt.
12821 2017-10-11  Richard Biener  <rguenther@suse.de>
12823         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12824         * tree-core.h (tree_base): Add chrec_var union member.
12825         * tree.h (CHREC_VAR): Remove.
12826         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12827         * tree-chrec.h (build_polynomial_chrec): Adjust.
12828         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12829         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12831 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12833         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12834         * match.pd: ... here.
12835         ((T) X == (T) Y): Relax condition.
12837 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12839         PR tree-optimization/82472
12840         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12841         comment.
12842         (break_alias_scc_partitions): Update postorder number.
12844 2017-10-11  Martin Liska  <mliska@suse.cz>
12846         PR sanitizer/82490
12847         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12848         variable.
12849         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12851 2017-10-11  Martin Liska  <mliska@suse.cz>
12853         * print-rtl.c (print_insn): Move declaration of idbuf
12854         to same scope as name.
12856 2017-10-11  Martin Liska  <mliska@suse.cz>
12858         Revert r253637:
12860         PR sanitizer/82484
12861         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12862         volatile arguments.
12864 2017-10-11  Martin Liska  <mliska@suse.cz>
12866         PR sanitizer/82484
12867         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12868         volatile arguments.
12870 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
12872         * config.gcc (default_gnu_indirect_function): Default to yes for
12873         arm*-*-linux* with glibc.
12875 2017-10-11  Richard Biener  <rguenther@suse.de>
12877         * tree-scalar-evolution.c (get_scalar_evolution): Handle
12878         default-defs and types we do not want to analyze.
12879         (interpret_loop_phi): Replace unreachable code with an assert.
12880         (compute_scalar_evolution_in_loop): Remove and inline ...
12881         (analyze_scalar_evolution_1): ... here, replacing condition with
12882         what makes the intent clearer.  Remove handling of cases
12883         get_scalar_evolution now handles.
12885 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
12887         PR rtl-optimization/81434
12888         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
12889         comment for main loop.  In sched_group_found if, also add checks for
12890         pass and min_cost_group.
12892 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
12894         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
12895         (rs6000_insn_cost): New function.
12896         * config/rs6000/rs6000.md (cost): New attribute.
12898 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
12899             H.J. Lu  <hongjiu.lu@intel.com>
12901         PR target/79565
12902         PR target/82483
12903         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
12904         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
12905         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
12906         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
12907         to OPTION_MASK_ISA_AVX512VL - builtins that have both
12908         OPTION_MASK_ISA_MMX and some other bit set require both
12909         mmx and the ISAs without the mmx bit.
12910         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
12911         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
12912         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
12913         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
12914         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
12915         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
12916         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
12917         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
12918         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
12919         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
12920         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
12921         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
12922         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
12923         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
12924         Add OPTION_MASK_ISA_MMX.
12926 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
12928         * config.gcc (armv7*-*-freebsd*): New target.
12929         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
12931 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
12933         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
12934         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
12935         spot in the file.
12937 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
12939         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
12940         a template parameter.
12941         (WIDE_INT_REF_FOR): Update accordingly.
12942         * tree.h (wi::int_traits <const_tree>): Delete.
12943         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
12944         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
12945         (wi::tree_to_wide_ref): New typedef.
12946         (wi::to_wide): New function.
12947         * calls.c (get_size_range): Use wi::to_wide when operating on
12948         trees as wide_ints.
12949         * cgraph.c (cgraph_node::create_thunk): Likewise.
12950         * config/i386/i386.c (ix86_data_alignment): Likewise.
12951         (ix86_local_alignment): Likewise.
12952         * dbxout.c (stabstr_O): Likewise.
12953         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
12954         * expr.c (const_vector_from_tree): Likewise.
12955         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
12956         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
12957         (fold_negate_expr_1, int_const_binop_1, const_binop)
12958         (fold_convert_const_int_from_real, optimize_bit_field_compare)
12959         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
12960         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
12961         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
12962         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
12963         (fold_not_const, round_up_loc): Likewise.
12964         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
12965         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
12966         (alloca_call_type): Likewise.
12967         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
12968         * godump.c (go_output_typedef): Likewise.
12969         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
12970         * internal-fn.c (get_min_precision): Likewise.
12971         * ipa-cp.c (ipcp_store_vr_results): Likewise.
12972         * ipa-polymorphic-call.c
12973         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
12974         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
12975         (ipa_modify_call_arguments): Likewise.
12976         * match.pd: Likewise.
12977         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
12978         * print-tree.c (print_node_brief, print_node): Likewise.
12979         * stmt.c (expand_case): Likewise.
12980         * stor-layout.c (layout_type): Likewise.
12981         * tree-affine.c (tree_to_aff_combination): Likewise.
12982         * tree-cfg.c (group_case_labels_stmt): Likewise.
12983         * tree-data-ref.c (dr_analyze_indices): Likewise.
12984         (prune_runtime_alias_test_list): Likewise.
12985         * tree-dump.c (dequeue_and_dump): Likewise.
12986         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
12987         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
12988         * tree-pretty-print.c (dump_generic_node): Likewise.
12989         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
12990         (simple_iv_with_niters): Likewise.
12991         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
12992         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
12993         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
12994         * tree-ssa-loop-niter.c (split_to_var_and_offset)
12995         (refine_value_range_using_guard, number_of_iterations_ne_max)
12996         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
12997         (get_cst_init_from_scev, record_nonwrapping_iv)
12998         (scev_var_range_cant_overflow): Likewise.
12999         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13000         * tree-ssa-pre.c (compute_avail): Likewise.
13001         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13002         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13003         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13004         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13005         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13006         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13007         (dump_case_nodes, try_switch_expansion): Likewise.
13008         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13009         (vect_do_peeling): Likewise.
13010         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13011         * tree-vect-stmts.c (vectorizable_load): Likewise.
13012         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13013         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13014         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13015         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13016         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13017         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13018         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13019         (evrp_dom_walker::before_dom_children): Likewise.
13020         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13021         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13022         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13023         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13024         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13025         (get_range_pos_neg): Likewise.
13026         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13027         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13028         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13029         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13030         * config/avr/avr.c (avr_fold_builtin): Likewise.
13031         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13032         * config/msp430/msp430.c (msp430_attr): Likewise.
13033         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13034         * config/powerpcspe/powerpcspe-c.c
13035         (altivec_resolve_overloaded_builtin): Likewise.
13036         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13037         (rs6000_expand_ternop_builtin): Likewise.
13038         * config/rs6000/rs6000-c.c
13039         (altivec_resolve_overloaded_builtin): Likewise.
13040         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13041         (rs6000_expand_ternop_builtin): Likewise.
13042         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13044 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13046         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13047         when copying loop nest with only one inner loop.
13049 2017-10-10  Richard Biener  <rguenther@suse.de>
13051         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13052         blocks if SCEV is active.
13053         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13054         dead code.
13055         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13056         (scev_initialize): Assert we are not yet initialized.
13058 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13060         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13061         inner loop's exit stmt by making it always exit the loop, otherwise
13062         we would generate an infinite empty loop.
13064 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13066         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13067         renaming variables in new preheader if it's deleted.
13069 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13071         * tree-loop-distribution.c (struct partition): Remove unused field
13072         loops of the structure.
13073         (partition_alloc, partition_free): Ditto.
13074         (build_rdg_partition_for_vertex): Ditto.
13076 2017-10-09  Jeff Law  <law@redhat.com>
13078         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13079         return type to match prototype and documentation.
13081 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13083         * config/rs6000/rs6000.c (processor_costs): Move to ...
13084         * config/rs6000/rs6000.h: ... here.
13085         (rs6000_cost): Declare.
13087 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13089         * except.c (setjmp_fn): New global variable.
13090         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13091         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13092         if DONT_USE_BUILTIN_SETJMP is defined.
13094 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13096         * target.def (insn_cost): New hook.
13097         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13098         * doc/tm.texi: Regenerate.
13099         * rtlanal.c (insn_cost): Use the new hook.
13101 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13103         * combine.c (combine_validate_cost): Compute the new insn_cost,
13104         not just pattern_cost.
13105         (try_combine): Adjust comment.
13107 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13109         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13110         insn_cost.
13111         * combine.c (uid_insn_cost): Adjust comment.
13112         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13113         of insn_rtx_cost
13114         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13115         * dse.c (find_shift_sequence): Ditto.
13116         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13117         (bb_valid_for_noce_process_p): Use pattern_cost.
13118         * rtl.h (insn_rtx_cost): Delete.
13119         (pattern_cost): New prototype.
13120         (insn_cost): New prototype.
13121         * rtlanal.c (insn_rtx_cost): Rename to...
13122         (pattern_cost): ... this.
13123         (insn_cost): New.
13125 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13127         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13128         (*jcc<mode>_0_r_i387): Ditto.
13129         (*jccxf_r_i387): Ditto.
13130         (*jcc<mode>_r_i387): Ditto.
13131         (*jccu<mode>_r_i387): Ditto.
13132         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13133         (*jcc): Rename from *jcc_1.
13135 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13137         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13138         deferred rescans after the lvx/stvx recombination pre-pass.
13140 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13142         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13143         memory operation instruction support.
13144         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13145         (rs6000-ibm-aix[789]*): Likewise.
13146         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13147         Document new functions.
13149 2017-10-09  Richard Biener  <rguenther@suse.de>
13151         PR tree-optimization/82397
13152         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13153         equality only for semantically equal trees.
13155 2017-10-09  Richard Biener  <rguenther@suse.de>
13157         PR tree-optimization/82449
13158         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13159         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13160         allow constant addresses.
13161         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13162         are linear.
13164 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13166         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13167         flags.
13169 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13171         PR target/82463
13172         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13173         definitions.
13175 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13177         PR target/82465
13178         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13180 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13182         PR target/82464
13183         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13184         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13186 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13188         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13189         (WI_BINARY_PREDICATE_RESULT): Likewise.
13190         (wi::binary_traits::operator_result): New type.
13191         (wi::binary_traits::predicate_result): Likewise.
13192         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13193         (generic_wide_int::operator==, generic_wide_int::operator!=)
13194         (generic_wide_int::operator&, generic_wide_int::and_not)
13195         (generic_wide_int::operator|, generic_wide_int::or_not)
13196         (generic_wide_int::operator^, generic_wide_int::operator+
13197         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13198         Delete.
13199         (operator~, unary operator-, operator==, operator!=, operator&)
13200         (operator|, operator^, operator+, binary operator-, operator*): New
13201         functions.
13202         * expr.c (get_inner_reference): Use wi::bit_and_not.
13203         * fold-const.c (fold_binary_loc): Likewise.
13204         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13205         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13206         (bit_value_binop): Likewise.
13207         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13208         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13209         (extract_range_from_binary_expr_1): Likewise.
13210         (masked_increment): Likewise.
13211         (simplify_bit_ops_using_ranges): Likewise.
13213 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13215         PR hsa/82416
13216         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13217         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13218         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13219         COMPLEX types.
13220         (hsa_fixup_mov_insn_type): New function.
13221         (hsa_op_with_type::get_in_type): Use it.
13222         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13223         immediates in an assert.
13224         (hsa_op_with_type::extend_int_to_32bit): New method.
13225         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13226         types.  Convert to dest type if necessary.
13227         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13228         (reg_for_gimple_ssa): Pass false as min32int to
13229         hsa_type_for_scalar_tree_type.
13230         (gen_hsa_addr): Fixup type when creating addresable temporary.
13231         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13232         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13233         necessary.  Call hsa_fixup_mov_insn_type.
13234         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13235         extend operands and convert to dest type if necessary.
13236         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13237         to dest type if necessary.
13238         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13239         if conversion nt necessary and size matches.
13240         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13241         to dest type if necessary.
13242         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13243         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13244         necessary.
13245         (gen_hsa_clrsb): Likewise.
13246         (gen_hsa_ffs): Likewise.
13247         (gen_hsa_divmod): Extend operands and convert to dest type if
13248         necessary.
13249         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13251 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13253         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13254         Remove empty default arguments.  Use a brace block as output
13255         statement.
13256         (conditional return): Ditto.
13257         (jump): Ditto.
13258         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13259         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13260         Ditto.
13261         (group_ending_nop): Ditto.
13262         (doloop_end): Ditto.
13263         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13264         (splitters for those): Ditto.
13266 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13268         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13269         a conditional jump (and the compare for it) so that pc_rtx is the
13270         last operand.
13271         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13272         for the deleted and renamed ctr<mode>_internal[234] patterns.
13273         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13274         Delete second conditional return pattern.
13275         (ctr<mode>_internal2): Delete this second bdnz pattern.
13276         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13277         (ctr<mode>_internal4): Delete this second bdz pattern.
13279 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13281         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13282         (always_initialized_rtx_for_ssa_name_p): New predicate.
13283         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13284         (finish_out_of_ssa): Free new field of SA.
13285         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13286         * tree-ssa-coalesce.c: Include tree-ssa.h.
13287         (get_parm_default_def_partitions): Remove extern keyword.
13288         (get_undefined_value_partitions): New function.
13289         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13290         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13291         uninitialized bits.
13292         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13294 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13296         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13298 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13300         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13301         for targets that preffer 128bit.
13303 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13305         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13307 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13309         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13310         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13311         prevent DSE.
13312         (thumb_set_return_address): Likewise.
13314 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13316         * common/config/arm/arm-common.c (arm_except_unwind_info):
13317         Handle DWARF2_UNWIND_INFO.
13319 2017-10-07  Michael Collison <michael.collison@arm.com>
13321         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13322         New pattern.
13324 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13326         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13327         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13328         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13329         defined, force the creation of a new block for a dispatch label.
13331 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13333         * invoke.texi (Wsuggest-attribute=cold): Document.
13334         * common.opt (Wsuggest-attribute=cold): New
13335         * ipa-pure-const.c (warn_function_cold): New function.
13336         * predict.c (compute_function_frequency): Use it.
13337         * predict.h (warn_function_cold): Declare.
13339 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13341         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13342         Update profile.
13344 2017-10-06  Martin Liska  <mliska@suse.cz>
13346         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13347         keyword for member functions.
13348         (struct sanopt_tree_couple): New struct.
13349         (struct sanopt_tree_couple_hash): New function.
13350         (struct sanopt_ctx): Add new hash_map.
13351         (has_dominating_ubsan_ptr_check): New function.
13352         (record_ubsan_ptr_check_stmt): Likewise.
13353         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13354         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13355         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13357 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13359         PR target/82440
13360         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13361         aarch64_simd_valid_immediate on CONST_VECTORs.
13362         (aarch64_reg_or_bic_imm): Likewise.
13364 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13366         PR rtl-optimization/82396
13367         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13369 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13371         * graphite-dependences.c (scop_get_reads): Move code to...
13372         (scop_get_must_writes): Move code to...
13373         (scop_get_may_writes): Move code to...
13374         (scop_get_reads_and_writes): ... here.
13375         (scop_get_dependences): Call scop_get_reads_and_writes.
13377 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13379         PR tree-optimization/82434
13380         * fold-const.h (can_native_encode_type_p,
13381         can_native_encode_string_p): Remove.
13382         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13383         don't encode anything, just return what would be otherwise returned.
13384         (native_encode_fixed, native_encode_complex, native_encode_vector):
13385         Likewise.
13386         (native_encode_string): Likewise.  Inline by hand
13387         can_native_encode_string_p.
13388         (can_native_encode_type_p): Remove.
13389         (can_native_encode_string_p): Remove.
13390         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13391         STRING_CSTs using can_native_encode_string_p, test all
13392         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13393         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13394         argument from native_encode_expr.
13395         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13396         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13397         but just 2.
13399 2017-10-06  Richard Biener  <rguenther@suse.de>
13401         PR tree-optimization/82397
13402         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13403         operand_equal_p but rely on data_ref_compare_tree for detecting
13404         equalities.
13405         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13406         to match up with dr_group_sort_cmp.
13408 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13410         PR target/82322
13411         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13412         builtin.
13413         * config/s390/s390-builtin-types.def: Regenerate.
13415 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13417         PR target/82317
13418         * config/s390/s390-builtin-types.def: Regenerate.
13419         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13420         Change flag from B_VXE to B_VX.
13421         (s390_vec_min_dbl): Remove B_VXE flag.
13423 2017-10-06  Richard Biener  <rguenther@suse.de>
13425         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13426         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13427         translate_isl_ast_to_gimple::is_valid_rename,
13428         translate_isl_ast_to_gimple::get_rename,
13429         translate_isl_ast_to_gimple::get_def_bb_for_const,
13430         translate_isl_ast_to_gimple::get_new_name,
13431         translate_isl_ast_to_gimple::collect_all_ssa_names,
13432         translate_isl_ast_to_gimple::copy_loop_phi_args,
13433         translate_isl_ast_to_gimple::collect_all_ssa_names,
13434         translate_isl_ast_to_gimple::copy_loop_phi_args,
13435         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13436         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13437         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13438         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13439         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13440         translate_isl_ast_to_gimple::copy_cond_phi_args,
13441         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13442         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13443         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13444         translate_isl_ast_to_gimple::rename_uses,
13445         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13446         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13447         (set_rename_for_each_def): Likewise.
13448         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13449         here.  Handle rewriting SCEV analyzable uses here.
13450         (copy_bb_and_scalar_dependences): Generate code for PHI
13451         copy-in/outs.
13452         (graphite_regenerate_ast_isl): Adjust.
13453         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13454         (add_write, add_read): New functions.
13455         (build_cross_bb_scalars_def): Use it and simplify.
13456         (build_cross_bb_scalars_use): Likewise.
13457         (graphite_find_cross_bb_scalar_vars): Inline into...
13458         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13459         simulating out-of-SSA.  Compute liveout and add dependencies.
13460         (build_scops): Force an empty entry block.
13461         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13462         members.
13463         (sese_build_liveouts): Declare.
13464         (sese_trivially_empty_bb_p): Likewise.
13465         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13466         compute liveout and debug_liveout.
13467         (sese_bad_liveouts_use): Remove.
13468         (sese_reset_debug_liveouts_bb): Likewise.
13469         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13470         (sese_build_liveouts): Build liveout and debug_liveout and store
13471         it in region.
13472         (new_sese_info): Adjust.
13473         (free_sese_info): Likewise.
13474         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13475         do not build liveout here.
13476         (move_sese_in_condition): Adjust region entry.
13477         (scev_analyzable_p): Match up with chrec_apply requirements.
13478         (sese_trivially_empty_bb_p): New.
13479         * tree-into-ssa.c (get_reaching_def): Properly support generating
13480         default-defs for incremental rewrite of anonymous names.
13482 2017-10-06  Richard Biener  <rguenther@suse.de>
13484         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13485         precision do not perform modulo reduction.
13487 2017-10-06  Richard Biener  <rguenther@suse.de>
13489         PR tree-optimization/82436
13490         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13491         conservatively choose the vectorization factor when checking
13492         whether we can perform the required load permutation.
13493         (vect_transform_slp_perm_load): Assert when we may not fail.
13495 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13497         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13498         message for incompatible -msdata=* and -mcall-* options.
13500 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13502         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13503         rate for post-reload scheduling.
13505 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13507         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13509 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13511         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13512         to improve monte carlo in scimark.
13514 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13516         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13517         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13518         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13519         pentium4_cost, nocona_cost): Set reassociation width to 1.
13520         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13521         width to 2 for fp operations and 1 otherwise.
13522         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13523         for int and fp.
13524         (atom_cost): Set reassociation width to 2.
13525         (slm_cost, generic_cost): Set fp reassociation width
13526         to 2 and 1 otherwise.
13527         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13528         (core_cost): Set fp reassociation width to 4 and vector to 2.
13529         (ix86_reassociation_width): Rewrite using cost table; special case
13530         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13531         and TARGET_AVX128_OPTIMAL.
13532         * config/i386/i386.h (processor_costs): Add
13533         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13534         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13535         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13536         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13537         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13538         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13540 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13542         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13544 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13546         * config/arm/arm.c (arm_test_fpu_data): New.
13547         (arm_run_selftests): Call arm_test_fpu_data.
13549 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13551         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13552         decl.
13553         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13555 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13557         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13558         check type for aarch64_simd_valid_immediate.
13559         (aarch64_output_simd_mov_immediate): Update prototype.
13560         (aarch64_simd_valid_immediate): Update prototype.
13561         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13562         support for ORR-immediate.
13563         (and<mode>3): modified pattern to add support for BIC-immediate.
13564         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13565         now checks for valid immediate for BIC and ORR based on new enum
13566         argument.
13567         (aarch64_output_simd_mov_immediate): Function now used to output
13568         BIC/ORR imm as well based on new enum argument.
13569         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13570         (Db) : Likewise.
13571         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13572         (aarch64_reg_or_bic_imm): Likewise.
13574 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13576         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13577         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13579 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13581         Revert r253399:
13583         PR rtl-optimization/82396
13584         * haifa-sched.c (autopref_multipass_init): Simplify
13585         initialization.
13586         (autopref_rank_data): Simplify sort order.
13587         * sched-int.h (autopref_multipass_data_): Remove
13588         multi_mem_insn_p, min_offset and max_offset.
13590 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13592         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13594 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13596         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13597         vect_doubleint_cvt.
13599 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13601         * doc/sourcebuild.texi: Document vect_long_mult.
13603 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13605         PR tree-optimization/82413
13606         * fold-const.c (build_range_check): Use widest_int when comparing
13607         the maximum ETYPE value with HIGH.
13609 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13611         PR rtl-optimization/82396
13612         * haifa-sched.c (autopref_multipass_init): Simplify
13613         initialization.
13614         (autopref_rank_data): Simplify sort order.
13615         * sched-int.h (autopref_multipass_data_): Remove
13616         multi_mem_insn_p, min_offset and max_offset.
13618 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13620         PR tree-optimization/82381
13621         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13622         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13623         is not.
13625         PR tree-optimization/82374
13626         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13627         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13628         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13629         current_function_decl to the new decl.
13631 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13633         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13634         helper macro for IEEE float128 hardware built-in functions.
13635         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13636         semantics.
13637         (TRUNCF128_ODD): Likewise.
13638         (ADDF128_ODD): Likewise.
13639         (SUBF128_ODD): Likewise.
13640         (MULF128_ODD): Likewise.
13641         (DIVF128_ODD): Likewise.
13642         (FMAF128_ODD): Likewise.
13643         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13644         UNSPEC_TRUNC_ROUND_TO_ODD.
13645         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13646         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13647         floating point round to odd instructions.
13648         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13649         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13650         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13651         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13652         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13653         (trunc<mode>sf2_hw): Change the truncate with round to odd
13654         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13655         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13656         to odd hardware instructions.
13657         (sub<mode>3_odd): Likewise.
13658         (mul<mode>3_odd): Likewise.
13659         (div<mode>3_odd): Likewise.
13660         (sqrt<mode>2_odd): Likewise.
13661         (fma<mode>4_odd): Likewise.
13662         (fms<mode>4_odd): Likewise.
13663         (nfma<mode>4_odd): Likewise.
13664         (nfms<mode>4_odd): Likewise.
13665         (trunc<mode>df2_odd): Change the truncate with round to odd
13666         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13667         function.
13668         * doc/extend.texi (PowerPC built-in functions): Update documentation
13669         for existing IEEE float128-bit built-in functions.  Add built-in
13670         functions that generate the IEEE 128-bit floating point round to
13671         odd instructions.
13673 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13675         PR rtl-optimization/77729
13676         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13677         to (X&(C1&~C2))|C2 transformations.
13679 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13681         PR tree-optimization/82363
13682         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13683         mismatch, mark lacc written regardless of racc.
13685 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13687         PR tree-optimization/82381
13688         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13689         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13690         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13691         fallthrough into reassoc_stmt_dominates_stmt_p.
13693         PR target/82386
13694         * combine.c (combine_instructions): Don't combine in unreachable
13695         basic blocks.
13697 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13699         PR target/80210
13700         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13701         function to not use the have_cpu variable.  Do not set cpu_index,
13702         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13703         or the default cpu.
13704         (rs6000_valid_attribute_p): Remove duplicate initializations of
13705         old_optimize and func_optimize.
13706         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13707         (rs6000_activate_target_options): Make global.
13708         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13709         prototype.
13711 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13713         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13714         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13715         Set *poffset to 0 and *psize and *pmax_size to -1 if
13716         *poffset + *psize overflows in HOST_WIDE_INT.
13718         PR tree-optimization/82387
13719         PR tree-optimization/82388
13720         PR tree-optimization/82389
13721         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13722         instead of live_bytes non-NULL.
13724 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13726         PR target/41076
13727         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13728         alternative.
13730 2017-10-02  Richard Biener  <rguenther@suse.de>
13732         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13733         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13734         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13736 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13738         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13739         requested precision matches the type's.
13740         * calls.c (alloc_max_size): Calculate the new candidate size as
13741         a widest_int and use wi::to_widest when comparing it with the
13742         current candidate size.
13743         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13744         zero rather than integer_zero_node.
13745         * match.pd: Check for a no-op conversion before using wi::add
13746         rather than after.  Use tree_to_uhwi when summing small shift
13747         counts into an unsigned int.
13749 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13750             Alan Hayward  <alan.hayward@arm.com>
13751             David Sherwood  <david.sherwood@arm.com>
13753         PR target/71307
13754         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13755         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13756         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13757         POINTER_AND_FP_REGS.
13759 2017-10-02  Richard Biener  <rguenther@suse.de>
13761         PR tree-optimization/82355
13762         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13763         a mapping for the enclosing loop but avoid generating one for
13764         the loop tree root.
13765         (copy_bb_and_scalar_dependences): Remove premature codegen
13766         error on PHIs in blocks duplicated into multiple places.
13767         * graphite-scop-detection.c
13768         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13769         in the region use it as loop and nest to analyze the DR in.
13770         (try_generate_gimple_bb): Likewise.
13771         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13772         (add_loop_constraints): For blocks in a loop not in the region
13773         create a dimension with a single iteration.
13774         * sese.h (gbb_loop_at_index): Remove assert.
13776 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13778         * omp-expand.c (adjust_context_scope): New function.
13779         (expand_parallel_call): Call adjust_context_scope.
13781 2017-10-01  Jeff Law  <law@redhat.com>
13783         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13784         dom_opt_dom_walker class with direct access to private members.
13785         Add comments.  Call test_for_singularity.
13786         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13787         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13788         m_dummy_cond anymore.
13789         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13790         class ctor.
13791         (pass_dominator:execute): Build the dummy_cond here and pass it
13792         to the dom_opt_dom_walker ctor.
13793         (test_for_singularity): New function.
13795 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13796             Maya Rashish  <coypu@sdf.org>
13798         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13799         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13800         (alpha*-*-netbsd*) Use nbsd_tm_file.
13801         (arm*-*-netbsdelf*) Likewise.
13802         (i[34567]86-*-netbsdelf*) Likewise.
13803         (x86_64-*-netbsd*) Likewise.
13804         (mips*-*-netbsd*) Likewise.
13805         (powerpc-*-netbsd*) Likewise.
13806         (sh*-*-netbsd*) Likewise.
13807         (sparc-*-netbsdelf*) Likewise.
13808         (sparc64-*-netbsd*) Likewise.
13809         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13810         to tm_defines.
13811         (vax-*-netbsdelf*) Likewise.
13812         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13813         (UINT_FAST8_TYPE) Likewise.
13814         (INT_FAST16_TYPE) Check CHAR_FAST16.
13815         (UINT_FAST16_TYPE) Likewise.
13817 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13819         PR target/82361
13820         * config/i386/i386.md
13821         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13822         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13823         *divmodsi4_zext_2): New define_insn_and_split.
13824         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13825         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13826         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13827         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13828         New define_insn_and_split.
13829         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13830         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13831         operands[1] having DImode when mode is SImode.
13833         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13834         always SImode for DIV and MOD in REG_EQUAL notes.
13836 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13838         PR middle-end/82319
13839         * match.pd: Fix handling of NaNs in pattern.
13841 2017-09-29  Jeff Law  <law@redhat.com>
13843         * sbitmap.c (bitmap_bit_in_range_p): New function.
13844         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13845         * tree-ssa-dse.c (live_bytes_read): New function.
13846         (dse_classify_store): Ignore reads of dead bytes.
13848         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13849         typos and whitespace errors.
13850         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13851         * config/s390/s390.c (s390_emit_prologue): Likewise.
13853 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13855         PR target/81481
13856         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13857         with a symbol for LRA.
13859 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13861         PR rtl-optimization/82338
13862         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13864 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
13866         * genmodes.c (calc_wider_mode): Suppress qsort macro.
13867         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
13868         (qsort_chk): Declare.
13869         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
13870         (qsort_chk): New function.
13872 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13874         PR tree-optimization/82337
13875         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
13876         phi definition if the PHI result appears in an abnormal PHI.
13877         (find_basis_for_base_expr): Don't record a basis if the LHS of the
13878         basis appears in an abnormal PHI.
13880 2017-09-29  Richard Biener  <rguenther@suse.de>
13882         * graphite-isl-ast-to-gimple.c
13883         (translate_isl_ast_to_gimple::set_codegen_error): New function.
13884         (binary_op_to_tree): Use it.
13885         (get_rename_from_scev): Likewise.
13886         (copy_loop_phi_nodes): Likewise.
13887         (copy_bb_and_scalar_dependences): Likewise.
13888         (translate_pending_phi_nodes): Likewise.
13890 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
13892         PR target/82339
13893         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
13894         for movabsq $(i32 << shift), r64.
13896 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
13898         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
13899         index when encoding %esp as %rsp to avoid 0x67 prefix.
13901 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
13903         * config/i386/i386.md (*movsf_internal, *movdf_internal):
13904         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
13906 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
13908         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
13909         Extensions with more than 16 double VFP registers.
13910         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
13911         to_clear_mask and all code related to it.  Replace the remaining
13912         entry by a sbitmap and adapt code accordingly.
13914 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
13916         * brig-builtins.def: Change pure attributes to const.
13918 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
13920         * config.gcc (default_gnu_indirect_function): Default to yes for
13921         sparc*-*-linux* with glibc.
13923 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
13925         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
13926         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
13927         when creating .init_array and .fini_array sections with priority
13928         specified.
13930 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
13932         PR target/71727
13933         * config/aarch64/aarch64.c
13934         (aarch64_builtin_support_vector_misalignment): Always return false
13935         when misalignment is unknown.
13937 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13939         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
13940         this function to return false if the definition used by the swap
13941         instruction is artificial, or if the memory address from which the
13942         constant value is loaded is not represented by a base address held
13943         in a register or if the base address register is a frame or stack
13944         pointer.  Additionally, return false if the base address of the
13945         loaded constant is a SYMBOL_REF but is not considered to be a
13946         constant.
13947         (replace_swapped_load_constant): New function.
13948         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
13949         loaded constant vector with a load of a swapped constant vector.
13951 2017-09-27  Carl Love  <cel@us.ibm.com>
13953         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13954         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13955         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13956         fctiw instruction.
13958 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
13960         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
13961         first, always call autopref_rank_data otherwise.
13963 2017-09-27  Richard Biener  <rguenther@suse.de>
13965         * graphite-scop-detection.c (find_scop_parameters): Move
13966         loop bound handling ...
13967         (gather_bbs::before_dom_children): ... here, avoiding the need
13968         to build scop_info->loop_nest.
13969         (record_loop_in_sese): Remove.
13970         * sese.h (sese_info_t::loop_nest): Remove.
13971         * sese.c (new_sese_info): Do not allocate loop_nest.
13972         (free_sese_info): Do not free loop_nest.
13974 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
13976         PR c++/82159
13977         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
13978         lhs from calls if the lhs has addressable type.
13980 2017-09-27  Richard Biener  <rguenther@suse.de>
13982         * graphite.h (scop::max_alias_set): New member.
13983         * graphite-scop-detection.c: Remove references to non-existing
13984         --param in comments.
13985         (build_alias_sets): Record the maximum alias set used for drs.
13986         (build_scops): Support zero as unlimited for
13987         --param graphite-max-arrays-per-scop.
13988         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
13989         and inline into ...
13990         (build_poly_sr_1): ... here.  Compute alias set based on the
13991         maximum alias set used for drs rather than
13992         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
13994 2017-09-27  Richard Biener  <rguenther@suse.de>
13996         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
13997         --param loop-block-tile-size=0 to disable tiling.
13999 2017-09-27  Richard Biener  <rguenther@suse.de>
14001         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14002         (graphite-max-nb-scop-params): Document special value zero.
14003         * domwalk.h (dom_walker::STOP): New symbolical constant.
14004         (dom_walker::dom_walker): Add optional parameter for bb to
14005         RPO mapping.
14006         (dom_walker::~dom_walker): Declare.
14007         (dom_walker::before_dom_children): Document STOP return value.
14008         (dom_walker::m_user_bb_to_rpo): New member.
14009         (dom_walker::m_bb_to_rpo): Likewise.
14010         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14011         mapping here if not provided by the user.
14012         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14013         provided by the user.
14014         (dom_walker::STOP): Define.
14015         (dom_walker::walk): Do not compute bb to RPO mapping here.
14016         Support STOP return value from before_dom_children to stop
14017         walking.
14018         * graphite-optimize-isl.c (optimize_isl): If the schedule
14019         is the same still generate code if -fgraphite-identity
14020         or -floop-parallelize-all are given.
14021         * graphite-scop-detection.c: Include cfganal.h.
14022         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14023         mapping.
14024         (gather_bbs::before_dom_children): Return STOP for BBs
14025         not in the region.
14026         (build_scops): Compute bb to RPO mapping and pass it to
14027         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14028         as not limiting the number of params.
14029         * graphite.c (graphite_initialize): Remove limit on the
14030         number of basic-blocks in a function.
14031         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14032         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14033         default value of 10.
14035 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14037         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14038         Adjust code to eliminate needing to do the shift right 32-bits
14039         operation after XSCVDPSPN.
14041 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14043         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14044         ((X / Y) != 0 -> X >= Y): Likewise.
14046 2017-09-26  Carl Love  <cel@us.ibm.com>
14048         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14049         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14050         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14051         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14052         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14053         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14054         definitions and overloading.
14055         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14056         statement for P9V_BUILTIN_XST_LEN_R.
14057         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14058         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14059         define_expand and define_insn for the instructions and builtins.
14060         * doc/extend.texi: Update the built-in documentation file for the new
14061         built-in functions.
14062         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14063         define_insn for the instructions
14065 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14067         PR target/39570
14068         * gcc/config/netbsd-protos.h: New file.
14069         * gcc/config/netbsd.c: New file.
14070         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14071         * gcc/config/t-netbsd: New file.
14072         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14073         (tmake_file) Add t-netbsd.
14074         (extra_objs) Add netbsd.o.
14076 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14078         PR fortran/82143
14079         PR fortran/82324
14080         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14082 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14084         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14085         sign extension from a vector register to a GPR by doing a 32-bit
14086         direct move and then an EXTSW.
14087         (extendsi<mode>2 splitter): Likewise.
14088         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14089         right or vector extract after doing XSCVDPSPN.  Use
14090         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14091         the GPRs.
14092         (movdi_from_sf_zero_ext): Likewise.
14093         (reload_gpr_from_vsxsf): Likewise.
14094         (p8_mfvsrd_4_disf): Delete, no longer used.
14095         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14096         then needing to move the SFmode to a GPR to use the XSCVDPSP
14097         instruction instead of FRSP and XSCVDPSPN.
14098         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14099         it is adjacent to the other XSCVSPDP insns.
14100         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14101         SFmode to be in traditional Altivec registers.
14102         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14103         (vsx_xscvspdpn): Likewise.
14104         (vsx_xscvspdpn_scalar): Likewise.
14106 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14108         * tree-sra.c (compare_access_positions): Put integral types first,
14109         stabilize sorting of integral types, remove conditions putting
14110         non-full-precision integers last.
14111         (sort_and_splice_var_accesses): Disable scalarization if a
14112         non-integert would be represented by a non-full-precision integer.
14114 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14116         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14117         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14118         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14119         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14120         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14121         conditionals inside the function instead of around it.  Call
14122         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14123         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14125 2017-09-26  Richard Biener  <rguenther@suse.de>
14127         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14128         fold in ...
14129         (scop_detection::build_scop_breadth): ... this.  Removed.
14130         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14131         (scop_detection::harmful_stmt_in_bb): Likewise.
14132         (scop_detection::graphite_can_represent_stmt): Likewise.
14133         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14134         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14135         (scop_detection::can_represent_loop_1): ... this.  Removed.
14136         (scop_detection::harmful_loop_in_region): Simplify after inlining
14137         the above and remove more quadraticness.
14138         (build_scops): Adjust.
14139         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14140         quadraticness.
14142 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14144         PR target/82267
14145         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14146         REGNO (base) == SP_REG if base is a REG.
14148         PR middle-end/35691
14149         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14150         if it is different SSA_NAME.
14151         (optimize_range_tests_cmp_bitwise): New function.
14152         (optimize_range_tests): Call it.
14154 2017-09-26  Richard Biener  <rguenther@suse.de>
14156         PR tree-optimization/82321
14157         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14158         for the def being inside the loop.
14160 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14162         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14163         assembler output.
14164         * config/s390/s390-builtins.def: Fix constraint on op4.
14166 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14168         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14169         independent expanders.
14170         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14171         ("vec_ordered", "vec_unordered"): New expanders.
14173 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14175         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14176         for SFmode.
14178 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14180         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14181         vec_unpacks_lo_v16qi.
14182         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14184 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14186         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14187         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14188         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14190 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14192         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14193         predicate.
14194         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14195         and V16QI.
14196         ("*vec_slb<mode>"): New insn pattern.
14197         ("vec_shr_<mode>"): New expander.
14198         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14199         and force the shift count operand to V16QImode.
14200         ("vec_srb<mode>"): Set shift count mode to V16QI.
14202 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14204         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14205         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14206         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14208 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14210         PR target/82175
14211         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14213 2017-09-26  Richard Biener  <rguenther@suse.de>
14215         PR tree-optimization/82320
14216         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14217         isn't a change.
14219 2017-09-25  Jeff Law  <law@redhat.com>
14221         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14222         prototype for new argument.
14223         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14224         mostly extracted from rs6000_emit_allocate_stack.
14225         (rs6000_emit_probe_stack_range_stack_clash): New function.
14226         (rs6000_emit_allocate_stack): Call
14227         rs6000_emit_probe_stack_range_stack_clash as needed.
14228         (rs6000_emit_probe_stack_range): Add additional argument
14229         to call to gen_probe_stack_range{si,di}.
14230         (output_probe_stack_range): New.
14231         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14232         (output_probe_stack_range_stack_clash): New.
14233         (rs6000_emit_prologue): Emit notes into dump file as requested.
14234         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14235         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14236         Add additional operand and pass it to output_probe_stack_range.
14238 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14240         PR tree-optimization/82163
14241         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14242         (checking_verify_loop_closed_ssa): New parameter.
14243         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14244         (check_loop_closed_ssa_stmt): Delete.
14245         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14246         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14247         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14248         changed loops.
14250 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14252         * brig-builtins.def: Treat HSAIL barrier builtins as
14253         setjmp/longjump style functions.
14255 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14257         * target.def (constant_alignment): New hook.
14258         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14259         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14260         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14261         * doc/tm.texi: Regenerate.
14262         * targhooks.h (default_constant_alignment): Declare.
14263         (constant_alignment_word_strings): Likewise.
14264         * targhooks.c (default_constant_alignment): New function.
14265         (constant_alignment_word_strings): Likewise.
14266         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14267         instead of CONSTANT_ALIGNMENT.
14268         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14269         (force_const_mem): Likewise.
14270         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14271         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14272         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14273         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14274         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14275         definition.
14276         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14277         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14278         constant_alignment_word_strings.
14279         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14280         (CONSTANT_ALIGNMENT): Likewise.
14281         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14282         (arm_constant_alignment): New function.
14283         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14284         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14285         constant_alignment_word_strings.
14286         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14287         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14288         constant_alignment_word_strings.
14289         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14290         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14291         (cris_constant_alignment): New function.
14292         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14293         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14294         (epiphany_constant_alignment): New function.
14295         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14296         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14297         constant_alignment_word_strings.
14298         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14299         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14300         constant_alignment_word_strings.
14301         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14302         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14303         constant_alignment_word_strings.
14304         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14305         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14306         * config/i386/i386.c (ix86_constant_alignment): Make static.
14307         Use the same interface as the target hook.
14308         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14309         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14310         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14311         constant_alignment_word_strings.
14312         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14313         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14314         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14315         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14316         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14317         constant_alignment_word_strings.
14318         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14319         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14320         constant_alignment_word_strings.
14321         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14322         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14323         constant_alignment_word_strings.
14324         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14325         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14326         New function.
14327         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14328         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14329         * config/mips/mips.c (mips_constant_alignment): New function.
14330         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14331         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14332         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14333         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14334         (mmix_constant_alignment): Make static.  Use the same interface
14335         as the target hook.
14336         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14337         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14338         constant_alignment_word_strings.
14339         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14340         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14341         constant_alignment_word_strings.
14342         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14343         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14344         constant_alignment_word_strings.
14345         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14346         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14347         (rs6000_constant_alignment): New function.
14348         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14349         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14350         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14351         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14352         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14353         (rs6000_constant_alignment): New function.
14354         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14355         * config/s390/s390.c (s390_constant_alignment): New function.
14356         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14357         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14358         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14359         constant_alignment_word_strings.
14360         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14361         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14362         (sparc_constant_alignment): New function.
14363         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14364         * config/spu/spu.c (spu_constant_alignment): New function.
14365         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14366         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14367         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14368         constant_alignment_word_strings.
14369         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14370         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14371         constant_alignment_word_strings.
14372         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14373         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14374         constant_alignment_word_strings.
14375         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14376         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14377         (visium_constant_alignment): New function.
14378         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14379         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14380         (xtensa_constant_alignment): New function.
14381         * system.h (CONSTANT_ALIGNMENT): Poison.
14383 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14385         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14386         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14387         (rs6000_builtin_valid_without_lhs): New helper function.
14388         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14389         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14391 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14393         * target.h (vec_perm_indices): Use unsigned short rather than
14394         unsigned char.
14395         (auto_vec_perm_indices): Likewise.
14396         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14397         Use unsigned int rather than unsigned char.
14398         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14400 2017-09-25  Richard Biener  <rguenther@suse.de>
14402         * cfgloop.h (sort_sibling_loops): Declare.
14403         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14404         (sort_sibling_loops): New function sorting the sibling loop list
14405         in RPO order.
14406         * graphite.c (graphite_transform_loops): Sort sibling loops.
14408 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14410         * target.def (vec_perm_const_ok): Change sel parameter to
14411         vec_perm_indices.
14412         * optabs-query.c (can_vec_perm_p): Update accordingly.
14413         * doc/tm.texi: Regenerate.
14414         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14415         auto_vec_perm_indices and remove separate nelt field.
14416         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14417         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14418         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14419         (aarch64_expand_vec_perm_const): Update accordingly.
14420         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14421         to vec_perm_indices.
14422         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14423         auto_vec_perm_indices and remove separate nelt field.
14424         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14425         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14426         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14427         accordingly.
14428         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14429         to vec_perm_indices.
14430         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14431         sel to vec_perm_indices.
14432         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14433         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14434         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14435         Likewise.
14436         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14437         Likewise.
14439 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14441         PR debug/82155
14442         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14443         on the FUNCTION_DECL function context if it has a DIE that is a
14444         declaration.
14446 2017-09-25  Richard Biener  <rguenther@suse.de>
14448         PR tree-optimization/82285
14449         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14450         enumeral types.
14452 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14454         PR target/80035
14455         PR target/81069
14456         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14457         noreturn function.
14459 2017-09-25  Richard Biener  <rguenther@suse.de>
14461         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14462         ISL errors other than isl_error_quota happen.  Dump if the
14463         schedule is the same.
14464         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14465         errors instead of aborting inside ISL.
14467 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14469         PR target/80556
14470         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14471         of libgcc_eh for m64.
14472         * config/i386/darwin64.h: Likewise.
14474 2017-09-25  Richard Biener  <rguenther@suse.de>
14476         PR middle-end/82144
14477         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14478         attribute for incomplete types nor twice for complete ones.
14480 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14482         PR target/82267
14483         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14484         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14485         register.
14487 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14489         PR bootstrap/82306
14490         * config/i386/i386.opt (mprefer-avx256): Use
14491         ix86_target_flags variable.
14492         * config/i386/i386.c (ix86_target_string): Move
14493         -mprefer-avx256 to flag2_opts.
14495 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14497         PR middle-end/35691
14498         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14499         and x != -1 | y != -1 into (x & y) != -1.
14501 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14503         * config.gcc: Add new case statement to set
14504         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14505         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14506         s390x-*-linux* case statements.   Added aarch64 to the list of
14507         supported architectures.
14509 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14511         PR tree-optimization/82289
14512         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14513         STMT_VINFO_RELEVANT_P.
14515 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14516             Alan Hayward  <alan.hayward@arm.com>
14517             David Sherwood  <david.sherwood@arm.com>
14519         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14520         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14521         (extract_range_from_binary_expr_1): Don't call
14522         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14524 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14525             Alan Hayward  <alan.hayward@arm.com>
14526             David Sherwood  <david.sherwood@arm.com>
14528         * target.def (preferred_vector_alignment): New hook.
14529         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14530         hook.
14531         * doc/tm.texi: Regenerate.
14532         * targhooks.h (default_preferred_vector_alignment): Declare.
14533         * targhooks.c (default_preferred_vector_alignment): New function.
14534         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14535         Expand commentary.
14536         (DR_TARGET_ALIGNMENT): New macro.
14537         (aligned_access_p): Update commentary.
14538         (vect_known_alignment_in_bytes): New function.
14539         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14540         function.
14541         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14542         Calculate the misalignment based on the target alignment rather than
14543         the vector size.
14544         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14545         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14546         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14547         the target alignment, rather than masking the element misalignment
14548         with the number of elements in a vector.  Also use the target
14549         alignment when calculating the maximum number of peels.
14550         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14551         instead of TYPE_ALIGN_UNIT.
14552         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14553         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14554         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14555         (vect_create_data_ref_ptr): Likewise.
14556         (vect_setup_realignment): Realign by ANDing with
14557         -DR_TARGET_MISALIGNMENT.
14558         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14559         the number of peels based on DR_TARGET_ALIGNMENT.
14560         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14561         with the guaranteed alignment boundary when deciding whether
14562         overrun is OK.
14563         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14564         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14565         (ensure_base_align): Remove stmt_info parameter.  Get the
14566         target base alignment from DR_TARGET_ALIGNMENT.
14567         (vectorizable_store): Update call accordingly.   Interpret
14568         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14569         TYPE_ALIGN_UNIT.
14570         (vectorizable_load): Likewise.
14572 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14573             Alan Hayward  <alan.hayward@arm.com>
14574             David Sherwood  <david.sherwood@arm.com>
14576         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14577         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14578         (vect_enhance_data_refs_alignment): Likewise.
14580 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14582         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14583         error.  Only quit immediately if parsing is complete.
14584         (BEGIN): Initialize fatal_err and parse_done.
14585         (begin fpu, end fpu): Check number of arguments.
14586         (begin arch, end arch): Likewise.
14587         (begin cpu, end cpu): Likewise.
14588         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14589         (optalias): Likewise.
14591 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14593         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14594         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14595         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14596         * config/arm/arm.c (arm_option_override): Use lower case for feature
14597         bit names.
14598         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14599         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14600         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14601         (isa_pfx): Delete.
14602         (print_isa_bits_for): New function.
14603         (gen_isa): New function.
14604         (gen_comm_data): Use print_isa_bits_for.
14605         (define feature): New keyword.
14606         (define fgroup): New keyword.
14607         * config/arm/t-arm (TM_H): Remove.
14608         (GTM_H): Add arm-isa.h.
14609         (arm-isa.h): Add rule to generate file.
14610         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14611         case for feature bit names.
14613 2017-09-22  Richard Biener  <rguenther@suse.de>
14615         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14616         single caller.
14617         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14618         print of no dependency loops ...
14619         * graphite.c (graphite_transform_loops): ... here.
14620         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14621         loops.
14622         (same_close_phi_node, remove_duplicate_close_phi,
14623         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14624         (canonicalize_loop_closed_ssa): ... here and simplify.
14625         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14626         (optimize_isl): Use dump_printf_loc to tell when we stopped
14627         optimizing because of an ISL timeout.
14629 2017-09-22  Richard Biener  <rguenther@suse.de>
14631         PR tree-optimization/82291
14632         * tree-if-conv.c (predicate_mem_writes): Make sure to
14633         remove writes in blocks predicated with false.
14635 2017-09-22  Richard Biener  <rguenther@suse.de>
14637         * sese.c: Include cfganal.h.
14638         (if_region_set_false_region): Remove.
14639         (create_if_region_on_edge): Likewise.
14640         (move_sese_in_condition): Re-implement without destroying
14641         dominators.
14643 2017-09-22  Richard Biener  <rguenther@suse.de>
14645         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14646         Verify both BBs contain loop PHI nodes before dispatching to
14647         copy_loop_phi_args.
14648         (graphite_regenerate_ast_isl): Do not recompute dominators,
14649         do not verify three times.  Restructure for clarity.
14650         * graphite-scop-detection.c (same_close_phi_node,
14651         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14652         defined_in_loop_p, canonicalize_loop_closed_ssa,
14653         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14654         checking and SSA rewrite, move to ...
14655         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14656         (graphite_initialize): Do not pass in ctx, do not reset the
14657         SCEV cache, compute only dominators.
14658         (graphite_transform_loops): Allocate ISL ctx after
14659         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14660         Maintain post-dominators only around build_scops.
14661         * sese.c (if_region_set_false_region): Make static.  Free
14662         and recompute dominators.
14663         (move_sese_in_condition): Assert we don't get called with
14664         post-dominators computed.
14665         * sese.h (if_region_set_false_region): Remove.
14667 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14669         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14670         mode attribute for TARGET_AVX512VL.
14672 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14674         * config/i386/i386.opt (mprefer-avx256): New option.
14675         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14676         to flag_opts.
14677         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14678         for TARGET_PREFER_AVX256.
14679         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14681 2017-09-21  Jeff Law  <law@redhat.com>
14683         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14684         Fix dump output if the only stack space is for pushed registers.
14686 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14688         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14689         of insn_cost.
14691 2017-09-21  Martin Sebor  <msebor@redhat.com>
14693         PR c/81882
14694         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14695         code (in C++) or code that triggers warnings.
14697 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14699         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14701 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14703         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14704         * sched-rgn.c: Ditto.
14705         * sel-sched-ir.c: Ditto.
14707 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14709         * toplev.h (set_random_seed): Adjust return type.
14710         * toplev.c (init_local_tick): Move eager initialization of random_seed
14711         to get_random_seed.  Adjust comment.
14712         (init_random_seed): Inline to get_random_seed, delete.
14713         (get_random_seed): Initialize random_seed lazily.
14714         (set_random_seed): Do not return previous value.
14715         (print_switch_value): Do not call get_random_seed.
14717 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14719         * cgraph.c (delete_function_version): New, broken out from...
14720         (cgraph_node::delete_function_version): ...here.  Rename to
14721         cgraph_node::delete_function_version_by_decl.  Update all uses.
14722         (cgraph_node::remove): Call delete_function_version.
14724 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14726         PR sanitizer/81715
14727         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14728         VAR_DECLs to which addressable non-volatile parameters are mapped
14729         and for id->retvar after the return value assignment.  Clear
14730         id->retval and id->retbnd after inlining.
14732 2017-09-21  Richard Biener  <rguenther@suse.de>
14734         PR tree-optimization/82276
14735         PR tree-optimization/82244
14736         * tree-vrp.c (build_assert_expr_for): Set
14737         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14738         has it set.
14739         (remove_range_assertions): Revert earlier change.
14741 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14743         PR target/71951
14744         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14746 2017-09-21  Richard Biener  <rguenther@suse.de>
14748         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14749         Restore valid IL after code generation errors.
14750         * graphite.c (graphite_transform_loops): Diagnose code
14751         generation issues as MSG_MISSED_OPTIMIZATION and continue
14752         with processing SCOPs.
14754 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14755             Alan Hayward  <alan.hayward@arm.com>
14756             David Sherwood  <david.sherwood@arm.com>
14758         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14759         rather than choosing between plus_constant and gen_rtx_<CODE>.
14760         * expr.c (emit_push_insn): Likewise.
14761         (expand_expr_real_2): Likewise.
14763 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14764             Alan Hayward  <alan.hayward@arm.com>
14765             David Sherwood  <david.sherwood@arm.com>
14767         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14769 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14770             Alan Hayward  <alan.hayward@arm.com>
14771             David Sherwood  <david.sherwood@arm.com>
14773         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14774         calling tree_to_uhwi.
14776 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14777             Alan Hayward  <alan.hayward@arm.com>
14778             David Sherwood  <david.sherwood@arm.com>
14780         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14781         INTEGER_CST rather than a negative test for ADDR_EXPR.
14783 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14784             Alan Hayward  <alan.hayward@arm.com>
14785             David Sherwood  <david.sherwood@arm.com>
14787         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14788         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14790 2017-09-21  Richard Biener  <rguenther@suse.de>
14792         PR tree-optimization/71351
14793         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14794         graphite_create_new_loop_guard): Remove, fold remaining parts
14795         into caller ...
14796         (translate_isl_ast_node_for): ... here and simplify.
14798 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14800         PR target/82260
14801         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14802         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14803         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14804         alternative always use QI mode, for -Os imov (=R,R) alternative
14805         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14806         ignore -Os.
14808 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14809             Jeff Law  <law@redhat.com>
14811         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14812         (allocate_stack_space): New function, partially extracted from
14813         s390_emit_prologue.
14814         (s390_emit_prologue): Track offset to most recent stack probe.
14815         Code to allocate space moved into allocate_stack_space.
14816         Dump actions when no stack is allocated.
14817         (s390_prologue_plus_offset): New function.
14818         (s390_emit_stack_probe): Likewise.
14820 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14822         * common.opt (Wa, Wl, Wp, g, gz=): Add
14823         RejectNegative.
14824         (gno-column-info): Remove.
14825         (gcolumn-info): Drop RejectNegative.
14826         (gno-): New prefix.
14827         (gno-record-gcc-switches): Remove.
14828         (grecord-gcc-switches): Drop RejectNegative.
14829         (gno-split-dwarf): Remove.
14830         (gsplit-dwarf): Drop RejectNegative.
14831         (gno-strict-dwarf): Remove.
14832         (gstrict-dwarf): Drop RejectNegative.
14833         * config/darwin.opt (gfull, gused): Add RejectNegative.
14834         * dwarf2out.c (gen_producer_string): Drop
14835         gno-record-gcc-switches handler.
14836         * optc-gen.awk: Add g to prefixes with negative forms.
14837         * opts-common.c (remapping_prefix_p): New.
14838         (find_opt): Check it.
14839         (generate_canonical_option): Test g prefix.
14840         (option_map): Add -gno- mapping.
14841         (add_misspelling_candidates): Check remapping_prefix_p.
14843 2017-09-20  Jeff Law  <law@redhat.com>
14845         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14846         thinko in stack clash protection support.
14848         * explow.c (compute_stack_clash_protection_loop_data): Use
14849         CONST_INT_P instead of explicit test.  Verify object is a
14850         CONST_INT_P before looking at INTVAL.
14851         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14852         instead of explicit test.
14854 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
14856         PR target/77687
14857         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14858         address instead of to r1 and r11.
14860 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
14862         * config.gcc: Support "knm".
14863         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14864         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14865         PROCESSOR_KNM.
14866         * config/i386/i386.c (m_KNM): Define.
14867         (processor_target_table): Add "knm".
14868         (PTA_KNM): Define.
14869         (ix86_option_override_internal): Add "knm".
14870         (ix86_issue_rate): Add PROCESSOR_KNM.
14871         (ix86_adjust_cost): Ditto.
14872         (ia32_multipass_dfa_lookahead): Ditto.
14873         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
14874         (fold_builtin_cpu): Add M_INTEL_KNM.
14875         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
14876         (processor_type): Add PROCESSOR_KNM.
14877         * config/i386/x86-tune.def: Add m_KNM.
14878         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
14880 2017-09-20  Richard Biener  <rguenther@suse.de>
14882         PR tree-optimization/80213
14883         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
14884         are allowed in empty BBs as well.
14885         (canonicalize_loop_closed_ssa): Also look for other complex
14886         edges.
14887         (scop_detection::get_sese): Include the loop-closed PHI block
14888         in loop SESEs.
14889         (scop_detection::merge_sese): Remove code adding extra blocks.
14890         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
14891         (build_scops): Assert the final returned scop is invalid.
14893 2017-09-20  Richard Biener  <rguenther@suse.de>
14895         PR tree-optimization/82264
14896         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
14897         for GIMPLE_CONDs.
14898         (vn_phi_lookup): Likewise.
14899         (vn_phi_insert): Likewise.
14901 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
14903         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
14904         that fits into uhwi or shwi, add DW_AT_const_value regardless
14905         of early_dwarf without going through RTL, using add_AT_unsigned
14906         or add_AT_int.
14908         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
14909         (DEBUG_LTO_ABBREV_SECTION): Likewise.
14910         (DEBUG_LTO_MACINFO_SECTION): Likewise.
14911         (DEBUG_MACRO_SECTION): Likewise.
14912         (DEBUG_LTO_MACRO_SECTION): Likewise.
14913         (DEBUG_STR_DWO_SECTION): Likewise.
14914         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
14915         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
14916         (DEBUG_LTO_DWO_LINE_SECTION): Define.
14917         (DEBUG_LTO_LINE_STR_SECTION): Define.
14918         (init_sections_and_labels): Initialize debug_line_str_section
14919         variable.  Initialize debug_loc_section for -gdwarf-5 to
14920         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
14922 2017-09-20  Richard Biener  <rguenther@suse.de>
14924         * graphite-sese-to-poly.c (extract_affine): Properly handle
14925         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
14927 2017-09-20  Richard Biener  <rguenther@suse.de>
14929         PR tree-optimization/81373
14930         * graphite-scop-detection.c (build_cross_bb_scalars_def):
14931         Force SESE live-out defs to be handled even if they are
14932         scev_analyzable_p.
14934 2017-09-19  Jeff Law  <law@redhat.com>
14936         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
14937         nothing for stack adjustments with REG_STACK_CHECK.
14938         * sched-deps.c (parse_add_or_inc): Reject insns with
14939         REG_STACK_CHECK from dependency breaking.
14940         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
14941         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
14942         * reg-notes.def (STACK_CHECK): New note.
14944         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
14945         (ix86_expand_prologue): Dump stack clash info as needed.
14946         Call ix86_adjust_stack_and_probe_stack_clash as needed.
14948         * function.c (dump_stack_clash_frame_info): New function.
14949         * function.h (dump_stack_clash_frame_info): Prototype.
14950         (enum stack_clash_probes): New enum.
14952         * config/alpha/alpha.c (alpha_expand_prologue): Also check
14953         flag_stack_clash_protection.
14954         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
14955         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
14956         (arm_frame_pointer_required): Likewise.
14957         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
14958         (ia64_expand_prologue): Likewise.
14959         * config/mips/mips.c (mips_expand_prologue): Likewise.
14960         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
14961         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
14962         (sparc_flat_expand_prologue): Likewise.
14963         * config/spu/spu.c (spu_expand_prologue): Likewise.
14965         * explow.c: Include "params.h".
14966         (anti_adjust_stack_and_probe_stack_clash): New function.
14967         (get_stack_check_protect): Likewise.
14968         (compute_stack_clash_protection_loop_data): Likewise.
14969         (emit_stack_clash_protection_loop_start): Likewise.
14970         (emit_stack_clash_protection_loop_end): Likewise.
14971         (allocate_dynamic_stack_space): Use get_stack_check_protect.
14972         Use anti_adjust_stack_and_probe_stack_clash.
14973         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
14974         (emit_stack_clash_protection_loop_start): Likewise.
14975         (emit_stack_clash_protection_loop_end): Likewise.
14976         * rtl.h (get_stack_check_protect): Prototype.
14977         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
14978         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
14979         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
14980         Prototype.
14981         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
14982         Add @hook.
14983         * doc/tm.texi: Rebuilt.
14984         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
14985         get_stack_check_protect.
14986         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
14987         * config/arm/arm.c (arm_expand_prologue): Likewise.
14988         (arm_frame_pointer_required): Likewise.
14989         * config/i386/i386.c (ix86_expand_prologue): Likewise.
14990         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
14991         * config/mips/mips.c (mips_expand_prologue): Likewise.
14992         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
14993         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
14994         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
14995         (sparc_flat_expand_prologue): Likewise.
14997         * common.opt (-fstack-clash-protection): New option.
14998         * flag-types.h (enum stack_check_type): Note difference between
14999         -fstack-check= and -fstack-clash-protection.
15000         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15001         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15002         * toplev.c (process_options): Issue warnings/errors for cases
15003         not handled with -fstack-clash-protection.
15004         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15005         (-fstack-check): Note additional problem with -fstack-check=generic.
15006         Note that -fstack-check is primarily for Ada and refer users
15007         to -fstack-clash-protection for stack-clash-protection.
15008         Document new params for stack clash protection.
15010 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15012         * config/i386/i386.c (ix86_split_long_move): Do not handle
15013         address used for LEA in a special way.
15015 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15017         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15019 2017-09-19  Martin Sebor  <msebor@redhat.com>
15021         PR c/81854
15022         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15023         of incompatible types.
15025 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15027         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15028         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15029         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15030         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15032 2017-09-19  Richard Biener  <rguenther@suse.de>
15034         PR tree-optimization/82244
15035         * tree-vrp.c (remove_range_assertions): Do not propagate
15036         a constant to abnormals but replace the assert with a copy.
15038 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15040         PR rtl-optimization/57878
15041         PR rtl-optimization/68988
15042         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15043         avoidance test involving non_reload_pseudos.  Move frequency test
15044         below the general fragmentation avoidance test.
15046 2017-09-19  Richard Biener  <rguenther@suse.de>
15048         PR tree-optimization/69728
15049         * graphite-sese-to-poly.c (schedule_error): New global.
15050         (add_loop_schedule): Handle empty domain by failing the
15051         schedule.
15052         (build_original_schedule): Handle schedule_error.
15054 2017-09-19  Richard Biener  <rguenther@suse.de>
15056         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15057         Do not iterate to sibling loops but only to siblings of inner
15058         loops.
15060 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15062         PR target/81613
15063         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15064         that the registers are different.
15066 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15068         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15069         to processor_model and "amdfam17h" to arch_names_table.
15070         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15072 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15074         PR c/82234
15075         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15077 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15078             Alan Hayward  <alan.hayward@arm.com>
15079             David Sherwood  <david.sherwood@arm.com>
15081         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15082         with a vec_info *.
15083         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15084         accordingly.
15085         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15086         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15087         vect_schedule_slp_instance.
15088         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15089         Update call to vect_slp_analyze_node_operations.  Simplify return
15090         value.
15091         (vect_slp_analyze_bb_1): Update call accordingly.
15092         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15093         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15094         (vect_schedule_slp): Update call accordingly.
15096 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15097             Alan Hayward  <alan.hayward@arm.com>
15098             David Sherwood  <david.sherwood@arm.com>
15100         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15101         with types that aren't in fact scalar.
15103 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15105         * tree-vect-slp.c (vect_record_max_nunits): New function,
15106         split out from...
15107         (vect_build_slp_tree_1): ...here.
15108         (vect_build_slp_tree_2): Call it for phis too.
15110 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15112         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15113         to vect_get_vec_def_for_operand when getting the mask operand.
15115 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15116             Alan Hayward  <alan.hayward@arm.com>
15117             David Sherwood  <david.sherwood@arm.com>
15119         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15120         bitstart.
15122 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15123             Alan Hayward  <alan.hayward@arm.com>
15124             David Sherwood  <david.sherwood@arm.com>
15126         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15127         calculation for vector booleans.
15129 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15130             Alan Hayward  <alan.hayward@arm.com>
15131             David Sherwood  <david.sherwood@arm.com>
15133         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15134         split out from...
15135         (vect_transform_stmt): ...here.
15136         (vect_analyze_stmt): Use it instead of calling
15137         vectorizable_live_operation directly.
15139 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15141         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15142         non-SIMT targets in acc vector loops.
15144 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15146         * configure.ac: Add arc and check if assembler supports gdwarf2.
15147         * configure: Regenerate.
15149 2017-09-18  Richard Biener  <rguenther@suse.de>
15151         PR tree-optimization/82220
15152         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15153         epilogue niters from the min_profitable_iters compute.
15155 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15157         PR target/82145
15158         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15159         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15160         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15162 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15164         PR target/81361
15165         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15166         switching to a new text section.
15168 2017-09-18  Richard Biener  <rguenther@suse.de>
15170         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15171         Simplify.
15172         (build_alias_set): Reject aliases with no access function.
15174 2017-09-18  Richard Biener  <rguenther@suse.de>
15176         PR tree-optimization/79622
15177         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15178         handle PHIs.
15179         (build_cross_bb_scalars_use): Likewise.
15181 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15183         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15185 2017-09-18  Alan Modra  <amodra@gmail.com>
15187         PR target/81996
15188         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15189         stack_pointer_rtx for count 0.  Update comments.  Break up
15190         large rtl expression.
15192 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15194         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15195         Increase to 20 bytes.
15196         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15197         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15198         or avx version of the stub.
15200 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15202         PR target/82166
15203         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15204         compute the minimum stack alignment.  Also update preferred stack
15205         boundary for leaf functions.
15207 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15209         PR tree-optimization/82228
15210         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15211         of ncopies.
15213 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15215         * common/config/nds32/nds32-common.c
15216         (nds32_option_optimization_table): Refine formatting.
15217         (nds32_option_optimization_table): Use -fsched-pressure and
15218         -fomit-frame-pointer for specific optimization level.
15220 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15222         * config/nds32/nds32.c: Refine formatting and comments.
15223         * config/nds32/nds32.h: Likewise.
15224         * config/nds32/nds32.md: Likewise.
15225         * config/nds32/nds32-cost.c: Likewise.
15226         * config/nds32/nds32-isr.c: Likewise.
15227         * config/nds32/nds32-md-auxiliary.c: Likewise.
15228         * config/nds32/nds32-multiple.md: Likewise.
15229         * config/nds32/nds32-predicates.c: Likewise.
15231 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15232             Jakub Jelinek  <jakub@redhat.com>
15234         Add support for -std=c++2a.
15235         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15236         or -std=gnu+2a.
15237         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15239 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15241         PR target/82066
15242         * doc/extend.texi (Common Function Attributes): Add 
15243         references to ARM, AArch64, and S/390 specific attributes.
15244         (Function Specific Option Pragmas): Add AArch64 and S/390
15245         to list of back ends that support the target pragma.
15247 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15249         * doc/standards.texi: Fix C++17 description.  Update URLs for
15250         C++11 & 14.
15252 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15254         * common.opt (Wcast-align=strict): New warning option.
15255         * doc/invoke.texi: Document -Wcast-align=strict. 
15257 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15259         * cgraph.h (cgraph_thunk_info): Add comments.
15260         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15261         assert for VIRTUAL_* arguments stricter.
15263 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15265         PR tree-optimization/71026
15266         * match.pd: Move RDIV patterns from fold-const.c
15267         * fold-const.c (distribute_real_division): Removed.
15268         (fold_binary_loc): Remove calls to distribute_real_divison.
15270 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15272         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15273         c++1z and gnu++1z as deprecated.  Change other references to
15274         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15275         Change -Wc++1z-compat to -Wc++17-compat.
15276         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15277         * dwarf2out.c (highest_c_language): Handle C++17.
15278         (gen_compile_unit_die): Likewise.
15280 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15282         PR rtl-optimization/82192
15283         * combine.c (make_extraction): Don't look through non-paradoxical
15284         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15285         inner's mode.
15287 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15288             Alan Hayward  <alan.hayward@arm.com>
15289             David Sherwood  <david.sherwood@arm.com>
15291         * target.def (function_arg_offset): New hook.
15292         * targhooks.h (default_function_arg_offset): Declare.
15293         * targhooks.c (default_function_arg_offset): New function.
15294         * function.c (locate_and_pad_parm): Use
15295         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15296         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15297         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15298         * doc/tm.texi: Regenerate.
15299         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15300         * config/spu/spu.c (spu_function_arg_offset): New function.
15301         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15302         * system.h (FUNCTION_ARG_OFFSET): Poison.
15304 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15305             Alan Hayard  <alan.hayward@arm.com>
15306             David Sherwood  <david.sherwood@arm.com>
15308         * target.def (truly_noop_truncation): New hook.
15309         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15310         than TRULY_NOOP_TRUNCATION.
15311         * hooks.h (hook_bool_uint_uint_true): Declare.
15312         * hooks.c (hook_bool_uint_uint_true): New function.
15313         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15314         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15315         * doc/tm.texi: Regenerate.
15316         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15317         rather than TRULY_NOOP_TRUNCATION in comments.
15318         (simplify_comparison): Likewise.
15319         (record_truncated_value): Likewise.
15320         * expmed.c (extract_bit_field_1): Likewise.
15321         (extract_split_bit_field): Likewise.
15322         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15323         instead of TRULY_NOOP_TRUNCATION.
15324         * function.c (assign_parm_setup_block): Likewise.
15325         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15326         * rtlhooks.c: Include target.h.
15327         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15328         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15329         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15330         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15331         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15332         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15333         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15334         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15335         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15336         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15337         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15338         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15339         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15340         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15341         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15342         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15343         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15344         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15345         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15346         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15347         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15348         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15349         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15350         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15351         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15352         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15353         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15354         rather than TRULY_NOOP_TRUNCATION in comments.
15355         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15356         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15357         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15358         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15359         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15360         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15361         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15362         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15363         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15364         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15365         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15366         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15367         rather than TRULY_NOOP_TRUNCATION in comments.
15368         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15369         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15370         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15371         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15372         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15373         TRULY_NOOP_TRUNCATION condition.
15374         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15375         (TRULY_NOOP_TRUNCATION): Delete.
15376         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15377         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15378         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15379         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15380         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15381         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15382         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15383         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15384         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15385         rather than TRULY_NOOP_TRUNCATION in comments.
15386         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15387         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15388         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15389         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15390         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15391         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15393 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15395         PR target/67591
15396         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15397         (*cmp_ior): Likewise.
15398         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15399         (*ior_scc_scc_cmp): Likewise.
15400         (*and_scc_scc): Likewise.
15401         (*and_scc_scc_cmp): Likewise.
15403 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15404             Alan Hayard  <alan.hayward@arm.com>
15405             David Sherwood  <david.sherwood@arm.com>
15407         * target.def (can_change_mode_class): New hook.
15408         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15409         (hard_regno_nregs): Likewise.
15410         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15411         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15412         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15413         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15414         (LOAD_EXTEND_OP): Update accordingly.
15415         * doc/tm.texi: Regenerate.
15416         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15417         CANNOT_CHANGE_MODE_CLASS.
15418         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15419         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15420         * combine.c (simplify_set): Update accordingly.
15421         * emit-rtl.c (validate_subreg): Likewise.
15422         * recog.c (general_operand): Likewise.
15423         * regcprop.c (mode_change_ok): Likewise.
15424         * reload1.c (choose_reload_regs): Likewise.
15425         (inherit_piecemeal_p): Likewise.
15426         * rtlanal.c (simplify_subreg_regno): Likewise.
15427         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15428         instead of CANNOT_CHANGE_MODE_CLASS.
15429         (reload_cse_simplify_operands): Likewise.
15430         * reload.c (push_reload): Use targetm.can_change_mode_class
15431         instead of CANNOT_CHANGE_MODE_CLASS.
15432         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15433         REG_CANNOT_CHANGE_MODE_P.
15434         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15435         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15436         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15437         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15438         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15439         (arm_can_change_mode_class): New function.
15440         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15441         than CANNOT_CHANGE_MODE_CLASS in comments.
15442         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15443         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15444         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15445         (ix86_can_change_mode_class): ...this new function, inverting the
15446         sense of the return value.
15447         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15448         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15449         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15450         (ia64_can_change_mode_class): New function.
15451         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15452         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15453         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15454         (m32c_can_change_mode_class): ...this new function, inverting the
15455         sense of the return value.
15456         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15457         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15458         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15459         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15460         (mips_can_change_mode_class): ...this new function, inverting the
15461         sense of the return value.
15462         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15463         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15464         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15465         (msp430_can_change_mode_class): New function.
15466         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15467         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15468         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15469         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15470         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15471         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15472         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15473         (pa_cannot_change_mode_class): Replace with...
15474         (pa_can_change_mode_class): ...this new function, inverting the
15475         sense of the return value.
15476         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15477         than CANNOT_CHANGE_MODE_CLASS in comments.
15478         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15479         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15480         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15481         (pdp11_cannot_change_mode_class): Replace with...
15482         (pdp11_can_change_mode_class): ...this new function, inverting the
15483         sense of the return value.
15484         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15485         * config/powerpcspe/powerpcspe-protos.h
15486         (rs6000_cannot_change_mode_class_ptr): Delete.
15487         * config/powerpcspe/powerpcspe.c
15488         (rs6000_cannot_change_mode_class_ptr): Delete.
15489         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15490         (rs6000_option_override_internal): Assign to
15491         targetm.can_change_mode_class instead of
15492         rs6000_cannot_change_mode_class_ptr.
15493         (rs6000_cannot_change_mode_class): Replace with...
15494         (rs6000_can_change_mode_class): ...this new function, inverting the
15495         sense of the return value.
15496         (rs6000_debug_cannot_change_mode_class): Replace with...
15497         (rs6000_debug_can_change_mode_class): ...this new function.
15498         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15499         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15500         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15501         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15502         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15503         Delete.
15504         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15505         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15506         (rs6000_option_override_internal): Assign to
15507         targetm.can_change_mode_class instead of
15508         rs6000_cannot_change_mode_class_ptr.
15509         (rs6000_cannot_change_mode_class): Replace with...
15510         (rs6000_can_change_mode_class): ...this new function, inverting the
15511         sense of the return value.
15512         (rs6000_debug_cannot_change_mode_class): Replace with...
15513         (rs6000_debug_can_change_mode_class): ...this new function.
15514         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15515         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15516         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15517         (s390_can_change_mode_class): ...this new function, inverting the
15518         sense of the return value.
15519         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15520         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15521         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15522         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15523         (sh_cannot_change_mode_class): Replace with...
15524         (sh_can_change_mode_class): ...this new function, inverting the
15525         sense of the return value.
15526         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15527         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15528         (sparc_can_change_mode_class): New function.
15529         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15530         * config/spu/spu.c (spu_can_change_mode_class): New function.
15531         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15532         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15533         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15534         (visium_can_change_mode_class): New function.
15535         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15537 2017-09-15  Richard Biener  <rguenther@suse.de>
15539         PR tree-optimization/82217
15540         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15541         but not undefined case.
15543 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15545         PR target/82145
15546         * postreload.c (reload_cse_simplify_operands): Skip
15547         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15549 2017-09-15  Richard Biener  <rguenther@suse.de>
15551         PR tree-optimization/68823
15552         * graphite-scop-detection.c (build_alias_set): If we have a
15553         possible dependence check whether we can handle them by just
15554         looking at the DRs DR_ACCESS_FNs.
15555         (build_scops): If build_alias_set fails, fail the SCOP.
15557 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15559         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15560         to support float128 built-in functions that require the ISA 3.0
15561         hardware.
15562         (BU_FLOAT128_3_HW): Likewise.
15563         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15564         built-in functions.
15565         (FMAF128): Likewise.
15566         (FMAQ): Likewise.
15567         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15568         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15569         floating point instructions.
15570         (rs6000_invalid_builtin): Likewise.
15571         (rs6000_builtin_mask_names): Likewise.
15572         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15573         (RS6000_BTM_FLOAT128_HW): Likewise.
15574         (RS6000_BTM_COMMON): Likewise.
15575         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15576         function.
15577         * doc/extend.texi (RS/6000 built-in functions): Document the
15578         IEEE 128-bit floating point square root and fused multiply-add
15579         built-in functions.
15581 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15583         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15584         reg (r2) isn't in the set of registers defined in the prologue.
15586 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15587             Alan Hayward  <alan.hayward@arm.com>
15588             David Sherwood  <david.sherwood@arm.com>
15590         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15591         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15592         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15593         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15594         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15595         accordingly.
15596         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15597         max_vectorization_factor.
15598         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15600 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15601             Alan Hayward  <alan.hayward@arm.com>
15602             David Sherwood  <david.sherwood@arm.com>
15604         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15605         (vect_worthwhile_without_simd_p): Declare.
15606         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15607         (vectorizable_reduction): Use it.
15608         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15609         (vectorizable_operation): Likewise.
15611 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15612             Alan Hayward  <alan.hayward@arm.com>
15613             David Sherwood  <david.sherwood@arm.com>
15615         * tree-vectorizer.h (vect_get_num_copies): New function.
15616         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15617         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15618         (vectorizable_induction): Likewise.
15619         (vectorizable_live_operation): Likewise.
15620         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15621         (vectorizable_bswap): Likewise.
15622         (vectorizable_call): Likewise.
15623         (vectorizable_conversion): Likewise.
15624         (vectorizable_assignment): Likewise.
15625         (vectorizable_shift): Likewise.
15626         (vectorizable_operation): Likewise.
15627         (vectorizable_store): Likewise.
15628         (vectorizable_load): Likewise.
15629         (vectorizable_condition): Likewise.
15630         (vectorizable_comparison): Likewise.
15631         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15633 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15634             Alan Hayward  <alan.hayward@arm.com>
15635             David Sherwood  <david.sherwood@arm.com>
15637         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15638         of vect_init_vector.
15640 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15641             Alan Hayward  <alan.hayward@arm.com>
15642             David Sherwood  <david.sherwood@arm.com>
15644         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15645         an inline wrapper that provides a location.
15646         (gimple_build_vector): Likewise.
15647         * gimple-fold.c (gimple_build_vector_from_val): New function.
15648         (gimple_build_vector): Likewise.
15649         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15650         functions to build the initial value.  Always return a gimple value.
15651         (get_initial_defs_for_reduction): Likewise.  Only compute
15652         neutral_vec once.
15653         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15654         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15655         (vectorizable_induction): Use gimple_build_vector rather than
15656         vect_init_vector.
15658 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15659             Alan Hayward  <alan.hayward@arm.com>
15660             David Sherwood  <david.sherwood@arm.com>
15662         * target.h (vec_perm_indices): New typedef.
15663         (auto_vec_perm_indices): Likewise.
15664         * optabs-query.h: Include target.h
15665         (can_vec_perm_p): Take a vec_perm_indices *.
15666         * optabs-query.c (can_vec_perm_p): Likewise.
15667         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15668         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15669         * tree-vect-generic.c (lower_vec_perm): Likewise.
15670         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15671         (vect_grouped_load_supported): Likewise.
15672         (vect_shift_permute_load_chain): Likewise.
15673         (vect_permute_store_chain): Use auto_vec_perm_indices.
15674         (vect_permute_load_chain): Likewise.
15675         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15676         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15677         Update uses of can_vec_perm_p.
15678         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15679         mode with a number of elements.  Take a vec_perm_indices *.
15680         (vect_create_epilog_for_reduction): Update accordingly.
15681         Use auto_vec_perm_indices.
15682         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15683         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15684         (vect_transform_slp_perm_load): Likewise.
15685         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15686         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15687         (vect_gen_perm_mask_checked): Likewise.
15688         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15689         (vect_gen_perm_mask_checked): Likewise.
15690         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15691         (vectorizable_store): Likewise.
15692         (vectorizable_load): Likewise.
15693         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15694         (vectorizable_bswap): Likewise.
15696 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15697             Alan Hayward  <alan.hayward@arm.com>
15698             David Sherwood  <david.sherwood@arm.com>
15700         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15701         * tree.c (build_vector): Likewise.
15702         (build_vector_from_ctor): Update accordingly.
15703         (build_vector_from_val): Likewise.
15704         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15705         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15706         * tree-vect-generic.c (add_rshift): Likewise.
15707         (expand_vector_divmod): Likewise.
15708         (optimize_vector_constructor): Likewise.
15709         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15710         (vect_transform_slp_perm_load): Likewise.
15711         (vect_schedule_slp_instance): Likewise.
15712         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15713         (vectorizable_call): Likewise.
15714         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15715         * expmed.c (make_tree): Likewise.
15716         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15717         a vector passed to build_vector.
15718         (fold_convert_const): Likewise.
15719         (exact_inverse): Likewise.
15720         (fold_ternary_loc): Likewise.
15721         (fold_relational_const): Likewise.
15722         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15723         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15724         (const_unop): Likewise.  Store the reduction accumulator in a
15725         variable rather than an array.
15726         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15727         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15728         the new vector, rather than constructing it after the input arrays.
15729         (native_interpret_vector): Use auto_vec<tree> when building
15730         a vector passed to build_vector.  Add elements in order.
15731         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15732         auto_vec<tree> when building a vector passed to build_vector.
15733         (vect_create_epilog_for_reduction): Likewise.
15734         (vectorizable_induction): Likewise.
15735         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15736         case statements.
15737         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15738         to a vec<tree> *.
15739         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15740         passed to build_vector.
15742 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15743             Alan Hayward  <alan.hayward@arm.com>
15744             David Sherwood  <david.sherwood@arm.com>
15746         * tree-core.h (tree_base::u): Add an "nelts" field.
15747         (tree_vector): Use VECTOR_CST_NELTS as the length.
15748         * tree.c (tree_size): Likewise.
15749         (make_vector): Initialize VECTOR_CST_NELTS.
15750         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15751         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15752         TYPE_VECTOR_SUBPARTS.
15753         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15754         as the number of units, setting it from VECTOR_CST_NELTS.
15755         (const_vector_from_tree): Likewise.
15756         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15757         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15758         (fold_negate_expr_1): Likewise.
15759         (fold_convert_const): Likewise.
15760         (const_binop): Likewise.  Differentiate the number of output and
15761         input elements.
15762         (const_unop): Likewise.
15763         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15764         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15765         in cases that did the opposite.
15767 2017-09-14  Richard Biener  <rguenther@suse.de>
15769         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15770         to VN_TOP.
15772 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15774         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15776 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15778         PR target/81325
15779         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15780         if and where to split a bb, except for splitting before debug insn
15781         sequences followed by non-label real insn.  Delete debug insns
15782         in between basic blocks.
15784         * combine.c (make_compound_operation_int): Formatting fixes.
15786         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15787         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15788         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15789         * config/sol2.h (LINK_EH_SPEC): Likewise.
15790         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15791         * config/s390/linux.h (LINK_SPEC): Likewise.
15792         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15793         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15794         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15795         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15796         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15797         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15798         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15799         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15801 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15803         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15804         support.
15805         (ENDFILE_LINUX_SPEC): Likewise.
15806         (LINK_EH_SPEC): Likewise.
15807         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15808         (LINK_OS_LINUX_SPEC32): Likewise.
15809         (LINK_OS_LINUX_SPEC64): Likewise.
15810         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15811         (LINK_OS_LINUX_SPEC): Likewise.
15813 2017-09-13  Martin Liska  <mliska@suse.cz>
15815         PR middle-end/82154
15816         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15817         CASE_HIGH is NULL_TREE.
15819 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15820             Alan Hayward  <alan.hayward@arm.com>
15821             David Sherwood  <david.sherwood@arm.com>
15823         * target.def (secondary_memory_needed): New hook.
15824         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15825         instead of SECONDARY_MEMORY_NEEDED.
15826         (secondary_memory_needed_mode): Likewise.
15827         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15828         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15829         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15830         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15831         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15832         * doc/tm.texi: Regenerate.
15833         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15834         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15835         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15836         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15837         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15838         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15839         mode argument first and change the reg_class arguments to reg_class_t.
15840         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15841         Make static.  Update the call to inline_secondary_memory_needed.
15842         (ix86_register_move_cost): Update the call to
15843         inline_secondary_memory_needed.
15844         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15845         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15846         definition.
15847         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15848         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15849         in comment.
15850         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15851         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15852         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15853         and match hook interface.  Add comment from mips.h.
15854         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15855         * config/mmix/mmix.md (truncdfsf2): Refer to
15856         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15857         in comment.
15858         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15859         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15860         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15861         (pa_secondary_memory_needed): New function.
15862         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15863         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15864         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15865         (pdp11_secondary_memory_needed): Make static and match hook interface.
15866         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
15867         * config/powerpcspe/powerpcspe-protos.h
15868         (rs6000_secondary_memory_needed_ptr): Delete.
15869         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
15870         Delete.
15871         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15872         (rs6000_option_override_internal): Assign to
15873         targetm.secondary_memory_needed rather than
15874         rs6000_secondary_memory_needed_ptr.
15875         (rs6000_secondary_memory_needed): Match hook interface.
15876         (rs6000_debug_secondary_memory_needed): Likewise.
15877         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
15878         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
15879         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
15880         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15881         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
15882         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
15883         Delete.
15884         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
15885         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15886         (rs6000_option_override_internal): Assign to
15887         targetm.secondary_memory_needed rather than
15888         rs6000_secondary_memory_needed_ptr.
15889         (rs6000_secondary_memory_needed): Match hook interface.
15890         (rs6000_debug_secondary_memory_needed): Likewise.
15891         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
15892         * config/s390/s390.c (s390_secondary_memory_needed): New function.
15893         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15894         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
15895         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15896         (sparc_secondary_memory_needed): New function.
15897         * lra-constraints.c (check_and_process_move): Refer to
15898         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15899         in comment.
15900         (curr_insn_transform): Likewise.
15901         (process_alt_operands): Use targetm.secondary_memory_needed
15902         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15903         (check_secondary_memory_needed_p): Likewise.
15904         (choose_split_class): Likewise.
15905         * reload.c: Unconditionally include code that was previously
15906         conditional on SECONDARY_MEMORY_NEEDED.
15907         (push_secondary_reload): Use targetm.secondary_memory_needed
15908         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15909         (push_reload): Likewise.
15910         * reload1.c: Unconditionally include code that was previously
15911         conditional on SECONDARY_MEMORY_NEEDED.
15912         (choose_reload_regs): Use targetm.secondary_memory_needed
15913         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15914         (gen_reload): Likewise.
15915         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
15917 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15918             Alan Hayward  <alan.hayward@arm.com>
15919             David Sherwood  <david.sherwood@arm.com>
15921         * target.def (secondary_memory_needed_mode): New hook:
15922         * targhooks.c (default_secondary_memory_needed_mode): Declare.
15923         * targhooks.h (default_secondary_memory_needed_mode): New function.
15924         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
15925         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
15926         * doc/tm.texi: Regenerate.
15927         * lra-constraints.c (check_and_process_move): Use
15928         targetm.secondary_memory_needed_mode instead of
15929         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
15930         (curr_insn_transform): Likewise.
15931         * reload.c (get_secondary_mem): Likewise.
15932         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15933         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
15934         function.
15935         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15936         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15937         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
15938         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15939         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
15940         Delete.
15941         * config/powerpcspe/powerpcspe-protos.h
15942         (rs6000_secondary_memory_needed_mode): Delete.
15943         * config/powerpcspe/powerpcspe.c
15944         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15945         (rs6000_secondary_memory_needed_mode): Make static.
15946         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15947         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
15948         Delete.
15949         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
15950         Redefine.
15951         (rs6000_secondary_memory_needed_mode): Make static.
15952         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15953         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
15954         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15955         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15956         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
15957         Redefine.
15958         (sparc_secondary_memory_needed_mode): New function.
15959         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
15961 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
15963         * config/aarch64/constraints.md (Umq): New constraint.
15964         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
15965         Change to use Umq.
15966         (mov<mode>): Update condition.
15968 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15970         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
15971         when bitposition is the same.
15973 2017-09-13  Richard Biener  <rguenther@suse.de>
15975         * dwarf2out.c (output_die_symbol): Remove.
15976         (output_die): Do not output a DIEs symbol.
15978 2017-09-13  Richard Biener  <rguenther@suse.de>
15980         PR middle-end/82128
15981         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
15982         default-def to avoid breaking iterator update with the weird
15983         interaction with cgraph_update_edges_for_call_stmt_node.
15985 2017-09-13  Richard Biener  <rguenther@suse.de>
15987         * tree-cfg.c (verify_gimple_assign_binary): Add verification
15988         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
15989         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
15990         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
15992 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
15994         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
15995         Disable pc relative literal load irrespective of
15996         TARGET_FIX_ERR_A53_84341 for default.
15998 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16000         * config/sparc/sparc.c (output_return): Output the source location of
16001         the insn in the delay slot, if any.
16002         (output_sibcall): Likewise.
16004 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16006         PR driver/81498
16007         * common.opt (-static-pie): New alias.
16008         (shared): Negate static-pie.
16009         (-no-pie): Update help text.
16010         (-pie): Likewise.
16011         (static-pie): New option.
16012         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16013         -static-pie support.
16014         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16015         (LINK_EH_SPEC): Likewise.
16016         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16017         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16018         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16019         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16020         (init_gcc_specs): Likewise.
16021         (init_spec): Likewise.
16022         (display_help): Update help message for -pie.
16023         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16024         -static-pie.
16026 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16028         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16029         (movdi_aarch64): Likewise.
16030         (movti_aarch64): Likewise.
16032 2017-09-12 Simon Wright <simon@pushface.org>
16034         PR target/80204
16035         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16036         calculation of the minor version, always output as 0.
16038 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16040         PR target/82112
16041         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16042         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16043         on it early, rather than manual conversion late.  For
16044         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16045         instead of performing manual conversion.
16047 2017-09-12  Carl Love  <cel@us.ibm.com>
16049         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16050         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16051         vmulouw, vmulosw.
16052         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16053         VMULOSW): Add definitions.
16054         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16055         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16056         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16057         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16058         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16060 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16062         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16063         types correctly.
16064         (movti_aarch64): Likewise.
16065         (movdf_aarch64): Likewise.
16066         (movtf_aarch64): Likewise.
16067         (load_pairdi): Likewise.
16068         (store_pairdi): Likewise.
16069         (load_pairdf): Likewise.
16070         (store_pairdf): Likewise.
16071         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16072         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16073         (ldr_got_small_<mode>): Likewise.
16074         (ldr_got_small_28k_<mode>): Likewise.
16075         (ldr_got_tiny): Likewise.
16076         * config/aarch64/iterators.md (ldst_sz): New.
16077         (ldpstp_sz): Likewise.
16078         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16079         to store_16.
16080         (thunderx_load): Split load_8 to load_16.
16081         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16082         load_8 to load_16.
16083         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16084         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16085         (xgene1_store_pair): Split store_8 to store_16.
16086         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16087         (falkor_st_0_st_sd): Split store_8 to store_16.
16089 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16091         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16092         and store1/2/3/4 to store_4/8/12/16.
16093         * config/aarch64/aarch64.md: Update for rename.
16094         * config/arm/arm.md: Likewise.: Likewise.
16095         * config/arm/arm.c: Likewise.
16096         * config/arm/thumb1.md: Likewise.
16097         * config/arm/thumb2.md: Likewise.
16098         * config/arm/vfp.md: Likewise.
16099         * config/arm/arm-generic.md: Likewise.
16100         * config/arm/arm1020e.md: Likewise.
16101         * config/arm/arm1026ejs.md: Likewise.
16102         * config/arm/arm1136jfs.md: Likewise.
16103         * config/arm/arm926ejs.md: Likewise.
16104         * config/arm/cortex-a15.md: Likewise.
16105         * config/arm/cortex-a17.md: Likewise.
16106         * config/arm/cortex-a5.md: Likewise.
16107         * config/arm/cortex-a53.md: Likewise.
16108         * config/arm/cortex-a57.md: Likewise.
16109         * config/arm/cortex-a7.md: Likewise.
16110         * config/arm/cortex-a8.md: Likewise.
16111         * config/arm/cortex-a9.md: Likewise.
16112         * config/arm/cortex-m4.md: Likewise.
16113         * config/arm/cortex-m7.md: Likewise.
16114         * config/arm/cortex-r4.md: Likewise.
16115         * config/arm/exynos-m1.md: Likewise.
16116         * config/arm/fa526.md: Likewise.
16117         * config/arm/fa606te.md: Likewise.
16118         * config/arm/fa626te.md: Likewise.
16119         * config/arm/fa726te.md: Likewise.
16120         * config/arm/fmp626.md: Likewise.
16121         * config/arm/iwmmxt.md: Likewise.
16122         * config/arm/ldmstm.md: Likewise.
16123         * config/arm/marvell-pj4.md: Likewise.
16124         * config/arm/xgene1.md: Likewise.
16125         * config/aarch64/thunderx.md: Likewise.
16126         * config/aarch64/thunderx2t99.md: Likewise.
16127         * config/aarch64/falkor.md: Likewise.
16129 2017-09-12  Martin Liska  <mliska@suse.cz>
16131         * attribs.c (private_lookup_attribute): New function.
16132         * attribs.h (private_lookup_attribute): Declared here.
16133         (lookup_attribute): Called from this place.
16135 2017-09-12  Richard Biener  <rguenther@suse.de>
16137         PR tree-optimization/82157
16138         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16139         stmts with side-effects.
16141 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16142             Alan Hayward  <alan.hayward@arm.com>
16143             David Sherwood <david.sherwood@arm.com>
16145         * target.def (hard_regno_nregs): New hook.
16146         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16147         * targhooks.h (default_hard_regno_nregs): Declare.
16148         * targhooks.c (default_hard_regno_nregs): New function.
16149         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16150         (TARGET_HARD_REGNO_NREGS): ...this hook.
16151         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16152         (CLASS_MAX_NREGS): Likewise.
16153         * doc/tm.texi: Regenerate.
16154         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16155         instead of HARD_REGNO_NREGS.
16156         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16157         HARD_REGNO_NREGS in the comment.
16158         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16159         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16160         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16161         Return an unsigned int.
16162         (TARGET_HARD_REGNO_NREGS): Redefine.
16163         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16164         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16165         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16166         (arc_hard_regno_nregs): New function.
16167         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16168         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16169         (arm_hard_regno_nregs): New function.
16170         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16171         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16172         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16173         (TARGET_HARD_REGNO_NREGS): Redefine.
16174         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16175         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16176         (HARD_REGNO_NREGS): Delete.
16177         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16178         (cr16_hard_regno_nregs): New function.
16179         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16180         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16181         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16182         (cris_hard_regno_nregs): New function.
16183         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16184         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16185         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16186         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16187         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16188         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16189         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16190         (frv_hard_regno_nregs): Make static.  Take and return an
16191         unsigned int.
16192         (frv_class_max_nregs): Remove outdated copy of documentation.
16193         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16194         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16195         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16196         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16197         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16198         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16199         (TARGET_HARD_REGNO_NREGS): Redefine.
16200         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16201         (CLASS_MAX_NREGS): Update comment.
16202         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16203         (ia64_hard_regno_nregs): New function.
16204         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16205         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16206         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16207         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16208         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16209         an unsigned int.
16210         (m32c_hard_regno_nregs): Likewise.  Make static.
16211         (TARGET_HARD_REGNO_NREGS): Redefine.
16212         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16213         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16214         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16215         (m68k_hard_regno_nregs): New function.
16216         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16217         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16218         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16219         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16220         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16221         Take and return an unsigned int.
16222         (TARGET_HARD_REGNO_NREGS): Redefine.
16223         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16224         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16225         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16226         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16227         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16228         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16229         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16230         (msp430_hard_regno_nregs): Make static.  Take and return an
16231         unsigned int.
16232         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16233         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16234         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16235         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16236         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16237         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16238         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16239         (TARGET_HARD_REGNO_NREGS): Redefine.
16240         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16241         (PA_HARD_REGNO_NREGS): ...this.
16242         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16243         (PA_HARD_REGNO_NREGS): ...this.
16244         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16245         (pa_hard_regno_nregs): New function.
16246         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16247         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16248         (pdp11_hard_regno_nregs): New function.
16249         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16250         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16251         (rs6000_hard_regno_nregs_hook): New function.
16252         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16253         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16254         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16255         Take and return an unsigned int.  Move earlier in file.
16256         (TARGET_HARD_REGNO_NREGS): Redefine.
16257         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16258         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16259         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16260         (rl78_hard_regno_nregs): Make static.  Take and return an
16261         unsigned int.
16262         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16263         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16264         (rs6000_hard_regno_nregs_hook): New function.
16265         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16266         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16267         (TARGET_HARD_REGNO_NREGS): Redefine.
16268         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16269         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16270         instead of HARD_REGNO_NREGS.
16271         (s390_hard_regno_nregs): New function.
16272         (s390_hard_regno_mode_ok): Add comment from s390.h.
16273         (TARGET_HARD_REGNO_NREGS): Redefine.
16274         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16275         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16276         (sh_hard_regno_nregs): New function.
16277         (sh_pass_in_reg_p): Use it.
16278         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16279         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16280         (sparc_hard_regno_nregs): New function.
16281         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16282         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16283         (spu_function_arg_advance): Use it, supplying a valid register number.
16284         (TARGET_HARD_REGNO_NREGS): Redefine.
16285         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16286         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16287         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16288         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16289         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16290         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16291         (CLASS_MAX_NREGS): Remove copy of old documentation.
16292         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16293         (visium_hard_regno_nregs): New function.
16294         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16295         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16296         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16297         xtensa_hard_regno_nregs): New function.
16298         * system.h (HARD_REGNO_NREGS): Poison.
16300 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16302         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16303         hard_regno_nregs instead of HARD_REGNO_NREGS.
16304         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16305         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16306         (c6x_expand_epilogue): Likewise.
16307         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16308         (frv_read_iacc_argument): Likewise.
16309         * config/sh/sh.c: Include regs.h.
16310         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16311         (regs_used): Likewise.
16312         (output_stack_adjust): Likewise.
16313         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16314         * expmed.c: Include regs.h.
16315         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16316         * ree.c: Include regs.h.
16317         (combine_reaching_defs): Use hard_regno_nregs instead of
16318         HARD_REGNO_NREGS.
16319         (add_removable_extension): Likewise.
16321 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16323         * regs.h (hard_regno_nregs): Turn into a function.
16324         (end_hard_regno): Update accordingly.
16325         * caller-save.c (setup_save_areas): Likewise.
16326         (save_call_clobbered_regs): Likewise.
16327         (replace_reg_with_saved_mem): Likewise.
16328         (insert_restore): Likewise.
16329         (insert_save): Likewise.
16330         * combine.c (can_change_dest_mode): Likewise.
16331         (move_deaths): Likewise.
16332         (distribute_notes): Likewise.
16333         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16334         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16335         (rs6000_split_multireg_move): Likewise.
16336         (rs6000_register_move_cost): Likewise.
16337         (rs6000_memory_move_cost): Likewise.
16338         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16339         (rs6000_split_multireg_move): Likewise.
16340         (rs6000_register_move_cost): Likewise.
16341         (rs6000_memory_move_cost): Likewise.
16342         * cselib.c (cselib_reset_table): Likewise.
16343         (cselib_lookup_1): Likewise.
16344         * emit-rtl.c (set_mode_and_regno): Likewise.
16345         * function.c (aggregate_value_p): Likewise.
16346         * ira-color.c (setup_profitable_hard_regs): Likewise.
16347         (check_hard_reg_p): Likewise.
16348         (calculate_saved_nregs): Likewise.
16349         (assign_hard_reg): Likewise.
16350         (improve_allocation): Likewise.
16351         (calculate_spill_cost): Likewise.
16352         * ira-emit.c (modify_move_list): Likewise.
16353         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16354         (ira_hard_reg_in_set_p): Likewise.
16355         * ira.c (setup_reg_mode_hard_regset): Likewise.
16356         (clarify_prohibited_class_mode_regs): Likewise.
16357         (check_allocation): Likewise.
16358         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16359         (lra_setup_reg_renumber): Likewise.
16360         (setup_try_hard_regno_pseudos): Likewise.
16361         (spill_for): Likewise.
16362         (assign_hard_regno): Likewise.
16363         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16364         * lra-constraints.c (in_class_p): Likewise.
16365         (lra_constraint_offset): Likewise.
16366         (simplify_operand_subreg): Likewise.
16367         (lra_constraints): Likewise.
16368         (split_reg): Likewise.
16369         (split_if_necessary): Likewise.
16370         (invariant_p): Likewise.
16371         (inherit_in_ebb): Likewise.
16372         * lra-lives.c (process_bb_lives): Likewise.
16373         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16374         (get_hard_regs): Likewise.
16375         (do_remat): Likewise.
16376         * lra-spills.c (assign_spill_hard_regs): Likewise.
16377         * mode-switching.c (create_pre_exit): Likewise.
16378         * postreload.c (reload_combine_recognize_pattern): Likewise.
16379         * recog.c (peep2_find_free_register): Likewise.
16380         * regcprop.c (kill_value_regno): Likewise.
16381         (set_value_regno): Likewise.
16382         (copy_value): Likewise.
16383         (maybe_mode_change): Likewise.
16384         (find_oldest_value_reg): Likewise.
16385         (copyprop_hardreg_forward_1): Likewise.
16386         * regrename.c (check_new_reg_p): Likewise.
16387         (regrename_do_replace): Likewise.
16388         * reload.c (push_reload): Likewise.
16389         (combine_reloads): Likewise.
16390         (find_dummy_reload): Likewise.
16391         (operands_match_p): Likewise.
16392         (find_reloads): Likewise.
16393         (find_equiv_reg): Likewise.
16394         (reload_adjust_reg_for_mode): Likewise.
16395         * reload1.c (count_pseudo): Likewise.
16396         (count_spilled_pseudo): Likewise.
16397         (find_reg): Likewise.
16398         (clear_reload_reg_in_use): Likewise.
16399         (free_for_value_p): Likewise.
16400         (allocate_reload_reg): Likewise.
16401         (choose_reload_regs): Likewise.
16402         (reload_adjust_reg_for_temp): Likewise.
16403         (emit_reload_insns): Likewise.
16404         (delete_output_reload): Likewise.
16405         * rtlanal.c (subreg_get_info): Likewise.
16406         * sched-deps.c (sched_analyze_reg): Likewise.
16407         * sel-sched.c (init_regs_for_mode): Likewise.
16408         (mark_unavailable_hard_regs): Likewise.
16409         (choose_best_reg_1): Likewise.
16410         (verify_target_availability): Likewise.
16411         * valtrack.c (dead_debug_insert_temp): Likewise.
16412         * var-tracking.c (track_loc_p): Likewise.
16413         (emit_note_insn_var_location): Likewise.
16414         * varasm.c (make_decl_rtl): Likewise.
16415         * reginfo.c (choose_hard_reg_mode): Likewise.
16416         (init_reg_modes_target): Refer directly to
16417         this_target_regs->x_hard_regno_nregs.
16419 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16421         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16422         instead of hard_regno_nregs.
16424 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16426         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16427         end_hard_regno instead of hard_regno_nregs.
16428         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16429         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16430         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16431         * ira-color.c (improve_allocation): Likewise.
16432         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16433         * lra-lives.c (mark_regno_live): Likewise.
16434         (mark_regno_dead): Likewise.
16435         * lra-remat.c (operand_to_remat): Likewise.
16436         * lra.c (collect_non_operand_hard_regs): Likewise.
16437         * postreload.c (reload_combine_note_store): Likewise.
16438         (move2add_valid_value_p): Likewise.
16439         * reload.c (regno_clobbered_p): Likewise.
16441 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16443         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16444         hard_regno_nregs.
16445         * config/v850/v850.c (v850_reorg): Likewise.
16446         * reload.c (refers_to_regno_for_reload_p): Likewise.
16447         (find_equiv_reg): Likewise.
16448         * reload1.c (reload_reg_reaches_end_p): Likewise.
16450 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16452         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16453         hard_regno_nregs.
16454         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16455         * config/arm/arm.c (output_move_neon): Likewise.
16456         (arm_attr_length_move_neon): Likewise.
16457         (neon_split_vcombine): Likewise.
16458         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16459         (c6x_mark_reg_written): Likewise.
16460         (c6x_dwarf_register_span): Likewise.
16461         * config/i386/i386.c (ix86_save_reg): Likewise.
16462         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16463         (rws_access_reg): Likewise.
16464         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16465         * mode-switching.c (create_pre_exit): Likewise.
16466         * ree.c (combine_reaching_defs): Likewise.
16467         (add_removable_extension): Likewise.
16468         * regcprop.c (find_oldest_value_reg): Likewise.
16469         (copyprop_hardreg_forward_1): Likewise.
16470         * reload.c (reload_inner_reg_of_subreg): Likewise.
16471         (push_reload): Likewise.
16472         (combine_reloads): Likewise.
16473         (find_dummy_reload): Likewise.
16474         (reload_adjust_reg_for_mode): Likewise.
16475         * reload1.c (find_reload_regs): Likewise.
16476         (forget_old_reloads_1): Likewise.
16477         (reload_reg_free_for_value_p): Likewise.
16478         (reload_adjust_reg_for_temp): Likewise.
16479         (emit_reload_insns): Likewise.
16480         (delete_output_reload): Likewise.
16481         * sel-sched.c (choose_best_reg_1): Likewise.
16482         (choose_best_pseudo_reg): Likewise.
16484 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16485             Alan Hayward  <alan.hayward@arm.com>
16486             David Sherwood <david.sherwood@arm.com>
16488         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16489         * target.def (slow_unaligned_access): New hook.
16490         * targhooks.h (default_slow_unaligned_access): Declare.
16491         * targhooks.c (default_slow_unaligned_access): New function.
16492         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16493         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16494         * doc/tm.texi: Regenerate.
16495         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16496         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16497         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16498         definition.
16499         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16500         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16501         Redefine.
16502         (rs6000_slow_unaligned_access): New function.
16503         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16504         (expand_block_compare): Likewise.
16505         (expand_strn_compare): Likewise.
16506         (rs6000_rtx_costs): Likewise.
16507         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16508         (riscv_slow_unaligned_access): Likewise.
16509         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16510         (riscv_slow_unaligned_access_p): ...this and make static.
16511         (riscv_option_override): Update accordingly.
16512         (riscv_slow_unaligned_access): New function.
16513         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16514         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16515         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16516         (rs6000_slow_unaligned_access): New function.
16517         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16518         (rs6000_rtx_costs): Likewise.
16519         * config/rs6000/rs6000-string.c (expand_block_compare)
16520         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16521         of SLOW_UNALIGNED_ACCESS.
16522         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16523         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16524         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16525         of SLOW_UNALIGNED_ACCESS.
16526         * expmed.c (simple_mem_bitfield_p): Likewise.
16527         * expr.c (alignment_for_piecewise_move): Likewise.
16528         (emit_group_load_1): Likewise.
16529         (emit_group_store): Likewise.
16530         (copy_blkmode_from_reg): Likewise.
16531         (emit_push_insn): Likewise.
16532         (expand_assignment): Likewise.
16533         (store_field): Likewise.
16534         (expand_expr_real_1): Likewise.
16535         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16536         * lra-constraints.c (simplify_operand_subreg): Likewise.
16537         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16538         * gimple-ssa-store-merging.c: Likewise in block comment at start
16539         of file.
16540         * tree-ssa-strlen.c: Include target.h.
16541         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16542         of SLOW_UNALIGNED_ACCESS.
16543         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16545 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16547         PR rtl-optimization/82185
16548         * expmed.c (emit_store_flag_int): Only test tem if it has been
16549         initialized.
16551 2017-09-12  Richard Biener  <rguenther@suse.de>
16553         PR middle-end/82149
16554         * match.pd ((FTYPE) N CMP CST): Fix typo.
16556 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16558         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16559         attribute.
16560         (mips_near_type_p): Add 'short_call' attribute as a synonym
16561         for 'near'.
16562         * doc/extend.texi (short_call): Document new function attribute.
16564 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16566         PR target/82112
16567         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16568         assertion check that in the condition.
16569         (get_atomic_generic_size): Likewise.  Before testing if parameter
16570         has pointer type, if it has array type, call for C++
16571         default_conversion to perform array-to-pointer conversion.
16573 2017-09-12  Richard Biener  <rguenther@suse.de>
16575         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16576         for operations we cannot scalarize.
16578 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16580         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16581         vectors heap vectors.  Clean up comments.
16582         Make visited_bbs a reference.
16583         (profitable_jump_thread_path): Make GC
16584         vectors heap vectors.  Clean up comments.
16585         Misc cleanups.
16586         (convert_and_register_jump_thread_path): Make GC vectors heap
16587         vectors.
16588         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16589         Make visited_bbs a reference.
16590         (handle_phi): Abstract common code to to
16591         register_jump_thread_path_if_profitable.
16592         Rename VAR_BB to DEF_BB.
16593         Update comments.
16594         Make GC vectors heap vectors.
16595         Make visited_bbs a reference.
16596         (handle_assignment): Same.
16597         (register_jump_thread_path_if_profitable): New.
16598         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16599         DEF_BB.
16600         Make GC vectors heap vectors.  Clean up comments.
16601         Make visited_bbs a reference.
16602         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16603         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16604         comment.
16606 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16608         PR target/82181
16609         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16610         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16612 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16614         Revert r251800 and r251799.
16616 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16618         PR hsa/82119
16619         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16620         arguments in advance.
16621         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16622         use it to find predecessor edges.
16623         (naive_outof_ssa): Collect vector of predecessors.
16625 2017-09-08  Jason Merrill  <jason@redhat.com>
16627         PR c++/70029 - ICE with ref-qualifier and -flto
16628         * langhooks.h (struct lang_hooks_for_types): Add
16629         copy_lang_qualifiers.
16630         * attribs.c (build_type_attribute_qual_variant): Use it.
16631         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16632         NULL.
16633         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16634         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16636 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16638         PR target/81988
16639         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16640         (*mulsi3_sp64): New instruction.
16641         (mulsi3): New expander.
16643 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16645         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16647 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16649         * sancov.c: Include memmodel.h.
16651 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16653         PR target/80897
16654         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16655         large offsets.
16657 2017-09-07  Carl Love  <cel@us.ibm.com>
16659         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16660         the sldi instruction.
16662 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16664         * sancov.c: Include tm_p.h.
16666 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16668         PR target/81979
16669         * output.h (switch_to_other_text_partition): New declaration.
16670         * varasm.c (switch_to_other_text_partition): New function.
16671         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16672         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16673         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16674         to the other text partition before emitting LCL label and switch back
16675         after emitting the word after it.
16677 2017-09-07  Richard Biener  <rguenther@suse.de>
16679         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16680         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16681         critical edges here, after loop init.
16682         (pass_data_pre): Remove PROP_no_crit_edges flags.
16683         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16684         for valueization of call args to avoid leaking VN_TOP.
16685         (visit_use): Assert we do not visit default defs.
16686         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16687         Use error_mark_node to more easily detect leaking VN_TOP.
16688         All default-defs are varying, not VN_TOP.  Mark them visited.
16689         (run_scc_vn): Make code match comment.
16691 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16693         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16694         OPTION_MASK_FLOAT128_KEYWORD.
16695         (POWERPC_MASKS): Likewise.
16696         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16697         support for the -mfloat128-type option, and make -mfloat128
16698         default on PowerPC Linux systems.  Define or undefine
16699         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16700         Define __float128 to be __ieee128 if IEEE 128-bit support is
16701         enabled, or undefine it.
16702         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16703         Delete defining __FLOAT128_TYPE__.
16704         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16705         -mfloat128-type option and make -mfloat128 default on PowerPC
16706         Linux systems.
16707         (TARGET_FLOAT128_TYPE): Likewise.
16708         (-mfloat128-type): Likewise.
16709         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16710         Delete the -mfloat128-type option and make -mfloat128 default on
16711         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16712         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16713         128-bit floating point is enabled.  Change tests from using
16714         -mfloat128-type to -mfloat128.
16715         (rs6000_mangle_type): Use the correct mangling for the __float128
16716         type even if normal long double is restricted to 64-bits.
16717         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16718         systems.
16719         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16720         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16721         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16722         -mfloat128-type.
16723         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16724         documentation for -mfloat128.
16726 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16728         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16730 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16731             Jakub Jelinek  <jakub@redhat.com>
16733         * asan.c (initialize_sanitizer_builtins): Add
16734         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16735         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16736         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16737         BT_FN_VOID_UINT64_PTR variables.
16738         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16739         (BT_FN_VOID_UINT16_UINT16): Likewise.
16740         (BT_FN_VOID_UINT32_UINT32): Likewise.
16741         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16742         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16743         (BT_FN_VOID_UINT64_PTR): Likewise.
16744         * common.opt (flag_sanitize_coverage): New variable.
16745         (fsanitize-coverage=trace-pc): Remove.
16746         (fsanitize-coverage=): Add.
16747         * flag-types.h (enum sanitize_coverage_code): New enum.
16748         * fold-const.c (fold_range_test): Disable non-short-circuit
16749         optimization if flag_sanitize_coverage.
16750         (fold_truth_andor): Likewise.
16751         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16752         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16753         (coverage_sanitizer_opts): New array.
16754         (get_closest_sanitizer_option): Add OPTS argument, handle also
16755         OPT_fsanitize_coverage_.
16756         (parse_sanitizer_options): Adjusted to also handle
16757         OPT_fsanitize_coverage_.
16758         (common_handle_option): Add OPT_fsanitize_coverage_.
16759         * sancov.c (instrument_comparison, instrument_switch): New function.
16760         (sancov_pass): Add trace-cmp support.
16761         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16762         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16763         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16764         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16765         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16766         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16767         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16768         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16769         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16770         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16772 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16774         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16775         error.  Only quit immediately if parsing is complete.
16776         (BEGIN): Initialize fatal_err and parse_done.
16777         (begin fpu, end fpu): Check number of arguments.
16778         (begin arch, end arch): Likewise.
16779         (begin cpu, end cpu): Likewise.
16780         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16781         (optalias): Likewise.
16783 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16785         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16786         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16787         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16788         * config/arm/arm.c (arm_option_override): Use lower case for feature
16789         bit names.
16790         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16791         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16792         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16793         (isa_pfx): Delete.
16794         (print_isa_bits_for): New function.
16795         (gen_isa): New function.
16796         (gen_comm_data): Use print_isa_bits_for.
16797         (define feature): New keyword.
16798         (define fgroup): New keyword.
16799         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16800         (arm-isa.h): Add rule to generate file.
16801         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16802         case for feature bit names.
16804 2017-09-06  Richard Biener  <rguenther@suse.de>
16806         * tree-ssa-pre.c (NECESSARY): Remove.
16807         (create_expression_by_pieces): Do not touch pass-local flags.
16808         (insert_into_preds_of_block): Likewise.
16809         (do_pre_regular_insertion): Likewise.
16810         (eliminate_insert): Likewise.
16811         (eliminate_dom_walker::before_dom_children): Likewise.
16812         (fini_eliminate): Do not look at inserted_exprs.
16813         (mark_operand_necessary): Remove.
16814         (remove_dead_inserted_code): Replace with simple work-list
16815         algorithm based on inserted_exprs and SSA uses.
16816         (pass_pre::execute): Re-order fini_eliminate and
16817         remove_dead_inserted_code.
16819 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16821         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16822         for VxWorks 7.  Adjust surrounding comments.
16824 2017-09-06  Richard Biener  <rguenther@suse.de>
16826         * gimple-ssa-strength-reduction.c
16827         (find_candidates_dom_walker::before_dom_children): Also allow
16828         pointer types.
16830 2017-09-06  Richard Biener  <rguenther@suse.de>
16832         PR tree-optimization/82108
16833         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16834         for gap in the non-permutation SLP case.
16836 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16838         PR tree-optimization/82078
16839         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16840         add_access_to_work_queue...
16841         (build_accesses_from_assign): ...here.
16842         (propagate_all_subaccesses): Make sure racc is the group
16843         representative, if there is one.
16845 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16847         PR middle-end/82095
16848         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16849         NULL DECL_INITIAL.
16851 2017-09-06  Richard Biener  <rguenther@suse.de>
16853         * gimple-ssa-strength-reduction.c
16854         (find_candidates_dom_walker::before_doom_children): Use a
16855         type and not a mode check.
16857 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16859         PR target/77308
16860         * config/arm/predicates.md (arm_general_adddi_operand): Create new
16861         non-vfp predicate.
16862         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16864 2017-09-05  Jeff Law  <law@redhat.com>
16866         PR tree-optimization/64910
16867         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
16868         cases where we have 3 or more operands.
16870 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
16872         PR middle-end/81768
16873         * omp-low.c (lower_omp_for): Recompute tree invariant if
16874         gimple_omp_for_initial/final is ADDR_EXPR.
16876         PR middle-end/81768
16877         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
16878         into gimple val before gimplification fo the COND_EXPR.
16880 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
16882         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
16883         REGION_COPY argument.
16884         (thread_through_all_blocks): Remove unused argument to
16885         duplicate_thread_path.
16887 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16888             Alan Hayward  <alan.hayward@arm.com>
16889             David Sherwood  <david.sherwood@arm.com>
16891         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
16892         Take a scalar_mode rather than a machine_mode.
16893         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16894         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
16895         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16896         (aarch64_gen_adjusted_ldpstp): Likewise.
16897         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
16899 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16900             Alan Hayward  <alan.hayward@arm.com>
16901             David Sherwood  <david.sherwood@arm.com>
16903         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
16904         Take a scalar_int_mode instead of a machine_mode.
16905         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16906         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16907         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16908         (aarch64_simd_attr_length_rglist): Delete.
16909         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
16910         a scalar_int_mode instead of a machine_mode.
16911         (aarch64_add_offset): Likewise.
16912         (aarch64_internal_mov_immediate): Likewise
16913         (aarch64_add_constant_internal): Likewise.
16914         (aarch64_add_constant): Likewise.
16915         (aarch64_movw_imm): Likewise.
16916         (aarch64_rtx_arith_op_extract_p): Likewise.
16917         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16918         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16919         Remove assert that the mode isn't a vector.
16920         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16921         (aarch64_expand_mov_immediate): Update calls after above changes.
16922         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
16923         (aarch64_and_bitmask_imm): Check for scalar integer modes.
16924         (aarch64_move_imm): Likewise.
16925         (aarch64_can_const_movi_rtx_p): Likewise.
16926         (aarch64_strip_extend): Likewise.
16927         (aarch64_extr_rtx_p): Likewise.
16928         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
16929         a CONST_INT when the mode parameter is VOIDmode.
16930         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
16932 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16934         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
16935         * stor-layout.c (bitwise_mode_for_mode): Likewise.
16936         (bitwise_type_for_mode): Update accordingly.
16938 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16940         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
16941         * stor-layout.c (mode_for_size_tree): Likewise.
16942         (mode_for_array): Update accordingly.
16943         (layout_decl): Likewise.
16944         (compute_record_mode): Likewise.  Only set the mode once.
16946 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16948         * target.def (get_mask_mode): Change return type to opt_mode.
16949         Expand commentary.
16950         * doc/tm.texi: Regenerate.
16951         * targhooks.h (default_get_mask_mode): Return an opt_mode.
16952         * targhooks.c (default_get_mask_mode): Likewise.
16953         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16954         * optabs-query.c (can_vec_mask_load_store_p): Update use of
16955         targetm.get_mask_mode.
16956         * tree.c (build_truth_vector_type): Likewise.
16958 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16960         * machmode.h (mode_for_vector): Return an opt_mode.
16961         * stor-layout.c (mode_for_vector): Likewise.
16962         (mode_for_int_vector): Update accordingly.
16963         (layout_type): Likewise.
16964         * config/i386/i386.c (emit_memmov): Likewise.
16965         (ix86_expand_set_or_movmem): Likewise.
16966         (ix86_expand_vector_init): Likewise.
16967         (ix86_get_mask_mode): Likewise.
16968         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
16969         Likewise.
16970         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
16971         * expmed.c (extract_bit_field_1): Likewise.
16972         * expr.c (expand_expr_real_2): Likewise.
16973         * optabs-query.c (can_vec_perm_p): Likewise.
16974         (can_vec_mask_load_store_p): Likewise.
16975         * optabs.c (expand_vec_perm): Likewise.
16976         * targhooks.c (default_get_mask_mode): Likewise.
16977         * tree-vect-stmts.c (vectorizable_store): Likewise.
16978         (vectorizable_load): Likewise.
16979         (get_vectype_for_scalar_type_and_size): Likewise.
16981 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16983         * machmode.h (mode_for_int_vector): New function.
16984         * stor-layout.c (mode_for_int_vector): Likewise.
16985         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
16986         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
16987         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
16988         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
16989         (s390_expand_vcond): Likewise.
16991 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16993         * machmode.h (opt_machine_mode): New type.
16994         (opt_mode<T>): Allow construction from anything that can be
16995         converted to a T.
16996         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
16997         (mode_for_size): Return an opt_machine_mode.
16998         * stor-layout.c (mode_for_size): Likewise.
16999         (mode_for_size_tree): Update call accordingly.
17000         (bitwise_mode_for_mode): Likewise.
17001         (make_fract_type): Likewise.
17002         (make_accum_type): Likewise.
17003         * caller-save.c (replace_reg_with_saved_mem): Update call
17004         accordingly.
17005         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17006         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17007         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17008         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17009         * expmed.c (extract_bit_field_1): Likewise.
17010         * reload.c (get_secondary_mem): Likewise.
17011         * varasm.c (assemble_integer): Likewise.
17012         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17013         early-out.
17015 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17017         * machmode.h (decimal_float_mode_for_size): New function.
17018         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17019         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17020         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17021         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17022         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17023         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17025 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17027         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17028         (get_builtin_sync_mode): Likewise.
17029         (expand_ifn_atomic_compare_exchange): Likewise.
17030         (expand_builtin_atomic_clear): Likewise.
17031         (expand_builtin_atomic_test_and_set): Likewise.
17032         (fold_builtin_atomic_always_lock_free): Likewise.
17033         * calls.c (compute_argument_addresses): Likewise.
17034         (emit_library_call_value_1): Likewise.
17035         (store_one_arg): Likewise.
17036         * combine.c (combine_instructions): Likewise.
17037         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17038         * config/arm/arm.c (arm_function_value): Likewise.
17039         (aapcs_allocate_return_reg): Likewise.
17040         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17041         * config/i386/i386.c (construct_container): Likewise.
17042         (ix86_gimplify_va_arg): Likewise.
17043         (ix86_expand_sse_cmp): Likewise.
17044         (emit_memmov): Likewise.
17045         (emit_memset): Likewise.
17046         (expand_small_movmem_or_setmem): Likewise.
17047         (ix86_expand_pextr): Likewise.
17048         (ix86_expand_pinsr): Likewise.
17049         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17050         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17051         Likewise.
17052         * config/mips/mips.c (mips_function_value_1) Likewise.
17053         (mips_block_move_straight): Likewise.
17054         (mips_expand_ins_as_unaligned_store): Likewise.
17055         * config/powerpcspe/powerpcspe.c
17056         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17057         (rs6000_darwin64_record_arg_flush): Likewise.
17058         * config/rs6000/rs6000.c
17059         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17060         (rs6000_darwin64_record_arg_flush): Likewise.
17061         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17062         (sparc_function_value_1): Likewise.
17063         * config/spu/spu.c (adjust_operand): Likewise.
17064         (spu_emit_branch_or_set): Likewise.
17065         (arith_immediate_p): Likewise.
17066         * emit-rtl.c (gen_lowpart_common): Likewise.
17067         * expr.c (expand_expr_real_1): Likewise.
17068         * function.c (assign_parm_setup_block): Likewise.
17069         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17070         * reload1.c (alter_reg): Likewise.
17071         * stor-layout.c (mode_for_vector): Likewise.
17072         (layout_type): Likewise.
17074 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17076         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17077         (spu_convert_move): Likewise.
17078         * lower-subreg.c (resolve_simple_move): Likewise.
17080 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17082         PR target/81833
17083         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17084         define_insn to a define_expand.
17085         (altivec_vsum2sws_direct): New define_insn.
17086         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17088 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17090         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17091         max_insns_skipped.
17093 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17095         PR target/59501
17096         PR target/81624
17097         PR target/81769
17098         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17099         realign stack if stack alignment needed is less than incoming
17100         stack boundary.
17102 2017-09-05  Marek Polacek  <polacek@redhat.com>
17104         PR sanitizer/82072
17105         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17106         check earlier.
17108 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17110         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17112 2017-09-05  Richard Biener  <rguenther@suse.de>
17114         PR tree-optimization/82084
17115         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17117 2017-09-05  Richard Biener  <rguenther@suse.de>
17119         PR tree-optimization/82102
17120         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17122 2017-09-05  Martin Liska  <mliska@suse.cz>
17124         PR tree-optimization/82032
17125         * tree-cfg.c (generate_range_test): New function.
17126         * tree-cfg.h (generate_range_test): Declared here.
17127         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17128         (cleanup_control_expr_graph): Use it.
17129         * tree-switch-conversion.c (try_switch_expansion): Remove
17130         assert.
17131         (emit_case_nodes): Use generate_range_test.
17133 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17135         PR target/82098
17136         * config/i386/i386.md (*<btsc><mode>_mask): Add
17137         TARGET_USE_BT to insn constraint.
17138         (*btr<mode>_mask): Ditto.
17140 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17142         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17143         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17145 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17147         PR target/77308
17148         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17149         TARGET_NEON and TARGET_IWMMXT.
17150         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17151         TARGET_NEON and TARGET_IWMMXT.
17152         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17154 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17156         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17157         (ix86_rewrite_tls_address): Ditto.
17158         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17159         (ix86_rewrite_tls_address_1): Ditto.
17160         (ix86_rewrite_tls_address): Ditto.
17161         * config/i386/predicates.md (tls_address_pattern): New predicate.
17162         * config/i386/i386.md (TLS address splitter): New splitter.
17164 2017-09-04  Richard Biener  <rguenther@suse.de>
17166         PR tree-optimization/82084
17167         * fold-const.h (can_native_encode_string_p): Declare.
17168         * fold-const.c (can_native_encode_string_p): Factor out from ...
17169         (native_encode_string): ... here.
17170         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17171         vectorizing stores from constants we later cannot handle.
17173 2017-09-04  Marek Polacek  <polacek@redhat.com>
17175         PR c/81783
17176         * doc/invoke.texi: Update -Wtautological-compare documentation.
17178 2017-09-04  Jeff Law  <law@redhat.com>
17180         PR tree-optimization/64910
17181         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17182         swap the first and last operand if the last is a constant.
17184 2017-09-04  Marek Polacek  <polacek@redhat.com>
17186         PR sanitizer/82072
17187         * convert.c (do_narrow): When sanitizing signed integer overflows,
17188         bail out for signed types.
17189         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17191 2017-09-04  Richard Biener  <rguenther@suse.de>
17193         PR tree-optimization/82060
17194         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17195         Move devirtualization after stmt folding and before EH/AB/noreturn
17196         cleanup to get the stmt refs canonicalized.  Use a bool instead
17197         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17198         NOPs generated by folding for removal.
17200 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17201             Alan Hayward  <alan.hayward@arm.com>
17202             David Sherwood  <david.sherwood@arm.com>
17204         * coretypes.h (pad_direction): New enum.
17205         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17206         (FUNCTION_ARG_PADDING): Likewise.
17207         * target.def (function_arg_padding): New hook.
17208         * targhooks.h (default_function_arg_padding): Declare.
17209         * targhooks.c (default_function_arg_padding): New function.
17210         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17211         (TARGET_FUNCTION_ARG_PADDING): ...this.
17212         * doc/tm.texi: Regenerate.
17213         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17214         instead of direction.
17215         (compute_argument_addresses): Likewise.
17216         (load_register_parameters): Likewise.
17217         (emit_library_call_value_1): Likewise.
17218         (store_one_arg): Use targetm.calls.function_arg_padding instead
17219         of FUNCTION_ARG_PADDING.
17220         (must_pass_in_stack_var_size_or_pad): Likewise.
17221         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17222         (emit_group_store): Likewise.
17223         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17224         instead of FUNCTION_ARG_PADDING.
17225         (emit_push_insn): Likewise, and propagate enum change throughout
17226         function.
17227         * function.h (direction): Delete.
17228         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17229         of direction.
17230         * function.c (assign_parm_find_stack_rtl): Likewise.
17231         (assign_parm_setup_block_p): Likewise.
17232         (assign_parm_setup_block): Likewise.
17233         (gimplify_parameters): Likewise.
17234         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17235         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17236         function.
17237         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17238         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17239         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17240         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17241         (aarch64_function_arg_padding): ...this new function.
17242         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17243         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17244         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17245         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17246         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17247         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17248         (arm_pad_arg_upward): Replace with...
17249         (arm_function_arg_padding): ...this new function.
17250         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17251         of direction.
17252         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17253         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17254         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17255         (ia64_hpux_function_arg_padding): Replace with...
17256         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17257         instead of direction.  Check for TARGET_HPUX.
17258         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17259         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17260         (iq2000_function_arg_padding): New function.
17261         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17262         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17263         (mips_function_arg_padding): ...this new function.
17264         (mips_pad_reg_upward): Update accordingly.
17265         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17266         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17267         targetm.calls.function_arg_padding.
17268         (FUNCTION_ARG_PADDING): Delete.
17269         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17270         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17271         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17272         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17273         (nios2_block_reg_padding): Return pad_direction instead of direction.
17274         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17275         instead of direction.
17276         (nios2_function_arg_padding): Likewise.  Make static.
17277         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17278         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17279         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17280         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17281         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17282         (pa_function_arg_padding): Make static.  Return pad_direction instead
17283         of direction.
17284         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17285         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17286         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17287         instead of direction.  Use targetm.calls.function_arg_padding.
17288         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17289         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17290         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17291         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17292         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17293         Redefine.
17294         (function_arg_padding): Rename to...
17295         (rs6000_function_arg_padding): ...this.  Make static.  Return
17296         pad_direction instead of direction.
17297         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17298         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17299         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17300         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17301         instead of direction.  Use targetm.calls.function_arg_padding.
17302         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17303         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17304         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17305         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17306         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17307         (function_arg_padding): Rename to...
17308         (rs6000_function_arg_padding): ...this.  Make static.  Return
17309         pad_direction instead of direction.
17310         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17311         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17312         * config/s390/s390.c (s390_function_arg_padding): New function.
17313         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17314         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17315         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17316         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17317         (function_arg_padding): Rename to...
17318         (sparc_function_arg_padding): ...this.  Make static.  Return
17319         pad_direction instead of direction.
17320         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17321         * config/spu/spu.c (spu_function_arg_padding): New function.
17322         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17323         * system.h (FUNCTION_ARG_PADDING): Poison.
17325 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17326             Alan Hayward  <alan.hayward@arm.com>
17327             David Sherwood  <david.sherwood@arm.com>
17329         * target.def (modes_tieable_p): New hook.
17330         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17331         (TARGET_MODES_TIEABLE_P): ...this.
17332         * doc/tm.texi.in: Regenerate.
17333         * hooks.h (hook_bool_mode_mode_true): Declare.
17334         * hooks.c (hook_bool_mode_mode_true): New function.
17335         * combine.c (subst): Use targetm.modes_tieable_p instead of
17336         MODES_TIEABLE_P.
17337         * dse.c (find_shift_sequence): Likewise.
17338         * expmed.c (extract_low_bits): Likewise.
17339         * lower-subreg.c: Include target.h.
17340         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17341         MODES_TIEABLE_P.
17342         * rtlanal.c (rtx_cost): Likewise.
17343         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17344         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17345         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17346         (TARGET_MODES_TIEABLE_P): Redefine.
17347         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17348         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17349         (TARGET_MODES_TIEABLE_P): Redefine.
17350         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17351         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17352         (arc_modes_tieable_p): New function.
17353         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17354         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17355         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17356         (arm_modes_tieable_p): Make static.
17357         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17358         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17359         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17360         (TARGET_MODES_TIEABLE_P): Redefine.
17361         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17362         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17363         (TARGET_MODES_TIEABLE_P): Redefine.
17364         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17365         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17366         (cr16_modes_tieable_p): New function.
17367         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17368         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17369         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17370         (TRULY_NOOP_TRUNCATION): Update comment.
17371         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17372         (TRULY_NOOP_TRUNCATION): Update comment.
17373         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17374         (frv_modes_tieable_p): New function.
17375         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17376         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17377         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17378         (TARGET_MODES_TIEABLE_P): Redefine.
17379         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17380         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17381         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17382         (TARGET_MODES_TIEABLE_P): Redefine.
17383         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17384         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17385         (ia64_modes_tieable_p): New function.
17386         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17387         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17388         (iq2000_modes_tieable_p): New function.
17389         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17390         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17391         (lm32_modes_tieable_p): New function.
17392         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17393         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17394         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17395         (TARGET_MODES_TIEABLE_P): Redefine.
17396         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17397         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17398         (m32r_modes_tieable_p): New function.
17399         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17400         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17401         (m68k_modes_tieable_p): New function.
17402         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17403         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17404         (mcore_modes_tieable_p): New function.
17405         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17406         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17407         function.
17408         (TARGET_MODES_TIEABLE_P): Redefine.
17409         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17410         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17411         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17412         (TARGET_MODES_TIEABLE_P): Redefine.
17413         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17414         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17415         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17416         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17417         (mn10300_modes_tieable_p): ...this and make static.
17418         (TARGET_MODES_TIEABLE_P): Redefine.
17419         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17420         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17421         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17422         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17423         (msp430_modes_tieable_p): Make static.
17424         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17425         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17426         (TARGET_MODES_TIEABLE_P): Redefine.
17427         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17428         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17429         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17430         (TARGET_MODES_TIEABLE_P): Redefine.
17431         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17432         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17433         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17434         (TARGET_MODES_TIEABLE_P): Redefine.
17435         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17436         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17437         (pdp11_modes_tieable_p): New function.
17438         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17439         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17440         (rs6000_modes_tieable_p): New function.
17441         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17442         * config/powerpcspe/powerpcspe.md: Update comment.
17443         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17444         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17445         (TARGET_MODES_TIEABLE_P): Redefine.
17446         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17447         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17448         (rl78_modes_tieable_p): New function.
17449         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17450         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17451         (rs6000_modes_tieable_p): New function.
17452         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17453         * config/rs6000/rs6000.md: Update comment.
17454         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17455         * config/rx/rx.c (rx_modes_tieable_p): New function.
17456         (TARGET_MODES_TIEABLE_P): Redefine.
17457         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17458         * config/s390/s390.c (s390_modes_tieable_p): New function.
17459         (TARGET_MODES_TIEABLE_P): Redefine.
17460         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17461         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17462         (sh_modes_tieable_p): New function.
17463         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17464         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17465         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17466         (sparc_modes_tieable_p): Make static.
17467         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17468         * config/spu/spu.c (spu_modes_tieable_p): New function.
17469         (TARGET_MODES_TIEABLE_P): Redefine.
17470         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17471         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17472         (TARGET_MODES_TIEABLE_P): Redefine.
17473         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17474         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17475         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17476         * config/v850/v850.c (v850_modes_tieable_p): New function.
17477         (TARGET_MODES_TIEABLE_P): Redefine.
17478         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17479         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17480         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17481         (visium_modes_tieable_p): New function.
17482         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17483         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17484         (xtensa_modes_tieable_p): New function.
17485         * system.h (MODES_TIEABLE_P): Poison.
17487 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17488             Alan Hayward  <alan.hayward@arm.com>
17489             David Sherwood  <david.sherwood@arm.com>
17491         * target.def (hard_regno_mode_ok): New hook.
17492         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17493         (TARGET_HARD_REGNO_MODE_OK): ...this.
17494         * doc/tm.texi.in: Regenerate.
17495         * hooks.h (hook_bool_uint_mode_true): Declare.
17496         * hooks.c (hook_bool_uint_mode_true): New function.
17497         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17498         HARD_REGNO_MODE_OK.
17499         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17500         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17501         instead of HARD_REGNO_MODE_OK.
17502         * caller-save.c: Include target.h.
17503         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17504         HARD_REGNO_MODE_OK.
17505         * combine.c (can_combine_p): Likewise.
17506         (combinable_i3pat): Likewise.
17507         (can_change_dest_mode): Likewise.
17508         * expr.c (init_expr_target): Likewise.
17509         (convert_move): Likewise.
17510         (convert_modes): Likewise.
17511         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17512         (setup_prohibited_mode_move_regs): Likewise.
17513         * ira.h (target_ira): Likewise.
17514         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17515         * lra-constraints.c (process_alt_operands): Likewise.
17516         (split_reg): Likewise.
17517         * recog.c (peep2_find_free_register): Likewise.
17518         * ree.c (combine_reaching_defs): Likewise.
17519         * regcprop.c (maybe_mode_change): Likewise.
17520         * reginfo.c (init_reg_sets_1): Likewise.
17521         (choose_hard_reg_mode): Likewise.
17522         (simplifiable_subregs): Likewise.
17523         * regrename.c (check_new_reg_p): Likewise.
17524         * reload.c (find_valid_class): Likewise.
17525         (find_valid_class_1): Likewise.
17526         (reload_inner_reg_of_subreg): Likewise.
17527         (push_reload): Likewise.
17528         (combine_reloads): Likewise.
17529         (find_dummy_reload): Likewise.
17530         (find_reloads): Likewise.
17531         * reload1.c (find_reg): Likewise.
17532         (set_reload_reg): Likewise.
17533         (allocate_reload_reg): Likewise.
17534         (choose_reload_regs): Likewise.
17535         (reload_adjust_reg_for_temp): Likewise.
17536         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17537         (simplify_subreg_regno): Likewise.
17538         * sel-sched.c (init_regs_for_mode): Likewise.
17539         * varasm.c (make_decl_rtl): Likewise.
17540         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17541         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17542         HARD_REGNO_MODE_OK.
17543         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17544         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17545         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17546         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17547         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17548         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17549         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17550         (arc_mode_class): Delete.
17551         (HARD_REGNO_MODE_OK): Delete.
17552         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17553         (arc_hard_regno_mode_ok): Rename old array to...
17554         (arc_hard_regno_mode_ok_modes): ...this.
17555         (arc_conditional_register_usage): Update accordingly.
17556         (arc_mode_class): Make static.
17557         (arc_hard_regno_mode_ok): New function.
17558         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17559         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17560         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17561         (arm_hard_regno_mode_ok): Make static.
17562         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17563         HARD_REGNO_MODE_OK.
17564         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17565         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17566         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17567         return a bool.
17568         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17569         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17570         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17571         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17572         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17573         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17574         * config/bfin/predicates.md (valid_reg_operand): Use
17575         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17576         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17577         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17578         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17579         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17580         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17581         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17582         (cr16_hard_regno_mode_ok): Make static and return a bool.
17583         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17584         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17585         (cris_hard_regno_mode_ok): New function.
17586         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17587         (epiphany_mode_class): Delete.
17588         (HARD_REGNO_MODE_OK): Delete.
17589         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17590         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17591         (hard_regno_mode_ok): Rename to...
17592         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17593         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17594         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17595         HARD_REGNO_MODE_OK.
17596         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17597         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17598         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17599         (frv_hard_regno_mode_ok): Make static and return a bool.
17600         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17601         HARD_REGNO_MODE_OK.
17602         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17603         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17604         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17605         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17606         and return a bool.
17607         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17608         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17609         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17610         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17611         return a bool.
17612         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17613         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17614         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17615         (ia64_hard_regno_mode_ok): New function.
17616         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17617         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17618         (iq2000_hard_regno_mode_ok): New function.
17619         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17620         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17621         (lm32_hard_regno_mode_ok): New function.
17622         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17623         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17624         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17625         instead of HARD_REGNO_MODE_OK.
17626         (m32c_hard_regno_ok): Rename to...
17627         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17628         (m32c_cannot_change_mode_class): Update accordingly.
17629         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17630         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17631         (m32r_mode_class): Delete.
17632         (HARD_REGNO_MODE_OK): Delete.
17633         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17634         (m32r_hard_regno_mode_ok): Rename to...
17635         (m32r_hard_regno_modes): ...this.
17636         (m32r_mode_class): Make static.
17637         (m32r_hard_regno_mode_ok): New function.
17638         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17639         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17640         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17641         (m68k_hard_regno_mode_ok): Make static.
17642         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17643         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17644         (mcore_hard_regno_mode_ok): New function.
17645         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17646         (HARD_REGNO_MODE_OK): Delete.
17647         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17648         Rename to...
17649         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17650         (microblaze_hard_regno_mode_ok): New function.
17651         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17652         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17653         (mips_hard_regno_mode_ok): Delete.
17654         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17655         (mips_hard_regno_mode_ok_p): ...this and make static.
17656         (mips_hard_regno_mode_ok_p): Rename to...
17657         (mips_hard_regno_mode_ok_uncached): ...this.
17658         (mips_hard_regno_mode_ok): New function.
17659         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17660         of HARD_REGNO_MODE_OK.
17661         (mips_option_override): Update after above name changes.
17662         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17663         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17664         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17665         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17666         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17667         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17668         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17669         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17670         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17671         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17672         (msp430_hard_regno_mode_ok): Make static and return a bool.
17673         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17674         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17675         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17676         and return a bool.
17677         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17678         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17679         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17680         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17681         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17682         (PA_HARD_REGNO_MODE_OK): ...this
17683         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17684         (PA_HARD_REGNO_MODE_OK): ...this.
17685         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17686         (pa_hard_regno_mode_ok): New function.
17687         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17688         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17689         (pdp11_hard_regno_mode_ok): New function.
17690         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17691         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17692         Delete.
17693         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17694         Make static.
17695         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17696         (rs6000_hard_regno_mode_ok): Rename to...
17697         (rs6000_hard_regno_mode_ok_uncached): ...this.
17698         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17699         (rs6000_hard_regno_mode_ok): New function.
17700         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17701         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17702         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17703         (riscv_hard_regno_mode_ok): ...this and make static.
17704         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17705         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17706         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17707         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17708         (rl78_hard_regno_mode_ok): Make static and return bool.
17709         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17710         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17711         Delete.
17712         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17713         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17714         (rs6000_hard_regno_mode_ok): Rename to...
17715         (rs6000_hard_regno_mode_ok_uncached): ...this.
17716         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17717         (rs6000_hard_regno_mode_ok): New function.
17718         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17719         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17720         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17721         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17722         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17723         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17724         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17725         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17726         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17727         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17728         (sh_hard_regno_mode_ok): Make static.
17729         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17730         instead of HARD_REGNO_MODE_OK.
17731         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17732         (sparc_mode_class): Delete.
17733         (HARD_REGNO_MODE_OK): Delete.
17734         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17735         (hard_regno_mode_classes): Make static.
17736         (sparc_mode_class): Likewise.
17737         (sparc_hard_regno_mode_ok): New function.
17738         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17739         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17740         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17741         function.
17742         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17743         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17744         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17745         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17746         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17747         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17748         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17749         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17750         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17751         (visium_hard_regno_mode_ok): New function.
17752         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17753         instead of HARD_REGNO_MODE_OK.
17754         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17755         (HARD_REGNO_MODE_OK): Delete.
17756         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17757         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17758         (xtensa_option_override): Update accordingly.
17759         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17760         (xtensa_hard_regno_mode_ok): New function.
17761         * system.h (HARD_REGNO_MODE_OK): Poison.
17763 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17764             Alan Hayward  <alan.hayward@arm.com>
17765             David Sherwood  <david.sherwood@arm.com>
17767         * target.def (hard_regno_call_part_clobbered): New hook.
17768         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17769         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17770         * doc/tm.texi: Regenerate.
17771         * hooks.h (hook_bool_uint_mode_false): Declare.
17772         * hooks.c (hook_bool_uint_mode_false): New function.
17773         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17774         * cselib.c (cselib_process_insn): Use
17775         targetm.hard_regno_call_part_clobbered instead of
17776         HARD_REGNO_CALL_PART_CLOBBERED.
17777         * ira-conflicts.c (ira_build_conflicts): Likewise.
17778         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17779         * lra-constraints.c (need_for_call_save_p): Likewise.
17780         * lra-lives.c: Include target.h.
17781         (check_pseudos_live_through_calls): Use
17782         targetm.hard_regno_call_part_clobbered instead of
17783         HARD_REGNO_CALL_PART_CLOBBERED.
17784         * regcprop.c: Include target.h.
17785         (copyprop_hardreg_forward_1): Use
17786         targetm.hard_regno_call_part_clobbered instead of
17787         HARD_REGNO_CALL_PART_CLOBBERED.
17788         * reginfo.c (choose_hard_reg_mode): Likewise.
17789         * regrename.c (check_new_reg_p): Likewise.
17790         * reload.c (find_equiv_reg): Likewise.
17791         * reload1.c (emit_reload_insns): Likewise.
17792         * sched-deps.c (deps_analyze_insn): Likewise.
17793         * sel-sched.c (init_regs_for_mode): Likewise.
17794         (mark_unavailable_hard_regs): Likewise.
17795         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17796         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17797         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17798         New function.
17799         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17800         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17801         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17802         Delete.
17803         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17804         and return a bool.
17805         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17806         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17807         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17808         function.
17809         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17810         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17811         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17812         function.
17813         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17814         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17815         Delete.
17816         * config/powerpcspe/powerpcspe.c
17817         (rs6000_hard_regno_call_part_clobbered): New function.
17818         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17819         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17820         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17821         New function.
17822         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17823         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17824         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17825         function.
17826         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17827         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17828         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17830 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17831             Alan Hayward  <alan.hayward@arm.com>
17832             David Sherwood  <david.sherwood@arm.com>
17834         * rtl.h (subreg_memory_offset): Declare.
17835         * emit-rtl.c (subreg_memory_offset): New function.
17836         * expmed.c (store_bit_field_1): Use it.
17837         * expr.c (undefined_operand_subword_p): Likewise.
17838         * simplify-rtx.c (simplify_subreg): Likewise.
17840 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17842         PR rtl-optimization/57448
17843         PR target/67458
17844         PR target/81316
17845         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17846         using atomic_load pattern.
17847         (expand_atomic_store): Likewise.
17849 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17851         PR sanitizer/81981
17852         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17853         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17854         handling.  Use replace_call_with_value with NULL instead of
17855         gsi_replace, unlink_stmt_vdef and release_defs.
17857         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17858         instead of tab.
17860         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17862 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17864         PR bootstrap/82045
17865         * rtl.h (emit_library_call_value_1): Declare.
17866         (emit_library_call): Replace declaration with a series of overloads.
17867         Remove the parameter count argument.
17868         (emit_library_call_value): Likewise.
17869         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
17870         with an "rtx_mode_t *".
17871         (emit_library_call_value): Delete.
17872         (emit_library_call): Likewise.
17873         * asan.c (asan_emit_stack_protection): Update calls accordingly.
17874         (asan_emit_allocas_unpoison): Likewise.
17875         * builtins.c (expand_builtin_powi): Likewise.
17876         (expand_asan_emit_allocas_unpoison): Likewise.
17877         * cfgexpand.c (expand_main_function): Likewise.
17878         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
17879         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
17880         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
17881         * config/arm/arm.c (arm_trampoline_init): Likewise.
17882         (arm_call_tls_get_addr): Likewise.
17883         (arm_expand_divmod_libfunc): Likewise.
17884         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
17885         (smulsi3_highpart): Likewise.
17886         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
17887         (c6x_expand_compare): Likewise.
17888         (c6x_expand_movmem): Likewise.
17889         * config/frv/frv.c (frv_trampoline_init): Likewise.
17890         * config/i386/i386.c (ix86_trampoline_init): Likewise.
17891         (ix86_expand_divmod_libfunc): Likewise.
17892         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
17893         (ia64_expand_compare): Likewise.
17894         (ia64_profile_hook): Likewise.
17895         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
17896         (nonlocal_goto): Likewise.
17897         (restore_stack_nonlocal): Likewise.
17898         * config/m32r/m32r.c (block_move_call): Likewise.
17899         (m32r_trampoline_init): Likewise.
17900         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
17901         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
17902         (m68k_call_m68k_read_tp): Likewise.
17903         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
17904         (microblaze_expand_divide): Likewise.
17905         * config/mips/mips.h (mips_args): Likewise.
17906         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
17907         (MIPS_ICACHE_SYNC): Likewise.
17908         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
17909         (nios2_trampoline_init): Likewise.
17910         * config/pa/pa.c (hppa_tls_call): Likewise.
17911         (pa_trampoline_init): Likewise.
17912         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
17913         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
17914         (expand_strn_compare): Likewise.
17915         (rs6000_generate_compare): Likewise.
17916         (rs6000_expand_float128_convert): Likewise.
17917         (output_profile_hook): Likewise.
17918         (rs6000_trampoline_init): Likewise.
17919         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
17920         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
17921         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
17922         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
17923         (rs6000_generate_compare): Likewise.
17924         (rs6000_expand_float128_convert): Likewise.
17925         (output_profile_hook): Likewise.
17926         (rs6000_trampoline_init): Likewise.
17927         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
17928         * config/sh/sh.c (sh_trampoline_init): Likewise.
17929         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
17930         (sparc_emit_float_lib_cmp): Likewise.
17931         (sparc32_initialize_trampoline): Likewise.
17932         (sparc64_initialize_trampoline): Likewise.
17933         (sparc_profile_hook): Likewise.
17934         * config/spu/spu.c (ea_load_store): Likewise.
17935         * config/spu/spu.md (floatunssidf2): Likewise.
17936         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
17937         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
17938         * config/visium/visium.c (expand_block_move_4): Likewise.
17939         (expand_block_move_2): Likewise.
17940         (expand_block_move_1): Likewise.
17941         (expand_block_set_4): Likewise.
17942         (expand_block_set_2): Likewise.
17943         (expand_block_set_1): Likewise.
17944         (visium_trampoline_init): Likewise.
17945         (visium_profile_hook): Likewise.
17946         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
17947         (xtensa_setup_frame_addresses): Likewise.
17948         (xtensa_trampoline_init): Likewise.
17949         * except.c (sjlj_emit_function_enter): Likewise.
17950         (sjlj_emit_function_exit): Likewise.
17951         * explow.c (allocate_dynamic_stack_space): Likewise.
17952         (probe_stack_range): Likewise.
17953         * expr.c (convert_mode_scalar): Likewise.
17954         * optabs.c (expand_binop): Likewise.
17955         (expand_twoval_binop_libfunc): Likewise.
17956         (expand_unop): Likewise.
17957         (prepare_cmp_insn): Likewise.
17958         (prepare_float_lib_cmp): Likewise.
17959         (expand_float): Likewise.
17960         (expand_fix): Likewise.
17961         (expand_fixed_convert): Likewise.
17962         (maybe_emit_sync_lock_test_and_set): Likewise.
17963         (expand_atomic_compare_and_swap): Likewise.
17964         (expand_mem_thread_fence): Likewise.
17965         (expand_atomic_fetch_op): Likewise.
17967 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
17969         * doc/generic.texi (OpenACC): Adjust URL.
17970         * doc/invoke.texi (C Dialect Options): Ditto.
17972 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
17974         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
17975         predicate for operand 1.  Add (m,<S>) constraint.
17976         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
17977         Prevent memory operand 1 with register operand 2.
17979 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
17981         PR rtl-optimization/82024
17982         * combine.c (try_combine): If the combination result is a PARALLEL,
17983         and we only need to retain the SET in there that would be placed
17984         at I2, check that we can place that at I3 instead, before doing so.
17986 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
17988         PR target/81766
17989         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
17990         instead of void.
17991         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
17992         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
17993         and label.
17995 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
17996             Jeff Law  <law@redhat.com>
17998         * varasm.c (bss_initializer_p): Do not put constants into .bss
17999         (categorize_decl_for_section): Handle bss_initializer_p returning
18000         false when DECL_INITIAL is NULL.
18002 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18004         PR target/82012
18005         * config/s390/s390.c (s390_can_inline_p): New function.
18007 2017-09-01  Jeff Law  <law@redhat.com>
18009         PR tree-optimization/82052
18010         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18011         Always initialize the returned slot after a hash table miss
18012         when INSERT is true.
18014 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18016         * config/s390/s390.md (mem_signal_fence): Remove.
18017         * doc/md.texi (mem_signal_fence): Remove.
18018         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18019         Update comments.
18020         * target-insns.def (mem_signal_fence): Remove.
18022 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18024         PR sanitizer/81902
18025         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18027         PR sanitizer/81923
18028         * asan.c (create_odr_indicator): Strip name encoding from assembler
18029         name before appending it after __odr_asan_.
18031 2017-09-01  Martin Liska  <mliska@suse.cz>
18033         PR tree-optimization/82059
18034         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18035         frequency only when an edge is redirected.
18037 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18039         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18040         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18041         (arc_conditional_register_usage): Remove ARC600 lp_count
18042         exception.
18043         (arc_file_start): Emit Tag_ARC_CPU_variation.
18044         (arc_can_use_doloop_p): New conditions to use ZOLs.
18045         (hwloop_fail): New function.
18046         (hwloop_optimize): Likewise.
18047         (hwloop_pattern_reg): Likewise.
18048         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18049         (arc_reorg_loops): New function, calls reorg_loops.
18050         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18051         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18052         hwloop_optimize.
18053         (arc_loop_hazard): Remove function, functionality moved into
18054         hwloop_optimize.
18055         (arc_hazard): Remove arc_loop_hazard call.
18056         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18057         into hwloop_optimize.
18058         (arc_label_align): Remove ZOL handling.
18059         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18060         * config/arc/arc.md (doloop_begin): Remove pattern.
18061         (doloop_begin_i): Likewise.
18062         (doloop_end_i): Likewise.
18063         (doloop_fallback): Likewise.
18064         (doloop_fallback_m): Likewise.
18065         (doloop_end): Reimplement expand.
18066         (arc_lp): New pattern for LP instruction.
18067         (loop_end): New pattern.
18068         (loop_fail): Likewise.
18069         (decrement_and_branch_until_zero): Likewise.
18070         * config/arc/arc.opt (mlpc-width): New option.
18071         * doc/invoke.texi (mlpc-width): Document option.
18073 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18075         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18076         (arc_ccfsm_advance): Fix checking for delay slots.
18077         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18079 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18081         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18082         immediates.
18083         (movhi_insn): Update store instruction constraint which are saving
18084         6-bit short immediates.
18085         (movsi_insn): Consider also short scaled load operations.
18086         (zero_extendhisi2_i): Use Usd constraint instead of T.
18087         (extendhisi2_i): Add q constraint.
18088         (arc_clzsi2): Add type and length attributes.
18089         (arc_ctzsi2): Likewise.
18090         * config/arc/constraints.md (Usc): Update constraint, the
18091         assembler can parse two relocations for a single instruction.
18093 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18095         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18096         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18098 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18100         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18101         match as powerpc-wrs-vxworks*.
18103 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18105         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18106         register constraint for by-element operand.
18107         (aarch64_mls_elt_merge<mode>): Likewise.
18109 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18111         * config/arc/arc.c (arc_can_follow_jump): Check for short
18112         branches.
18114 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18116         * config.gcc: Use g.opt for arc.
18117         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18118         functionality moved to ...
18119         (legitimate_scaled_address_p): New function, ...here.
18120         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18121         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18122         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18123         condition.
18124         (arc_override_options): Handle G option.
18125         (arc_output_pic_addr_const): Correct function definition.
18126         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18127         (arc_decl_anon_ns_mem_p): Delete.
18128         (arc_in_small_data_p): Overhaul this function to take into
18129         consideration the value given via G option.
18130         (arc_rewrite_small_data_1): Renamed and corrected old
18131         arc_rewrite_small_data function.
18132         (arc_rewrite_small_data): New function.
18133         (small_data_pattern): Don't use pic_offset_table_rtx.
18134         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18135         * config/arc/simdext.md (movmisalignv2hi): Use
18136         prepare_move_operands function.
18137         (mov*): Likewise.
18138         (movmisalign*): Likewise.
18139         * doc/invoke.texi (ARC options): Document -G option.
18141 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18143         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18144         prototype.
18145         * config/arc/arc.c (arc_print_operand): Output scalled address for
18146         sdata whenever is possible.
18147         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18148         load/stores are available.
18149         (compact_sda_memory_operand): Check for the alignment required by
18150         code density instructions.
18151         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18152         constraint.
18153         * config/arc/constraints.md (Usd): Update constraint.
18154         (Us0): New constraint.
18155         (Usc): Update constraint.
18157 2017-08-31  Richard Biener  <rguenther@suse.de>
18159         PR middle-end/82054
18160         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18161         function only once.
18163 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18165         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18166         Resize type_signature.
18168 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18169             Alan Hayward  <alan.hayward@arm.com>
18170             David Sherwood  <david.sherwood@arm.com>
18172         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18173         subregs whose inner modes can be stored in GPRs.
18174         (aarch64_classify_index): Likewise.
18176 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18177             Alan Hayward  <alan.hayward@arm.com>
18178             David Sherwood  <david.sherwood@arm.com>
18180         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18181         (V_INT_EQUIV): ...this.
18182         (v_cmp_result): Rename to...
18183         (v_int_equiv): ...this.
18184         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18185         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18186         (copysign<mode>3): Likewise.
18187         (aarch64_simd_bsl<mode>_internal): Likewise.
18188         (aarch64_simd_bsl<mode>): Likewise.
18189         (vec_cmp<mode><mode>): Likewise.
18190         (vcond<mode><mode>): Likewise.
18191         (vcond<v_cmp_mixed><mode>): Likewise.
18192         (vcondu<mode><v_cmp_mixed>): Likewise.
18193         (aarch64_cm<optab><mode>): Likewise.
18194         (aarch64_cmtst<mode>): Likewise.
18195         (aarch64_fac<optab><mode>): Likewise.
18196         (vec_perm_const<mode>): Likewise.
18197         (vcond_mask_<mode><v_cmp_result>): Rename to...
18198         (vcond_mask_<mode><v_int_equiv>): ...this.
18199         (vec_cmp<mode><v_cmp_result>): Rename to...
18200         (vec_cmp<mode><v_int_equiv>): ...this.
18202 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18203             Alan Hayward  <alan.hayward@arm.com>
18204             David Sherwood  <david.sherwood@arm.com>
18206         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18207         vector modes.
18208         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18209         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18210         (UNSPEC_LD4_DREG): New unspecs.
18211         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18212         (aarch64_ld2<mode>_dreg_be): Replace with...
18213         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18214         unspec.
18215         (aarch64_ld3<mode>_dreg_le)
18216         (aarch64_ld3<mode>_dreg_be): Replace with...
18217         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18218         unspec.
18219         (aarch64_ld4<mode>_dreg_le)
18220         (aarch64_ld4<mode>_dreg_be): Replace with...
18221         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18222         unspec.
18224 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18226         PR tree-optimization/81987
18227         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18228         insert an initializer in a location not dominated by the stride
18229         definition.
18231 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18233         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18234         on the entire header of the finally block in the fallthru case.
18236 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18238         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18240 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18242         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18243         rs6000_emit_move_from_cr and call renamed function.
18244         (rs6000_emit_prologue): Call renamed functions.
18245         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18246         movesi_from_cr, remove volatile CRs.
18248 2017-08-30  Jon Beniston  <jon@beniston.com>
18249             Richard Biener  <rguenther@suse.de>
18251         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18252         of VECTOR_MODE_P check.
18253         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18254         element vector types.
18256 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18258         * df.h (df_read_modify_subreg_p): Remove in favor of...
18259         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18260         const_rtx instead of an rtx.
18261         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18262         * df-problems.c (df_word_lr_mark_ref): Likewise.
18263         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18264         (mark_pseudo_reg_dead): Likewise.
18265         (mark_ref_dead): Likewise.
18266         * reginfo.c (init_subregs_of_mode): Likewise.
18267         * sched-deps.c (sched_analyze_1): Likewise.
18268         * df-scan.c (df_def_record_1): Likewise.
18269         (df_uses_record): Likewise.
18270         (df_read_modify_subreg_p): Remove in favor of...
18271         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18272         const_rtx instead of an rtx.
18274 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18275             Alan Hayward  <alan.hayward@arm.com>
18276             David Sherwood  <david.sherwood@arm.com>
18278         * rtl.h (partial_subreg_p): New function.
18279         * caller-save.c (save_call_clobbered_regs): Use it.
18280         * calls.c (expand_call): Likewise.
18281         * combine.c (combinable_i3pat): Likewise.
18282         (simplify_set): Likewise.
18283         (make_extraction): Likewise.
18284         (make_compound_operation_int): Likewise.
18285         (gen_lowpart_or_truncate): Likewise.
18286         (force_to_mode): Likewise.
18287         (make_field_assignment): Likewise.
18288         (reg_truncated_to_mode): Likewise.
18289         (record_truncated_value): Likewise.
18290         (move_deaths): Likewise.
18291         * cse.c (record_jump_cond): Likewise.
18292         (cse_insn): Likewise.
18293         * cselib.c (cselib_lookup_1): Likewise.
18294         * expmed.c (extract_bit_field_using_extv): Likewise.
18295         * function.c (assign_parm_setup_reg): Likewise.
18296         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18297         * ira-build.c (create_insn_allocnos): Likewise.
18298         * lra-coalesce.c (merge_pseudos): Likewise.
18299         * lra-constraints.c (match_reload): Likewise.
18300         (simplify_operand_subreg): Likewise.
18301         (curr_insn_transform): Likewise.
18302         * lra-lives.c (process_bb_lives): Likewise.
18303         * lra.c (new_insn_reg): Likewise.
18304         (lra_substitute_pseudo): Likewise.
18305         * regcprop.c (mode_change_ok): Likewise.
18306         (maybe_mode_change): Likewise.
18307         (copyprop_hardreg_forward_1): Likewise.
18308         * reload.c (push_reload): Likewise.
18309         (find_reloads): Likewise.
18310         (find_reloads_subreg_address): Likewise.
18311         * reload1.c (alter_reg): Likewise.
18312         (eliminate_regs_1): Likewise.
18313         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18315 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18317         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18318         back to if statements, including unpack.
18320 2017-08-30  Martin Liska  <mliska@suse.cz>
18322         PR inline-asm/82001
18323         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18324         Rename to ...
18325         (func_checker::compare_asm_inputs_outputs): ... this function.
18326         (func_checker::compare_gimple_asm): Use the function to compare
18327         also ASM constrains.
18328         * ipa-icf-gimple.h: Rename the function.
18330 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18331             Alan Hayward  <alan.hayward@arm.com>
18332             David Sherwood  <david.sherwood@arm.com>
18334         * coretypes.h (complex_mode): New type.
18335         * gdbhooks.py (build_pretty_printer): Handle it.
18336         * machmode.h (complex_mode): New class.
18337         (complex_mode::includes_p): New function.
18338         (is_complex_int_mode): Likewise.
18339         (is_complex_float_mode): Likewise.
18340         * genmodes.c (get_mode_class): Handle complex mode classes.
18341         * function.c (expand_function_end): Use is_complex_int_mode.
18343 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18344             Alan Hayward  <alan.hayward@arm.com>
18345             David Sherwood  <david.sherwood@arm.com>
18347         * coretypes.h (scalar_mode_pod): New typedef.
18348         * gdbhooks.py (build_pretty_printer): Handle it.
18349         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18350         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18351         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18352         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18353         as_a <scalar_mode>.
18355 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18356             Alan Hayward  <alan.hayward@arm.com>
18357             David Sherwood  <david.sherwood@arm.com>
18359         * machmode.h (mode_for_vector): Take a scalar_mode instead
18360         of a machine_mode.
18361         * stor-layout.c (mode_for_vector): Likewise.
18362         * explow.c (promote_mode): Use as_a <scalar_mode>.
18363         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18365 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18366             Alan Hayward  <alan.hayward@arm.com>
18367             David Sherwood  <david.sherwood@arm.com>
18369         * target.def (preferred_simd_mode): Take a scalar_mode
18370         instead of a machine_mode.
18371         * targhooks.h (default_preferred_simd_mode): Likewise.
18372         * targhooks.c (default_preferred_simd_mode): Likewise.
18373         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18374         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18375         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18376         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18377         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18378         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18379         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18380         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18381         Likewise.
18382         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18383         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18384         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18385         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18386         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18387         * doc/tm.texi: Regenerate.
18388         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18389         non-scalar modes.
18391 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18392             Alan Hayward  <alan.hayward@arm.com>
18393             David Sherwood  <david.sherwood@arm.com>
18395         * target.def (scalar_mode_supported_p): Take a scalar_mode
18396         instead of a machine_mode.
18397         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18398         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18399         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18400         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18401         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18402         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18403         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18404         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18405         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18406         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18407         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18408         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18409         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18410         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18411         Likewise.
18412         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18413         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18414         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18415         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18416         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18417         Likewise.
18418         * doc/tm.texi: Regenerate.
18420 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18421             Alan Hayward  <alan.hayward@arm.com>
18422             David Sherwood  <david.sherwood@arm.com>
18424         * coretypes.h (opt_scalar_mode): New typedef.
18425         * gdbhooks.py (build_pretty_printers): Handle it.
18426         * machmode.h (mode_iterator::get_2xwider): Add overload for
18427         opt_mode<T>.
18428         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18429         over scalar modes.
18430         * expr.c (convert_mode_scalar): Likewise.
18431         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18432         * optabs.c (expand_float): Likewise.
18433         (expand_fix): Likewise.
18434         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18436 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18437             Alan Hayward  <alan.hayward@arm.com>
18438             David Sherwood  <david.sherwood@arm.com>
18440         * optabs.c (expand_float): Explicitly check for scalars before
18441         using a branching expansion.
18442         (expand_fix): Likewise.
18444 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18445             Alan Hayward  <alan.hayward@arm.com>
18446             David Sherwood  <david.sherwood@arm.com>
18448         * expr.c (convert_mode): Split scalar handling out into...
18449         (convert_mode_scalar): ...this new function.  Treat the modes
18450         as scalar_modes.
18452 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18453             Alan Hayward  <alan.hayward@arm.com>
18454             David Sherwood  <david.sherwood@arm.com>
18456         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18457         and scalar_mode.
18458         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18460 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18461             Alan Hayward  <alan.hayward@arm.com>
18462             David Sherwood  <david.sherwood@arm.com>
18464         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18465         rather than a machine_mode.
18466         (fixed_from_string): Likewise.
18467         (fixed_convert): Likewise.
18468         (fixed_convert_from_int): Likewise.
18469         (fixed_convert_from_real): Likewise.
18470         (real_convert_from_fixed): Likewise.
18471         * fixed-value.c (fixed_from_double_int): Likewise.
18472         (fixed_from_string): Likewise.
18473         (fixed_convert): Likewise.
18474         (fixed_convert_from_int): Likewise.
18475         (fixed_convert_from_real): Likewise.
18476         (real_convert_from_fixed): Likewise.
18477         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
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         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18484         of separate mode class checks.  Do not allow vector modes here.
18485         (immed_wide_int_const): Use as_a <scalar_mode>.
18486         * explow.c (trunc_int_for_mode): Likewise.
18487         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18488         (wi::shwi): Likewise.
18489         (wi::min_value): Likewise.
18490         (wi::max_value): Likewise.
18491         * dwarf2out.c (loc_descriptor): Likewise.
18492         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18493         for CONST_WIDE_INT.
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         * tree.h (SCALAR_TYPE_MODE): New macro.
18500         * expr.c (expand_expr_addr_expr_1): Use it.
18501         (expand_expr_real_2): Likewise.
18502         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18503         (fold_convert_const_fixed_from_int): Likewise.
18504         (fold_convert_const_fixed_from_real): Likewise.
18505         (native_encode_fixed): Likewise
18506         (native_encode_complex): Likewise
18507         (native_encode_vector): Likewise.
18508         (native_interpret_fixed): Likewise.
18509         (native_interpret_real): Likewise.
18510         (native_interpret_complex): Likewise.
18511         (native_interpret_vector): Likewise.
18512         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18513         (simd_clone_adjust_argument_types): Likewise.
18514         (simd_clone_init_simd_arrays): Likewise.
18515         (simd_clone_adjust): Likewise.
18516         * stor-layout.c (layout_type): Likewise.
18517         * tree.c (build_minus_one_cst): Likewise.
18518         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18519         * tree-inline.c (estimate_move_cost): Likewise.
18520         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18521         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18522         (vectorizable_reduction): Likewise.
18523         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18524         (vect_recog_mixed_size_cond_pattern): Likewise.
18525         (check_bool_pattern): Likewise.
18526         (adjust_bool_pattern): Likewise.
18527         (search_type_for_mask_1): Likewise.
18528         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18529         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18530         (vectorizable_load): Likewise.
18531         (vectorizable_store): Likewise.
18532         * ubsan.c (ubsan_encode_value): Likewise.
18533         * varasm.c (output_constant): Likewise.
18535 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18536             Alan Hayward  <alan.hayward@arm.com>
18537             David Sherwood  <david.sherwood@arm.com>
18539         * coretypes.h (scalar_mode): New class.
18540         * machmode.h (scalar_mode): Likewise.
18541         (scalar_mode::includes_p): New function.
18542         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18543         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18544         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18545         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18546         * expmed.c (store_bit_field_1): Likewise.
18547         (extract_bit_field_1): Likewise.
18548         * expr.c (write_complex_part): Likewise.
18549         (read_complex_part): Likewise.
18550         (emit_move_complex_push): Likewise.
18551         (expand_expr_real_2): Likewise.
18552         * function.c (assign_parm_setup_reg): Likewise.
18553         (assign_parms_unsplit_complex): Likewise.
18554         * optabs.c (expand_binop): Likewise.
18555         * rtlanal.c (subreg_get_info): Likewise.
18556         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18557         * varasm.c (output_constant_pool_2): Likewise.
18559 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18560             Alan Hayward  <alan.hayward@arm.com>
18561             David Sherwood  <david.sherwood@arm.com>
18563         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18564         assertion.
18565         (expmed_mult_highpart_optab): Likewise.
18566         (expmed_mult_highpart): Likewise.
18568 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18569             Alan Hayward  <alan.hayward@arm.com>
18570             David Sherwood  <david.sherwood@arm.com>
18572         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18573         instead of a machine_mode.
18574         (builtin_memset_read_str): Likewise.
18575         * builtins.c (c_readstr): Likewise.
18576         (builtin_memcpy_read_str): Likewise.
18577         (builtin_strncpy_read_str): Likewise.
18578         (builtin_memset_read_str): Likewise.
18579         (builtin_memset_gen_str): Likewise.
18580         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18581         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18582         instead of a machine_mode.
18583         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18584         variables.
18585         (make_extraction): Likewise.
18586         (try_widen_shift_mode): Take and return scalar_int_modes instead
18587         of machine_modes.
18588         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18589         a scalar_int_mode instead of a machine_mode.
18590         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18591         (avr_addr_space_pointer_mode): Likewise.
18592         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18593         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18594         (msp430_unwind_word_mode): Likewise.
18595         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18596         (spu_addr_space_pointer_mode): Likewise.
18597         (spu_addr_space_address_mode): Likewise.
18598         (spu_libgcc_cmp_return_mode): Likewise.
18599         (spu_libgcc_shift_count_mode): Likewise.
18600         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18601         (rl78_addr_space_pointer_mode): Likewise.
18602         (fl78_unwind_word_mode): Likewise.
18603         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18604         machine_mode.
18605         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18606         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18607         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18608         (mips_valid_pointer_mode): Likewise.
18609         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18610         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18611         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18612         of a machine_mode.
18613         (ft32_addr_space_address_mode): Likewise.
18614         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18615         scalar_int_mode instead of a machine_mode.
18616         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18617         of a machine_mode.
18618         (m32c_addr_space_address_mode): Likewise.
18619         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18620         (rs6000_eh_return_filter_mode): Likewise.
18621         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18622         (rs6000_eh_return_filter_mode): Likewise.
18623         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18624         (s390_libgcc_shift_count_mode): Likewise.
18625         (s390_unwind_word_mode): Likewise.
18626         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18627         machine_mode.
18628         * target.def (mode_rep_extended): Likewise.
18629         (valid_pointer_mode): Likewise.
18630         (addr_space.valid_pointer_mode): Likewise.
18631         (eh_return_filter_mode): Return a scalar_int_mode rather than
18632         a machine_mode.
18633         (libgcc_cmp_return_mode): Likewise.
18634         (libgcc_shift_count_mode): Likewise.
18635         (unwind_word_mode): Likewise.
18636         (addr_space.pointer_mode): Likewise.
18637         (addr_space.address_mode): Likewise.
18638         * doc/tm.texi: Regenerate.
18639         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18640         a machine_mode.
18641         (do_jump): Use scalar_int_mode for local variables.
18642         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18643         rather than a machine_mode.
18644         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18645         (scompare_loc_descriptor_wide): Likewise.
18646         (scompare_loc_descriptor_narrow): Likewise.
18647         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18648         variables.
18649         * except.c (sjlj_emit_dispatch_table): Likewise.
18650         (expand_builtin_eh_copy_values): Likewise.
18651         * explow.c (convert_memory_address_addr_space_1): Likewise.
18652         Take a scalar_int_mode rather than a machine_mode.
18653         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18654         than a machine_mode.
18655         (memory_address_addr_space): Use scalar_int_mode for local variables.
18656         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18657         rather than a machine_mode.
18658         * expmed.c (mask_rtx): Likewise.
18659         (init_expmed_one_conv): Likewise.
18660         (expand_mult_highpart_adjust): Likewise.
18661         (extract_high_half): Likewise.
18662         (expmed_mult_highpart_optab): Likewise.
18663         (expmed_mult_highpart): Likewise.
18664         (expand_smod_pow2): Likewise.
18665         (expand_sdiv_pow2): Likewise.
18666         (emit_store_flag_int): Likewise.
18667         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18668         variables.
18669         (extract_low_bits): Likewise.
18670         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18671         a machine_mode.
18672         * expr.c (pieces_addr::adjust):  Likewise.
18673         (can_store_by_pieces): Likewise.
18674         (store_by_pieces): Likewise.
18675         (clear_by_pieces_1): Likewise.
18676         (expand_expr_addr_expr_1): Likewise.
18677         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18678         (expand_expr_real_1): Likewise.
18679         (try_casesi): Likewise.
18680         * final.c (shorten_branches): Likewise.
18681         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18682         type of "mode" to machine_mode.
18683         * internal-fn.c (expand_arith_overflow_result_store): Take a
18684         scalar_int_mode rather than a machine_mode.
18685         (expand_mul_overflow): Use scalar_int_mode for local variables.
18686         * loop-doloop.c (doloop_modify): Likewise.
18687         (doloop_optimize): Likewise.
18688         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18689         than a machine_mode.
18690         (expand_doubleword_shift_condmove): Likewise.
18691         (expand_doubleword_shift): Likewise.
18692         (expand_doubleword_clz): Likewise.
18693         (expand_doubleword_popcount): Likewise.
18694         (expand_doubleword_parity): Likewise.
18695         (expand_absneg_bit): Use scalar_int_mode for local variables.
18696         (prepare_float_lib_cmp): Likewise.
18697         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18698         rather than a machine_mode.
18699         (convert_memory_address_addr_space): Likewise.
18700         (get_mode_bounds): Likewise.
18701         (get_address_mode): Return a scalar_int_mode rather than a
18702         machine_mode.
18703         * rtlanal.c (get_address_mode): Likewise.
18704         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18705         than a machine_mode.
18706         * targhooks.c (default_mode_rep_extended): Likewise.
18707         (default_valid_pointer_mode): Likewise.
18708         (default_addr_space_valid_pointer_mode): Likewise.
18709         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18710         than a machine_mode.
18711         (default_libgcc_cmp_return_mode): Likewise.
18712         (default_libgcc_shift_count_mode): Likewise.
18713         (default_unwind_word_mode): Likewise.
18714         (default_addr_space_pointer_mode): Likewise.
18715         (default_addr_space_address_mode): Likewise.
18716         * targhooks.h (default_eh_return_filter_mode): Likewise.
18717         (default_libgcc_cmp_return_mode): Likewise.
18718         (default_libgcc_shift_count_mode): Likewise.
18719         (default_unwind_word_mode): Likewise.
18720         (default_addr_space_pointer_mode): Likewise.
18721         (default_addr_space_address_mode): Likewise.
18722         (default_mode_rep_extended): Take a scalar_int_mode rather than
18723         a machine_mode.
18724         (default_valid_pointer_mode): Likewise.
18725         (default_addr_space_valid_pointer_mode): Likewise.
18726         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18727         local variables.
18728         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18729         rather than a machine_mode.
18730         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18731         for local variables.
18732         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18733         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18734         than a machine_mode.
18736 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18737             Alan Hayward  <alan.hayward@arm.com>
18738             David Sherwood  <david.sherwood@arm.com>
18740         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18741         the mode argument to scalar_int_mode.
18742         (do_jump_by_parts_zero_rtx): Likewise.
18743         (do_jump_by_parts_equality_rtx): Likewise.
18744         (do_jump_by_parts_greater): Take a mode argument.
18745         (do_jump_by_parts_equality): Likewise.
18746         (do_jump_1): Update calls accordingly.
18748 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18749             Alan Hayward  <alan.hayward@arm.com>
18750             David Sherwood  <david.sherwood@arm.com>
18752         * is-a.h (safe_dyn_cast): New function.
18753         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18754         (jump_table_for_label): Likewise.
18755         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18756         instead of an rtx_insn *.
18757         (shorten_branches): Use dyn_cast instead of LABEL_P and
18758         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18759         rtx_jump_table_data::get_data_mode.
18760         (final_scan_insn): Likewise.
18762 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18763             Alan Hayward  <alan.hayward@arm.com>
18764             David Sherwood  <david.sherwood@arm.com>
18766         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18767         trying to combine a full-register integer set with a subreg
18768         integer set.
18770 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18771             Alan Hayward  <alan.hayward@arm.com>
18772             David Sherwood  <david.sherwood@arm.com>
18774         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18775         that is always either address_mode or pointer_mode.
18777 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18778             Alan Hayward  <alan.hayward@arm.com>
18779             David Sherwood  <david.sherwood@arm.com>
18781         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18782         when the two are known to be equal.
18784 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18785             Alan Hayward  <alan.hayward@arm.com>
18786             David Sherwood  <david.sherwood@arm.com>
18788         * simplify-rtx.c (simplify_const_unary_operation): Use
18789         is_a <scalar_int_mode> instead of checking for a nonzero
18790         precision.  Forcibly convert op_mode to a scalar_int_mode
18791         in that case.  More clearly differentiate the operand and
18792         result modes and use the former when deciding what the value
18793         of a count-bits operation should be.  Use is_int_mode instead
18794         of checking for a MODE_INT.  Remove redundant check for whether
18795         this mode has a zero precision.
18797 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18798             Alan Hayward  <alan.hayward@arm.com>
18799             David Sherwood  <david.sherwood@arm.com>
18801         * optabs.c (widen_leading): Change the type of the mode argument
18802         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18803         (widen_bswap): Likewise.
18804         (expand_parity): Likewise.
18805         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18806         (expand_ffs): Likewise.
18807         (epand_unop): Check for scalar integer modes before calling the
18808         above routines.
18810 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18811             Alan Hayward  <alan.hayward@arm.com>
18812             David Sherwood  <david.sherwood@arm.com>
18814         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18815         Expand commentary.
18816         (expand_expr_real_1): Update call accordingly.
18818 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18819             Alan Hayward  <alan.hayward@arm.com>
18820             David Sherwood  <david.sherwood@arm.com>
18822         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18823         value_mode arguments.  Use scalar_int_mode internally.
18824         (store_bit_field_1): Rename the new integer mode from imode
18825         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18826         to store_split_bit_field, store_bit_field_using_insv and
18827         store_fixed_bit_field.
18828         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18829         Use scalar_int_mode internally.  Use a bit count rather than a mode
18830         when calculating the largest bit size for get_best_mode.
18831         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18832         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18833         Remove assertion that OP0 has a scalar integer mode.
18834         (store_split_bit_field): Add op0_mode and value_mode arguments.
18835         Update calls to extract_fixed_bit_field.
18836         (extract_bit_field_using_extv): Add an op0_mode argument.
18837         Use scalar_int_mode internally.
18838         (extract_bit_field_1): Rename the new integer mode from imode to
18839         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18840         extract_split_bit_field, extract_bit_field_using_extv and
18841         extract_fixed_bit_field.
18842         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18843         to extract_split_bit_field and extract_fixed_bit_field_1.
18844         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18845         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18846         on the target mode.
18847         (extract_split_bit_field): Add an op0_mode argument.  Update call
18848         to extract_fixed_bit_field.
18850 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18851             Alan Hayward  <alan.hayward@arm.com>
18852             David Sherwood  <david.sherwood@arm.com>
18854         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18855         * explow.c (hard_function_value): Likewise.
18856         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
18857         convert_to_mode call outside the loop.
18858         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18859         for the mode iterator.  Require the mode specified by max_pieces
18860         to exist.
18861         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18862         mode iterator.
18863         (copy_blkmode_to_reg): Likewise.
18864         (set_storage_via_setmem): Likewise.
18865         * optabs.c (prepare_cmp_insn): Likewise.
18866         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
18867         * stor-layout.c (finish_bitfield_representative): Likewise.
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         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
18874         * expr.c (convert_move): Use them.
18875         (convert_modes): Likewise.
18876         (store_expr_with_bounds): Likewise.
18878 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18879             Alan Hayward  <alan.hayward@arm.com>
18880             David Sherwood  <david.sherwood@arm.com>
18882         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
18883         parameter for the mode of "x".  Remove the "known_x", "known_mode"
18884         and "known_ret" arguments.  Change the type of the mode argument
18885         to scalar_int_mode.
18886         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
18887         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
18888         (reg_num_sign_bit_copies_for_combine): Likewise.
18889         * rtlanal.c (nonzero_bits1): Likewise.
18890         (num_sign_bit_copies1): Likewise.
18891         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
18892         (reg_num_sign_bit_copies_general): Likewise.
18893         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
18894         (reg_nonzero_bits_general): Likewise.
18896 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18897             Alan Hayward  <alan.hayward@arm.com>
18898             David Sherwood  <david.sherwood@arm.com>
18900         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
18901         than in subroutines.  Return 1 for non-integer modes.
18902         (cached_num_sign_bit_copies): Change the type of the mode parameter
18903         to scalar_int_mode.
18904         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
18905         classes.  Handle CONST_INT_P first and then check whether X also
18906         has a scalar integer mode.  Check the same thing for inner registers
18907         of a SUBREG and for values that are being extended or truncated.
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         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
18914         in subroutines.  Return the mode mask for non-integer modes.
18915         (cached_nonzero_bits): Change the type of the mode parameter
18916         to scalar_int_mode.
18917         (nonzero_bits1): Likewise.  Remove early exit for other mode
18918         classes.  Handle CONST_INT_P first and then check whether X
18919         also has a scalar integer mode.
18921 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18922             Alan Hayward  <alan.hayward@arm.com>
18923             David Sherwood  <david.sherwood@arm.com>
18925         * expr.c (widest_int_mode_for_size): Make the comment match the code.
18926         Return a scalar_int_mode and assert that the size is greater than
18927         one byte.
18928         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
18929         (op_by_pieces_d::op_by_pieces_d): Likewise.
18930         (op_by_pieces_d::run): Likewise.
18931         (can_store_by_pieces): Likewise.
18933 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18934             Alan Hayward  <alan.hayward@arm.com>
18935             David Sherwood  <david.sherwood@arm.com>
18937         * combine.c (extract_left_shift): Add a mode argument and update
18938         recursive calls.
18939         (make_compound_operation_int): Change the type of the mode parameter
18940         to scalar_int_mode and update the call to extract_left_shift.
18942 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18943             Alan Hayward  <alan.hayward@arm.com>
18944             David Sherwood  <david.sherwood@arm.com>
18946         * combine.c (simplify_and_const_int): Change the type of the mode
18947         parameter to scalar_int_mode.
18948         (simplify_and_const_int_1): Likewise.  Update recursive call.
18950 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18951             Alan Hayward  <alan.hayward@arm.com>
18952             David Sherwood  <david.sherwood@arm.com>
18954         * combine.c (simplify_compare_const): Check that the mode is a
18955         scalar_int_mode (rather than VOIDmode) before testing its
18956         precision.
18957         (simplify_comparison): Move COMPARISON_P handling out of the
18958         loop and restrict the latter part of the loop to scalar_int_modes.
18959         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
18960         and when considering SUBREG_REGs.  Use is_int_mode instead of
18961         checking GET_MODE_CLASS against MODE_INT.
18963 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18964             Alan Hayward  <alan.hayward@arm.com>
18965             David Sherwood  <david.sherwood@arm.com>
18967         * combine.c (try_widen_shift_mode): Move check for equal modes to...
18968         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
18969         shift_unit_mode and for modes involved in scalar shifts.
18971 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18972             Alan Hayward  <alan.hayward@arm.com>
18973             David Sherwood  <david.sherwood@arm.com>
18975         * combine.c (force_int_to_mode): New function, split out from...
18976         (force_to_mode): ...here.  Keep xmode up-to-date and use it
18977         instead of GET_MODE (x).
18979 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18980             Alan Hayward  <alan.hayward@arm.com>
18981             David Sherwood  <david.sherwood@arm.com>
18983         * optabs-query.h (extraction_insn::struct_mode): Change type to
18984         opt_scalar_int_mode and update comment.
18985         (extraction_insn::field_mode): Change type to scalar_int_mode.
18986         (extraction_insn::pos_mode): Likewise.
18987         * combine.c (make_extraction): Update accordingly.
18988         * optabs-query.c (get_traditional_extraction_insn): Likewise.
18989         (get_optab_extraction_insn): Likewise.
18990         * recog.c (simplify_while_replacing): Likewise.
18991         * expmed.c (narrow_bit_field_mem): Change the type of the mode
18992         parameter to opt_scalar_int_mode.
18994 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18995             Alan Hayward  <alan.hayward@arm.com>
18996             David Sherwood  <david.sherwood@arm.com>
18998         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
18999         to a scalar_int_mode instead of a machine_mode.
19000         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19001         (get_best_mode): Return a boolean and use a pointer argument to store
19002         the selected mode.  Replace the limit mode parameter with a bit limit.
19003         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19004         for the values returned by bit_field_mode_iterator::next_mode.
19005         (store_bit_field): Update call to get_best_mode.
19006         (store_fixed_bit_field): Likewise.
19007         (extract_fixed_bit_field): Likewise.
19008         * expr.c (optimize_bitfield_assignment_op): Likewise.
19009         * fold-const.c (optimize_bit_field_compare): Likewise.
19010         (fold_truth_andor_1): Likewise.
19011         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19012         Update for new type of m_mode.
19013         (get_best_mode): As above.
19015 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19016             Alan Hayward  <alan.hayward@arm.com>
19017             David Sherwood  <david.sherwood@arm.com>
19019         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19020         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19021         (store_bit_field): Check is_a <scalar_int_mode> before calling
19022         strict_volatile_bitfield_p.
19023         (extract_bit_field): Likewise.
19025 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19026             Alan Hayward  <alan.hayward@arm.com>
19027             David Sherwood  <david.sherwood@arm.com>
19029         * target.def (cstore_mode): Return a scalar_int_mode.
19030         * doc/tm.texi: Regenerate.
19031         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19032         * targhooks.h (default_cstore_mode): Likewise.
19033         * targhooks.c (default_cstore_mode): Likewise, using a forced
19034         conversion.
19035         * expmed.c (emit_cstore): Expect the target of the cstore to be
19036         a scalar_int_mode.
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         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19043         scalar_int_mode.
19044         (niter_desc): Likewise mode.
19045         (iv_analyze): Add a mode parameter.
19046         (biv_p): Likewise.
19047         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19048         and change its type to scalar_int_mode.
19049         * loop-iv.c: Update commentary at head of file.
19050         (iv_constant): Pass the mode paraeter before the rtx it describes
19051         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19052         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19053         (iv_extend): Likewise.
19054         (shorten_into_mode): Likewise.
19055         (iv_add): Use scalar_int_mode.
19056         (iv_mult): Likewise.
19057         (iv_shift): Likewise.
19058         (canonicalize_iv_subregs): Likewise.
19059         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19060         it describes and change its mode to scalar_int_mode.   Also change
19061         the type of the returned inner_mode to scalar_int_mode.
19062         (get_biv_step): Likewise, turning outer_mode from a pointer
19063         into a direct parameter.  Update call to get_biv_step_1.
19064         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19065         iv_constant and get_biv_step.
19066         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19067         and change its type to scalar_int_mode.  Don't initialise iv->mode
19068         to VOIDmode and remove later checks for its still being VOIDmode.
19069         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19070         is_a <scalar_int_mode> when changing the mode under consideration.
19071         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19072         Update call to iv_analyze_expr.
19073         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19074         inner register is not also a scalar_int_mode.  Update call to
19075         iv_analyze_biv.
19076         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19077         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19078         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19079         separate mode class checks.  Update calls to iv_analyze.  Remove
19080         fix-up of VOIDmodes after iv_analyze_biv.
19081         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19082         don't have a scalar_int_mode.  Update call to biv_p.
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         * cfgexpand.c (convert_debug_memory_address): Use
19089         as_a <scalar_int_mode>.
19090         * combine.c (expand_compound_operation): Likewise.
19091         (make_extraction): Likewise.
19092         (change_zero_ext): Likewise.
19093         (simplify_comparison): Likewise.
19094         * cse.c (cse_insn): Likewise.
19095         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19096         (mem_loc_descriptor): Likewise.
19097         (loc_descriptor): Likewise.
19098         * expmed.c (init_expmed_one_mode): Likewise.
19099         (synth_mult): Likewise.
19100         (emit_store_flag_1): Likewise.
19101         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19102         of a comparison with size.
19103         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19104         (reduce_to_bit_field_precision): Likewise.
19105         * function.c (expand_function_end): Likewise.
19106         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19107         * loop-doloop.c (doloop_modify): Likewise.
19108         * optabs.c (expand_binop): Likewise.
19109         (expand_unop): Likewise.
19110         (expand_copysign_absneg): Likewise.
19111         (prepare_cmp_insn): Likewise.
19112         (maybe_legitimize_operand): Likewise.
19113         * recog.c (const_scalar_int_operand): Likewise.
19114         * rtlanal.c (get_address_mode): Likewise.
19115         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19116         (simplify_cond_clz_ctz): Likewise.
19117         * tree-nested.c (get_nl_goto_field): Likewise.
19118         * tree.c (build_vector_type_for_mode): Likewise.
19119         * var-tracking.c (use_narrower_mode): Likewise.
19121 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19122             Alan Hayward  <alan.hayward@arm.com>
19123             David Sherwood  <david.sherwood@arm.com>
19125         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19126         * builtins.c (expand_builtin_signbit): Use it.
19127         * cfgexpand.c (expand_debug_expr): Likewise.
19128         * dojump.c (do_jump): Likewise.
19129         (do_compare_and_jump): Likewise.
19130         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19131         * expmed.c (make_tree): Likewise.
19132         * expr.c (expand_expr_real_2): Likewise.
19133         (expand_expr_real_1): Likewise.
19134         (try_casesi): Likewise.
19135         * fold-const-call.c (fold_const_call_ss): Likewise.
19136         * fold-const.c (unextend): Likewise.
19137         (extract_muldiv_1): Likewise.
19138         (fold_single_bit_test): Likewise.
19139         (native_encode_int): Likewise.
19140         (native_encode_string): Likewise.
19141         (native_interpret_int): Likewise.
19142         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19143         * internal-fn.c (expand_addsub_overflow): Likewise.
19144         (expand_neg_overflow): Likewise.
19145         (expand_mul_overflow): Likewise.
19146         (expand_arith_overflow): Likewise.
19147         * match.pd: Likewise.
19148         * stor-layout.c (layout_type): Likewise.
19149         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19150         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19151         * tree-ssanames.c (get_range_info): Likewise.
19152         * tree-switch-conversion.c (array_value_type) Likewise.
19153         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19154         (vect_recog_divmod_pattern): Likewise.
19155         (vect_recog_mixed_size_cond_pattern): Likewise.
19156         * tree-vrp.c (extract_range_basic): Likewise.
19157         (simplify_float_conversion_using_ranges): Likewise.
19158         * tree.c (int_fits_type_p): Likewise.
19159         * ubsan.c (instrument_bool_enum_load): Likewise.
19160         * varasm.c (mergeable_string_section): Likewise.
19161         (narrowing_initializer_constant_valid_p): Likewise.
19162         (output_constant): Likewise.
19164 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19165             Alan Hayward  <alan.hayward@arm.com>
19166             David Sherwood  <david.sherwood@arm.com>
19168         * machmode.h (NARROWEST_INT_MODE): New macro.
19169         * expr.c (alignment_for_piecewise_move): Use it instead of
19170         GET_CLASS_NARROWEST_MODE (MODE_INT).
19171         (push_block): Likewise.
19172         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19173         Likewise.
19174         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19176 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19177             Alan Hayward  <alan.hayward@arm.com>
19178             David Sherwood  <david.sherwood@arm.com>
19180         * postreload.c (move2add_valid_value_p): Change the type of the
19181         mode parameter to scalar_int_mode.
19182         (move2add_use_add2_insn): Add a mode parameter and use it instead
19183         of GET_MODE (reg).
19184         (move2add_use_add3_insn): Likewise.
19185         (reload_cse_move2add): Update accordingly.
19187 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19188             Alan Hayward  <alan.hayward@arm.com>
19189             David Sherwood  <david.sherwood@arm.com>
19191         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19192         double-word mode.
19193         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19194         * optabs.c (expand_unop): Likewise.
19196 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19197             Alan Hayward  <alan.hayward@arm.com>
19198             David Sherwood  <david.sherwood@arm.com>
19200         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19201         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19202         (popcount_loc_descriptor): Likewise.
19203         (bswap_loc_descriptor): Likewise.
19204         (rotate_loc_descriptor): Likewise.
19205         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19206         calling the functions above.
19208 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19209             Alan Hayward  <alan.hayward@arm.com>
19210             David Sherwood  <david.sherwood@arm.com>
19212         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19213         checks.
19214         (try_combine): Likewise.
19215         (simplify_if_then_else): Likewise.
19216         * cse.c (cse_insn): Likewise.
19217         * dwarf2out.c (mem_loc_descriptor): Likewise.
19218         * emit-rtl.c (gen_lowpart_common): Likewise.
19219         * simplify-rtx.c (simplify_truncation): Likewise.
19220         (simplify_binary_operation_1): Likewise.
19221         (simplify_const_relational_operation): Likewise.
19222         (simplify_ternary_operation): Likewise.
19223         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19225 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19226             Alan Hayward  <alan.hayward@arm.com>
19227             David Sherwood  <david.sherwood@arm.com>
19229         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19230         * reload.c (push_reload): Likewise.
19231         (find_reloads): Likewise.
19233 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19234             Alan Hayward  <alan.hayward@arm.com>
19235             David Sherwood  <david.sherwood@arm.com>
19237         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19238         (make_compound_operation_int): Likewise.
19239         (change_zero_ext): Likewise.
19240         * expr.c (convert_move): Likewise.
19241         (convert_modes): Likewise.
19242         * fwprop.c (forward_propagate_subreg): Likewise.
19243         * loop-iv.c (get_biv_step_1): Likewise.
19244         * optabs.c (widen_operand): Likewise.
19245         * postreload.c (move2add_valid_value_p): Likewise.
19246         * recog.c (simplify_while_replacing): Likewise.
19247         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19248         (simplify_binary_operation_1): Likewise.  Remove redundant
19249         mode equality check.
19251 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19252             Alan Hayward  <alan.hayward@arm.com>
19253             David Sherwood  <david.sherwood@arm.com>
19255         * combine.c (combine_simplify_rtx): Add checks for
19256         is_a <scalar_int_mode>.
19257         (simplify_if_then_else): Likewise.
19258         (make_field_assignment): Likewise.
19259         (simplify_comparison): Likewise.
19260         * ifcvt.c (noce_try_bitop): Likewise.
19261         * loop-invariant.c (canonicalize_address_mult): Likewise.
19262         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19264 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19265             Alan Hayward  <alan.hayward@arm.com>
19266             David Sherwood  <david.sherwood@arm.com>
19268         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19269         is_a <scalar_int_mode> instead of != BLKmode.
19271 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19272             Alan Hayward  <alan.hayward@arm.com>
19273             David Sherwood  <david.sherwood@arm.com>
19275         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19276         instead of != VOIDmode.
19277         * combine.c (if_then_else_cond): Likewise.
19278         (change_zero_ext): Likewise.
19279         * dwarf2out.c (mem_loc_descriptor): Likewise.
19280         (loc_descriptor): Likewise.
19281         * rtlanal.c (canonicalize_condition): Likewise.
19282         * simplify-rtx.c (simplify_relational_operation_1): 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         * simplify-rtx.c (simplify_binary_operation_1): Use
19289         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19291 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19292             Alan Hayward  <alan.hayward@arm.com>
19293             David Sherwood  <david.sherwood@arm.com>
19295         * wide-int.h (int_traits<unsigned char>) New class.
19296         (int_traits<unsigned short>) Likewise.
19297         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19298         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19299         SCALAR_INT_MODE_P.
19300         * combine.c (set_nonzero_bits_and_sign_copies): Use
19301         is_a <scalar_int_mode>.
19302         (find_split_point): Likewise.
19303         (combine_simplify_rtx): Likewise.
19304         (simplify_logical): Likewise.
19305         (expand_compound_operation): Likewise.
19306         (expand_field_assignment): Likewise.
19307         (make_compound_operation): Likewise.
19308         (extended_count): Likewise.
19309         (change_zero_ext): Likewise.
19310         (simplify_comparison): Likewise.
19311         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19312         (ucompare_loc_descriptor): Likewise.
19313         (minmax_loc_descriptor): Likewise.
19314         (mem_loc_descriptor): Likewise.
19315         (loc_descriptor): Likewise.
19316         * expmed.c (init_expmed_one_mode): Likewise.
19317         * lra-constraints.c (lra_constraint_offset): Likewise.
19318         * optabs.c (prepare_libcall_arg): Likewise.
19319         * postreload.c (move2add_note_store): Likewise.
19320         * reload.c (operands_match_p): Likewise.
19321         * rtl.h (load_extend_op): Likewise.
19322         * rtlhooks.c (gen_lowpart_general): Likewise.
19323         * simplify-rtx.c (simplify_truncation): Likewise.
19324         (simplify_unary_operation_1): Likewise.
19325         (simplify_binary_operation_1): Likewise.
19326         (simplify_const_binary_operation): Likewise.
19327         (simplify_const_relational_operation): Likewise.
19328         (simplify_subreg): Likewise.
19329         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19330         * var-tracking.c (adjust_mems): Likewise.
19331         (prepare_call_arguments): Likewise.
19333 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19334             Alan Hayward  <alan.hayward@arm.com>
19335             David Sherwood  <david.sherwood@arm.com>
19337         * machmode.h (is_int_mode): New fuction.
19338         * combine.c (find_split_point): Use it.
19339         (combine_simplify_rtx): Likewise.
19340         (simplify_if_then_else): Likewise.
19341         (simplify_set): Likewise.
19342         (simplify_shift_const_1): Likewise.
19343         (simplify_comparison): Likewise.
19344         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19345         * cse.c (notreg_cost): Likewise.
19346         (cse_insn): Likewise.
19347         * cselib.c (cselib_lookup_1): Likewise.
19348         * dojump.c (do_jump_1): Likewise.
19349         (do_compare_rtx_and_jump): Likewise.
19350         * dse.c (get_call_args): Likewise.
19351         * dwarf2out.c (rtl_for_decl_init): Likewise.
19352         (native_encode_initializer): Likewise.
19353         * expmed.c (emit_store_flag_1): Likewise.
19354         (emit_store_flag): Likewise.
19355         * expr.c (convert_modes): Likewise.
19356         (store_field): Likewise.
19357         (expand_expr_real_1): Likewise.
19358         * fold-const.c (fold_read_from_constant_string): Likewise.
19359         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19360         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19361         * optabs.c (expand_binop): Likewise.
19362         (expand_unop): Likewise.
19363         (expand_abs_nojump): Likewise.
19364         (expand_one_cmpl_abs_nojump): Likewise.
19365         * simplify-rtx.c (mode_signbit_p): Likewise.
19366         (val_signbit_p): Likewise.
19367         (val_signbit_known_set_p): Likewise.
19368         (val_signbit_known_clear_p): Likewise.
19369         (simplify_relational_operation_1): Likewise.
19370         * tree.c (vector_type_mode): Likewise.
19372 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19373             Alan Hayward  <alan.hayward@arm.com>
19374             David Sherwood  <david.sherwood@arm.com>
19376         * machmode.h (smallest_mode_for_size): Fix formatting.
19377         (smallest_int_mode_for_size): New function.
19378         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19379         instead of smallest_mode_for_size.
19380         * combine.c (make_extraction): Likewise.
19381         * config/arc/arc.c (arc_expand_movmem): Likewise.
19382         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19383         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19384         * config/s390/s390.c (s390_expand_insv): Likewise.
19385         * config/sparc/sparc.c (assign_int_registers): Likewise.
19386         * config/spu/spu.c (spu_function_value): Likewise.
19387         (spu_function_arg): Likewise.
19388         * coverage.c (get_gcov_type): Likewise.
19389         (get_gcov_unsigned_t): Likewise.
19390         * dse.c (find_shift_sequence): Likewise.
19391         * expmed.c (store_bit_field_1): Likewise.
19392         * expr.c (convert_move): Likewise.
19393         (store_field): Likewise.
19394         * internal-fn.c (expand_arith_overflow): Likewise.
19395         * optabs-query.c (get_best_extraction_insn): Likewise.
19396         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19397         * stor-layout.c (layout_type): Likewise.
19398         (initialize_sizetypes): Likewise.
19399         * targhooks.c (default_get_mask_mode): Likewise.
19400         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19402 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19403             Alan Hayward  <alan.hayward@arm.com>
19404             David Sherwood  <david.sherwood@arm.com>
19406         * machmode.h (opt_mode::else_blk): New function.
19407         (int_mode_for_mode): Declare.
19408         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19409         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19410         return type.
19411         * cfgexpand.c (expand_debug_expr): Likewise.
19412         * combine.c (gen_lowpart_or_truncate): Likewise.
19413         (gen_lowpart_for_combine): Likewise.
19414         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19415         * config/avr/avr.c (avr_to_int_mode): Likewise.
19416         (avr_out_plus_1): Likewise.
19417         (avr_out_plus): Likewise.
19418         (avr_out_round): Likewise.
19419         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19420         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19421         (s390_expand_vcond): Likewise.
19422         * config/spu/spu.c (spu_split_immediate): Likewise.
19423         (spu_expand_mov): Likewise.
19424         * dse.c (get_stored_val): Likewise.
19425         * expmed.c (store_bit_field_1): Likewise.
19426         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19427         int_mode_for_size.
19428         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19429         (extract_low_bits): Likewise.
19430         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19431         handling rather than repeating the check.
19432         (emit_group_store): Likewise.
19433         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19434         * optabs.c (expand_absneg_bit): Likewise.
19435         (expand_copysign_absneg): Likewise.
19436         (expand_copysign_bit): Likewise.
19437         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19438         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19439         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19440         * var-tracking.c (prepare_call_arguments):  Likewise.
19441         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19442         int_mode_for_mode instead of mode_for_size.
19443         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19445 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19446             Alan Hayward  <alan.hayward@arm.com>
19447             David Sherwood  <david.sherwood@arm.com>
19449         * machmode.h (int_mode_for_size): New function.
19450         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19451         instead of mode_for_size.
19452         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19453         explicit.
19454         * combine.c (expand_field_assignment): Use int_mode_for_size
19455         instead of mode_for_size.
19456         (make_extraction): Likewise.
19457         (simplify_shift_const_1): Likewise.
19458         (simplify_comparison): Likewise.
19459         * dojump.c (do_jump): Likewise.
19460         * dwarf2out.c (mem_loc_descriptor): Likewise.
19461         * emit-rtl.c (init_derived_machine_modes): Likewise.
19462         * expmed.c (flip_storage_order): Likewise.
19463         (convert_extracted_bit_field): Likewise.
19464         * expr.c (copy_blkmode_from_reg): Likewise.
19465         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19466         * internal-fn.c (expand_mul_overflow): Likewise.
19467         * lower-subreg.c (simple_move): Likewise.
19468         * optabs-libfuncs.c (init_optabs): Likewise.
19469         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19470         * tree.c (vector_type_mode): Likewise.
19471         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19472         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19473         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19474         * tree-vect-stmts.c (vectorizable_load): Likewise.
19475         (vectorizable_store): Likewise.
19477 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19478             Alan Hayward  <alan.hayward@arm.com>
19479             David Sherwood  <david.sherwood@arm.com>
19481         * coretypes.h (pod_mode): New type.
19482         (scalar_int_mode_pod): New typedef.
19483         * machmode.h (pod_mode): New class.
19484         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19485         * genmodes.c (emit_mode_int_n): Update accordingly.
19486         * lower-subreg.h (target_lower_subreg): Change type to
19487         scalar_int_mode_pod.
19488         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19489         scalar_int_mode_pod.
19491 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19492             Alan Hayward  <alan.hayward@arm.com>
19493             David Sherwood  <david.sherwood@arm.com>
19495         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19496         machine_mode to scalar_int_mode.
19497         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19498         (rs6000_option_override_internal): Remove cast to int.
19499         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19500         machine_mode to scalar_int_mode.
19501         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19502         (rs6000_option_override_internal): Remove cast to int.
19503         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19504         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19505         to machine_mode.
19506         * config/s390/s390.c (s390_expand_builtin): Likewise.
19507         * coretypes.h (scalar_int_mode): New type.
19508         (opt_scalar_int_mode): New typedef.
19509         * machmode.h (scalar_int_mode): New class.
19510         (scalar_int_mode::includes_p): New function.
19511         (byte_mode): Change type to scalar_int_mode.
19512         (word_mode): Likewise.
19513         (ptr_mode): Likewise.
19514         * emit-rtl.c (byte_mode): Likewise.
19515         (word_mode): Likewise.
19516         (ptr_mode): Likewise.
19517         (init_derived_machine_modes): Update accordingly.
19518         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19519         and MODE_PARTIAL_INT.
19520         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19521         opt_scalar_int_mode.
19523 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19524             Alan Hayward  <alan.hayward@arm.com>
19525             David Sherwood  <david.sherwood@arm.com>
19527         * target.def (libgcc_floating_mode_supported_p): Take a
19528         scalar_float_mode.
19529         * doc/tm.texi: Regenerate.
19530         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19531         scalar_float_mode.
19532         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19533         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19534         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         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19541         * doc/tm.texi: Regenerate.
19542         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19543         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19544         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19545         * targhooks.h (default_floatn_mode): Likewise.
19546         * targhooks.c (default_floatn_mode): Likewise.
19547         * tree.c (build_common_tree_nodes): Update accordingly.
19549 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19550             Alan Hayward  <alan.hayward@arm.com>
19551             David Sherwood  <david.sherwood@arm.com>
19553         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19554         (mode_iterator::iterate_p): Likewise.
19555         (mode_iterator::get_wider): Likewise.
19556         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19558 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19559             Alan Hayward  <alan.hayward@arm.com>
19560             David Sherwood  <david.sherwood@arm.com>
19562         * coretypes.h (opt_scalar_float_mode): New typedef.
19563         * machmode.h (float_mode_for_size): New function.
19564         * emit-rtl.c (double_mode): Delete.
19565         (init_emit_once): Use float_mode_for_size.
19566         * stor-layout.c (layout_type): Likewise.
19567         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19569 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19570             Alan Hayward  <alan.hayward@arm.com>
19571             David Sherwood  <david.sherwood@arm.com>
19573         * output.h (assemble_real): Take a scalar_float_mode.
19574         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19575         * config/arm/arm.md (consttable_4): Likewise.
19576         (consttable_8): Likewise.
19577         (consttable_16): Likewise.
19578         * config/mips/mips.md (consttable_float): Likewise.
19579         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19580         * varasm.c (assemble_real): Take a scalar_float_mode.
19581         (output_constant_pool_2): Update accordingly.
19582         (output_constant): Likewise.
19584 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19585             Alan Hayward  <alan.hayward@arm.com>
19586             David Sherwood  <david.sherwood@arm.com>
19588         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19589         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19590         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19591         (native_encode_real): Likewise.
19592         (native_interpret_real): Likewise.
19593         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19594         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19596 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19597             Alan Hayward  <alan.hayward@arm.com>
19598             David Sherwood  <david.sherwood@arm.com>
19600         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19601         <scalar_float_mode>.  Simplify.
19602         (gen_extend_conv_libfunc): Likewise.
19604 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19605             Alan Hayward  <alan.hayward@arm.com>
19606             David Sherwood  <david.sherwood@arm.com>
19608         * coretypes.h (scalar_float_mode): New type.
19609         * machmode.h (mode_traits::from_int): Use machine_mode if
19610         USE_ENUM_MODES is defined.
19611         (is_a): New function.
19612         (as_a): Likewise.
19613         (dyn_cast): Likewise.
19614         (scalar_float_mode): New class.
19615         (scalar_float_mode::includes_p): New function.
19616         (is_float_mode): Likewise.
19617         * gdbhooks.py (MachineModePrinter): New class.
19618         (build_pretty_printer): Use it for scalar_float_mode.
19619         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19620         (format_helper::format_helper): Turn into a template.
19621         * genmodes.c (get_mode_class): New function.
19622         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19623         or machine_mode if none.
19624         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19625         as_a <scalar_float_mode>.
19626         * dwarf2out.c (mem_loc_descriptor): Likewise.
19627         (insert_float): Likewise.
19628         (add_const_value_attribute): Likewise.
19629         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19630         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19631         (expand_unop): Update accordingly.
19632         (expand_abs_nojump): Likewise.
19633         (expand_copysign_absneg): Take a scalar_float_mode.
19634         (expand_copysign_bit): Likewise.
19635         (expand_copysign): Update accordingly.
19637 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19638             Alan Hayward  <alan.hayward@arm.com>
19639             David Sherwood  <david.sherwood@arm.com>
19641         * coretypes.h (opt_mode): New class.
19642         * machmode.h (opt_mode): Likewise.
19643         (opt_mode::else_void): New function.
19644         (opt_mode::require): Likewise.
19645         (opt_mode::exists): Likewise.
19646         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19647         (GET_MODE_2XWIDER_MODE): Likewise.
19648         (mode_iterator::get_wider): Update accordingly.
19649         (mode_iterator::get_2xwider): Likewise.
19650         (mode_iterator::get_known_wider): Likewise, turning into a template.
19651         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19652         forcing a wider mode to exist.
19653         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19654         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19655         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19656         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19657         * lower-subreg.c (init_lower_subreg): Likewise.
19658         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19659         on the final iteration.
19660         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19661         a wider mode exists before asking for a move pattern.
19662         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19663         forcing a wider mode to exist.
19664         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19665         returning false if no such mode exists.
19666         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19667         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19668         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19669         Avoid checking for a MODE_INT if we already know the mode is not a
19670         SCALAR_INT_MODE_P.
19671         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19672         forcing a wider mode to exist.
19673         (expmed_mult_highpart_optab): Likewise.
19674         (expmed_mult_highpart): Likewise.
19675         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19676         using else_void.
19677         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19678         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19679         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19680         * internal-fn.c (expand_mul_overflow): Update use of
19681         GET_MODE_2XWIDER_MODE.
19682         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19683         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19684         GET_MODE_WIDER_MODE.
19685         (convert_plusminus_to_widen): Likewise.
19686         * tree-switch-conversion.c (array_value_type): Likewise.
19687         * var-tracking.c (emit_note_insn_var_location): Likewise.
19688         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19689         Return false inside rather than outside the loop if no wider mode
19690         exists
19691         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19692         and GET_MODE_2XWIDER_MODE
19693         (can_compare_p): Use else_void.
19694         * gdbhooks.py (OptMachineModePrinter): New class.
19695         (build_pretty_printer): Use it for opt_mode.
19697 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19698             Alan Hayward  <alan.hayward@arm.com>
19699             David Sherwood  <david.sherwood@arm.com>
19701         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19702         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19704 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19705             Alan Hayward  <alan.hayward@arm.com>
19706             David Sherwood  <david.sherwood@arm.com>
19708         * machmode.h (mode_traits): New structure.
19709         (get_narrowest_mode): New function.
19710         (mode_iterator::start): Likewise.
19711         (mode_iterator::iterate_p): Likewise.
19712         (mode_iterator::get_wider): Likewise.
19713         (mode_iterator::get_known_wider): Likewise.
19714         (mode_iterator::get_2xwider): Likewise.
19715         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19716         (FOR_EACH_MODE): Likewise.
19717         (FOR_EACH_MODE_FROM): Likewise.
19718         (FOR_EACH_MODE_UNTIL): Likewise.
19719         (FOR_EACH_WIDER_MODE): Likewise.
19720         (FOR_EACH_2XWIDER_MODE): Likewise.
19721         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19722         * combine.c (simplify_comparison): Likewise
19723         * config/i386/i386.c (type_natural_mode): Likewise.
19724         * cse.c (cse_insn): Likewise.
19725         * dse.c (find_shift_sequence): Likewise.
19726         * emit-rtl.c (init_derived_machine_modes): Likewise.
19727         (init_emit_once): Likewise.
19728         * explow.c (hard_function_value): Likewise.
19729         * expmed.c (extract_fixed_bit_field_1): Likewise.
19730         (extract_bit_field_1): Likewise.
19731         (expand_divmod): Likewise.
19732         (emit_store_flag_1): Likewise.
19733         * expr.c (init_expr_target): Likewise.
19734         (convert_move): Likewise.
19735         (alignment_for_piecewise_move): Likewise.
19736         (widest_int_mode_for_size): Likewise.
19737         (emit_block_move_via_movmem): Likewise.
19738         (copy_blkmode_to_reg): Likewise.
19739         (set_storage_via_setmem): Likewise.
19740         (compress_float_constant): Likewise.
19741         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19742         * optabs-query.c (get_best_extraction_insn): Likewise.
19743         * optabs.c (expand_binop): Likewise.
19744         (expand_twoval_unop): Likewise.
19745         (expand_twoval_binop): Likewise.
19746         (widen_leading): Likewise.
19747         (widen_bswap): Likewise.
19748         (expand_parity): Likewise.
19749         (expand_unop): Likewise.
19750         (prepare_cmp_insn): Likewise.
19751         (prepare_float_lib_cmp): Likewise.
19752         (expand_float): Likewise.
19753         (expand_fix): Likewise.
19754         (expand_sfix_optab): Likewise.
19755         * postreload.c (move2add_use_add2_insn): Likewise.
19756         * reg-stack.c (reg_to_stack): Likewise.
19757         * reginfo.c (choose_hard_reg_mode): Likewise.
19758         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19759         * stor-layout.c (mode_for_size): Likewise.
19760         (smallest_mode_for_size): Likewise.
19761         (mode_for_vector): Likewise.
19762         (finish_bitfield_representative): Likewise.
19763         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19764         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19765         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19766         * var-tracking.c (prepare_call_arguments): Likewise.
19768 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19769             Alan Hayward  <alan.hayward@arm.com>
19770             David Sherwood  <david.sherwood@arm.com>
19772         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19773         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19774         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19775         * machmode.h (mode_size): Move earlier in file.
19776         (mode_precision): Likewise.
19777         (mode_inner): Likewise.
19778         (mode_nunits): Likewise.
19779         (mode_unit_size): Likewise.
19780         (unit_unit_precision): Likewise.
19781         (mode_wider): Likewise.
19782         (mode_2xwider): Likewise.
19783         (machine_mode): New class.
19784         (mode_to_bytes): New function.
19785         (mode_to_bits): Likewise.
19786         (mode_to_precision): Likewise.
19787         (mode_to_inner): Likewise.
19788         (mode_to_unit_size): Likewise.
19789         (mode_to_unit_precision): Likewise.
19790         (mode_to_nunits): Likewise.
19791         (GET_MODE_SIZE): Use mode_to_bytes.
19792         (GET_MODE_BITSIZE): Use mode_to_bits.
19793         (GET_MODE_PRECISION): Use mode_to_precision.
19794         (GET_MODE_INNER): Use mode_to_inner.
19795         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19796         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19797         (GET_MODE_NUNITS): Use mode_to_nunits.
19798         * system.h (ALWAYS_INLINE): New macro.
19799         * config/powerpcspe/powerpcspe-c.c
19800         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19801         int for arg1_mode and arg2_mode.
19803 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19804             Alan Hayward  <alan.hayward@arm.com>
19805             David Sherwood  <david.sherwood@arm.com>
19807         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19808         Prefix mode names with E_ in case statements.
19809         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19810         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19811         (aarch64_split_simd_move): Likewise.
19812         (aarch64_gen_storewb_pair): Likewise.
19813         (aarch64_gen_loadwb_pair): Likewise.
19814         (aarch64_gen_store_pair): Likewise.
19815         (aarch64_gen_load_pair): Likewise.
19816         (aarch64_get_condition_code_1): Likewise.
19817         (aarch64_constant_pool_reload_icode): Likewise.
19818         (get_rsqrte_type): Likewise.
19819         (get_rsqrts_type): Likewise.
19820         (get_recpe_type): Likewise.
19821         (get_recps_type): Likewise.
19822         (aarch64_gimplify_va_arg_expr): Likewise.
19823         (aarch64_simd_container_mode): Likewise.
19824         (aarch64_emit_load_exclusive): Likewise.
19825         (aarch64_emit_store_exclusive): Likewise.
19826         (aarch64_expand_compare_and_swap): Likewise.
19827         (aarch64_gen_atomic_cas): Likewise.
19828         (aarch64_emit_bic): Likewise.
19829         (aarch64_emit_atomic_swap): Likewise.
19830         (aarch64_emit_atomic_load_op): Likewise.
19831         (aarch64_evpc_trn): Likewise.
19832         (aarch64_evpc_uzp): Likewise.
19833         (aarch64_evpc_zip): Likewise.
19834         (aarch64_evpc_ext): Likewise.
19835         (aarch64_evpc_rev): Likewise.
19836         (aarch64_evpc_dup): Likewise.
19837         (aarch64_gen_ccmp_first): Likewise.
19838         (aarch64_gen_ccmp_next): Likewise.
19839         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19840         (alpha_emit_xfloating_libcall): Likewise.
19841         (emit_insxl): Likewise.
19842         (alpha_arg_type): Likewise.
19843         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19844         (arc_preferred_simd_mode): Likewise.
19845         (arc_secondary_reload): Likewise.
19846         (get_arc_condition_code): Likewise.
19847         (arc_print_operand): Likewise.
19848         (arc_legitimate_constant_p): Likewise.
19849         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19850         * config/arc/arc.md (casesi_load): Likewise.
19851         (casesi_compact_jump): Likewise.
19852         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19853         (cc_use_register): Likewise.
19854         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19855         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19856         (arm_init_iwmmxt_builtins): Likewise.
19857         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19858         (neon_expand_vector_init): Likewise.
19859         (arm_attr_length_move_neon): Likewise.
19860         (maybe_get_arm_condition_code): Likewise.
19861         (arm_emit_vector_const): Likewise.
19862         (arm_preferred_simd_mode): Likewise.
19863         (arm_output_iwmmxt_tinsr): Likewise.
19864         (thumb1_output_casesi): Likewise.
19865         (thumb2_output_casesi): Likewise.
19866         (arm_emit_load_exclusive): Likewise.
19867         (arm_emit_store_exclusive): Likewise.
19868         (arm_expand_compare_and_swap): Likewise.
19869         (arm_evpc_neon_vuzp): Likewise.
19870         (arm_evpc_neon_vzip): Likewise.
19871         (arm_evpc_neon_vrev): Likewise.
19872         (arm_evpc_neon_vtrn): Likewise.
19873         (arm_evpc_neon_vext): Likewise.
19874         (arm_validize_comparison): Likewise.
19875         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
19876         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
19877         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
19878         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
19879         (c6x_preferred_simd_mode): Likewise.
19880         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
19881         (epiphany_rtx_costs): Likewise.
19882         * config/epiphany/predicates.md (proper_comparison_operator):
19883         Likewise.
19884         * config/frv/frv.c (condexec_memory_operand): Likewise.
19885         (frv_emit_move): Likewise.
19886         (output_move_single): Likewise.
19887         (output_condmove_single): Likewise.
19888         (frv_hard_regno_mode_ok): Likewise.
19889         (frv_matching_accg_mode): Likewise.
19890         * config/h8300/h8300.c (split_adds_subs): Likewise.
19891         (h8300_rtx_costs): Likewise.
19892         (h8300_print_operand): Likewise.
19893         (compute_mov_length): Likewise.
19894         (output_logical_op): Likewise.
19895         (compute_logical_op_length): Likewise.
19896         (compute_logical_op_cc): Likewise.
19897         (h8300_shift_needs_scratch_p): Likewise.
19898         (output_a_shift): Likewise.
19899         (compute_a_shift_length): Likewise.
19900         (compute_a_shift_cc): Likewise.
19901         (expand_a_rotate): Likewise.
19902         (output_a_rotate): Likewise.
19903         * config/i386/i386.c (classify_argument): Likewise.
19904         (function_arg_advance_32): Likewise.
19905         (function_arg_32): Likewise.
19906         (function_arg_64): Likewise.
19907         (function_value_64): Likewise.
19908         (ix86_gimplify_va_arg): Likewise.
19909         (ix86_legitimate_constant_p): Likewise.
19910         (put_condition_code): Likewise.
19911         (split_double_mode): Likewise.
19912         (ix86_avx256_split_vector_move_misalign): Likewise.
19913         (ix86_expand_vector_logical_operator): Likewise.
19914         (ix86_split_idivmod): Likewise.
19915         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
19916         (ix86_build_const_vector): Likewise.
19917         (ix86_build_signbit_mask): Likewise.
19918         (ix86_match_ccmode): Likewise.
19919         (ix86_cc_modes_compatible): Likewise.
19920         (ix86_expand_branch): Likewise.
19921         (ix86_expand_sse_cmp): Likewise.
19922         (ix86_expand_sse_movcc): Likewise.
19923         (ix86_expand_int_sse_cmp): Likewise.
19924         (ix86_expand_vec_perm_vpermi2): Likewise.
19925         (ix86_expand_vec_perm): Likewise.
19926         (ix86_expand_sse_unpack): Likewise.
19927         (ix86_expand_int_addcc): Likewise.
19928         (ix86_split_to_parts): Likewise.
19929         (ix86_vectorize_builtin_gather): Likewise.
19930         (ix86_vectorize_builtin_scatter): Likewise.
19931         (avx_vpermilp_parallel): Likewise.
19932         (inline_memory_move_cost): Likewise.
19933         (ix86_tieable_integer_mode_p): Likewise.
19934         (x86_maybe_negate_const_int): Likewise.
19935         (ix86_expand_vector_init_duplicate): Likewise.
19936         (ix86_expand_vector_init_one_nonzero): Likewise.
19937         (ix86_expand_vector_init_one_var): Likewise.
19938         (ix86_expand_vector_init_concat): Likewise.
19939         (ix86_expand_vector_init_interleave): Likewise.
19940         (ix86_expand_vector_init_general): Likewise.
19941         (ix86_expand_vector_set): Likewise.
19942         (ix86_expand_vector_extract): Likewise.
19943         (emit_reduc_half): Likewise.
19944         (ix86_emit_i387_round): Likewise.
19945         (ix86_mangle_type): Likewise.
19946         (ix86_expand_round_sse4): Likewise.
19947         (expand_vec_perm_blend): Likewise.
19948         (canonicalize_vector_int_perm): Likewise.
19949         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
19950         (expand_vec_perm_1): Likewise.
19951         (expand_vec_perm_interleave3): Likewise.
19952         (expand_vec_perm_even_odd_pack): Likewise.
19953         (expand_vec_perm_even_odd_1): Likewise.
19954         (expand_vec_perm_broadcast_1): Likewise.
19955         (ix86_vectorize_vec_perm_const_ok): Likewise.
19956         (ix86_expand_vecop_qihi): Likewise.
19957         (ix86_expand_mul_widen_hilo): Likewise.
19958         (ix86_expand_sse2_abs): Likewise.
19959         (ix86_expand_pextr): Likewise.
19960         (ix86_expand_pinsr): Likewise.
19961         (ix86_preferred_simd_mode): Likewise.
19962         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
19963         * config/i386/sse.md (*andnot<mode>3): Likewise.
19964         (<mask_codefor><code><mode>3<mask_name>): Likewise.
19965         (*<code><mode>3): Likewise.
19966         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
19967         (ia64_expand_atomic_op): Likewise.
19968         (ia64_arg_type): Likewise.
19969         (ia64_mode_to_int): Likewise.
19970         (ia64_scalar_mode_supported_p): Likewise.
19971         (ia64_vector_mode_supported_p): Likewise.
19972         (expand_vec_perm_broadcast): Likewise.
19973         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
19974         (iq2000_function_arg_advance): Likewise.
19975         (iq2000_function_arg): Likewise.
19976         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
19977         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
19978         (m68k_libcall_value): Likewise.
19979         (m68k_function_value): Likewise.
19980         (sched_attr_op_type): Likewise.
19981         * config/mcore/mcore.c (mcore_output_move): Likewise.
19982         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
19983         Likewise.
19984         (microblaze_function_arg): Likewise.
19985         * config/mips/mips.c (mips16_build_call_stub): Likewise.
19986         (mips_print_operand): Likewise.
19987         (mips_mode_ok_for_mov_fmt_p): Likewise.
19988         (mips_vector_mode_supported_p): Likewise.
19989         (mips_preferred_simd_mode): Likewise.
19990         (mips_expand_vpc_loongson_even_odd): Likewise.
19991         (mips_expand_vec_unpack): Likewise.
19992         (mips_expand_vi_broadcast): Likewise.
19993         (mips_expand_vector_init): Likewise.
19994         (mips_expand_vec_reduc): Likewise.
19995         (mips_expand_msa_cmp): Likewise.
19996         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
19997         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
19998         (cc_flags_for_mode): Likewise.
19999         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20000         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20001         (nds32_output_casesi_pc_relative): Likewise.
20002         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20003         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20004         (nvptx_gen_unpack): Likewise.
20005         (nvptx_gen_pack): Likewise.
20006         (nvptx_gen_shuffle): Likewise.
20007         (nvptx_gen_wcast): Likewise.
20008         (nvptx_preferred_simd_mode): Likewise.
20009         * config/pa/pa.c (pa_secondary_reload): Likewise.
20010         * config/pa/predicates.md (base14_operand): Likewise.
20011         * config/powerpcspe/powerpcspe-c.c
20012         (altivec_resolve_overloaded_builtin): Likewise.
20013         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20014         Likewise.
20015         (rs6000_preferred_simd_mode): Likewise.
20016         (output_vec_const_move): Likewise.
20017         (rs6000_expand_vector_extract): Likewise.
20018         (rs6000_split_vec_extract_var): Likewise.
20019         (reg_offset_addressing_ok_p): Likewise.
20020         (rs6000_legitimate_offset_address_p): Likewise.
20021         (rs6000_legitimize_address): Likewise.
20022         (rs6000_emit_set_const): Likewise.
20023         (rs6000_const_vec): Likewise.
20024         (rs6000_emit_move): Likewise.
20025         (spe_build_register_parallel): Likewise.
20026         (rs6000_darwin64_record_arg_recurse): Likewise.
20027         (swap_selector_for_mode): Likewise.
20028         (spe_init_builtins): Likewise.
20029         (paired_init_builtins): Likewise.
20030         (altivec_init_builtins): Likewise.
20031         (do_load_for_compare): Likewise.
20032         (rs6000_generate_compare): Likewise.
20033         (rs6000_expand_float128_convert): Likewise.
20034         (emit_load_locked): Likewise.
20035         (emit_store_conditional): Likewise.
20036         (rs6000_output_function_epilogue): Likewise.
20037         (rs6000_handle_altivec_attribute): Likewise.
20038         (rs6000_function_value): Likewise.
20039         (emit_fusion_gpr_load): Likewise.
20040         (emit_fusion_p9_load): Likewise.
20041         (emit_fusion_p9_store): Likewise.
20042         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20043         (fusion_gpr_mem_load): Likewise.
20044         (fusion_addis_mem_combo_load): Likewise.
20045         (fusion_addis_mem_combo_store): Likewise.
20046         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20047         (fusion_gpr_mem_load): Likewise.
20048         (fusion_addis_mem_combo_load): Likewise.
20049         (fusion_addis_mem_combo_store): Likewise.
20050         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20051         Likewise.
20052         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20053         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20054         (rs6000_preferred_simd_mode): Likewise.
20055         (output_vec_const_move): Likewise.
20056         (rs6000_expand_vector_extract): Likewise.
20057         (rs6000_split_vec_extract_var): Likewise.
20058         (reg_offset_addressing_ok_p): Likewise.
20059         (rs6000_legitimate_offset_address_p): Likewise.
20060         (rs6000_legitimize_address): Likewise.
20061         (rs6000_emit_set_const): Likewise.
20062         (rs6000_const_vec): Likewise.
20063         (rs6000_emit_move): Likewise.
20064         (rs6000_darwin64_record_arg_recurse): Likewise.
20065         (swap_selector_for_mode): Likewise.
20066         (paired_init_builtins): Likewise.
20067         (altivec_init_builtins): Likewise.
20068         (rs6000_expand_float128_convert): Likewise.
20069         (emit_load_locked): Likewise.
20070         (emit_store_conditional): Likewise.
20071         (rs6000_output_function_epilogue): Likewise.
20072         (rs6000_handle_altivec_attribute): Likewise.
20073         (rs6000_function_value): Likewise.
20074         (emit_fusion_gpr_load): Likewise.
20075         (emit_fusion_p9_load): Likewise.
20076         (emit_fusion_p9_store): Likewise.
20077         * config/rx/rx.c (rx_gen_move_template): Likewise.
20078         (flags_from_mode): Likewise.
20079         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20080         (s390_slb_comparison): Likewise.
20081         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20082         (s390_vector_mode_supported_p): Likewise.
20083         (s390_cc_modes_compatible): Likewise.
20084         (s390_match_ccmode_set): Likewise.
20085         (s390_canonicalize_comparison): Likewise.
20086         (s390_emit_compare_and_swap): Likewise.
20087         (s390_branch_condition_mask): Likewise.
20088         (s390_rtx_costs): Likewise.
20089         (s390_secondary_reload): Likewise.
20090         (__SECONDARY_RELOAD_CASE): Likewise.
20091         (s390_expand_cs): Likewise.
20092         (s390_preferred_simd_mode): Likewise.
20093         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20094         * config/sh/sh.c (sh_print_operand): Likewise.
20095         (dump_table): Likewise.
20096         (sh_secondary_reload): Likewise.
20097         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20098         * config/sh/sh.md (casesi_worker_1): Likewise.
20099         (casesi_worker_2): Likewise.
20100         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20101         (fcc_comparison_operator): Likewise.
20102         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20103         (emit_soft_tfmode_cvt): Likewise.
20104         (sparc_preferred_simd_mode): Likewise.
20105         (output_cbranch): Likewise.
20106         (sparc_print_operand): Likewise.
20107         (sparc_expand_vec_perm_bmask): Likewise.
20108         (vector_init_bshuffle): Likewise.
20109         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20110         (spu_vector_mode_supported_p): Likewise.
20111         (spu_expand_insv): Likewise.
20112         (spu_emit_branch_or_set): Likewise.
20113         (spu_handle_vector_attribute): Likewise.
20114         (spu_builtin_splats): Likewise.
20115         (spu_builtin_extract): Likewise.
20116         (spu_builtin_promote): Likewise.
20117         (spu_expand_sign_extend): Likewise.
20118         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20119         (tilegx_simd_int): Likewise.
20120         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20121         (tilepro_simd_int): Likewise.
20122         * config/v850/v850.c (const_double_split): Likewise.
20123         (v850_print_operand): Likewise.
20124         (ep_memory_offset): Likewise.
20125         * config/vax/vax.c (vax_rtx_costs): Likewise.
20126         (vax_output_int_move): Likewise.
20127         (vax_output_int_add): Likewise.
20128         (vax_output_int_subtract): Likewise.
20129         * config/visium/predicates.md (visium_branch_operator): Likewise.
20130         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20131         (visium_print_operand_address): Likewise.
20132         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20133         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20134         (xtensa_expand_conditional_branch): Likewise.
20135         (xtensa_copy_incoming_a7): Likewise.
20136         (xtensa_output_literal): Likewise.
20137         * dfp.c (decimal_real_maxval): Likewise.
20138         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20140 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20141             Alan Hayward  <alan.hayward@arm.com>
20142             David Sherwood  <david.sherwood@arm.com>
20144         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20145         (mode_nunits_inline): Likewise.
20146         (mode_inner_inline): Likewise.
20147         (mode_unit_size_inline): Likewise.
20148         (mode_unit_precision_inline): Likewise.
20149         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20150         unprefixed name.
20151         (emit_mode_wider): Add an E_ prefix to mode names.
20152         (emit_mode_complex): Likewise.
20153         (emit_mode_inner): Likewise.
20154         (emit_mode_adjustments): Likewise.
20155         (emit_mode_int_n): Likewise.
20156         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20157         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20158         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20159         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20160         (CRC32_BUILTIN, ENTRY): Likewise.
20161         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20162         (aarch64_pop_regs): Likewise.
20163         (aarch64_process_components): Likewise.
20164         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20165         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20166         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20167         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20168         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20169         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20170         Likewise.
20171         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20172         (pcmpistr): Likewise.
20173         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20174         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20175         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20176         Likewise.
20177         * config/rl78/rl78.c (mduc_regs): Likewise.
20178         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20179         (htm_expand_builtin): Likewise.
20180         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20181         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20182         * config/xtensa/xtensa.c (print_operand): Likewise.
20183         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20184         (NUM_MODE_VECTOR_INT): Likewise.
20185         * genoutput.c (null_operand): Likewise.
20186         (output_operand_data): Likewise.
20187         * genrecog.c (print_parameter_value): Likewise.
20188         * lra.c (debug_operand_data): Likewise.
20190 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20192         * dwarf2out.c (scompare_loc_descriptor_wide)
20193         (scompare_loc_descriptor_narrow): New functions, split out from...
20194         (scompare_loc_descriptor): ...here.
20195         * expmed.c (emit_store_flag_int): New function, split out from...
20196         (emit_store_flag): ...here.
20198 2017-08-30  Richard Biener  <rguenther@suse.de>
20200         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20201         (dwarf2out_early_finish): Move setting of AT_pubnames from
20202         early debug output to early finish.
20204 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20206         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20207         and -mdata-region to the assembler.
20209 2017-08-30  Richard Biener  <rguenther@suse.de>
20211         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20212         attributes.
20213         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20214         (dwarf2out_early_global_decl): Only generate a DIE for the
20215         abstract origin if it doesn't already exist or is a declaration DIE.
20216         (resolve_addr): Do not add the linkage name twice when
20217         generating a stub DIE for the DW_TAG_GNU_call_site target.
20219 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20221         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20222         Use machine_mode rather than int for arg1_mode.
20224 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20226         PR target/82015
20227         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20228         that the second argument of the built-in functions to unpack
20229         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20230         switch statement instead a lot of if statements.
20231         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20232         Allow 64-bit values to be in Altivec registers as well as
20233         traditional floating point registers.
20234         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20236 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20238         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20239         MAX_REGS_PER_ADDRESS == 1.
20241 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20243         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20244         * config/i386/i386.c (ix86_option_override_internal): Simplify
20245         setting of opts->x_flag_entry.
20247 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20248             Jakub Jelinek  <jakub@redhat.com>
20249             Richard Biener  <rguenther@suse.de>
20251         PR tree-optimization/81503
20252         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20253         folded constant fits in the target type; reorder tests for clarity.
20255 2017-08-29  Martin Liska  <mliska@suse.cz>
20257         * passes.def: Include pass_lower_switch.
20258         * stmt.c (dump_case_nodes): Remove and move to
20259         tree-switch-conversion.
20260         (case_values_threshold): Likewise.
20261         (expand_switch_as_decision_tree_p): Likewise.
20262         (emit_case_decision_tree): Likewise.
20263         (expand_case): Likewise.
20264         (balance_case_nodes): Likewise.
20265         (node_has_low_bound): Likewise.
20266         (node_has_high_bound): Likewise.
20267         (node_is_bounded): Likewise.
20268         (emit_case_nodes): Likewise.
20269         (struct simple_case_node): New struct.
20270         (add_case_node): Remove.
20271         (emit_case_dispatch_table): Use vector instead of case_list.
20272         (reset_out_edges_aux): Remove.
20273         (compute_cases_per_edge): Likewise.
20274         (expand_case): Build list of simple_case_node.
20275         (expand_sjlj_dispatch_table): Use it.
20276         * tree-switch-conversion.c (struct case_node): Moved from
20277         stmt.c and adjusted.
20278         (emit_case_nodes): Likewise.
20279         (node_has_low_bound): Likewise.
20280         (node_has_high_bound): Likewise.
20281         (node_is_bounded): Likewise.
20282         (case_values_threshold): Likewise.
20283         (reset_out_edges_aux): Likewise.
20284         (compute_cases_per_edge): Likewise.
20285         (add_case_node): Likewise.
20286         (dump_case_nodes): Likewise.
20287         (balance_case_nodes): Likewise.
20288         (expand_switch_as_decision_tree_p): Likewise.
20289         (emit_jump): Likewise.
20290         (emit_case_decision_tree): Likewise.
20291         (try_switch_expansion): Likewise.
20292         (do_jump_if_equal): Likewise.
20293         (emit_cmp_and_jump_insns): Likewise.
20294         (fix_phi_operands_for_edge): New function.
20295         (record_phi_operand_mapping): Likewise.
20296         (class pass_lower_switch): New pass.
20297         (pass_lower_switch::execute): New function.
20298         (make_pass_lower_switch): Likewise.
20299         (conditional_probability):
20300         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20301         * tree-pass.h: Add make_pass_lower_switch.
20303 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20305         PR target/80993
20306         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20307         handlers as used.
20309 2017-08-29  Richard Biener  <rguenther@suse.de>
20311         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20312         we do not add a DW_AT_inline attribute twice.
20313         (gen_subprogram_die): Remove code setting DW_AT_inline on
20314         DECL_ABSTRACT_P nodes.
20316 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20318         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20319         calls to internal functions.
20320         (gimplify_modify_expr): Likewise.
20321         * tree-call-cdce.c (use_internal_fn): Likewise.
20322         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20323         (convert_to_divmod): Set the nothrow flag.
20324         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20325         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20326         (vectorizable_call): Likewise.
20327         (vectorizable_store): Likewise.
20328         (vectorizable_load): Likewise.
20329         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20330         (vect_recog_mask_conversion_pattern): Likewise.
20332 2017-08-29  Martin Liska  <mliska@suse.cz>
20334         PR other/39851
20335         * gcc.c (driver_handle_option): Add new argument.
20336         * opts-common.c (handle_option): Pass
20337         target_option_override_hook.
20338         * opts-global.c (lang_handle_option): Add new option.
20339         (set_default_handlers):  Add new argument.
20340         (decode_options): Likewise.
20341         * opts.c (target_handle_option): Likewise.
20342         (common_handle_option): Call target_option_override_hook.
20343         * opts.h (struct cl_option_handler_func): Add hook for
20344         target option override.
20345         (struct cl_option_handlers): Likewise.
20346         (set_default_handlers): Add new argument.
20347         (decode_options): Likewise.
20348         (common_handle_option): Likewise.
20349         (target_handle_option): Likewise.
20350         * toplev.c (toplev::main): Pass targetm.target_option.override
20351         hook.
20353 2017-08-29  Richard Biener  <rguenther@suse.de>
20354         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20356         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20357         life to the active subtree.
20359 2017-08-28  Jeff Law  <law@redhat.com>
20361         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20362         derive_equivalences.
20363         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20364         Code moved into....
20365         (edge_info::derive_equivalences): New private member function
20367         * tree-ssa-dom.c (class edge_info): Changed from a struct
20368         to a class.  Add ctor/dtor, methods and data members.
20369         (edge_info::edge_info): Renamed from allocate_edge_info.
20370         Initialize additional members.
20371         (edge_info::~edge_info): New.
20372         (free_dom_edge_info): Delete the edge info.
20373         (record_edge_info): Use new class & associated member functions.
20374         Tighten forms for testing for edge equivalences.
20375         (record_temporary_equivalences): Iterate over the simple
20376         equivalences rather than assuming there's only one per edge.
20377         (cprop_into_successor_phis): Iterate over the simple
20378         equivalences rather than assuming there's only one per edge.
20379         (optimize_stmt): Use operand_equal_p rather than pointer
20380         equality for mini-DSE code.
20382 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20384         * gcc.c (execute): Fold SIGPIPE handling into switch
20385         statement.  Adjust internal error message.
20387 2017-08-28  Richard Biener  <rguenther@suse.de>
20389         PR debug/81993
20390         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20391         Do nothing for removed DIEs.
20393 2017-08-28  Richard Biener  <rguenther@suse.de>
20395         PR tree-optimization/81977
20396         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20397         memcpy.
20399 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20401         PR target/80640
20402         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20403         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20404         using targetm.gen_mem_thread_fence.
20406 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20408         PR target/81995
20409         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20410         predicate to register_operand.  Reorder operands.
20411         (*btr<mode>): Ditto.
20412         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20413         (*btr<mode>_mask): Ditto.
20415 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20417         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20418         * config/rs6000/xmmintrin.h: New file.
20419         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20421 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20423         PR target/81504
20424         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20425         parameter and_insn and return it.
20426         (recombine_lvx_pattern): Insert a copy to ensure availability of
20427         the base register of the copied masking operation at the point of
20428         the instruction replacement.
20429         (recombine_stvx_pattern): Likewise.
20431 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20433         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20434         undocumented switches.
20435         (-mpower9-dform-vector): Likewise.
20436         (-mpower9-dform): Likewise.
20437         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20438         comments to delete references to -mpower9-dform* switches.
20439         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20440         Delete reference to -mpower9-dform* switches, test for
20441         -mpower9-vector instead.
20442         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20443         (OTHER_P9_VECTOR_MASKS): Likewise.
20444         (POWERPC_MASKS): Likewise.
20445         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20446         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20447         that checked for -mpower9-dform* consistancy with other options.
20448         Add test for -mpower9-misc to enable other power9 switches.
20449         (rs6000_init_hard_regno_mode_ok): Likewise.
20450         (rs6000_option_override_internal): Likewise.
20451         (rs6000_emit_prologue): Likewise.
20452         (rs6000_emit_epilogue): Likewise.
20453         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20454         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20455         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20456         -mpower9-vector.
20457         (emit_fusion_p9_store): Likewise.
20458         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20459         resetting these macros if the assembler does not support ISA 3.0
20460         instructions.
20461         (TARGET_P9_DFORM_VECTOR): Likewise.
20462         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20463         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20465 2017-08-25  Alan Modra  <amodra@gmail.com>
20467         PR rtl-optimization/81747
20468         * cse.c (cse_extended_basic_block): Don't attempt to record
20469         equivalences for degenerate conditional branches that branch
20470         to their fall-through.
20472 2017-08-24  Martin Sebor  <msebor@redhat.com>
20474         PR middle-end/81908
20475         * gimple-fold.c (size_must_be_zero_p): New function.
20476         (gimple_fold_builtin_memory_op): Call it.
20478 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20480         * config/rs6000/mm_malloc.h: New file.
20482 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20484         PR tree-optimization/81913
20485         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20486         analysis when either IVs in condition can wrap.
20488 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20490         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20491         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20493 2017-08-24  Richard Biener  <rguenther@suse.de>
20495         PR target/81921
20496         * targhooks.c (default_target_can_inline_p): Properly
20497         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20498         is present and always compare.
20499         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20500         infer -mfpmath=sse from TARGET_SSE_P.
20501         (ix86_can_inline_p): Properly use target_option_default_node when
20502         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20504 2017-08-24  Richard Biener  <rguenther@suse.de>
20506         PR debug/81936
20507         * dwarf2out.c (output_die): Handle flag_generate_offload like
20508         flag_generate_lto.
20509         (output_comp_unit): Likewise.
20510         (gen_array_type_die): Likewise.
20511         (dwarf2out_early_finish): Likewise.
20512         (note_variable_value_in_expr): Likewise.
20513         (dwarf2out_finish): Likewise.  Adjust assert.
20514         * cgraphunit.c (symbol_table::compile): Move setting of
20515         flag_generate_offload earlier ...
20516         (symbol_table::finalize_compilation_unit): ... here, before
20517         early debug finalization.
20519 2017-08-24  Richard Biener  <rguenther@suse.de>
20521         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20522         and ipa-fnsummary.h.
20523         (ix86_can_inline_p): When ix86_fpmath flags do not match
20524         check whether the callee uses FP math at all.
20526 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20528         PR middle-end/81931
20529         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20530         instead of TYPE_PRECISION.
20532 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20533             Alan Hayward  <alan.hayward@arm.com>
20534             David Sherwood  <david.sherwood@arm.com>
20536         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20538 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20540         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20541         Remove field.
20542         (ix86_frame::stack_realign_allocate): New field.
20543         (struct machine_frame_state): Modify comments.
20544         (machine_frame_state::sp_realigned_fp_end): New field.
20545         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20546         layout calculation.
20547         (sp_valid_at): Add assertion to assure no attempt to access invalid
20548         offset of a realigned stack.
20549         (fp_valid_at): Likewise.
20550         (choose_baseaddr): Modify comments.
20551         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20552         ix86_expand_prologue.
20553         (ix86_expand_prologue): Modify stack realignment and allocation.
20554         (ix86_expand_epilogue): Modify comments.
20555         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20556         avx2_runtime, avx512f, and avx512f_runtime.
20558 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20560         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20561         (mstackrealign): Do not init to -1.
20562         * config/i386/i386.c (ix86_option_override_internal):
20563         Check opts_set, not opts when setting default value of
20564         opts->x_ix86_force_align_arg_pointer.
20566 2017-08-23  Richard Biener  <rguenther@suse.de>
20568         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20569         lang_hooks.decl_printable_name.
20570         * print-rtl-function.c (print_rtx_function): Likewise.
20571         * tree-pretty-print.c (dump_function_header): Likewise.
20573 2017-08-23  Richard Biener  <rguenther@suse.de>
20575         PR lto/81940
20576         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20577         -g0 at compile-time.
20579 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20581         PR middle-end/19706
20582         * doc/sourcebuild.texi (Other hardware attributes):
20583         Document xorsign.
20585 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20587         PR middle-end/19706
20588         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20589         Fix single-use check.
20591 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20593         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20595 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20597         * doc/install.texi: Modify to add more details on running selected
20598         tests.
20600 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20602         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20603         is combined with -mabi=ms.
20604         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20605         ms_abi.
20607 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20609         PR tree-optimization/81488
20610         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20611         and cached_basis fields.
20612         (MAX_SPREAD): New constant.
20613         (alloc_cand_and_find_basis): Initialize new fields.
20614         (clear_visited): New function.
20615         (create_phi_basis_1): Rename from create_phi_basis, set visited
20616         and cached_basis fields.
20617         (create_phi_basis): New wrapper function.
20618         (phi_add_costs_1): Rename from phi_add_costs, add spread
20619         parameter, set visited field, short-circuit when limits reached.
20620         (phi_add_costs): New wrapper function.
20621         (record_phi_increments_1): Rename from record_phi_increments, set
20622         visited field.
20623         (record_phi_increments): New wrapper function.
20624         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20625         (phi_incr_cost): New wrapper function.
20626         (all_phi_incrs_profitable_1): Rename from
20627         all_phi_incrs_profitable, set visited field.
20628         (all_phi_incrs_profitable): New wrapper function.
20630 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20631             Alan Hayward  <alan.hayward@arm.com>
20632             David Sherwood  <david.sherwood@arm.com>
20634         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20635         that takes the outer and inner modes.
20636         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20637         comparison as the canonical test for a paradoxical subreg.
20638         * combine.c (simplify_set): Use paradoxical_subreg_p.
20639         (make_extraction): Likewise.
20640         (force_to_mode): Likewise.
20641         (rtx_equal_for_field_assignment_p): Likewise.
20642         (gen_lowpart_for_combine): Likewise.
20643         (simplify_comparison): Likewise.
20644         * cse.c (equiv_constant): Likewise.
20645         * expmed.c (store_bit_field_1): Likewise.
20646         * final.c (alter_subreg): Likewise.
20647         * fwprop.c (propagate_rtx): Likewise.
20648         (forward_propagate_subreg): Likewise.
20649         * ira-conflicts.c (ira_build_conflicts): Likewise.
20650         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20651         * lra-constraints.c (curr_insn_transform): Likewise.
20652         (split_reg): Likewise.
20653         * lra-eliminations.c (move_plus_up): Likewise.
20654         (lra_eliminate_regs_1): Likewise.
20655         * recog.c (general_operand): Likewise.
20656         * ree.c (combine_reaching_defs): Likewise.
20657         * reload.c (push_reload): Likewise.
20658         (find_reloads): Likewise.
20659         * reload1.c (elimination_effects): Likewise.
20660         (compute_reload_subreg_offset): Likewise.
20661         (choose_reload_regs): Likewise.
20662         * rtlanal.c (subreg_lsb_1): Likewise.
20663         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20664         (simplify_subreg): Likewise.
20665         * var-tracking.c (track_loc_p): Likewise.
20666         * emit-rtl.c (byte_lowpart_offset): Likewise.
20667         (paradoxical_subreg_p): Delete out-of-line definition.
20669 2017-08-22  Jeff Law  <law@redhat.com>
20671         PR tree-optimization/81741
20672         PR tree-optimization/71947
20673         * tree-ssa-dom.c: Include tree-inline.h.
20674         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20675         equivalences if one is more expensive to compute than the other.
20676         * tree-ssa-scopedtables.h (class const_or_copies): Make
20677         record_const_or_copy_raw method private.
20678         (class avail_exprs_stack): New method simplify_binary_operation.
20679         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20680         avail_exprs_stack::simplify_binary_operation as needed.
20681         (avail_exprs_stack::simplify_binary_operation): New function.
20683 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20685         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20686         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20687         (DOT_SYMBOLS): Likewise.
20688         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20689         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20690         (RS6000_ABI_NAME): Likewise.
20691         (TARGET_CMODEL): Likewise.
20692         (TOC_SECTION_ASM_OP): Likewise.
20693         (SET_CMODEL): New macro.
20694         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20696 2017-08-22  Richard Biener  <rguenther@suse.de>
20698         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20699         to allow for free-lang-data replacements similar to verify_type_variant.
20701 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20702             Alan Hayward  <alan.hayward@arm.com>
20703             David Sherwood  <david.sherwood@arm.com>
20705         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20706         VOIDmode for the LABEL_REF.
20708 2017-08-22  Richard Biener  <rguenther@suse.de>
20710         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20712 2017-08-22  Richard Biener  <rguenther@suse.de>
20714         * common.opt (feliminate-dwarf2-dups): Ignore.
20715         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20716         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20717         same_die_p_wrap, compute_section_prefix,
20718         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20719         (comdat_symbol_id, comdat_symbol_number): Likewise.
20720         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20721         Likewise.
20722         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20723         (output_die): Mark unreachable path unreachable.
20724         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20725         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20726         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20727         (dwarf2out_early_finish): Likewise.
20729 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20731         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20733 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20735         PR target/81910
20736         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20737         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20738         (avr_attribute_table) <io, io_low, address>: Initialize
20739         .decl_required with true.
20741 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20743         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20744         undocumented debugging options.
20745         (-mvsx-scalar-double): Likewise.
20746         (-mallow-df-permute): Likewise.
20747         (-mvectorize-builtins): Likewise.
20748         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20749         (rs6000_builtin_vectorized_function): Likewise.
20750         (rs6000_builtin_md_vectorized_function): Likewise.
20751         (rs6000_opt_vars): Likewise.
20753 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20755         PR target/46091
20756         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20757         (*btrq_imm): Rename from *btrq.
20758         (*btcq_imm): Rename from *btcq.
20759         (btsc): New code attribute.
20760         (*<btsc><mode>): New insn pattern.
20761         (*btr<mode>): Ditto.
20762         (*<btsc><mode>_mask): New insn_and_split pattern.
20763         (*btr<mode>_mask): Ditto.
20765 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20766             Alan Hayward  <alan.hayward@arm.com>
20767             David Sherwood  <david.sherwood@arm.com>
20769         * function.c (pad_below): Simplify padding calculation.
20771 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20772             Alan Hayward  <alan.hayward@arm.com>
20773             David Sherwood  <david.sherwood@arm.com>
20775         * target.def (function_prologue): Remove frame size argument.
20776         (function_epilogue): Likewise.
20777         * doc/tm.texi: Regenerate.
20778         * final.c (final_start_function): Update call to function_prologue.
20779         (final_end_function): Update call to function_epilogue.
20780         (default_function_pro_epilogue): Remove frame size argument.
20781         * output.h (default_function_pro_epilogue): Likewise.
20782         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20783         (arm_output_function_prologue): Likewise.
20784         * config/frv/frv.c (frv_function_prologue): Likewise.
20785         (frv_function_epilogue): Likewise.
20786         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20787         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20788         (ia64_output_function_epilogue): Likewise.
20789         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20790         (m32r_output_function_epilogue): Likewise.
20791         * config/microblaze/microblaze.c (microblaze_function_prologue)
20792         (microblaze_function_epilogue): Likewise.
20793         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20794         (mips_output_function_epilogue): Likewise.
20795         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20796         (mmix_target_asm_function_epilogue): Likewise.
20797         * config/msp430/msp430.c (msp430_start_function): Likewise.
20798         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20799         (nds32_asm_function_epilogue): Likewise.
20800         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20801         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20802         (pa_output_function_epilogue): Likewise.
20803         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20804         (rs6000_output_function_epilogue): Likewise.
20805         * config/rl78/rl78.c (rl78_start_function): Likewise.
20806         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20807         (rs6000_output_function_epilogue): Likewise.
20808         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20809         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20810         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20811         (sparc_asm_function_epilogue): Likewise.
20813 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20815         * tree.h (type_has_mode_precision_p): New function.
20816         * convert.c (convert_to_integer_1): Use it.
20817         * expr.c (expand_expr_real_2): Likewise.
20818         (expand_expr_real_1): Likewise.
20819         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20820         * match.pd: Likewise.
20821         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20822         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20823         * tree-tailcall.c (process_assignment): Likewise.
20824         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20825         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20826         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20827         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20828         (vectorizable_assignment): Likewise.
20829         (vectorizable_shift): Likewise.
20830         (vectorizable_operation): Likewise.
20831         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20833 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20835         * match.pd: Add pow (C, x) simplification.
20837 2017-08-21  Richard Biener  <rguenther@suse.de>
20839         PR tree-optimization/81900
20840         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20841         for blocks with abnormal predecessors.
20842         (compute_antic): Do not set visited flag prematurely.
20844 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20846         PR target/79883
20847         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20849 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20851         * stor-layout.h (vector_type_mode): Move to...
20852         * tree.h (vector_type_mode): ...here.
20853         * stor-layout.c (vector_type_mode): Move to...
20854         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
20856 2017-08-21  Richard Biener  <rguenther@suse.de>
20858         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20859         register_external_die hooks.
20860         (debug_false_tree_charstarstar_uhwistar): Declare.
20861         (debug_nothing_tree_charstar_uhwi): Likewise.
20862         * debug.c (do_nothing_debug_hooks): Adjust.
20863         (debug_false_tree_charstarstar_uhwistar): New do nothing.
20864         (debug_nothing_tree_charstar_uhwi): Likewise.
20865         * dbxout.c (dbx_debug_hooks): Adjust.
20866         (xcoff_debug_hooks): Likewise.
20867         * sdbout.c (sdb_debug_hooks): Likewise.
20868         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
20869         * dwarf2out.c (macinfo_label_base): New global.
20870         (dwarf2out_register_external_die): New function for the
20871         register_external_die hook.
20872         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
20873         (dwarf2_debug_hooks): Use them.
20874         (dwarf2_lineno_debug_hooks): Adjust.
20875         (struct die_struct): Add with_offset flag.
20876         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
20877         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
20878         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
20879         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
20880         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
20881         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
20882         defining section names for the early LTO debug variants.
20883         (reset_indirect_string): New helper.
20884         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
20885         (print_dw_val): Add support for offsetted symbol references.
20886         (get_ultimate_context): Split out from is_cxx.
20887         (is_cxx): Use get_ultimate_context.
20888         (is_fortran): Add decl overload.
20889         (compute_comp_unit_symbol): Split out worker from
20890         compute_section_prefix.
20891         (compute_section_prefix): Call compute_comp_unit_symbol and
20892         set comdat_type_p here.
20893         (output_die): Skip DIE symbol output for the LTO added one.
20894         Handle DIE symbol references with offset.
20895         (output_comp_unit): Guard section name mangling properly.
20896         For LTO debug sections emit a symbol at the section beginning
20897         which we use to refer to its DIEs.
20898         (add_abstract_origin_attribute): For DIEs registered via
20899         dwarf2out_register_external_die directly refer to the early
20900         DIE rather than indirectly through the shadow one we created.
20901         Remove obsolete call to dwarf2out_abstract_function for
20902         non-function/block origins.
20903         (gen_array_type_die): When generating early LTO debug do
20904         not emit DW_AT_string_length.
20905         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
20906         late when in LTO.  As suggested place a gcc_unreachable for
20907         the DECL_ABSTRACT_P case.
20908         (gen_subprogram_die): Avoid another specification DIE
20909         for early built declarations/definitions for the late LTO case.
20910         (gen_variable_die): Add type references for late duplicated VLA dies
20911         when in late LTO.
20912         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
20913         we have the abstract instance already.
20914         (process_scope_var): Adjust decl DIE contexts in LTO which
20915         first puts them in limbo.
20916         (gen_decl_die): Do not generate type DIEs late apart from
20917         types for VLAs or for decls we do not yet have a DIE.  Do not
20918         call dwarf2out_abstract_function late.
20919         (dwarf2out_early_global_decl): Make sure to create DIEs
20920         for abstract instances of a decl first.
20921         (dwarf2out_late_global_decl): Adjust comment.
20922         (output_macinfo_op): With multiple macro sections use
20923         macinfo_label_base to distinguish labels.
20924         (output_macinfo): Likewise.  Update macinfo_label_base.
20925         Pass in the line info label.
20926         (note_variable_value_in_expr): When generating LTO resolve
20927         all variable values here by generating DIEs as needed.
20928         (init_sections_and_labels): Add early LTO debug flag parameter
20929         and generate different sections and names if set.  Add generation
20930         counter for the labels so we can have multiple of them.
20931         (reset_dies): Helper to allow DIEs to be output multiple times.
20932         (dwarf2out_finish): When outputting DIEs to the fat part of an
20933         LTO object first reset DIEs.
20934         (dwarf2out_early_finish): Output early DIEs when generating LTO.
20935         (modified_type_die): Check for decl_ultimate_origin being self
20936         before recursing.
20937         (gen_type_die_with_usage): Likewise.
20938         (gen_typedef_die): Allow decl_ultimate_origin being self.
20939         (set_decl_abstract_flags): Remove.
20940         (set_block_abstract_flags): Likewise.
20941         (dwarf2out_abstract_function): Treat the early generated DIEs
20942         as the abstract copy and only add DW_AT_inline and
20943         DW_AT_artificial here and call set_decl_origin_self.
20944         If the DIE has an abstract origin don't do anything.
20945         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
20946         if we have none yet (Go fails to build one, PR78628).
20947         (variably_modified_type_p): Prevent endless recursion for Ada
20948         cyclic pointer types.
20949         * lto-streamer-in.c: Include debug.h.
20950         (dref_queue): New global.
20951         (lto_read_tree_1): Stream in DIE references.
20952         (lto_input_tree): Register DIE references.
20953         (input_function): Stream DECL_DEBUG_ARGS.
20954         * lto-streamer-out.c: Include debug.h.
20955         (lto_write_tree_1): Output DIE references.
20956         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
20957         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
20958         (output_function): Stream DECL_DEBUG_ARGS.
20959         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20960         Stream DECL_ABSTRACT_ORIGIN.
20961         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20962         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
20963         DECL_CONTEXT for file-scope decls.
20964         * lto-streamer.h (struct dref_entry): Declare.
20965         (dref_queue): Likewise.
20966         * cfgexpand.c (pass_expand::execute): Do not call the
20967         outlining_inline_function hook here.
20968         * lto-wrapper.c (debug_obj): New global.
20969         (tool_cleanup): Unlink it if required.
20970         (debug_objcopy): New function.
20971         (run_gcc): Handle early debug sections in the IL files by
20972         extracting them to separate files, partially linkin them and
20973         feeding the result back as result to the linker.
20974         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
20975         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
20976         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
20977         sections into a separate segment.
20978         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
20979         segments.
20980         (darwin_asm_dwarf_section): Likewise.
20981         (darwin_asm_output_dwarf_offset): Likewise.
20982         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
20984 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20985             Alan Hayward  <alan.hayward@arm.com>
20986             David Sherwood  <david.sherwood@arm.com>
20988         * read-md.h (md_reader::record_potential_iterator_use): Replace
20989         pointer argument with an rtx and an index.
20990         * read-rtl.c (iterator_group::apply_iterator): Likewise.
20991         (apply_mode_iterator): Likewise.
20992         (apply_code_iterator): Likewise.
20993         (apply_int_iterator): Likewise.
20994         (apply_subst_iterator): Likewise.
20995         (record_iterator_use): Likewise.
20996         (record_attribute_use): Likewise.
20997         (md_reader::record_potential_iterator_use): Likewise.  Update calls
20998         to record_iterator_use and apply_iterator.
20999         (iterator_use): Replace ptr with x and index.
21000         (attribute_use): Likewise.
21001         (apply_attribute_uses): Update calls to apply_iterator.
21002         (apply_iterators): Likewise.  Update initialization of iterator_use.
21003         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21004         and record_potential_iterator_use.
21005         (rtx_reader::read_rtx_operand): Likewise.
21007 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21008             Alan Hayward  <alan.hayward@arm.com>
21009             David Sherwood  <david.sherwood@arm.com>
21011         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21012         CONST_WIDE_INT.
21014 2017-08-21  Richard Biener  <rguenther@suse.de>
21016         PR middle-end/81884
21017         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21018         at struct end conservatively when comparing common bases.
21020 2017-08-21  Richard Biener  <rguenther@suse.de>
21022         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21023         (mem_ref_in_stmt): Remove.
21024         (determine_max_movement): Use ref index to get at the reference.
21025         (invariantness_dom_walker::before_dom_children): Deal with
21026         lim data already initialized.
21027         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21029 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21031         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21032         (TARGET_ISA_ROUND): Ditto.
21033         (TARGET_ROUND): Ditto.
21034         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21035         * config/i386/i386.md: Ditto.
21036         * config/i386/sse.md: Ditto.
21037         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21038         with OPTION_MASK_ISA_SSE4_1.
21040 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21042         PR target/81894
21043         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21044         __builtin_ia32_lzcnt_u16.
21046 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21048         PR target/80210
21049         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21050         (rs6000_set_current_function): Rewrite function to use it.
21052 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21054         PR c/53037
21055         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21056         and TYPE_WARN_IF_NOT_ALIGN.
21057         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21058         (handle_warn_if_not_align): New.
21059         (place_union_field): Call handle_warn_if_not_align.
21060         (place_field): Call handle_warn_if_not_align.
21061         Copy TYPE_WARN_IF_NOT_ALIGN.
21062         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21063         (layout_type): Likewise.
21064         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21065         spare to 18.
21066         (tree_decl_common): Add warn_if_not_align.
21067         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21068         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21069         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21070         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21071         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21072         * doc/extend.texi: Document warn_if_not_aligned attribute.
21073         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21075 2017-08-17  Martin Liska  <mliska@suse.cz>
21077         PR bootstrap/81864
21078         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21079         (get_data_dependence): Use it as pointer type.
21080         (distribute_loop): Likewise.
21082 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21084         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21085         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21086         (p8_vmrgow_<mode>_direct): New define_insn.
21087         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21088         handle endianness for vmrgew and vmrgow permute patterns.
21090 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21092         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21093         * config/rs6000/rs6000-cpus.def: Remove comment.
21094         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21095         (POWERPC_MASKS): Likewise.
21096         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21097         use of TARGET_VSX_TIMODE.
21098         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21099         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21100         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21101         (rs6000_option_override_internal): Remove dead code.
21102         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21103         (rs6000_legitimize_reload_address): Likewise.
21104         (rs6000_legitimate_address_p): Likewise.
21105         (rs6000_opt_masks): Delete "vsx-timode".
21106         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21107         from function comment.
21108         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21109         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21110         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21111         condition.
21112         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21113         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21114         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21115         (VSX_TI): Likewise.
21116         (VSX_M): Likewise.
21117         (define_peephole2): Likewise.
21119 2017-08-17  Martin Sebor  <msebor@redhat.com>
21121         PR c/81859
21122         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21123         past the end of an array.
21124         (test_pp_format): Add test cases.
21126 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21128         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21129         missing ECF_NOTHROW flags.
21131 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21133         PR target/72804
21134         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21135         operands residing in integer registers.
21136         (*vsx_le_perm_load_<mode>): Likewise.
21137         (*vsx_le_perm_store_<mode>): Likewise.
21138         (define_peephole2): Add peepholes to optimize the above.
21140 2017-08-17  Marek Polacek  <polacek@redhat.com>
21142         PR middle-end/81814
21143         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21144         to mimic what shorten_compare did.  Change the return type to bool.
21145         (fold_cond_expr_with_comparison): Update call to
21146         operand_equal_for_comparison_p.
21147         (fold_ternary_loc): Likewise.
21149 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21151         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21152         register.
21153         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21155 2017-08-17  Richard Biener  <rguenther@suse.de>
21157         * tree-ssa-structalias.c (solve_graph): When propagating
21158         to successors update the graphs succ edges and avoid duplicate work.
21160 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21162         PR target/81861
21163         * config/i386/i386.c (ix86_option_override_internal): Save target
21164         specific options after ix86_stack_protector_guard_reg was changed.
21166 2017-08-17  Richard Biener  <rguenther@suse.de>
21168         PR tree-optimization/81827
21169         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21170         (new_var_info): Initialize it conservatively.
21171         (get_call_vi): Mark register vars.
21172         (new_scalar_tmp_constraint_exp): Likewise.
21173         (handle_rhs_call): Likewise.
21174         (handle_const_call): Likewise.
21175         (create_function_info_for): Likewise.
21176         (solve_constraints): Sort varinfos to separate register from
21177         non-register vars to pack points-to solution bitmaps during
21178         iteration.
21180 2017-08-17  Marek Polacek  <polacek@redhat.com>
21182         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21184 2017-08-17  Richard Biener  <rguenther@suse.de>
21186         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21187         to true when overflow is undefined and we saturated the result.
21189 2017-08-17  Alan Modra  <amodra@gmail.com>
21191         PR target/80938
21192         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21193         Don't use store multiple if only one reg needs saving.
21194         (interesting_frame_related_regno): New function.
21195         (rs6000_frame_related): Don't emit frame info for regs that
21196         don't need saving.
21197         (rs6000_emit_epilogue): Likewise.
21199 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21201         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21202         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21203         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21204         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21205         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21206         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21207         (verify_type): Adjust for TYPE_BINFO move.
21208         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21209         process TYPE_BINFO directly.
21210         (hash_tree): Likewise.
21211         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21212         Likewise.
21213         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21214         Likewise.
21216 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21218         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21220 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21222         PR target/46091
21223         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21224         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21225         Add ix86_binary_operator_ok to insn constraint.
21226         (*iordi_1_bts): Ditto.
21227         (*xordi_1_btc): Ditto.
21228         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21229         Update corresponding peephole2 pattern.
21230         (*btrq): Ditto.
21231         (*btcq): Ditto.
21233 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21235         PR tree-optimization/81832
21236         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21237         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21239 2017-08-16  Marek Polacek  <polacek@redhat.com>
21241         PR middle/81695
21242         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21243         regarding size_zero_node.
21245 2017-08-16  Martin Liska  <mliska@suse.cz>
21247         PR target/81753
21248         * config.gcc: Respect previously set extra_objs in case
21249         of darwin target.
21251 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21253         PR tree-optimization/81835
21254         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21255         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21256         not depend on the phi.
21258 2017-08-16  Alan Modra  <amodra@gmail.com>
21260         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21261         dead code.
21263 2017-08-16  Alan Modra  <amodra@gmail.com>
21265         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21266         (save_reg_p): ..into this.  Update all callers.
21267         (first_reg_to_save): Simplify.
21269 2017-08-16  Alan Modra  <amodra@gmail.com>
21271         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21272         fixed regs.
21274 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21276         PR target/78460
21277         PR target/67712
21278         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21279         constant count if that count is less than 32.
21281 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21283         * gcc.c (execute): Emit friendlier message if inferior is killed
21284         by an external cause.
21286 2017-08-15  Richard Biener  <rguenther@suse.de>
21288         PR tree-optimization/81790
21289         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21290         CONSTRUCTORs from simplifying and VN.
21292 2017-08-14  Martin Sebor  <msebor@redhat.com>
21294         * builtin-attrs.def: Add comments.
21296 2017-08-14  Martin Sebor  <msebor@redhat.com>
21298         PR c/81117
21299         * doc/extend.texi (attribute nonstring): Document new attribute.
21301 2017-08-14  Martin Sebor  <msebor@redhat.com>
21303         PR c/81117
21304         * tree-diagnostic.c (default_tree_printer): Handle %G.
21305         * gimple-pretty-print.h (percent_G_format): Declare new function.
21306         * gimple-pretty-print.c (percent_G_format): Define.
21307         * tree-pretty-print.c (percent_K_format): Add argument.
21309 2017-08-14  Martin Sebor  <msebor@redhat.com>
21311         PR translation/79998
21312         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21313         Remove a stray space.
21315 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21317         PR target/46091
21318         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21319         (*iordi_1_bts): Ditto.
21320         (*xordi_1_btc): Ditto.
21322 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21324         PR target/79845
21325         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21326         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21327         Likewise.
21328         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21329         quoted strings, and make more translator-friendly.
21330         (darwin_rs6000_override_options): Likewise.
21331         (rs6000_option_override_internal): Likewise.
21332         (rs6000_return_in_memory): Fix overlong line.
21333         (init_cmulative_args): Use quoted strings, and make more
21334         translator-friendly.
21335         (rs6000_pass_by_reference): Fix overlong line.
21336         (def_builtin): Use quoted strings.
21337         (altivec_expand_predicate_builtin): Use quoted strings, and make
21338         more translator-friendly.
21339         (htm_expand_builtin): Use quoted strings.
21340         (cpu_expand_builtin): Use quoted strings, and make more
21341         translator-friendly.
21342         (altivec_expand_builtin): Likewise.
21343         (paired_expand_predicate_builtin): Likewise.
21344         (rs6000_invalid_builtin): Likewise.
21345         (builtin_function_type): Use quoted strings.
21346         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21347         more translator-friendly.
21348         (rs6000_trampoline_init): Likewise.
21349         (rs6000_handle_altivec_attribute): Likewise.
21350         (rs6000_inner_target_options): Use quoted strings.
21351         (rs6000_disable_incompatible_switches): Likewise.
21352         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21353         strings, and make more translator-friendly.
21354         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21356 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21358         PR tree-optimization/81799
21359         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21360         cond_expr to simple gimple operand.
21362 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21364         PR middle-end/46932
21365         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21367 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21369         PR target/81754
21370         PR target/81268
21371         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21372         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21373         TARGET_GASISR_PROLOGUES.
21374         * config/avr/avr.c (avr_option_override): Same.
21375         (avr_pass_pre_proep::execute): Same.
21377 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21379         PR target/81820
21380         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21381         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21383 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21385         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21386         define_insn_and_split.  Split to a memory load from 0 in
21387         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21388         using PTR mode iterator.
21389         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21390         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21391         (*add_tp_<mode>): Redefine as define_insn_and_split.
21392         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21393         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21394         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21395         Split to an add with a  memory load from 0 in
21396         DEFAULT_TLS_SEG_REG address space.
21398 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21400         * config/aarch64/aarch64-option-extensions.def (rdma):
21401         Fix feature string to what Linux prints out in /proc/cpuinfo.
21403 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21405         PR ada/79542
21406         * dwarf2out.c (modified_type_die): For C typedef types that have
21407         an ultimate origin, process the ultimate origin instead of the
21408         input type.
21409         (gen_typedef_die): Assert that input DECLs have no ultimate
21410         origin.
21411         (gen_type_die_with_usage): For typedef variants that have an
21412         ultimate origin, just call gen_decl_die on the original DECL.
21413         (process_scope_var): Avoid creating DIEs for local typedefs and
21414         concrete static variables.
21416 2017-08-12  Alan Modra  <amodra@gmail.com>
21418         PR target/81170
21419         PR target/81295
21420         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21421         match gnu-user.h startfile.
21422         (ENDFILE_LINUX_SPEC): Similarly.
21424 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21426         PR lto/81430
21427         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21428         Remove function.
21429         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21431 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21432         * config/aarch64/aarch64.md (mov<mode>): Change.
21433         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21434         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21435         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21437 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21439         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21440         for storage order barriers.
21442 2017-08-11  Martin Liska  <mliska@suse.cz>
21444         PR tree-opt/79987
21445         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21446         variables of void type.
21448 2017-08-11  Martin Liska  <mliska@suse.cz>
21450         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21451         TARGET_SUPPORTS_ALIASES.
21452         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21453         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21454         (optimize_weakref): Likewise.
21455         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21456         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21457         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21459 2017-08-11  Martin Liska  <mliska@suse.cz>
21461         PR ipa/81213
21462         * config/i386/i386.c (make_resolver_func): Do complete
21463         refactoring of the function.
21465 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21467         PR target/81708
21468         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21469         * config/i386/i386.c (ix86_stack_protect_guard): Use
21470         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21471         * doc/invoke.texi (x86 Options): Document
21472         -mstack-protector-guard-symbol= option.
21474 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21476         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21477         * config/i386/i386.c (ix86_split_stack_guard): New function.
21478         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21479         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21480         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21481         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21482         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21483         (split_stack_space_check): Call ix86_split_stack_guard.
21485 2017-08-10  Martin Sebor  <msebor@redhat.com>
21487         * print-tree.c (print_node): Print location using the established
21488         format %s:%i%i.
21489         Replace spaces with colons.
21490         (debug_raw, debug): Ditto.
21492 2017-08-10  Martin Sebor  <msebor@redhat.com>
21494         PR c++/81586
21495         * pretty-print.c (pp_format): Correct the handling of %s precision.
21497 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21499         PR target/81736
21500         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21501         to ...
21502         (ix86_finalize_stack_frame_flags): This.  Also clear
21503         frame_pointer_needed if -fno-omit-frame-pointer is used without
21504         stack access.
21505         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21506         with ix86_finalize_stack_frame_flags.
21507         (ix86_expand_epilogue): Likewise.
21508         (ix86_expand_split_stack_prologue): Likewise.
21509         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21511 2017-08-10  Martin Liska  <mliska@suse.cz>
21513         PR c++/81355
21514         * c-attribs.c (handle_target_attribute):
21515         Report warning for an empty string argument of target attribute.
21517 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21519         PR c/81687
21520         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21521         LABEL_DECLs.
21522         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21523         or DECL_NONLOCAL labels.
21524         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21525         or DECL_NONLOCAL labels here.
21527 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21529         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21530         to indicate when early gimple folding has been disabled.
21531         (rs6000_gimple_fold_builtin): Add debug content.
21532         (rs6000_invalid_builtin): Fix whitespace.
21533         (rs6000_expand_builtin): Fix whitespace.
21534         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21536 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21538         PR target/80938
21539         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21540         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21542 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21544         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21545         (qdf24xx): Likewise.
21546         * config/aarch64/aarch64.md: Include falkor.md.
21547         * config/aarch64/falkor.md: New.
21549 2017-08-09  Marek Polacek  <polacek@redhat.com>
21551         PR c/81233
21552         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21553         * diagnostic.c (emit_diagnostic): Add a comment.
21554         (emit_diagnostic_valist): New function.
21556 2017-08-09  Marek Polacek  <polacek@redhat.com>
21558         PR c/81417
21559         * input.c (make_location): New overload.
21560         * input.h (make_location): Declare.
21562 2017-08-08  Alan Modra  <amodra@gmail.com>
21563             H.J. Lu  <hongjiu.lu@intel.com>
21565         PR driver/81523
21566         * gcc.c (NO_PIE_SPEC): Delete.
21567         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21568         exclusion..
21569         (LINK_PIE_SPEC): ..to here.
21570         (LINK_COMMAND_SPEC): Support -no-pie.
21571         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21572         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21573         format.
21574         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21575         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21576         (ENDFILE_CRTEND_SPEC): Similarly.
21578 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21580         PR target/81708
21581         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21582         (mstack-protector-guard-offset=): Ditto.
21583         * config/i386/i386.c (ix86_option_override): Handle
21584         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21585         options.
21586         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21587         ix86_stack_protect_guard_offset variables.
21588         (TARGET_STACK_PROTECT_GUARD): Always define.
21589         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21590         and -mstack-protector-guard-offset= options.
21592 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21594         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21595         boundary case for the last candidate.
21597 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21599         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21600         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21602 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21604         PR middle-end/19706
21605         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21606         * config/aarch64/aarch64-builtins.c
21607         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21608         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21609         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21611 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21612             Andrew Pinski <pinskia@gmail.com>
21614         PR middle-end/19706
21615         * internal-fn.def (XORSIGN): New.
21616         * optabs.def (xorsign_optab): New.
21617         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21618         (convert_expand_mult_copysign): New.
21619         (pass_optimize_widening_mul::execute): Call
21620         convert_expand_mult_copysign.
21622 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21624         PR tree-optimization/81354
21625         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21626         Insert on edges rather than explicitly creating landing pads.
21627         (analyze_candidates_and_replace): Commit edge inserts.
21629 2017-08-08  Richard Biener  <rguenther@suse.de>
21631         PR middle-end/81719
21632         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21633         (expand_simple_operations): Also look through ADDR_EXPRs with
21634         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21636 2017-08-08  Richard Biener  <rguenther@suse.de>
21638         PR tree-optimization/81723
21639         * tree-vect-slp.c (struct bst_traits): New hash traits.
21640         (bst_fail): New global.
21641         (vect_build_slp_tree_2): New worker, split out from ...
21642         (vect_build_slp_tree): ... this now wrapping it with using
21643         bst_fail set to cache SLP tree build fails.  Properly handle
21644         max_tree_size.
21645         (vect_analyze_slp_instance): Allocate and free bst_fail.
21647 2017-08-08  Martin Liska  <mliska@suse.cz>
21649         PR tree-opt/81696
21650         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21651         LABEL_DECLs that can be from a different function.
21653 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21655         PR tree-optimization/81744
21656         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21657         loop's number of iterations.
21659 2017-08-08  Martin Liska  <mliska@suse.cz>
21661         * asan.c: Include header files.
21662         * attribs.c (build_decl_attribute_variant): New function moved
21663         from tree.[ch].
21664         (build_type_attribute_qual_variant): Likewise.
21665         (cmp_attrib_identifiers): Likewise.
21666         (simple_cst_list_equal): Likewise.
21667         (omp_declare_simd_clauses_equal): Likewise.
21668         (attribute_value_equal): Likewise.
21669         (comp_type_attributes): Likewise.
21670         (build_type_attribute_variant): Likewise.
21671         (lookup_ident_attribute): Likewise.
21672         (remove_attribute): Likewise.
21673         (merge_attributes): Likewise.
21674         (merge_type_attributes): Likewise.
21675         (merge_decl_attributes): Likewise.
21676         (merge_dllimport_decl_attributes): Likewise.
21677         (handle_dll_attribute): Likewise.
21678         (attribute_list_equal): Likewise.
21679         (attribute_list_contained): Likewise.
21680         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21681         (lookup_attribute_by_prefix): Likewise.
21682         * bb-reorder.c: Include header files.
21683         * builtins.c: Likewise.
21684         * calls.c: Likewise.
21685         * cfgexpand.c: Likewise.
21686         * cgraph.c: Likewise.
21687         * cgraphunit.c: Likewise.
21688         * convert.c: Likewise.
21689         * dwarf2out.c: Likewise.
21690         * final.c: Likewise.
21691         * fold-const.c: Likewise.
21692         * function.c: Likewise.
21693         * gimple-expr.c: Likewise.
21694         * gimple-fold.c: Likewise.
21695         * gimple-pretty-print.c: Likewise.
21696         * gimple.c: Likewise.
21697         * gimplify.c: Likewise.
21698         * hsa-common.c: Likewise.
21699         * hsa-gen.c: Likewise.
21700         * internal-fn.c: Likewise.
21701         * ipa-chkp.c: Likewise.
21702         * ipa-cp.c: Likewise.
21703         * ipa-devirt.c: Likewise.
21704         * ipa-fnsummary.c: Likewise.
21705         * ipa-inline.c: Likewise.
21706         * ipa-visibility.c: Likewise.
21707         * ipa.c: Likewise.
21708         * lto-cgraph.c: Likewise.
21709         * omp-expand.c: Likewise.
21710         * omp-general.c: Likewise.
21711         * omp-low.c: Likewise.
21712         * omp-offload.c: Likewise.
21713         * omp-simd-clone.c: Likewise.
21714         * opts-global.c: Likewise.
21715         * passes.c: Likewise.
21716         * predict.c: Likewise.
21717         * sancov.c: Likewise.
21718         * sanopt.c: Likewise.
21719         * symtab.c: Likewise.
21720         * toplev.c: Likewise.
21721         * trans-mem.c: Likewise.
21722         * tree-chkp.c: Likewise.
21723         * tree-eh.c: Likewise.
21724         * tree-into-ssa.c: Likewise.
21725         * tree-object-size.c: Likewise.
21726         * tree-parloops.c: Likewise.
21727         * tree-profile.c: Likewise.
21728         * tree-ssa-ccp.c: Likewise.
21729         * tree-ssa-live.c: Likewise.
21730         * tree-ssa-loop.c: Likewise.
21731         * tree-ssa-sccvn.c: Likewise.
21732         * tree-ssa-structalias.c: Likewise.
21733         * tree-ssa.c: Likewise.
21734         * tree-streamer-in.c: Likewise.
21735         * tree-vectorizer.c: Likewise.
21736         * tree-vrp.c: Likewise.
21737         * tsan.c: Likewise.
21738         * ubsan.c: Likewise.
21739         * varasm.c: Likewise.
21740         * varpool.c: Likewise.
21741         * tree.c: Remove functions moved to attribs.[ch].
21742         * tree.h: Likewise.
21743         * config/aarch64/aarch64.c: Add attrs.h header file.
21744         * config/alpha/alpha.c: Likewise.
21745         * config/arc/arc.c: Likewise.
21746         * config/arm/arm.c: Likewise.
21747         * config/avr/avr.c: Likewise.
21748         * config/bfin/bfin.c: Likewise.
21749         * config/c6x/c6x.c: Likewise.
21750         * config/cr16/cr16.c: Likewise.
21751         * config/cris/cris.c: Likewise.
21752         * config/darwin.c: Likewise.
21753         * config/epiphany/epiphany.c: Likewise.
21754         * config/fr30/fr30.c: Likewise.
21755         * config/frv/frv.c: Likewise.
21756         * config/ft32/ft32.c: Likewise.
21757         * config/h8300/h8300.c: Likewise.
21758         * config/i386/winnt.c: Likewise.
21759         * config/ia64/ia64.c: Likewise.
21760         * config/iq2000/iq2000.c: Likewise.
21761         * config/lm32/lm32.c: Likewise.
21762         * config/m32c/m32c.c: Likewise.
21763         * config/m32r/m32r.c: Likewise.
21764         * config/m68k/m68k.c: Likewise.
21765         * config/mcore/mcore.c: Likewise.
21766         * config/microblaze/microblaze.c: Likewise.
21767         * config/mips/mips.c: Likewise.
21768         * config/mmix/mmix.c: Likewise.
21769         * config/mn10300/mn10300.c: Likewise.
21770         * config/moxie/moxie.c: Likewise.
21771         * config/msp430/msp430.c: Likewise.
21772         * config/nds32/nds32-isr.c: Likewise.
21773         * config/nds32/nds32.c: Likewise.
21774         * config/nios2/nios2.c: Likewise.
21775         * config/nvptx/nvptx.c: Likewise.
21776         * config/pa/pa.c: Likewise.
21777         * config/pdp11/pdp11.c: Likewise.
21778         * config/powerpcspe/powerpcspe.c: Likewise.
21779         * config/riscv/riscv.c: Likewise.
21780         * config/rl78/rl78.c: Likewise.
21781         * config/rx/rx.c: Likewise.
21782         * config/s390/s390.c: Likewise.
21783         * config/sh/sh.c: Likewise.
21784         * config/sol2.c: Likewise.
21785         * config/sparc/sparc.c: Likewise.
21786         * config/spu/spu.c: Likewise.
21787         * config/stormy16/stormy16.c: Likewise.
21788         * config/tilegx/tilegx.c: Likewise.
21789         * config/tilepro/tilepro.c: Likewise.
21790         * config/v850/v850.c: Likewise.
21791         * config/vax/vax.c: Likewise.
21792         * config/visium/visium.c: Likewise.
21793         * config/xtensa/xtensa.c: Likewise.
21795 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21797         PR target/81593
21798         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21799         constraints since the -mupper-regs-* switches have been
21800         eliminated.
21801         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21802         into a vector from a double word element that was extracted from
21803         another vector, and eliminate extra XXPERMDI instructions.
21804         (vsx_concat_<mode>_2): Likewise.
21805         (vsx_concat_<mode>_3): Likewise.
21806         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21807         concat to allow optimizing inserts from previous extracts.
21809 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21811         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21812         memory reference to a SSP offset in TLS address space.
21813         (ix86_print_operand) <case '@'>: Remove.
21814         (ix86_print_operand_punct_valid_p): Remove '@' code.
21815         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21816         UNSPEC_SP_TLS_TEST.
21817         (stack_tls_protect_set_<mode>): Remove.
21818         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21819         (stack_tls_protect_test_<mode>): Remove.
21820         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21822 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21824         PR target/81755
21825         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21827 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21829         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21830         variable was referenced as multidir in command.
21832 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21834         PR c/69389
21835         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21836         BIT_FIELD_REF.
21838 2017-08-07  Martin Liska  <mliska@suse.cz>
21840         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21841         * config/rl78/rl78.c: Add include of attribs.h.
21842         * config/sh/sh.c: Likewise.
21843         * config/v850/v850.c: Likewise.
21845 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21847         PR middle-end/78266
21848         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21850 2017-08-07  Martin Liska  <mliska@suse.cz>
21852         * config/mips/mips.c: Include attribs.h.
21854 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21856         PR fortran/68829
21857         * doc/invoke.texi: Document change in behvaior for -Ofast for
21858         Fortran.
21860 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
21862         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21863         Use gen_frame_mem.
21864         (aarch64_pop_regs): Likewise.
21865         (aarch64_gen_load_pair): Likewise.
21866         (aarch64_save_callee_saves): Likewise.
21867         (aarch64_restore_callee_saves): Likewise.
21869 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21871         * config/i386/i386.c: Revert the last change.
21873 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21875         PR target/81736
21876         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21877         to ...
21878         (ix86_finalize_stack_frame_flags): This.  Also clear
21879         frame_pointer_needed if -fno-omit-frame-pointer is used without
21880         stack access.
21881         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21882         with ix86_finalize_stack_frame_flags.
21883         (ix86_expand_epilogue): Likewise.
21884         (ix86_expand_split_stack_prologue): Likewise.
21886 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21888         PR target/81743
21889         * config/i386/i386.c (get_builtin_code_for_version): Set priority
21890         to P_AES for Westmere.
21892 2017-08-07  Jonathan Yong  <10walls@gmail.com>
21894         * config/i386/mingw.opt (fset-stack-executable): Removed.
21895         * config/i386/cygming.opt (fset-stack-executable): Moved
21896         from mingw.opt.
21897         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
21899 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
21901         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
21903 2017-08-07  Marek Polacek  <polacek@redhat.com>
21905         PR middle-end/81737
21906         * fold-const.c (fold_indirect_ref_1): Check type_domain.
21908 2017-08-07  Martin Liska  <mliska@suse.cz>
21910         * attribs.h (canonicalize_attr_name): New function.
21911         (cmp_attribs): Move from c-format.c and adjusted.
21912         (is_attribute_p): Moved from tree.h.
21913         * tree-inline.c: Add new includes.
21914         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
21915         (private_is_attribute_p): Remove.
21916         (private_lookup_attribute): Likewise.
21917         (private_lookup_attribute_by_prefix): Simplify.
21918         (remove_attribute): Use is_attribute_p.
21919         * tree.h: Remove removed declarations.
21921 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21923         PR middle-end/81698
21924         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
21925         instead of computing it in the function.  Formatting fix.
21926         (expand_case): Don't rely on default_edge being the first edge,
21927         clear it if removing it, pass default_edge to
21928         emit_case_dispatch_table.
21929         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
21930         fix.
21932 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
21934         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
21935         insn in the function, emit NOP after the insn.
21937 2017-08-06  Tom de Vries  <tom@codesourcery.com>
21939         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
21940         and element loops.
21942 2017-08-06  Tom de Vries  <tom@codesourcery.com>
21944         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
21945         loop.
21947 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
21949         PR tree-optimization/57371
21950         * match.pd: New pattern.
21952 2017-08-04  Marek Polacek  <polacek@redhat.com>
21954         PR middle-end/81695
21955         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
21956         perform the computation in offset_int.
21958 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
21960         PR tree-optimization/81136
21961         * tree-vectorizer.h: Include tree-hash-traits.h.
21962         (vec_base_alignments): New typedef.
21963         (vec_info): Add a base_alignments field.
21964         (vect_record_base_alignments): Declare.
21965         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
21966         field.
21967         (DR_IS_CONDITIONAL_IN_STMT): New macro.
21968         (create_data_ref): Add an is_conditional_in_stmt argument.
21969         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
21970         the is_conditional_in_stmt field.
21971         (data_ref_loc): Add an is_conditional_in_stmt field.
21972         (get_references_in_stmt): Set the is_conditional_in_stmt field.
21973         (find_data_references_in_stmt): Update call to create_data_ref.
21974         (graphite_find_data_references_in_stmt): Likewise.
21975         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
21976         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
21977         (vect_record_base_alignment): New function.
21978         (vect_record_base_alignments): Likewise.
21979         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
21980         for nested statements even if we fail to compute a misalignment.
21981         Use pooled base alignments for unconditional references.
21982         (vect_find_same_alignment_drs): Compare base addresses instead
21983         of base objects.
21984         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
21985         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
21987 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
21989         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
21990         Add an explicit name for the enum.  Use auto_vec for slp_instances
21991         and grouped_stores.
21992         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
21993         for all vectors.
21994         (_bb_vec_info): Add a constructor and destructor.
21995         (vinfo_for_stmt): Return NULL for uids of -1 as well.
21996         (destroy_loop_vec_info): Delete.
21997         (vect_destroy_datarefs): Likewise.
21998         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
21999         (vec_info::vec_info): New function.
22000         (vec_info::~vec_info): Likewise.
22001         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22002         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22003         destroy_loop_vec_info.
22004         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22005         (_loop_vec_info::_loop_vec_info): ...this.
22006         (destroy_loop_vec_info): Replace with...
22007         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22008         the stmt_vec_infos.  Leave handling of vec_info information to its
22009         destructor.  Remove explicit vector releases.
22010         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22011         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22012         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22013         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22014         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22015         (destroy_bb_vec_info): Replace with...
22016         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22017         information to its destructor.
22018         (vect_slp_analyze_bb_1): Use new and delete instead of
22019         new_bb_vec_info and destroy_bb_vec_info.
22020         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22021         single delete.
22023 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22025         * tree-data-ref.h (subscript): Add access_fn field.
22026         (data_dependence_relation): Add could_be_independent_p.
22027         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22028         (same_access_functions): Move to tree-data-ref.c.
22029         * tree-data-ref.c (ref_contains_union_access_p): New function.
22030         (access_fn_component_p): Likewise.
22031         (access_fn_components_comparable_p): Likewise.
22032         (dr_analyze_indices): Add a reference to access_fn_component_p.
22033         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22034         DR_ACCESS_FN.
22035         (constant_access_functions): Likewise.
22036         (add_other_self_distances): Likewise.
22037         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22038         (initialize_data_dependence_relation): Use XCNEW and remove
22039         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22040         of access functions that have the same type.  Allow the
22041         subsequence to end with different bases in some circumstances.
22042         Record the chosen access functions in SUB_ACCESS_FN.
22043         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22044         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22045         (subscript_dependence_tester_1): Likewise dra and drb.
22046         (build_classic_dist_vector): Update calls accordingly.
22047         (subscript_dependence_tester): Likewise.
22048         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22049         DDR_COULD_BE_INDEPENDENT_P.
22050         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22051         comp_alias_ddrs instead of may_alias_ddrs.
22052         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22053         New function.
22054         (vect_analyze_data_ref_dependence): Use it if
22055         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22056         distance vectors if that fails.
22057         (dependence_distance_ge_vf): New function.
22058         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22059         LOOP_VINFO_MAY_ALIAS_DDRS.
22061 2017-08-04  Richard Biener  <rguenther@suse.de>
22063         PR middle-end/81705
22064         * fold-const.c (fold_binary_loc): Properly restrict
22065         minus_var0 && minus_var1 case when associating undefined overflow
22066         entities.
22068 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22070         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22072 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22074         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22075         Don't start diagnostic messages with a capital letter.
22076         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22077         Likewise.
22078         (rs6000_invalid_builtin): Likewise.
22079         (rs6000_trampoline_init): Likewise.
22081 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22083         PR target/81621
22084         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22085         after setting changeable df flags.
22087 2017-08-03  Richard Biener  <rguenther@suse.de>
22089         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22090         up if the use is in USE - X.
22092 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22094         * toplev.c (dumpfile.h): New include.
22095         (internal_error_reentered): New static function.  Use it...
22096         (internal_error_function): ...here to handle reentered internal_error.
22098 2017-08-03  Richard Biener  <rguenther@suse.de>
22100         PR middle-end/81148
22101         * fold-const.c (split_tree): Add minus_var and minus_con
22102         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22103         here but always use minus_*.
22104         (associate_trees): Assert we never associate with MINUS_EXPR
22105         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22106         when associating as MINUS_EXPR either.
22107         (fold_binary_loc): Track minus_var and minus_con.
22109 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22111         PR lto/81430
22112         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22113         ACCEL_COMPILER, apply finish_options on
22114         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22116 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22118         PR target/81662
22119         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22120         function_entry_patch_area_size > 0.
22122 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22124         PR driver/81650
22125         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22126         instead of 10??LU, perform unit multiplication in wide_int,
22127         don't change alloc_object_size_limit if the limit is larger
22128         than SSIZE_MAX.
22130         PR tree-optimization/81655
22131         PR tree-optimization/81588
22132         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22133         the case when ranges[i].low and high are 1 for unsigned type with
22134         precision 1.
22136         PR middle-end/81052
22137         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22138         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22140 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22142         * tree-vrp.h: Add include guard.
22144 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22146         PR target/81644
22147         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22148         (ud2): New insn pattern.
22149         * config/i386/i386.c (ix86_expand_epilogue):
22150         For naked functions, generate ud2 instead of trap insn.
22152 2017-08-02  Marek Polacek  <polacek@redhat.com>
22154         PR other/81667
22155         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22157 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22158             Cesar Philippidis  <cesar@codesourcery.com>
22160         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22161         Add missing edge probabilities.
22163 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22165         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22166         Correct endianness.
22168 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22170         PR middle-end/79499
22171         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22172         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22173         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22174         prologue_seq sequences - if any.
22176 2017-08-02  Richard Biener  <rguenther@suse.de>
22178         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22179         via vectors if supported, integer extracts via punning if supported
22180         or otherwise vector extracts.
22182 2017-08-02  Richard Biener  <rguenther@suse.de>
22184         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22185         into ...
22186         (bitmap_insert_into_set): ... this.
22188 2017-08-02  Richard Biener  <rguenther@suse.de>
22190         PR tree-optimization/81633
22191         Revert
22192         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22194         PR tree-optimization/71752
22195         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22197 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22199         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22200         (machine_function::call_ms2sysv_pad_out): Remove field.
22201         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22202         (ix86_compute_frame_layout): Likewise.
22204 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22206         PR target/81654
22207         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22208         attribute with interrupt attribute.
22210 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22212         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22213         BIT_INSERT_EXPR's operand 1
22214         to see if the types precision matches.
22216 2017-08-01  Martin Liska  <mliska@suse.cz>
22218         PR middle-end/70140
22219         * builtins.c (expand_builtin_memcpy_args): Remove.
22220         (expand_builtin_memcpy): Call newly added function
22221         expand_builtin_memory_copy_args.
22222         (expand_builtin_memcpy_with_bounds): Likewise.
22223         (expand_builtin_mempcpy): Remove last argument.
22224         (expand_builtin_mempcpy_with_bounds): Likewise.
22225         (expand_builtin_memory_copy_args): New function created from
22226         expand_builtin_mempcpy_args with small modifications.
22227         (expand_builtin_mempcpy_args): Remove.
22228         (expand_builtin_stpcpy): Remove unused argument.
22229         (expand_builtin): Likewise.
22230         (expand_builtin_with_bounds): Likewise.
22232 2017-08-01  Martin Liska  <mliska@suse.cz>
22234         Revert r250771
22235         Make mempcpy more optimal (PR middle-end/70140).
22237 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22239         PR target/81622
22240         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22241         __builtin_vec_cmpne verify both arguments are compatible vectors
22242         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22243         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22244         move computation of aligned to after checking the argument types.
22245         Formatting fixes.
22247         PR target/80846
22248         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22249         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22250         calls.
22252 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22253             Doug Rupp  <rupp@adacore.com>
22254             Olivier Hainque  <hainque@adacore.com>
22256         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22257         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22258         arm8 (arch v4).
22259         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22260         for TARGET_OS_CPP_BUILTIN.
22261         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22262         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22263         arm_arch7.
22264         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22265         passing required abi options to the assembler for EABI configurations.
22266         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22267         of .text.hot and .text.unlikely sections for kernel modules when
22268         using ARM style exceptions.
22269         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22270         options. Add EXTRA_CC1_SPEC.
22271         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22272         toolchain options.
22273         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22274         transition.
22275         (ARM_TARGET2_DWARF_FORMAT): Define.
22276         * config/arm/t-vxworks: Adjust multilib control to removal of the
22277         Diab command line options.
22279 2017-08-01  Martin Liska  <mliska@suse.cz>
22281         PR gcov-profile/81561
22282         * gcov.c (unblock): Make unblocking safe as we need to preserve
22283         index correspondence of blocks and block_lists.
22285 2017-08-01  Richard Biener  <rguenther@suse.de>
22287         PR tree-optimization/81181
22288         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22289         (compute_antic): ... end of iteration here.
22291 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22293         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22294         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22295         (ftree-slp-vectorize): Likewise.
22296         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22297         can no longer be set independent of flag_tree_loop_vectorize.
22298         * omp-general.c (emp_max_vf): Likewise.
22299         * opts.c (enable_fdo_optimizations): Remove references to
22300         flag_tree_vectorize, these are now implicit.
22301         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22302         and leave it for the options machinery.
22304 2017-08-01  Martin Liska  <mliska@suse.cz>
22306         PR middle-end/70140
22307         * builtins.c (expand_builtin_memcpy_args): Remove.
22308         (expand_builtin_memcpy): Call newly added function
22309         expand_builtin_memory_copy_args.
22310         (expand_builtin_memcpy_with_bounds): Likewise.
22311         (expand_builtin_mempcpy): Remove last argument.
22312         (expand_builtin_mempcpy_with_bounds): Likewise.
22313         (expand_builtin_memory_copy_args): New function created from
22314         expand_builtin_mempcpy_args with small modifications.
22315         (expand_builtin_mempcpy_args): Remove.
22316         (expand_builtin_stpcpy): Remove unused argument.
22317         (expand_builtin): Likewise.
22318         (expand_builtin_with_bounds): Likewise.
22320 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22322         PR target/81641
22323         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22324         print "ds:" only for immediates in generic address space.
22326 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22328         PR target/81639
22329         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22330         (ix86_function_ok_for_sibcall): Return false for naked functions.
22332 2017-08-01  Richard Biener  <rguenther@suse.de>
22334         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22335         (compute_antic): Seed worklist with exit block predecessors.
22336         * cfganal.c (dfs_find_deadend): For a cycle return the source
22337         of the edge closing it.
22339 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22341         * config/aarch64/aarch64.c
22342         (aarch64_can_const_movi_rtx_p): Move 0 check.
22344 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22346         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22347         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22348         above macro.
22349         * match.pd: Ditto in address comparison pattern.
22351 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22353         PR tree-optimization/81627
22354         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22355         closed ssa form for store-store chain.
22357 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22359         PR tree-optimization/81620
22360         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22361         for store-store chain.
22363 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22365         PR tree-optimization/81588
22366         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22367         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22368         swap rhs1 and rhs2 and comparison code unconditionally,
22369         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22370         ranges[i].in_p, instead invert comparison code ccode if
22371         opcode or oe->rank is BIT_IOR_EXPR.
22373         PR target/80846
22374         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22375         a direct optab to conversion optab.
22376         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22377         with GET_MODE_INNER as last argument instead of optab_handler.
22378         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22379         vector extraction if possible and optab is available.
22380         * expr.c (store_constructor): Use convert_optab_handler instead
22381         of optab_handler.  Use vector initialization from smaller
22382         vectors if possible and optab is available.
22383         * tree-vect-stmts.c (vectorizable_load): Likewise.
22384         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22385         now have two modes.
22386         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22387         of vec_init from half-sized vectors with the same element mode.
22388         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22389         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22390         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22391         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22392         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22393         after mode in gen_vec_extract* calls.
22394         (vec_extract<mode>): Renamed to ...
22395         (vec_extract<mode><ssescalarmodelower>): ... this.
22396         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22397         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22398         element mode after mode in gen_vec_init* calls.
22399         (VEC_INIT_HALF_MODE): New mode iterator.
22400         (vec_init<mode>): Renamed to ...
22401         (vec_init<mode><ssescalarmodelower>): ... this.
22402         (vec_init<mode><ssehalfvecmodelower>): New expander.
22403         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22404         (vec_extractv2sfsf): ... this.
22405         (vec_initv2sf): Renamed to ...
22406         (vec_initv2sfsf): ... this.
22407         (vec_extractv2si): Renamed to ...
22408         (vec_extractv2sisi): ... this.
22409         (vec_initv2si): Renamed to ...
22410         (vec_initv2sisi): ... this.
22411         (vec_extractv4hi): Renamed to ...
22412         (vec_extractv4hihi): ... this.
22413         (vec_initv4hi): Renamed to ...
22414         (vec_initv4hihi): ... this.
22415         (vec_extractv8qi): Renamed to ...
22416         (vec_extractv8qiqi): ... this.
22417         (vec_initv8qi): Renamed to ...
22418         (vec_initv8qiqi): ... this.
22419         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22420         (vec_init<mode>): Renamed to ...
22421         (vec_init<mode><VEC_base_l>): ... this.
22422         (vec_extract<mode>): Renamed to ...
22423         (vec_extract<mode><VEC_base_l>): ... this.
22424         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22425         (vec_initv2sfsf): ... this.
22426         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22427         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22428         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22429         element mode after mode in gen_vec_init* calls.
22430         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22431         (vec_init<mode><Vel>): ... this.
22432         (vec_extract<mode>): Renamed to ...
22433         (vec_extract<mode><Vel>): ... this.
22434         * config/aarch64/iterators.md (Vel): New mode attribute.
22435         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22436         Add element mode after mode in gen_vec_extract* calls.
22437         * config/s390/vector.md (non_vec_l): New mode attribute.
22438         (vec_extract<mode>): Renamed to ...
22439         (vec_extract<mode><non_vec_l>): ... this.
22440         (vec_init<mode>): Renamed to ...
22441         (vec_init<mode><non_vec_l>): ... this.
22442         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22443         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22444         vec_extract mode.
22445         * config/arm/iterators.md (V_elem_l): New mode attribute.
22446         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22447         (vec_extract<mode><V_elem_l>): ... this.
22448         (vec_extractv2di): Renamed to ...
22449         (vec_extractv2didi): ... this.
22450         (vec_init<mode>): Renamed to ...
22451         (vec_init<mode><V_elem_l>): ... this.
22452         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22453         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22454         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22455         Add element mode after gen_vec_extract* calls.
22456         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22457         (vec_init<mode><unitmode>): ... this.
22458         (vec_extract<mode>): Renamed to ...
22459         (vec_extract<mode><unitmode>): ... this.
22460         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22461         (vec_init<mode><unitmode>): ... this.
22462         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22463         (vec_initv2sfsf): ... this.
22464         (vec_extractv2sf): Renamed to ...
22465         (vec_extractv2sfsf): ... this.
22466         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22467         Add element mode after gen_vec_extract* calls.
22468         * config/mips/mips.md (unitmode): New mode iterator.
22469         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22470         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22471         * config/spu/spu.md (inner_l): New mode attribute.
22472         (vec_init<mode>): Renamed to ...
22473         (vec_init<mode><inner_l>): ... this.
22474         (vec_extract<mode>): Renamed to ...
22475         (vec_extract<mode><inner_l>): ... this.
22476         * config/sparc/sparc.md (veltmode): New mode iterator.
22477         (vec_init<VMALL:mode>): Renamed to ...
22478         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22479         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22480         (vec_initv2sisi): ... this.
22481         (vec_initv2sf): Renamed to ...
22482         (vec_initv2sfsf): ... this.
22483         (vec_extractv2sf): Renamed to ...
22484         (vec_extractv2sfsf): ... this.
22485         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22486         (vec_init<mode>): Renamed to ...
22487         (vec_init<mode><VEC_base_l>): ... this.
22488         (vec_extract<mode>): Renamed to ...
22489         (vec_extract<mode><VEC_base_l>): ... this.
22490         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22491         (vec_initv2sfsf): ... this.
22492         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22493         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22494         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22495         gen_vec_init* calls.
22497 2017-08-01  Richard Biener  <rguenther@suse.de>
22499         PR tree-optimization/81297
22500         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22501         TREE_OVERFLOW from INTEGER_CSTs.
22503 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22505         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22507 2017-07-31  Carl Love  <cel@us.ibm.com>
22509         * config/rs6000/rs6000-c: Add support for built-in functions
22510         vector signed char vec_xl_be (signed long long, signed char *);
22511         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22512         vector signed int vec_xl_be (signed long long, signed int *);
22513         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22514         vector signed long long vec_xl_be (signed long long, signed long long *);
22515         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22516         vector signed short vec_xl_be (signed long long, signed short *);
22517         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22518         vector double vec_xl_be (signed long long, double *);
22519         vector float vec_xl_be (signed long long, float *);
22520         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22521         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22522         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22523         for the builtins.
22524         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22525         (altivec_expand_builtin): Add switch statement to call
22526         altivec_expand_xl_be for each builtin.
22527         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22528         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22529         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22530         __builtin_vsx_le_be_v16qi.
22531         * doc/extend.texi: Update the built-in documentation file for the
22532         new built-in functions.
22534 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22536         PR target/25967
22537         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22538         New function.
22539         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22541 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22543         * config.gcc: Add z14.
22544         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22545         CPU model numbers for z13s and z14.
22546         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22547         arch12 with z14.
22548         * config/s390/s390-opts.h (enum processor_type): Rename
22549         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22550         * config/s390/s390.c (processor_table): Add field for CPU name to
22551         be passed to Binutils.
22552         (s390_asm_output_machine_for_arch): Use the new field in
22553         processor_table for Binutils.
22554         (s390_expand_builtin): Replace arch12 with z14.
22555         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22556         (s390_get_sched_attrmask): Likewise.
22557         (s390_get_unit_mask): Likewise.
22558         * config/s390/s390.opt: Add z14 to processor_type enum.
22560 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22562         PR hsa/81477
22563         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22564         regardless of optimization level.
22566 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22567             Martin Liska  <mliska@suse.cz>
22569         * predict.def: Remove old comment and adjust probability.
22570         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22571         PREDICT statements.
22573 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22575         PR target/25967
22576         * config/i386/i386.c (ix86_function_naked): New function.
22577         (ix86_can_use_return_insn_p): Return false for naked functions.
22578         (ix86_expand_prologue): Skip prologue for naked functions.
22579         (ix86_expand_epilogue): Skip epilogue for naked functions
22580         and emit trap instruction.
22581         (ix86_warn_func_return): New function.
22582         (ix86_attribute_table): Add "naked" attribute specification.
22583         (TARGET_WARN_FUNC_RETURN): Define.
22584         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22586 2017-07-31  Martin Liska  <mliska@suse.cz>
22588         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22589         (dump_gimple_bb_header): Always dump BB info.
22590         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22592 2017-07-31  Martin Liska  <mliska@suse.cz>
22594         PR sanitize/81530
22595         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22596         also with current_function_decl non-null equality.
22598 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22600         PR sanitizer/81604
22601         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22602         change type to the element type, instead add eltype variable and
22603         use it where we are interested in the element type.
22605         PR tree-optimization/81603
22606         * ipa-polymorphic-call.c
22607         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22608         offset arithmetic in offset_int, bail out if the resulting bit offset
22609         doesn't fit into shwi.
22611 2017-07-31  Martin Liska  <mliska@suse.cz>
22613         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22614         (gimplify_save_expr): Fix comment.
22616 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22618         PR target/79793
22619         * config/i386/i386.c (ix86_function_arg): Update arguments for
22620         exception handler.
22621         (ix86_compute_frame_layout): Set the initial stack offset to
22622         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22623         INCOMING_FRAME_SP_OFFSET.
22624         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22625         stack before exception handler returns.
22626         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22627         the 'ERROR_CODE' for exception handler.
22629 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22631         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22632         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22633         (ASM_OUTPUT_REG_POP): Ditto.
22634         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22635         instead of asm_fprintf to output pure string.
22637 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22639         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22640         to imported_module_or_decl hook.
22641         (debug_nothing_tree_tree_tree_bool): Remove.
22642         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22643         * debug.c (do_nothing_debug_hooks): Use
22644         debug_nothing_tree_tree_tree_bool_bool instead of
22645         debug_nothing_tree_tree_tree_bool.
22646         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22647         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22648         * sdbout.c (sdb_debug_hooks): Likewise.
22649         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22650         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22651         langhook wants it.
22652         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22653         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22654         attribute, don't add anything.
22656 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22658         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22659         (fold_build2_stat_loc): Likewise.
22660         (fold_build3_stat_loc): Likewise.
22661         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22662         (fold_build1_loc): Remove macro.
22663         (fold_build2_loc): Likewise.
22664         (fold_build3_loc): Likewise.
22666 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22668         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22669         (gimple_build_debug_bind_source_stat): Likewise.
22670         * gimple.h (gimple_build_debug_bind): Remove macro.
22671         (gimple_build_debug_bind_source): Likewise.
22673 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22675         * bitmap.c (bitmap_alloc): Adjust.
22676         (bitmap_gc_alloc): Likewise.
22677         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22679 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22681         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22682         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22683         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22684         (bitmap_gc_alloc_stat): Likewise.
22685         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22687 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22689         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22690         * rtl.h (shallow_copy_rtx): Remove macro.
22692 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22694         * emit-rtl.c (gen_raw_REG): Adjust.
22695         * gengenrtl.c (gendef): Likewise.
22696         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22697         * rtl.h (rtx_alloc): Remove macro.
22699 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22701         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22702         (build_tree_list_stat): Likewise.
22703         * tree.h (build_tree_list): Remove macro.
22704         (build_tree_list_vec): Likewise.
22706 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22708         * tree.c (make_vector_stat): Remove _stat from name.
22709         (build_vector_stat): Likewise.
22710         * tree.h (make_vector_stat): Remove macro.
22711         (build_vector_stat): Likewise.
22713 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22715         * tree.h (build_var_debug_value): Remove prototype.
22717 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22719         * tree.c (tree_cons_stat): Remove _stat from name.
22720         * tree.h (tree_cons): Remove macro.
22722 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22724         * tree.c (build_vl_exp_stat): Remove _stat from name.
22725         * tree.h (build_vl_exp): Remove macro.
22727 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22729         * tree.c (build_decl_stat): Remove _stat from name.
22730         * tree.h (build_decl): Remove macro.
22732 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22734         * gimple.c (gimple_build_with_ops_stat): Adjust.
22735         (gimple_alloc_stat): Remove _stat from name.
22736         * gimple.h (gimple_alloc): Remove macro.
22738 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22740         * tree.c (make_tree_vec_stat): Remove _stat from name.
22741         (grow_tree_vec_stat): Likewise.
22742         * tree.h (make_tree_vec_stat): Adjust prototype.
22743         (grow_tree_vec_stat): Likewise.
22744         (make_tree_vec): Remove macro.
22745         (grow_tree_vec): Likewise.
22747 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22749         * fold-const.c (fold_build1_stat_loc): Adjust.
22750         (fold_build2_stat_loc): Likewise.
22751         (fold_build3_stat_loc): Likewise.
22752         * tree.c (build0_stat): Remove _stat from name.
22753         (build1_stat): Likewise.
22754         (build2_stat): Likewise.
22755         (build3_stat): Likewise.
22756         (build4_stat): Likewise.
22757         (build5_stat): Likewise.
22758         * tree.h (build1_loc): Remove macro, and rename _stat function
22759         to this.
22760         (build2_loc): Likewise.
22761         (build3_loc): Likewise.
22762         (build4_loc): Likewise.
22763         (build5_loc): Likewise.
22765 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22767         * tree.c (make_int_cst_stat): Remove _stat from name.
22768         * tree.h (make_int_cst_stat): Adjust prototype.
22769         (make_int_cst): Remove macro.
22771 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22773         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22774         * tree.h (make_tree_binfo_stat): Adjust prototype.
22775         (make_tree_binfo): Remove.
22777 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22779         * tree.c (copy_node_stat): Rename to copy_node.
22780         (build_distinct_type_copy): Adjust.
22781         * tree.h (copy_node_stat): Adjust prototype.
22782         (copy_node): Remove macro.
22784 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22786         * tree.c (make_node_stat): rename to make_node.
22787         (build_tree_list_stat): Adjust.
22788         (build0_stat): Likewise.
22789         (build2_stat): Likewise.
22790         (build3_stat): Likewise.
22791         (build4_stat): Likewise.
22792         (build5_stat): Likewise.
22793         (build_decl_stat): Likewise.
22794         * tree.h (make_node_stat): Adjust prototype.
22795         (make_node): remove macro.
22797 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22799         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22800         (PPC_FEATURE2_SCV): Likewise.
22801         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22803 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22805         * config/aarch64/aarch64.c
22806         (aarch64_internal_mov_immediate): Add new special pattern.
22807         * config/aarch64/aarch64.md (*movdi_aarch64):
22808         Add reg/32bit const mov case.
22810 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22811             Richard Sandiford <richard.sandiford@linaro.org>
22813         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22814         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22815         Add integer and movi cases.
22816         (movi-split-hf-df-sf split, fp16): New.
22817         (enabled): Added TARGET_FP_F16INST.
22818         * config/aarch64/iterators.md (GPF_HF): New.
22819         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22821 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22823         * config/aarch64/aarch64.c
22824         (aarch64_simd_container_mode): Add prototype.
22825         (aarch64_expand_mov_immediate): Add HI support.
22826         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22827         (aarch64_can_const_movi_rtx_p): New.
22828         (aarch64_preferred_reload_class):
22829         Remove restrictions of using FP registers for certain SIMD operations.
22830         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22831         (aarch64_valid_floating_const): Add integer move validation.
22832         (aarch64_simd_imm_scalar_p): Remove.
22833         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22834         (aarch64_legitimate_constant_p): Expand list of supported cases.
22835         * config/aarch64/aarch64-protos.h
22836         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22837         (aarch64_reinterpret_float_as_int): New.
22838         (aarch64_simd_imm_scalar_p): Remove.
22839         * config/aarch64/constraints.md (Uvi): New.
22840         (Dd): Split into Ds and new Dd.
22841         * config/aarch64/aarch64.md (*movsi_aarch64):
22842         Add SIMD mov case.
22843         (*movdi_aarch64): Add SIMD mov case.
22845 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22847         * tree-predcom.c: (struct chain): Handle store-store chain in which
22848         stores for elimination only store loop invariant values.
22849         (execute_pred_commoning_chain): Ditto.
22850         (prepare_initializers_chain_store_elim): Ditto.
22851         (prepare_finalizers): Ditto.
22852         (is_inv_store_elimination_chain): New function.
22853         (initialize_root_vars_store_elim_1): New function.
22855 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22857         * tree-predcom.c: Revise general description of the pass.
22858         (enum chain_type): New enum type for store elimination.
22859         (struct chain): New field supporting store elimination.
22860         (struct component): Ditto.
22861         (dump_chain): Dump store-stores chain.
22862         (release_chain): Release resources.
22863         (split_data_refs_to_components): Compute and create component
22864         contains only stores for elimination.
22865         (get_chain_last_ref_at): New function.
22866         (make_invariant_chain): Initialization.
22867         (make_rooted_chain): Specify chain type in parameter and record it.
22868         (add_looparound_copies): Skip for store-stores chain.
22869         (determine_roots_comp): Compute type of chain and pass it to
22870         make_rooted_chain.
22871         (initialize_root_vars_store_elim_2): New function.
22872         (finalize_eliminated_stores): New function.
22873         (remove_stmt): Handle store for elimination.
22874         (execute_pred_commoning_chain): Execute predictive commoning on
22875         store-store chains.
22876         (determine_unroll_factor): Skip unroll for store-stores chain.
22877         (prepare_initializers_chain_store_elim): New function.
22878         (prepare_initializers_chain): Hanlde store-store chain.
22879         (prepare_finalizers_chain, prepare_finalizers): New function.
22880         (tree_predictive_commoning_loop): Return integer value indicating
22881         if loop is unrolled or lcssa form is corrupted.
22882         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
22884 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22886         * tree-predcom.c (initialize_root): Delete.
22887         (execute_pred_commoning_chain): Initialize root vars and replace
22888         reference of non-combined chain directly, rather than call above
22889         function.
22891 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22893         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
22894         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
22896 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22898         * tree-predcom.c (struct chain): New field init_seq.
22899         (release_chain): Release init_seq.
22900         (prepare_initializers_chain): Record intialization stmts in above
22901         field.
22902         (insert_init_seqs): New function.
22903         (tree_predictive_commoning_loop): Call insert_init_seqs.
22905 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22907         * tree-predcom.c (determine_roots_comp): Skip trivial components.
22909 2017-07-28  Richard Biener  <rguenther@suse.de>
22911         * match.pd: Remove superfluous :c.
22912         * genmatch.c (simplify::id): Add member.
22913         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
22914         Copy id.
22915         (current_id): New global.
22916         (dt_node::parent): Move from ...
22917         (dt_operand::parent): ... here.  Add for_id member.
22918         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
22919         (decision_tree::find_node): Relax order requirement when
22920         merging DT_TRUE nodes to ones inbetween the current simplify
22921         and the one we try to merge with.  Add diagnostic whenever
22922         we need to enforce pattern order by not merging.
22923         (decision_tree::insert): Set current_id.
22924         (decision_tree::print_node): Dump parent node and for_id.
22925         (parser::last_id): Add member.
22926         (parser::push_simplify): Assign unique id.
22927         (parser::parser): Initialize last_id.
22929 2017-07-28  Martin Liska  <mliska@suse.cz>
22931         PR sanitizer/81340
22932         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
22933         gimple_build_debug_bind.
22935 2017-07-28  Richard Biener  <rguenther@suse.de>
22937         PR tree-optimization/81502
22938         * match.pd: Add pattern combining BIT_INSERT_EXPR with
22939         BIT_FIELD_REF.
22940         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
22941         size/pos operands.
22942         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
22943         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
22944         for BIT_FIELD_REF args.
22945         * fold-const.c (make_bit_field_ref): Likewise.
22946         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22948 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
22950         PR sanitizer/80998
22951         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
22952         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
22953         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
22954         Or it into SANITIZER_UNDEFINED.
22955         * ubsan.c: Include gimple-fold.h and varasm.h.
22956         (ubsan_expand_ptr_ifn): New function.
22957         (instrument_pointer_overflow): New function.
22958         (maybe_instrument_pointer_overflow): New function.
22959         (instrument_object_size): Formatting fix.
22960         (pass_ubsan::execute): Call instrument_pointer_overflow
22961         and maybe_instrument_pointer_overflow.
22962         * internal-fn.c (expand_UBSAN_PTR): New function.
22963         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
22964         * sanitizer.def (__ubsan_handle_pointer_overflow,
22965         __ubsan_handle_pointer_overflow_abort): New builtins.
22966         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
22967         * internal-fn.def (UBSAN_PTR): New internal function.
22968         * opts.c (sanitizer_opts): Add pointer-overflow.
22969         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
22970         * fold-const.c (build_range_check): Compute pointer range check in
22971         integral type if pointer arithmetics would be needed.  Formatting
22972         fixes.
22974 2017-07-28  Martin Liska  <mliska@suse.cz>
22976         PR sanitizer/81460
22977         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
22978         parameters that are of a variable-length.
22980 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
22982         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
22983         rs6000/biarch64.h.
22984         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
22985         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
22986         (CRT_CALL_STATIC_FUNCTION): Likewise.
22987         (ASM_DEFAULT_SPEC): New define.
22988         (ASM_SPEC32): Likewise.
22989         (ASM_SPEC64): Likewise.
22990         (ASM_SPEC_COMMON): Likewise.
22991         (ASM_SPEC): Likewise.
22992         (INVALID_64BIT): Likewise.
22993         (LINK_OS_DEFAULT_SPEC): Likewise.
22994         (LINK_OS_SPEC32): Likewise.
22995         (LINK_OS_SPEC64): Likewise.
22996         (POWERPC_LINUX): Likewise.
22997         (PTRDIFF_TYPE): Likewise.
22998         (RESTORE_FP_PREFIX): Likewise.
22999         (RESTORE_FP_SUFFIX): Likewise.
23000         (SAVE_FP_PREFIX): Likewise.
23001         (SAVE_FP_SUFFIX): Likewise.
23002         (SIZE_TYPE): Likewise.
23003         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23004         (TARGET_64BIT): Likewise.
23005         (TARGET_64BIT): Likewise.
23006         (TARGET_AIX): Likewise.
23007         (WCHAR_TYPE_SIZE): Likewise.
23008         (WCHAR_TYPE): Undefine.
23009         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23010         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23011         (CPP_OS_RTEMS_SPEC): Delete.
23012         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23013         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23014         link_os_spec64.
23015         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23017 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23019         PR tree-optimization/81578
23020         * tree-parloops.c (build_new_reduction): Bail out if
23021         reduction_code isn't one of the standard OpenMP reductions.
23022         Move the details printing after that decision.
23024 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23026         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23027         related to reload_in_progress.
23028         (splat_input_operand): Likewise.
23029         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23030         Delete prototype.
23031         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23032         field.
23033         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23034         (TARGET_INSTANTIATE_DECLS): Likewise.
23035         (legitimate_indexed_address_p): Delete reload_in_progress code.
23036         (rs6000_debug_legitimate_address_p): Likewise.
23037         (rs6000_eliminate_indexed_memrefs): Likewise.
23038         (rs6000_emit_le_vsx_store): Likewise.
23039         (rs6000_emit_move_si_sf_subreg): Likewise.
23040         (rs6000_emit_move): Likewise.
23041         (register_to_reg_type): Likewise.
23042         (rs6000_pre_atomic_barrier): Likewise.
23043         (rs6000_machopic_legitimize_pic_address): Likewise.
23044         (rs6000_allocate_stack_temp): Likewise.
23045         (rs6000_address_for_fpconvert): Likewise.
23046         (rs6000_address_for_altivec): Likewise.
23047         (rs6000_secondary_memory_needed_rtx): Delete function.
23048         (rs6000_check_sdmode): Likewise.
23049         (rs6000_alloc_sdmode_stack_slot): Likewise.
23050         (rs6000_instantiate_decls): Likewise.
23051         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23052         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23053         Delete reload_in_progress.
23054         (*vec_reload_and_plus_<mptrsize>): Likewise.
23055         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23056         (vsx_div_v2di): Likewise.
23057         (vsx_udiv_v2di): Likewise.
23059 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23061         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23062         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23063         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23064         (rs6000_debug_reg_global): Delete print of LRA status.
23065         (rs6000_option_override_internal): Delete dead LRA related code.
23066         (rs6000_lra_p): Delete function.
23067         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23069 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23071         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23072         * config/riscv/rtems.h: New file.
23074 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23075             Sudakshina Das  <sudi.das@arm.com>
23077         * config/aarch64/aarch64.md
23078         (define_split for and<mode>3nr_compare): Move
23079         non aarch64_logical_operand to a register.
23080         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23081         register immediate operand to a register.
23082         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23084 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23086         PR middle-end/81564
23087         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23089 2017-07-27  Richard Biener  <rguenther@suse.de>
23091         PR tree-optimization/81573
23092         PR tree-optimization/81494
23093         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23094         multi defuse cycle case.
23096 2017-07-27  Richard Biener  <rguenther@suse.de>
23098         PR tree-optimization/81571
23099         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23100         PHIs.
23102 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23104         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23105         earlier and only if MASK_FPU is set.  Adjust formatting.
23107 2017-07-27  Martin Liska  <mliska@suse.cz>
23109         * opt-functions.awk: Add validation of value of Init.
23110         * optc-gen.awk: Pass new argument.
23112 2017-07-27  Martin Liska  <mliska@suse.cz>
23114         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23115         Fix wrong condition.
23117 2017-07-27  Martin Liska  <mliska@suse.cz>
23119         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23120         BBs and edges seen by autoFDO.
23122 2017-07-27  Richard Biener  <rguenther@suse.de>
23124         PR tree-optimization/81502
23125         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23126         with incompatible but same sized type.
23127         (execute_update_addresses_taken): Likewise.
23129 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23131         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23132         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23134 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23136         PR target/81534
23137         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23138         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23139         Change s_operand to memory_operand.
23141 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23143         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23144         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23145         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23146         Emit instructions rather than returning an expression.  Handle TFmode
23147         and KFmode by casting to TImode.
23148         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23149         (rs6000_emit_le_vsx_store): Likewise.
23150         * config/rs6000/vsx.md (VSX_TI): New iterator.
23151         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23152         (*vsx_le_undo_permute_<mode>): Likewise.
23153         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23154         emit the split sequence.
23155         (*vsx_le_perm_store_<mode>): Likewise.
23157 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23159         PR tree-optimization/81555
23160         PR tree-optimization/81556
23161         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23162         if true, force CHANGED for the recursive invocation.
23163         (reassociate_bb): Remember original length of ops array, pass
23164         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23166         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23167         attributes for noipa attribute.  For naked attribute use
23168         lookup_attribute first before lookup_attribute_spec.
23169         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23170         noipa attribute.
23171         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23172         for functions with noipa attribute.
23173         (cgraph_externally_visible_p): Return true for functions with noipa
23174         attribute.
23175         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23176         for functions with noipa attribute.
23177         * doc/extend.texi: Document noipa function attribute.
23178         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23179         also for functions with noipa attribute.
23180         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23182 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23184         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23185         vec_unalign_load_cost and vec_unalign_store_cost.
23187 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23189         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23190         -mvsx-small-integer option.
23191         (ISA_3_0_MASKS_IEEE): Likewise.
23192         (OTHER_VSX_VECTOR_MASKS): Likewise.
23193         (POWERPC_MASKS): Likewise.
23194         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23195         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23196         code, only testing for DImode being allowed in non-VSX floating
23197         point registers.
23198         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23199         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23200         another VSX test.
23201         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23202         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23203         TARGET_P8_VECTOR test.
23204         (rs6000_secondary_reload_simple_move): Likewise.
23205         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23206         since TARGET_P9_VECTOR was already tested.
23207         (rs6000_opt_masks): Remove -mvsx-small-integer.
23208         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23209         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23210         used.
23211         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23212         test for TARGET_VEXTRACTUB was used, and that uses
23213         TARGET_P9_VECTOR.
23214         (p9 extract splitter): Likewise.
23215         (vsx_extract_<mode>_di_p9): Likewise.
23216         (vsx_extract_<mode>_store_p9): Likewise.
23217         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23218         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23219         the elimination of TARGET_VSX_SMALL_INTEGER.
23220         (vsx_extract_<mode>_p8): Likewise.
23221         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23222         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23223         (vsx_set_<mode>_p9): Likewise.
23224         (vsx_set_v4sf_p9): Likewise.
23225         (vsx_set_v4sf_p9_zero): Likewise.
23226         (vsx_insert_extract_v4sf_p9): Likewise.
23227         (vsx_insert_extract_v4sf_p9_2): Likewise.
23228         * config/rs6000/rs6000.md (sign extend splitter): Change
23229         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23230         (floatsi<mode>2_lfiwax_mem): Likewise.
23231         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23232         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23233         since a test for TARGET_P9_VECTOR was used.
23234         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23235         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23236         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23237         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23238         TARGET_P8_VECTOR test.
23239         (fix_trunc<mode>si2_stfiwx): Likewise.
23240         (fix_trunc<mode>si2_internal): Likewise.
23241         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23242         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23243         used.
23244         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23245         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23246         TARGET_P8_VECTOR test.
23247         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23248         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23249         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23250         used.
23251         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23252         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23253         since a test for TARGET_P9_VECTOR was used.
23254         (splitter for loading small constants): Likewise.
23256 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23258         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23259         vec_fp_stmt_cost.
23261 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23263         PR target/81563
23264         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23265         (fp_valid_at): Likewise.
23267 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23269         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23270         (qdf24xx_addrcost_table): Likewise.
23271         (cortexa57_tunings): Update to use generic_branch_cost.
23272         (cortexa72_tunings): Likewise.
23273         (cortexa73_tunings): Likewise.
23274         (qdf24xx_tunings): Likewise.
23276 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23278         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23279         (thunderx2t99_branch_cost): Likewise.
23280         (cortexa35_tunings): Update to use generic_branch_cost.
23281         (cortexa53_tunings): Likewise.
23282         (cortexa57_tunings): Likewise.
23283         (cortexa72_tunings): Likewise.
23284         (cortexa73_tunings): Likewise.
23285         (thunderx2t99_tunings): Likewise.
23287 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23289         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23290         (sparc_option_override): Honour MASK_FSMULD.
23291         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23292         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23293         * config/sparc/sparc.opt (mfsmuld): New option.
23294         * doc/invoke.texi (mfsmuld): Document option.
23296 2017-07-26  Marek Polacek  <polacek@redhat.com>
23298         PR middle-end/70992
23299         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23301 2017-07-26  Richard Biener  <rguenther@suse.de>
23303         * gimple-match-head.c (do_valueize): Return OP if valueize
23304         returns NULL_TREE.
23305         (get_def): New helper to get at the def stmt of a SSA name
23306         if valueize allows.
23307         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23308         do_valueize to get at the def stmt.
23309         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23311 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23313         PR middle-end/46932
23314         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23316 2017-07-26  Martin Liska  <mliska@suse.cz>
23318         PR sanitize/81186
23319         * function.c (expand_function_start): Make expansion of
23320         nonlocal_goto_save_area after parm_birth_insn.
23322 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23324         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23325         from all CPU target flags enable members.
23327 2017-07-26  Richard Biener  <rguenther@suse.de>
23329         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23330         (decision_tree::gen): Make 'type' const.
23331         (write_predicate): Likewise.
23333 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23335         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23336         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23337         (rs6000_option_override_internal): Likewise.
23338         (rs6000_expand_vector_set): Likewise.
23339         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23340         (TARGET_UPPER_REGS_SF): Likewise.
23341         (TARGET_UPPER_REGS_DI): Likewise.
23342         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23343         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23344         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23345         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23346         (Splitters for DI constants in Altivec registers): Likewise.
23347         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23348         (vsx_set_v4sf_p9): Likewise.
23349         (vsx_set_v4sf_p9_zero): Likewise.
23350         (vsx_insert_extract_v4sf_p9): Likewise.
23351         (vsx_insert_extract_v4sf_p9_2): Likewise.
23353 2017-07-25  Carl Love  <cel@us.ibm.com>
23355         * doc/extend.texi: Update the built-in documentation file for the
23356         existing built-in functions
23357         vector signed char vec_cnttz (vector signed char);
23358         vector unsigned char vec_cnttz (vector unsigned char);
23359         vector signed short vec_cnttz (vector signed short);
23360         vector unsigned short vec_cnttz (vector unsigned short);
23361         vector signed int vec_cnttz (vector signed int);
23362         vector unsigned int vec_cnttz (vector unsigned int);
23363         vector signed long long vec_cnttz (vector signed long long);
23364         vector unsigned long long vec_cnttz (vector unsigned long long);
23366 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23368         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23369         accesses where the use is for the first operand of a BIT_INSERT.
23371 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23373         PR bootstrap/81521
23374         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23375         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23377 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23379         * config/i386/gstabs.h: Delete.
23380         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23382 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23384         * config/i386/i386.c (ix86_decompose_address): Do not check for
23385         register RTX when looking at index_reg or base_reg.
23386         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23388 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23390         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23391         to update EH info here.
23393 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23395         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23397 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23399         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23401 2017-07-25  Torsten Duwe  <duwe@suse.de>
23403         * common.opt: Introduce -fpatchable-function-entry
23404         command line option, and its variables function_entry_patch_area_size
23405         and function_entry_patch_area_start.
23406         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23407         including a two-value parser.
23408         * target.def (print_patchable_function_entry): New target hook.
23409         * targhooks.h (default_print_patchable_function_entry): New function.
23410         * targhooks.c (default_print_patchable_function_entry): Likewise.
23411         * toplev.c (process_options): Switch off IPA-RA if
23412         patchable function entries are being generated.
23413         * varasm.c (assemble_start_function): Look at the
23414         patchable-function-entry command line switch and current
23415         function attributes and maybe generate NOP instructions by
23416         calling the print_patchable_function_entry hook.
23417         * doc/extend.texi: Document patchable_function_entry attribute.
23418         * doc/invoke.texi: Document -fpatchable_function_entry
23419         command line option.
23420         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23421         New target hook.
23422         * doc/tm.texi: Re-generate.
23424 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23426         PR target/81532
23427         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23428         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23430 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23432         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23434 2017-07-25  Richard Biener  <rguenther@suse.de>
23436         PR tree-optimization/81455
23437         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23438         not walk in cycles when looking for guards.
23440 2017-07-25  Richard Biener  <rguenther@suse.de>
23442         PR tree-optimization/81529
23443         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23444         when optimizing backedge uses.
23446 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23448         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23449         character for AIX.
23450         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23451         to dl_section_ref.  On AIX, append an expression to subtract
23452         the size of the section length to dl_section_ref.
23454 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23456         * configure.ac: If any of the config.* scripts fail, exit 1.
23457         * configure: Regenerate.
23459 2017-07-25  Richard Biener  <rguenther@suse.de>
23461         PR middle-end/81546
23462         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23463         of immediate uses, be more verbose on errors.
23465 2017-07-25  Richard Biener  <rguenther@suse.de>
23467         PR tree-optimization/81510
23468         * tree-vect-loop.c (vect_is_simple_reduction): When the
23469         reduction stmt is not inside the loop bail out.
23471 2017-07-25  Richard Biener  <rguenther@suse.de>
23473         PR tree-optimization/81303
23474         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23475         profitability check against LOOP_VINFO_NITERSM1.
23477 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23479         * domwalk.c (cmp_bb_postorder): Simplify.
23480         (sort_bbs_postorder): New function.  Use it...
23481         (dom_walker::walk): ...here to optimize common cases.
23483 2017-07-25  Martin Liska  <mliska@suse.cz>
23485         PR ipa/81520
23486         * ipa-visibility.c (function_and_variable_visibility): Make the
23487         redirection just on target that supports aliasing.
23488         Fix GNU coding style.
23490 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23492         PR libgcc/61152
23493         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23494         Format changes.
23495         * config/arm/rtems.h: Likewise.
23496         * config/bfin/rtems.h: Likewise.
23497         * config/i386/rtemself.h: Likewise.
23498         * config/lm32/rtems.h: Likewise.
23499         * config/m32c/rtems.h: Likewise.
23500         * config/m68k/rtemself.h: Likewise.
23501         * config/microblaze/rtems.h: Likewise.
23502         * config/mips/rtems.h: Likewise.
23503         * config/moxie/rtems.h: Likewise.
23504         * config/nios2/rtems.h: Likewise.
23505         * config/powerpcspe/rtems.h: Likewise.
23506         * config/rs6000/rtems.h: Likewise.
23507         * config/rtems.h: Likewise.
23508         * config/sh/rtems.h: Likewise.
23509         * config/sh/rtemself.h: Likewise.
23510         * config/sparc/rtemself.h: Likewise.
23512 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23514         PR 81487
23515         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23516         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23517         * tree-ssa-structalias.c (alias_get_name): Same.
23519 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23521         PR target/81414
23522         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23523         instructions if no du chain is found.
23525 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23527         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23529 2017-07-25  Richard Biener  <rguenther@suse.de>
23531         PR middle-end/81505
23532         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23533         sticky.
23535 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23537         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23538         upper-regs options.
23539         (ISA_2_7_MASKS_SERVER): Likewise.
23540         (ISA_3_0_MASKS_IEEE): Likewise.
23541         (OTHER_P8_VECTOR_MASKS): Likewise.
23542         (OTHER_VSX_VECTOR_MASKS): Likewise.
23543         (POWERPC_MASKS): Likewise.
23544         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23545         duplicate list of options.
23546         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23547         explicit -mupper-regs options.
23548         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23549         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23550         alias for -mupper-regs-df.
23551         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23552         (rs6000_init_hard_regno_mode_ok): Likewise.
23553         (rs6000_option_override_internal): Likewise.
23554         (rs6000_opt_masks): Likewise.
23555         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23556         options in terms of whether -mvsx or -mpower8-vector was used.
23557         (TARGET_UPPER_REGS_DI): Likewise.
23558         (TARGET_UPPER_REGS_SF): Likewise.
23559         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23560         -mupper-regs-* options.
23562 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23564         * passes.c (emergency_dump_function): Print some empty lines and a
23565         header before the RTL dump.
23567 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23569         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23571 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23573         PR target/79041
23574         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23575         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23577 2017-07-24  Carl Love  <cel@us.ibm.com>
23579         * config/rs6000/rs6000-c.c: Add support for built-in functions
23580         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23581         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23582         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23583         vec_extract_fp_from_shortl): Add defines for the two builtins.
23584         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23585         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23586         new builtins.
23587         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23588         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23589         * doc/extend.texi: Update the built-in documentation file for the
23590         new built-in function.
23592 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23594         PR bootstrap/81521
23595         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23596         documentation.
23597         * doc/generic.texi: Likewise.
23598         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23599         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23601 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23603         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23604         (aarch64_mls_elt_merge<mode>): Likewise.
23606 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23608         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23609         having __cxa_atexit.
23611 2017-07-23  Michael Collison  <michael.collison@arm.com>
23613         * config/arm/arm.c (arm_option_override): Deprecate
23614         use of -mstructure-size-boundary.
23615         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23616         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23618 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23620         PR target/80695
23621         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23622         Reduce cost estimate for direct moves.
23624 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23626         PR target/80569
23627         * config/i386/i386.c (ix86_option_override_internal): Disable
23628         BMI, BMI2 and TBM instructions for -m16.
23630 2017-07-21  Carl Love  <cel@us.ibm.com>
23632         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23633         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23634         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23635         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23636         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23637         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23638         VMULOSW): New enum "unspec" values.
23639         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23640         altivec_vmulosw): New patterns.
23641         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23642         VMULOSW): Add definitions.
23644 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23646         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23647         (qdf24xx): Likewise.
23648         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23649         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23650         (AARCH64_FL_V8_1): Renumber.
23651         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23652         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23653         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23654         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23655         rdma to feature modifiers list.
23657 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23659         PR middle-end/56727
23660         * ipa-visibility (function_and_variable_visibility): Convert
23661         recursive PLT call to direct call if appropriate.
23663 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23665         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23666         operand 1 to see if the types precision matches.
23667         * fold-const.c (operand_equal_p): Likewise.
23669 2017-07-21  Richard Biener  <rguenther@suse.de>
23671         PR tree-optimization/81303
23672         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23673         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23674         (vect_peeling_hash_get_lowest_cost): Adjust.
23675         (vect_enhance_data_refs_alignment): Likewise.  Use
23676         vect_get_peeling_costs_all_drs to compute the penalty for no
23677         peeling to match up costs.
23679 2017-07-21  Richard Biener  <rguenther@suse.de>
23681         PR tree-optimization/81500
23682         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23683         we didn't identify a reduction path.
23685 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23686             Cesar Philippidis  <cesar@codesourcery.com>
23688         PR gcov-profile/81442
23689         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23690         probabilities.
23692 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23694         PR lto/81430
23695         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23696         function.
23697         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23698         nvptx_override_options_after_change.
23700 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23702         * dwarf2out.c (output_file_names): Avoid double testing for
23703         dwarf_version >= 5.
23705 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23707         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23709 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23711         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23712         hot/cold regions.
23713         (try_crossjump_to_edge): Do not punt on partitioned functions.
23715 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23717         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23718         Put all BBs reachable only via paths crossing cold region to cold
23719         region.
23720         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23722 2016-07-21  Richard Biener  <rguenther@suse.de>
23724         PR tree-optimization/81303
23725         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23726         into account prologue and epilogue iterations when raising
23727         min_profitable_iters to sth at least covering one vector iteration.
23729 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23731         * config/arm/arm.c (arm_test_cpu_arch_dat):
23732         Check for overlap.
23734 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23736         Remove TYPE_METHODS.
23737         * tree.h (TYPE_METHODS): Delete.
23738         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23739         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23740         (dbxout_type_methods): Scan TYPE_FIELDS.
23741         (dbxout_type): Don't check TYPE_METHODS here.
23742         * function.c (use_register_for_decl): Always ignore register for
23743         class types when not optimizing.
23744         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23745         * tree.c (free_lang_data_in_type): Stitch out member functions and
23746         templates from TYPE_FIELDS.
23747         (build_distinct_type_copy, verify_type_variant,
23748         verify_type): Member fns are on TYPE_FIELDS.
23749         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23750         * tree-pretty-print.c (dump_generic_node): Likewise.
23752 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23754         PR target/80846
23755         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23756         V2TImode and V4TImode.
23757         (ix86_expand_vector_extract): Likewise.
23758         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23759         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23760         (ssescalarmode): Handle V4TImode and V2TImode.
23761         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23762         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23763         (VEXTRACTI128_MODE): New mode iterator.
23764         (splitter for *vec_extractv?ti first element): New.
23765         (VEC_INIT_MODE): New mode iterator.
23766         (vec_init<mode>): Consolidate 3 expanders into one using
23767         VEC_INIT_MODE mode iterator.
23769 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23771         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23772         non_spilled_static_chain_regno_p.
23774 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23776         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23778 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23780         * bb-reorder.c (connect_traces): Allow copying of blocks within
23781         single partition.
23783 2017-07-20  Richard Biener  <rguenther@suse.de>
23785         * gimple.h (gimple_phi_result): Add gphi * overload.
23786         (gimple_phi_result_ptr): Likewise.
23787         (gimple_phi_arg): Likewise.  Adjust index assert to only
23788         allow actual argument accesses rather than all slots available
23789         by capacity.
23790         (gimple_phi_arg_def): Add gphi * overload.
23791         * tree-phinodes.c (make_phi_node): Initialize only actual
23792         arguments.
23793         (resize_phi_node): Clear memory not covered by old node,
23794         do not initialize excess argument slots.
23795         (reserve_phi_args_for_new_edge): Initialize new argument slot
23796         completely.
23798 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23800         PR tree-optimization/81388
23801         Revert r238585:
23802         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23804         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23805         by removing computation of may_be_zero.
23807 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23808             Tom de Vries  <tom@codesourcery.com>
23810         PR middle-end/81030
23811         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23812         when gimple level profile disagrees with what RTL expander did.
23814 2017-07-20  Richard Biener  <rguenther@suse.de>
23816         PR tree-optimization/61171
23817         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23818         (vect_analyze_stmt): Add slp instance parameter.
23819         (vectorizable_reduction): Likewise.
23820         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23821         (vect_is_simple_reduction): Deal with chains not detected
23822         as SLP reduction chain, specifically not properly associated
23823         chains containing a mix of plus/minus.
23824         (get_reduction_op): Remove.
23825         (get_initial_defs_for_reduction): Simplify, pass in whether
23826         this is a reduction chain, pass in the SLP node for the PHIs.
23827         (vect_create_epilog_for_reduction): Get the SLP instance as
23828         arg and adjust.
23829         (vectorizable_reduction): Get the SLP instance as arg.
23830         During analysis remember the SLP node with the PHIs in the
23831         instance.  Simplify getting at the vectorized reduction PHIs.
23832         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23833         through SLP instance.
23834         (vect_slp_analyze_operations): Likewise.
23835         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23836         (vect_transform_stmt): Likewise.
23838 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23840         PR tree-optimization/81489
23841         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23842         read of phi arg location to before loop that modifies phi.
23844 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23846         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23847         New pattern.
23849 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23851         PR middle-end/81331
23852         * except.c (execute): Fix ordering issue.
23854 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23856         PR rtl-optimization/81423
23857         * combine.c (make_compound_operation_int): Don't try to optimize
23858         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23860 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23862         PR rtl-optimization/81423
23863         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23864         with a constant that is -1 in the truncated to mode.
23866 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23868         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
23869         (determine_unlikely_bbs): ... here.
23870         * predict.h (propagate_unlikely_bbs_forward): Declare.
23871         * cfgexpand.c (pass_expand::execute): Use it.
23872         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
23873         unlikely edges.
23874         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
23875         propagate_unlikely_bbs_forward.
23877 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23879         PR middle-end/81331
23880         * except.c (maybe_add_nop_after_section_switch): New function.
23881         (execute): Use it.
23883 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23885         * gimple.h (gimple_phi_set_arg): Make assert more strict.
23887 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23889         * gimple.h (gimple_phi_arg): Make assert more strict.
23891 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
23893         * config.gcc (powerpc*-*-*): Add mmintrin.h.
23894         * config/rs6000/mmintrin.h: New file.
23895         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
23897 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
23899         PR tree-optimization/81346
23900         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
23902 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23904         * config/nvptx/nvptx.md (VECIM): Add V2DI.
23906 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23908         * config/nvptx/nvptx-modes.def: Add V2DImode.
23909         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
23910         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
23911         (nvptx_output_mov_insn): Handle lack of mov.b128.
23912         (nvptx_print_operand): Handle 'H' and 'L' codes.
23913         (nvptx_vector_mode_supported): Allow V2DImode.
23914         (nvptx_preferred_simd_mode): New function.
23915         (nvptx_data_alignment): New function.
23916         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
23917         nvptx_preferred_simd_mode.
23918         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
23919         64 to 128 bits.
23920         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
23922 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23924         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
23925         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
23926         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
23927         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
23928         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
23929         (mov<VECIM>_insn): New define_insn.
23930         (define_expand "mov<VECIM>): New define_expand.
23932 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23934         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
23936 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
23938         PR tree-optimization/81346
23939         * fold-const.h (fold_div_compare, range_check_type): Declare.
23940         * fold-const.c (range_check_type): New function.
23941         (build_range_check): Use range_check_type.
23942         (fold_div_compare): No longer static, rewritten into
23943         a match.pd helper function.
23944         (fold_comparison): Don't call fold_div_compare here.
23945         * match.pd (X / C1 op C2): New optimization using fold_div_compare
23946         as helper function.
23948 2017-07-19  Nathan Sidwell  <nathan@acm.org>
23950         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
23951         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
23952         * tree.c (find_decls_types_r, verify_type): Use
23953         TYPE_{MIN,MAX}_VALUE_RAW.
23954         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
23955         (hash_tree): Likewise.
23956         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
23957         Likewise.
23958         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
23959         Likewise.
23961 2017-07-18  Tom de Vries  <tom@codesourcery.com>
23963         PR middle-end/81464
23964         * omp-expand.c (expand_omp_for_static_chunk): Handle
23965         equal-argument loop exit phi.
23967 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
23969         PR target/81471
23970         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
23971         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
23972         operand 2 predicate.
23973         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
23974         operand 2 predicate.
23975         (ror,rol -> rorx splitters): Use const_int_operand as
23976         operand 2 predicate.
23978 2017-06-18  Richard Biener  <rguenther@suse.de>
23980         PR tree-optimization/81410
23981         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
23982         the gap in the ! slp_perm SLP case after each group.
23984 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23986         PR middle-end/81463
23987         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
23988         again.
23990 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23992         PR middle-end/81462
23993         * predict.c (set_even_probabilities): Cleanup; do not affect
23994         probabilities that are already known.
23995         (combine_predictions_for_bb): Call even when count is set.
23997 2017-07-18  Nathan Sidwell  <nathan@acm.org>
23999         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24000         TYPE_MAX_VALUE.
24002 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24004         PR target/81408
24005         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24006         optimization for loop niter analysis.
24008 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24010         PR target/81473
24011         * config/avr/avr.c (avr_optimize_casesi): Don't use
24012         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24014 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24016         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24017         body_cost_vec from _vect_peel_extended_info.
24018         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24019         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24020         npeel.
24022 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24024         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24026 2017-07-18  Richard Biener  <rguenther@suse.de>
24028         PR tree-optimization/80620
24029         PR tree-optimization/81403
24030         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24031         info when re-using a VN table entry.
24033 2017-07-18  Richard Biener  <rguenther@suse.de>
24035         PR tree-optimization/81418
24036         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24037         vectype_in.  Verify that with lane-reducing reduction operations
24038         we have a single def-use cycle.
24040 2017-07-17  Carl Love  <cel@us.ibm.com>
24042         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24044         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24045         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24046         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24047         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24048         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24049         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24050         VMULOSW): New enum "unspec" values.
24051         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24052         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24053         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24054         altivec_vmulosw): New patterns.
24055         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24056         VMULOSW): Add definitions.
24058 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24060         * config/alpha/alpha.c: Include predict.h.
24062 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24064         * tree-vrp.c (compare_assert_loc): Fix comparison function
24065         to return predictable results.
24067 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24069         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24070         option.
24071         (subdi3): Likewise.
24072         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24073         * doc/invoke.texi (mexpand-adddi): Update text.
24075 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24077         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24078         that also clobbers the CC register. The old expand code is moved
24079         to ...
24080         (*arc_clzsi2): ... here.
24081         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24082         the CC register. The old expand code is moved to ...
24083         (arc_ctzsi2): ... here.
24085 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24087         * config/arc/arc.opt (mindexed-loads): Use initial value
24088         TARGET_INDEXED_LOADS_DEFAULT.
24089         (mauto-modify-reg): Use initial value
24090         TARGET_AUTO_MODIFY_REG_DEFAULT.
24091         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24092         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24093         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24094         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24096 2017-07-17  Martin Liska  <mliska@suse.cz>
24098         PR sanitizer/81302
24099         * opts.c (finish_options): Do not allow -fgnu-tm
24100         w/ -fsanitize={kernel-,}address.  Say sorry.
24102 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24104         PR target/81369
24105         * tree-loop-distribution.c (classify_partition): Only assert on
24106         numer of iterations.
24107         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24108         (distribute_loop): Remove code handling loop with unknown niters.
24109         (pass_loop_distribution::execute): Skip loop with unknown niters.
24111 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24113         PR target/81369
24114         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24115         function sort_partitions_by_post_order.
24117 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24119         PR tree-optimization/81374
24120         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24121         the max index of basic blocks, rather than number of basic blocks.
24123 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24125         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24126         proto.
24127         (arc_legitimate_pic_operand_p): Likewise.
24128         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24129         function.
24130         (arc_needs_pcl_p): Likewise.
24131         (arc_legitimate_pc_offset_p): Likewise.
24132         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24133         function is also used in constrains.md.
24134         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24135         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24136         PLUS.  Only return true/false in known cases, otherwise assert.
24137         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24138         is already called in arc_legitimate_constant_p.
24139         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24140         pic addresses.
24141         (LEGITIMATE_PIC_OPERAND_P): Use
24142         arc_raw_symbolic_reference_mentioned_p function.
24143         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24144         function.
24145         (Cal): Likewise.
24146         (C32): Likewise.
24148 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24149         Andrew Burgess  <andrew.burgess@embecosm.com>
24151         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24152         (arc_return_address_register): New function.
24153         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24154         (arc_handle_fndecl_attribute): Add naked attribute.
24155         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24156         (TARGET_WARN_FUNC_RETURN): Likewise.
24157         (arc_allocate_stack_slots_for_args): New function.
24158         (arc_warn_func_return): Likewise.
24159         (machine_function): Change type fn_type.
24160         (arc_compute_function_type): Consider new naked function type,
24161         change function return type.
24162         (arc_must_save_register): Adapt to handle new
24163         arc_compute_function_type's return type.
24164         (arc_expand_prologue): Likewise.
24165         (arc_expand_epilogue): Likewise.
24166         (arc_return_address_regs): Delete.
24167         (arc_return_address_register): New function.
24168         (arc_epilogue_uses): Use above function.
24169         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24170         (arc_function_type): Change encoding, add naked type.
24171         (ARC_INTERRUPT_P): Change to handle the new encoding.
24172         (ARC_FAST_INTERRUPT_P): Likewise.
24173         (ARC_NORMAL_P): Define.
24174         (ARC_NAKED_P): Likewise.
24175         (arc_compute_function_type): Delete prototype.
24176         * config/arc/arc.md (in_ret_delay_slot): Use
24177         arc_return_address_register function.
24178         (simple_return): Likewise.
24179         (p_return_i): Likewise.
24181 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24183         PR tree-optimization/81428
24184         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24185         can't be built for those types.
24187 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24189         Remove stuff dead since r239246.
24191         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24192         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24193         (avr_inform_devices): Remove dead stuff.
24195 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24197         * config/arm/arm_neon.h: Fix softp typo.
24199 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24201         PR tree-optimization/81365
24202         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24203         aggregate moves onto bb predecessor edges, make sure there are no
24204         loads that could alias the lhs in between the start of bb and the
24205         loads from *phi.
24207 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24209         PR 80929
24210         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24211         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24212         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24214 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24216         PR tree-optimization/81396
24217         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24218         (init_symbolic_number): Initialize it to 1.
24219         (perform_symbolic_merge): Add n_ops from both operands into the new
24220         n_ops.
24221         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24222         without base_addr as useless if they need more than one operation.
24223         (bswap_replace): Handle !bswap case for NULL base_addr.
24225 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24227         PR target/81069
24228         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24229         as possible.
24231 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24233         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24234         conditional builtin define __FIX_LEON3FT_B2BST.
24236 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24238         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24239         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24240         with -mfix-ut700.
24242 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24244         PR rtl-optimization/81424
24245         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24246         to remove potential trapping from operands if -fnon-call-exceptions.
24248 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24250         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24251         profile_proability for scalling.
24252         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24254 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24256         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24258 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24260         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24261         fixpoint arithmetics.
24263 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24265         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24266         fixpoint arithmetics.
24268 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24270         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24271         fixpoint arithmetics.
24273 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24275         * profile-count.h (profile_probability::from_reg_br_prob_note,
24276         profile_probability::to_reg_br_prob_note): New functions.
24277         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24278         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24279         * predict.c (probability_reliable_p): Update.
24280         (edge_probability_reliable_p): Update.
24281         (br_prob_note_reliable_p): Update.
24282         (invert_br_probabilities): Update.
24283         (add_reg_br_prob_note): New function.
24284         (combine_predictions_for_insn): Update.
24285         * asan.c (asan_clear_shadow): Update.
24286         * cfgbuild.c (compute_outgoing_frequencies): Update.
24287         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24288         (update_br_prob_note): Update.
24289         (rtl_verify_edges): Update.
24290         (purge_dead_edges): Update.
24291         (fixup_reorder_chain): Update.
24292         * emit-rtl.c (try_split): Update.
24293         * ifcvt.c (cond_exec_process_insns): Update.
24294         (cond_exec_process_if_block): Update.
24295         (dead_or_predicable): Update.
24296         * internal-fn.c (expand_addsub_overflow): Update.
24297         (expand_neg_overflow): Update.
24298         (expand_mul_overflow): Update.
24299         * loop-doloop.c (doloop_modify): Update.
24300         * loop-unroll.c (compare_and_jump_seq): Update.
24301         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24302         * predict.h: Update.
24303         * reorg.c (mostly_true_jump): Update.
24304         * rtl.h: Update.
24305         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24306         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24307         * config/arc/arc.c: (emit_unlikely_jump): Update.
24308         * config/arm/arm.c: (emit_unlikely_jump): Update.
24309         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24310         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24311         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24312         (ix86_print_operand): Update.
24313         (ix86_split_fp_branch): Update.
24314         (predict_jump): Update.
24315         * config/ia64/ia64.c (ia64_print_operand): Update.
24316         * config/mmix/mmix.c (mmix_print_operand): Update.
24317         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24318         (rs6000_expand_split_stack_prologue): Update.
24319         * config/rs6000/rs6000.c: Update.
24320         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24321         (s390_expand_vec_movstr): Update.
24322         (s390_expand_cs_tdsi): Update.
24323         (s390_expand_split_stack_prologue): Update.
24324         * config/sh/sh.c (sh_print_operand): Update.
24325         (expand_cbranchsi4): Update.
24326         (expand_cbranchdi4): Update.
24327         * config/sparc/sparc.c (output_v9branch): Update.
24328         * config/spu/spu.c (get_branch_target): Update.
24329         (ea_load_store_inline): Update.
24330         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24331         * config/tilepro/tilepro.c: Update.
24333 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24335         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24336         (gimplify_save_expr): Likewise.
24338 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24340         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24342 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24344         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24345         TV_IPA_FNSUMMARY.
24346         * timevar.def (TV_IPA_FNSUMMARY): Define.
24348 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24350         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24351         to back store errata sensitive sequence from being generated.
24352         (sqrtdf2_fix): Likewise.
24354 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24356         * tree-ssa-threadupdate.c (compute_path_counts,
24357         update_joiner_offpath_counts): Use profile_probability.
24359 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24361         Revert:
24362         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24364         * config/arm/arm-c.c (arm_cpu_builtins): Define
24365         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24367 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24369         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24370         array entries to represent __ieee128 versions of the
24371         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24372         scalar_extract_sig, and scalar_insert_exp built-in functions.
24373         (altivec_resolve_overloaded_builtin): Add special case handling
24374         for the __builtin_scalar_insert_exp function, as represented by
24375         the P9V_BUILTIN_VEC_VSIEDP constant.
24376         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24377         exponent support for __ieee128 argument.
24378         (VSESQP): Add scalar extract signature support for __ieee128
24379         argument.
24380         (VSTDCNQP): Add scalar test negative support for __ieee128
24381         argument.
24382         (VSIEQP): Add scalar insert exponent support for __int128 argument
24383         with __ieee128 result.
24384         (VSIEQPF): Add scalar insert exponent support for __ieee128
24385         argument with __ieee128 result.
24386         (VSTDCQP): Add scalar test data class support for __ieee128
24387         argument.
24388         (VSTDCNQP): Add overload support for scalar test negative with
24389         __ieee128 argument.
24390         (VSTDCQP): Add overload support for scalar test data class
24391         __ieee128 argument.
24392         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24393         UNSPEC_VSX_SXSIGDP.
24394         (UNSPEC_VSX_SIEXPQP): New constant.
24395         (xsxexpqp): New insn for VSX scalar extract exponent quad
24396         precision.
24397         (xsxsigqp): New insn for VSX scalar extract significand quad
24398         precision.
24399         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24400         precision with floating point argument.
24401         (xststdcqp): New expand for VSX scalar test data class quad
24402         precision.
24403         (xststdcnegqp): New expand for VSX scalar test negative quad
24404         precision.
24405         (xststdcqp): New insn to match expansions for VSX scalar test data
24406         class quad precision and VSX scalar test negative quad precision.
24407         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24408         special case operand checking to enforce that second operand of
24409         VSX scalar test data class with quad precision argument is a 7-bit
24410         unsigned literal.
24411         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24412         prototypes and descriptions of __ieee128 versions of
24413         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24414         scalar_test_data_class, and scalar_test_neg built-in functions.
24416 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24418         PR tree-optimization/81162
24419         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24420         replace a negate with an add.
24422 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24424         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24426 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24428         * config/arm/arm-c.c (arm_cpu_builtins): Define
24429         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24431 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24433         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24434         (armv8-r): Set ARM Cortex-R52 as default CPU.
24435         * config/arm/arm-tables.opt: Regenerate.
24436         * config/arm/arm-tune.md: Regenerate.
24437         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24438         Cortex-R52.
24439         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24440         extension for -mcpu=cortex-r52.
24442 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24444         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24445         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24446         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24447         (fp-armv8): Define it as FP_ARMv8 only.
24448         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24449         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24450         TARGET_FPU_ARMV8.
24451         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24452         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24453         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24454         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24455         than TARGET_FPU_ARMV8.
24456         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24457         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24458         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24459         TARGET_FPU_ARMV8.
24460         * config/arm/neon.md (neon_vrint): Likewise.
24461         (neon_vcvt): Likewise.
24462         (neon_<fmaxmin_op><mode>): Likewise.
24463         (<fmaxmin><mode>3): Likewise.
24464         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24465         * config/arm/predicates.md (arm_cond_move_operator): Check against
24466         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24468 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24470         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24471         to top of function.
24473 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24475         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24476         loop in comment with memset.
24478 2017-07-14  Martin Liska  <mliska@suse.cz>
24480         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24481         * dwarf2out.c (is_java): Remove the function.
24482         (output_pubname): Remove usage of the function.
24483         (lower_bound_default): Remove usage of DW_LANG_Java.
24484         (gen_compile_unit_die): Likewise.
24485         * gcc.c: Remove compiler defaults for .java and .zip files.
24486         * gimple-expr.c (remove_suffix): Change as there's no longer
24487         extension than 4-letter one.
24488         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24489         (gimplify_save_expr): Likewise.
24490         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24491         as it's possible even for other languages than Java.
24492         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24493         * lto-opts.c (lto_write_options): Remove reference to Java.
24494         * opts.c (strip_off_ending): Update file extension handling.
24495         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24496         * tree-eh.c (lower_resx): Likewise.
24497         * tree.c (free_lang_data_in_type): Remove dead code.
24498         (find_decls_types_r): Likewise.
24499         (build_common_builtin_nodes): Remove Java from a comment.
24500         (verify_type): Remove dead code.
24501         * varasm.c (assemble_external): Remove Java from a comment.
24503 2017-07-14  Martin Liska  <mliska@suse.cz>
24505         * opts.c (finish_options): Add quotes.
24506         (common_handle_option): Likewise.
24508 2017-07-14  Martin Liska  <mliska@suse.cz>
24510         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24511         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24512         Remove N_SO_PASCAL.
24513         * dwarf2out.c (lower_bound_default): Do not handle
24514         DW_LANG_Pascal83.
24515         (gen_compile_unit_die): Likewise.
24516         * gcc.c: Remove default extension binding for GNU Pascal.
24517         * stmt.c: Remove Pascal language from a comment.
24518         * xcoffout.c: Likewise.
24520 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24522         PR c/81405
24523         * diagnostic-show-locus.c (fixit_cmp): New function.
24524         (layout::layout): Sort m_fixit_hints.
24525         (column_range::column_range): Assert that the values are valid.
24526         (struct char_span): New struct.
24527         (correction::overwrite): New method.
24528         (struct source_line): New struct.
24529         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24530         calls in terms of classes source_line and char_span, and
24531         correction::overwrite.
24532         (selftest::test_overlapped_fixit_printing_2): New function.
24533         (selftest::diagnostic_show_locus_c_tests): Call it.
24535 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24537         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24538         early if there is no lhs.
24540 2017-07-13  Martin Liska  <mliska@suse.cz>
24542         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24543         (gen_reference_type_die): Likewise.
24544         * stor-layout.c: Remove Pascal-related comment.
24546 2017-07-13  Martin Liska  <mliska@suse.cz>
24548         * opts.c (finish_options): Add quotes to error messages.
24549         (parse_sanitizer_options): Likewise.
24551 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24553         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24555 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24557         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24559 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24561         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24562         during expansion.
24563         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24565 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24567         PR target/81193
24568         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24569         provides the hardware capability bits, define the macro
24570         __BUILTIN_CPU_SUPPORTS__.
24571         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24572         if GLIBC does not provide the hardware capability bits.  Add a
24573         gcc_unreachable call if the built-in cpu function is neither
24574         __builtin_cpu_is nor __builtin_cpu_supports.
24575         (rs6000_get_function_versions_dispatcher): Change the warning
24576         that an old GLIBC is used which does not export the capability
24577         bits to be an error.
24578         * doc/extend.texi (target_clones attribute): Document the
24579         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24580         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24581         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24582         the macros defined by GCC if the newer GLIBC is available.
24584 2017-07-12  Jeff Law  <law@redhat.com>
24586         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24587         remaining includes slightly.
24588         * config/riscv/riscv-builtins.c: Include profile-count.h.
24590 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24592         PR target/79883
24593         * config/avr/avr.c (avr_set_current_function): In diagnostic
24594         messages: Quote keywords and (parts of) identifiers.
24595         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24596         "INTERUPT".
24598 2017-07-12  Carl Love  <cel@us.ibm.com>
24600         * config/rs6000/rs6000-c.c: Add support for built-in functions
24601         vector bool char vec_revb (vector bool char);
24602         vector bool short vec_revb (vector short char);
24603         vector bool int vec_revb (vector bool int);
24604         vector bool long long vec_revb (vector bool long long);
24605         * doc/extend.texi: Update the built-in documentation file for the
24606         new built-in functions.
24608 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24610         * config/s390/s390.md: Remove movcc splitter.
24612 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24614         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24615         load/store on condition.
24617 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24619         PR target/81407
24620         * config/avr/avr.c (avr_encode_section_info)
24621         [progmem && !TREE_READONLY]: Error if progmem object needs
24622         constructing.
24624 2017-07-11  Michael Collison  <michael.collison@arm.com>
24626         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24627         New pattern.
24629 2017-07-11  Carl Love  <cel@us.ibm.com>
24631         * config/rs6000/rs6000-c.c: Add support for builtins
24632         vector unsigned int vec_parity_lsbb (vector signed int);
24633         vector unsigned int vec_parity_lsbb (vector unsigned int);
24634         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24635         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24636         vector unsigned long long vec_parity_lsbb (vector signed long long);
24637         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24638         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24639         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24640         * doc/extend.texi: Update the built-in documentation file for the
24641         new built-in functions.
24643 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24645         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24646         (layout::m_primary_loc): New field.
24647         (layout::layout): Initialize new field.  Move location filtering
24648         logic from here to...
24649         (layout::maybe_add_location_range): ...this new method.  Add
24650         support for filtering to just the lines already specified by other
24651         locations.
24652         (layout::will_show_line_p): New method.
24653         (gcc_rich_location::add_location_if_nearby): New method.
24654         (selftest::test_add_location_if_nearby): New test function.
24655         (selftest::diagnostic_show_locus_c_tests): Call it.
24656         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24657         New method.
24659 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24661         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24662         (bb_first_real_insn): New function.
24663         (nvptx_single): Add extra initialization of broadcasted condition
24664         variables.
24666 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24668         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24670 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24672         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24673         Binutils doc as TEXI will mess them up.
24674         * doc/invoke.texi (AVR Options): Same here.
24676 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24678         * config/sparc/sparc.opt (mfix-ut700): New option.
24679         (mfix-gr712rc): Likewise.
24680         (sparc_fix_b2bst): New variable.
24681         * doc/invoke.texi (SPARC options): Document them.
24682         (ARM options): Fix warnings.
24683         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24684         instructions to prevent sequences that can trigger the store-store
24685         errata for certain LEON3FT processors.
24686         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24687         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24688         * config/sparc/sparc.md (fix_b2bst): New attribute.
24689         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24691 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24693         PR target/81375
24694         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24695         (rcpps): Ditto.
24696         (*rsqrtsf2_sse): Ditto.
24697         (rsqrtsf2): Ditto.
24698         (div<mode>3): Macroize insn from divdf3 and divsf3
24699         using MODEF mode iterator.
24701 2017-07-10  Martin Sebor  <msebor@redhat.com>
24703         PR tree-optimization/80397
24704         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24705         instead of testing for equality to INTEGER_TYPE.
24707 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24709         * config.gcc: Remove uclibc from arc target spec.
24711 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24713         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24715 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24717         PR lto/80838
24718         * lto-wrapper.c (remove_option): New function.
24719         (merge_and_complain): Merge PIC/PIE options more realistically.
24721 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24723         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24725         PR target/20296
24726         PR target/81268
24727         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24728         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24729         * config.in: Regenerate.
24730         * configure: Regenerate.
24731         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24732         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24733         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24734         (TARGET_GASISR_PROLOGUES): ...target mask.
24735         * common/config/avr/avr-common.c
24736         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24737         Set -mgas-isr-prologues.
24738         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24739         INSERT_PASS_BEFORE for it.
24740         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24741         * config/avr/avr.c (avr_option_override)
24742         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24743         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24744         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24745         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24746         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24747         and rtl_opt_pass.
24748         (make_avr_pass_pre_proep): New function.
24749         (emit_push_sfr) <treg>: Add argument to function and use it
24750         instead of TMP_REG.
24751         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24752         and set machine->gasisr.yes.
24753         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24754         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24755         __gcc_isr.n_pushed to .L__stack_usage.
24756         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24757         (avr_asm_final_postscan_insn): ...this new static function.
24758         * config/avr/avr.h (machine_function)
24759         <is_no_gccisr, use_L__stack_usage>: New fields.
24760         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24761         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24762         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24763         (gasisr, *gasisr): New expander and insn.
24764         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24765         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24766         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24768 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24770         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24771         in quoted strings.
24773 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24775         Move jump-tables out of .text again.
24777         PR target/81075
24778         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24779         (ASM_OUTPUT_ADDR_VEC): New function.
24780         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24781         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24782         INSN_ADDRESSes as asm comment.
24783         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24784         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24785         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24786         * config/avr/avr.md (*tablejump): Adjust comment.
24787         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24788         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24789         New detail.
24790         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24791         (avr_output_addr_vec): New proto.
24792         (avr_log_t) <insn_addresses>: New field.
24794 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24796         PR target/81313
24797         * config/i386/i386.c (ix86_function_arg_advance): Set
24798         outgoing_args_on_stack to true if there are outgoing arguments
24799         on stack.
24800         (ix86_function_arg): Likewise.
24801         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24802         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24803         * config/i386/i386.h (machine_function): Add
24804         outgoing_args_on_stack.
24806 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24808         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24809         supporting pthreds.
24810         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24812 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24814         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24815         (REAL_H): Remove $(MACHMODE_H).
24816         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24817         double-int.h.
24818         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24819         $(MACHMODE_H) and double-int.h.
24820         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24821         $(MACHMODE_H).
24822         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24823         double-int.h.
24825 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24827         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24828         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24830 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24832         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24833         Add warning if GCC was not configured to link against a GLIBC that
24834         exports the hardware capability bits.
24835         (make_resolver_func): Make resolver function private and not a
24836         COMDAT function.  Create the name with clone_function_name instead
24837         of make_unique_name.
24839         PR target/81348
24840         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24841         correct operand in doing the split.
24843 2017-07-07 Carl Love  <cel@us.ibm.com>
24845         * config/rs6000/rs6000-c: Add support for built-in function
24846         vector unsigned short vec_pack_to_short_fp32 (vector float,
24847                                                       vector float).
24848         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24849         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24850         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24851         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24852         (convert_4f32_8i16): Add define_expand.
24853         * doc/extend.texi: Update the built-in documentation file for the
24854         new built-in function.
24856 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24858         * config/sparc/m8.md: New file.
24859         * config/sparc/sparc.md: Include m8.md.
24861 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24863         * config/sparc/sparc.opt: New option -mvis4b.
24864         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
24865         (sparc_option_override): Handle VIS4B.
24866         (enum sparc_builtins): Define
24867         SPARC_BUILTIN_DICTUNPACK{8,16,32},
24868         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
24869         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
24870         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
24871         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
24872         (check_constant_argument): New function.
24873         (sparc_vis_init_builtins): Define builtins
24874         __builtin_vis_dictunpack{8,16,32},
24875         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
24876         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
24877         __builtin_vis_fpcmpde{8,16,32}shl and
24878         __builtin_vis_fpcmpur{8,16,32}shl.
24879         (sparc_expand_builtin): Check that the constant operands to
24880         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
24881         constant and in range.
24882         * config/sparc/sparc-c.c (sparc_target_macros): Handle
24883         TARGET_VIS4B.
24884         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
24885         (SPARC_IMM5_P): Likewise.
24886         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
24887         (enabled): Handle vis4b.
24888         (UNSPEC_DICTUNPACK): New unspec.
24889         (UNSPEC_FPCMPSHL): Likewise.
24890         (UNSPEC_FPUCMPSHL): Likewise.
24891         (UNSPEC_FPCMPDESHL): Likewise.
24892         (UNSPEC_FPCMPURSHL): Likewise.
24893         (cpu_feature): New CPU feature `vis4b'.
24894         (dictunpack{8,16,32}): New insns.
24895         (FPCSMODE): New mode iterator.
24896         (fpcscond): New code iterator.
24897         (fpcsucond): Likewise.
24898         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
24899         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
24900         (fpcmpde{8,16,32}{si,di}shl): Likewise.
24901         (fpcmpur{8,16,32}{si,di}shl): Likewise.
24902         * config/sparc/constraints.md: Define constraints `q' for unsigned
24903         2-bit integer constants and `t' for unsigned 5-bit integer
24904         constants.
24905         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
24906         predicate.
24907         (imm5_operand_dictunpack16): Likewise.
24908         (imm5_operand_dictunpack32): Likewise.
24909         (imm2_operand): Likewise.
24910         * doc/invoke.texi (SPARC Options): Document -mvis4b.
24911         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
24912         ditunpack* and fpcmp*shl builtins.
24914 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24916         * config.gcc: Handle m8 in --with-{cpu,tune} options.
24917         * config.in: Add HAVE_AS_SPARC6 define.
24918         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
24919         M8.
24920         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
24921         TARGET_CPU_m8.
24922         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
24923         (CPP_CPU_SPEC): Handle m8.
24924         (ASM_CPU_SPEC): Likewise.
24925         * config/sparc/sparc-opts.h (enum processor_type): Add
24926         PROCESSOR_M8.
24927         * config/sparc/sparc.c (m8_costs): New struct.
24928         (sparc_option_override): Handle TARGET_CPU_m8.
24929         (sparc32_initialize_trampoline): Likewise.
24930         (sparc64_initialize_trampoline): Likewise.
24931         (sparc_issue_rate): Likewise.
24932         (sparc_register_move_cost): Likewise.
24933         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
24934         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
24935         (ASM_CPU64_DEFAULT_SPEC): Likewise.
24936         (CPP_CPU_SPEC): Handle M8.
24937         (ASM_CPU_SPEC): Likewise.
24938         (AS_M8_FLAG): Define.
24939         * config/sparc/sparc.md: Add m8 to the cpu attribute.
24940         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
24941         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
24942         M8 instructions.
24943         * configure: Regenerate.
24944         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
24945         -mtune=m8.
24947 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24949         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
24950         subtypes.
24951         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
24952         ("*movdi_insn_sp32"): Do not set v3pipe.
24953         ("*movsi_insn"): Likewise.
24954         ("*movdi_insn_sp64"): Likewise.
24955         ("*movsf_insn"): Likewise.
24956         ("*movdf_insn_sp32"): Likewise.
24957         ("*movdf_insn_sp64"): Likewise.
24958         ("*zero_extendsidi2_insn_sp64"): Likewise.
24959         ("*sign_extendsidi2_insn"): Likewise.
24960         ("*mov<VM32:mode>_insn"): Likewise.
24961         ("*mov<VM64:mode>_insn_sp64"): Likewise.
24962         ("*mov<VM64:mode>_insn_sp32"): Likewise.
24963         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
24964         ("<vlop:code><VL:mode>3"): Likewise.
24965         ("*not_<vlop:code><VL:mode>3"): Likewise.
24966         ("*nand<VL:mode>_vis"): Likewise.
24967         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
24968         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
24969         ("one_cmpl<VL:mode>2"): Likewise.
24970         ("faligndata<VM64:mode>_vis"): Likewise.
24971         ("alignaddrsi_vis"): Likewise.
24972         ("alignaddrdi_vis"): Likweise.
24973         ("alignaddrlsi_vis"): Likewise.
24974         ("alignaddrldi_vis"): Likewise.
24975         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
24976         ("bmaskdi_vis"): Likewise.
24977         ("bmasksi_vis"): Likewise.
24978         ("bshuffle<VM64:mode>_vis"): Likewise.
24979         ("cmask8<P:mode>_vis"): Likewise.
24980         ("cmask16<P:mode>_vis"): Likewise.
24981         ("cmask32<P:mode>_vis"): Likewise.
24982         ("pdistn<P:mode>_vis"): Likewise.
24983         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
24985 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24987         * config/sparc/sparc.md ("subtype"): New insn attribute.
24988         ("*wrgsr_sp64"): Set insn subtype.
24989         ("*rdgsr_sp64"): Likewise.
24990         ("alignaddrsi_vis"): Likewise.
24991         ("alignaddrdi_vis"): Likewise.
24992         ("alignaddrlsi_vis"): Likewise.
24993         ("alignaddrldi_vis"): Likewise.
24994         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
24995         ("fexpand_vis"): Likewise.
24996         ("fpmerge_vis"): Likewise.
24997         ("faligndata<VM64:mode>_vis"): Likewise.
24998         ("bshuffle<VM64:mode>_vis"): Likewise.
24999         ("cmask8<P:mode>_vis"): Likewise.
25000         ("cmask16<P:mode>_vis"): Likewise.
25001         ("cmask32<P:mode>_vis"): Likewise.
25002         ("fchksm16_vis"): Likewise.
25003         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25004         ("fmean16_vis"): Likewise.
25005         ("fp<plusminus_insn>64_vis"): Likewise.
25006         ("<plusminus_insn>v8qi3"): Likewise.
25007         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25008         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25009         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25010         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25011         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25012         ("*movqi_insn"): Likewise.
25013         ("*movhi_insn"): Likewise.
25014         ("*movsi_insn"): Likewise.
25015         ("movsi_pic_gotdata_op"): Likewise.
25016         ("*movdi_insn_sp32"): Likewise.
25017         ("*movdi_insn_sp64"): Likewise.
25018         ("movdi_pic_gotdata_op"): Likewise.
25019         ("*movsf_insn"): Likewise.
25020         ("*movdf_insn_sp32"): Likewise.
25021         ("*movdf_insn_sp64"): Likewise.
25022         ("*zero_extendhisi2_insn"): Likewise.
25023         ("*zero_extendqihi2_insn"): Likewise.
25024         ("*zero_extendqisi2_insn"): Likewise.
25025         ("*zero_extendqidi2_insn"): Likewise.
25026         ("*zero_extendhidi2_insn"): Likewise.
25027         ("*zero_extendsidi2_insn_sp64"): Likewise.
25028         ("ldfsr"): Likewise.
25029         ("prefetch_64"): Likewise.
25030         ("prefetch_32"): Likewise.
25031         ("tie_ld32"): Likewise.
25032         ("tie_ld64"): Likewise.
25033         ("*tldo_ldub_sp32"): Likewise.
25034         ("*tldo_ldub1_sp32"): Likewise.
25035         ("*tldo_ldub2_sp32"): Likewise.
25036         ("*tldo_ldub_sp64"): Likewise.
25037         ("*tldo_ldub1_sp64"): Likewise.
25038         ("*tldo_ldub2_sp64"): Likewise.
25039         ("*tldo_ldub3_sp64"): Likewise.
25040         ("*tldo_lduh_sp32"): Likewise.
25041         ("*tldo_lduh1_sp32"): Likewise.
25042         ("*tldo_lduh_sp64"): Likewise.
25043         ("*tldo_lduh1_sp64"): Likewise.
25044         ("*tldo_lduh2_sp64"): Likewise.
25045         ("*tldo_lduw_sp32"): Likewise.
25046         ("*tldo_lduw_sp64"): Likewise.
25047         ("*tldo_lduw1_sp64"): Likewise.
25048         ("*tldo_ldx_sp64"): Likewise.
25049         ("*mov<VM32:mode>_insn"): Likewise.
25050         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25051         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25053 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25055         * config/sparc/sparc.md ("type"): New insn type viscmp.
25056         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25057         viscmp.
25058         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25059         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25060         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25061         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25062         viscmp.
25063         ("n7_vis_logical_11cycle"): Likewise.
25064         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25065         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25066         * config/sparc/niagara.md ("niag_vis"): Likewise.
25067         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25068         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25070 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25072         * config/sparc/sparc.md: New instruction type `bmask'.
25073         (bmaskdi_vis): Use the `bmask' type.
25074         (bmasksi_vis): Likewise.
25075         * config/sparc/ultra3.md (us3_array): Likewise.
25076         * config/sparc/niagara7.md (n7_array): Likewise.
25077         * config/sparc/niagara4.md (n4_array): Likewise.
25078         * config/sparc/niagara2.md (niag2_vis): Likewise.
25079         (niag3_vis): Likewise.
25080         * config/sparc/niagara.md (niag_vis): Likewise.
25082 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25084         * ipa-comdats.c: Remove optimize check from gate.
25085         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25086         for functions not optimized.
25087         (ipa_fn_summary_read): Skip optimize check.
25088         (ipa_fn_summary_write): Likewise.
25089         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25090         is optimized.
25091         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25092         uninlinable.
25093         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25094         (check_callers): Give up on caller which is not optimized.
25095         (inline_small_functions): Likewise.
25096         (ipa_inline): Do not give up when not optimizing.
25097         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25098         away unoptimizes cdtors.
25099         (whole_program_function_and_variable_visibility): Do
25100         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25101         * ipa.c (process_references): Do not check optimize.
25102         (symbol_table::remove_unreachable_nodes): Update optimize check.
25103         (set_writeonly_bit): Update optimize check.
25104         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25105         (pass_ipa_single_use::gate): Remove.
25107 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25109         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25110         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25111         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25112         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25113         permute_load, permute_store, adjust_extract, adjust_splat,
25114         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25115         replace_swap_with_copy, dump_swap_insn_table,
25116         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25117         recombine_lvx_pattern, recombine_stvx_pattern,
25118         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25119         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25120         to file rs6000-p8swap.c.
25121         * config/rs6000/rs6000-p8swap.c: New file.
25122         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25123         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25124         and rs6000*-*-* targets.
25126 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25128         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25130 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25132         * lto-wrapper.c (merge_and_complain): Do not merge
25133         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25134         fsigned_zeros, ftrapping_math, fwrapv.
25135         (append_compiler_options): Do not track these options.
25136         (append_linker_options): Likewie
25138 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25140         * cgraphunit.c (cgraph_node::finalize_function): When
25141         !flag_toplevel_reorde set no_reorder flag.
25142         (varpool_node::finalize_decl): Likewise.
25143         (symbol_table::compile): Drop no toplevel reorder path.
25145 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25147         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25148         edges; zero probability is not better than uninitialized.
25150 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25152         * asan.h (asan_sanitize_allocas_p): Declare.
25153         * asan.c (asan_sanitize_allocas_p): New function.
25154         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25155         (handle_builtin_alloca): Likewise.
25156         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25157         if !asan_sanitize_allocas_p.
25158         * params.def (asan-instrument-allocas): Add new option.
25159         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25160         * opts.c (common_handle_option): Disable allocas sanitization for
25161         KASan by default.
25163 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25165         * asan.c: Include gimple-fold.h.
25166         (get_last_alloca_addr): New function.
25167         (handle_builtin_stackrestore): Likewise.
25168         (handle_builtin_alloca): Likewise.
25169         (asan_emit_allocas_unpoison): Likewise.
25170         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25171         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25172         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25173         (instrument_builtin_call): Pass gimple iterator to
25174         get_mem_refs_of_builtin_call.
25175         (last_alloca_addr): New global.
25176         * asan.h (asan_emit_allocas_unpoison): Declare.
25177         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25178         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25179         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25180         if function calls alloca.
25181         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25182         * gimple-fold.h (replace_call_with_value): Declare.
25183         * internal-fn.c: Include asan.h.
25184         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25185         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25187 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25189         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25190         (C_SELFTEST_FLAGS): New.
25191         (CPP_SELFTEST_FLAGS): New.
25192         (SELFTEST_DEPS): New, from deps of s-selftest.
25193         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25194         (CPP_SELFTEST_DEPS): New.
25195         (selftest): Add dependency on s-selftest-c++.
25196         (s-selftest): Rename to...
25197         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25198         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25199         than SELFTEST_FLAGS.
25200         (selftest-gdb): Rename to...
25201         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25202         C_SELFTEST_FLAGS.
25203         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25204         (selftest-valgrind): Rename to...
25205         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25206         C_SELFTEST_FLAGS.
25207         (selftest-valgrind): Reintroduce as an alias for
25208         selftest-c-valgrind.
25209         (s-selftest-c++): New.
25210         (selftest-c++-gdb): New.
25211         (selftest-c++-valgrind): New.
25213 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25215         * gcc.c (process_command): When deciding if undefined variables
25216         should be ignored when processing specs, accept "gcc -v" as well.
25218 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25220         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25221         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25223 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25225         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25226         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25227         * config/arm/arm-tables.opt: Regenerate.
25228         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25229         enumerator.
25230         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25232 2017-07-06  Carl Love  <cel@us.ibm.com>
25234         * ChangeLog: Clean up from mid air collision
25236 2017-07-06  Carl Love  <cel@us.ibm.com>
25238         * config/rs6000/rs6000-c.c: Add support for built-in functions
25239         vector signed int vec_subc (vector signed int, vector signed int);
25240         vector signed __int128 vec_subc (vector signed __int128,
25241                                          vector signed __int128);
25242         vector unsigned __int128 vec_subc (vector unsigned __int128,
25243                                            vector unsigned __int128);
25244         vector signed int vec_sube (vector signed int, vector signed int,
25245                                     vector signed int);
25246         vector unsigned int vec_sube (vector unsigned int,
25247                                       vector unsigned int,
25248                                       vector unsigned int);
25249         vector signed __int128 vec_sube (vector signed __int128,
25250                                          vector signed __int128,
25251                                          vector signed__int128);
25252         vector unsigned __int128 vec_sube (vector unsigned __int128,
25253                                            vector unsigned __int128,
25254                                            vector unsigned __int128);
25255         vector signed int vec_subec (vector signed int, vector signed int,
25256                                      vector signed int);
25257         vector unsigned int vec_subec (vector unsigned int,
25258                                        vector unsigned int,
25259                                        vector unsigned int);
25260         vector signed __int128 vec_subec (vector signed __int128,
25261                                           vector signed __int128,
25262                                           vector signed__int128);
25263         vector unsigned __int128 vec_subec (vector unsigned __int128,
25264                                             vector unsigned __int128,
25265                                             vector unsigned __int128);
25266         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25267         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25268         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25269         BU_ALTIVEC_OVERLOAD_X definitions.
25270         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25271         * doc/extend.texi: Update the built-in documentation file for the new
25272         built-in functions.
25274 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25276         PR c++/79300
25277         * diagnostic-show-locus.c (layout::layout): Use start and finish
25278         spelling location for the start and finish of each range.
25279         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25280         Add unused aspect param.
25281         * input.c (expand_location_1): Add "aspect" param, and use it
25282         to access the correct part of the location.
25283         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25284         expand_location_1.
25285         (expand_location_to_spelling_point): Likewise.
25286         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25287         param, and pass it to expand_location_1.
25289 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25291         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25292         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25293         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25294         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25295         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25296         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25297         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25298         _mm_maskz_getmant_ss): New intrinsics.
25299         (__builtin_ia32_getexpss128_mask): Changed to ...
25300         __builtin_ia32_getexpss128_round ... this.
25301         (__builtin_ia32_getexpsd128_mask): Changed to ...
25302         __builtin_ia32_getexpsd128_round ... this.
25303         * config/i386/i386-builtin-types.def
25304         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25305         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25306         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25307         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25308         __builtin_ia32_getmantss_mask_round): New builtins.
25309         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25310         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25311         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25312         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25313         * config/i386/sse.md
25314         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25315         avx512f_sgetexp<mode><mask_scalar_name>
25316         <round_saeonly_scalar_name> ... this.
25317         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25318         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25319         vgetexp<ssescalarmodesuffix>
25320         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25321         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25322         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25323         avx512f_vgetmant<mode><mask_scalar_name>
25324         <round_saeonly_scalar_name> ... this.
25325         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25326         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25327         vgetmant<ssescalarmodesuffix>
25328         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25329         %0<mask_scalar_operand4>, %1, %2
25330         <round_saeonly_scalar_mask_op4>, %3} ... this.
25331         * config/i386/subst.md (mask_scalar_operand4,
25332         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25333         round_saeonly_scalar_nimm_predicate): New subst attributes.
25335 2017-07-06  Julia Koval  <julia.koval@intel.com>
25337         * config/i386/i386.c (ix86_erase_embedded_rounding):
25338         Remove code for old rounding pattern.
25340 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25342         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25344 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25346         * doc/sourcebuild.texi (Test Directives, Variants of
25347         dg-require-support): Add documentation for dg-require-stack-check.
25349 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25351         * config/i386/subst.md (mask_scalar, round_scalar,
25352         round_saeonly_scalar): New meta-templates.
25353         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25354         round_scalar_mask_operand3, round_scalar_mask_op3,
25355         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25356         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25357         round_saeonly_scalar_constraint,
25358         round_saeonly_scalar_prefix): New subst attribute.
25359         * config/i386/sse.md
25360         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25361         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25362         <round_scalar_name> ... this.
25363         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25364         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25365         <round_scalar_name> ... this.
25366         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25367         <sse>_vm<code><mode>3<mask_scalar_name>
25368         <round_saeonly_scalar_name> ... this.
25369         (v<plusminus_mnemonic><ssescalarmodesuffix>
25370         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25371         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25372         v<plusminus_mnemonic><ssescalarmodesuffix>
25373         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25374         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25375         (v<multdiv_mnemonic><ssescalarmodesuffix>
25376         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25377         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25378         v<multdiv_mnemonic><ssescalarmodesuffix>
25379         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25380         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25381         (v<maxmin_float><ssescalarmodesuffix>
25382         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25383         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25384         v<maxmin_float><ssescalarmodesuffix>
25385         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25386         %0<mask_scalar_operand3>, %1, %<iptr>2
25387         <round_saeonly_scalar_mask_op3>} ... this.
25389 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25391         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25392         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25394 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25395             Alan Hayward  <alan.hayward@arm.com>
25396             David Sherwood  <david.sherwood@arm.com>
25398         * combine.c (simplify_if_then_else): Remove "enum" before
25399         "machine_mode".
25400         * compare-elim.c (can_eliminate_compare): Likewise.
25401         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25402         Likewise.
25403         (aarch64_lookup_simd_builtin_type): Likewise.
25404         (aarch64_simd_builtin_type): Likewise.
25405         (aarch64_init_simd_builtin_types): Likewise.
25406         (aarch64_simd_expand_args): Likewise.
25407         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25408         Likewise.
25409         (aarch64_reverse_mask): Likewise.
25410         (aarch64_simd_emit_reg_reg_move): Likewise.
25411         (aarch64_gen_adjusted_ldpstp): Likewise.
25412         (aarch64_ccmp_mode_to_code): Likewise.
25413         (aarch64_operands_ok_for_ldpstp): Likewise.
25414         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25415         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25416         Likewise.
25417         (aarch64_min_divisions_for_recip_mul): Likewise.
25418         (aarch64_reassociation_width): Likewise.
25419         (aarch64_get_condition_code_1): Likewise.
25420         (aarch64_simd_emit_reg_reg_move): Likewise.
25421         (aarch64_simd_attr_length_rglist): Likewise.
25422         (aarch64_reverse_mask): Likewise.
25423         (aarch64_operands_ok_for_ldpstp): Likewise.
25424         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25425         (aarch64_gen_adjusted_ldpstp): Likewise.
25426         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25427         Likewise.
25428         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25429         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25430         (arm_lookup_simd_builtin_type): Likewise.
25431         (arm_simd_builtin_type): Likewise.
25432         (arm_init_simd_builtin_types): Likewise.
25433         (arm_expand_builtin_args): Likewise.
25434         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25435         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25436         (ft32_setup_incoming_varargs): Likewise.
25437         (ft32_function_arg): Likewise.
25438         (ft32_function_arg_advance): Likewise.
25439         (ft32_pass_by_reference): Likewise.
25440         (ft32_arg_partial_bytes): Likewise.
25441         (ft32_valid_pointer_mode): Likewise.
25442         (ft32_addr_space_pointer_mode): Likewise.
25443         (ft32_addr_space_legitimate_address_p): Likewise.
25444         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25445         Likewise.
25446         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25447         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25448         (iamcu_alignment): Likewise.
25449         (canonicalize_vector_int_perm): Likewise.
25450         (ix86_noce_conversion_profitable_p): Likewise.
25451         (ix86_mpx_bound_mode): Likewise.
25452         (ix86_operands_ok_for_move_multiple): Likewise.
25453         * config/microblaze/microblaze-protos.h
25454         (microblaze_expand_conditional_branch_reg): Likewise.
25455         * config/microblaze/microblaze.c
25456         (microblaze_expand_conditional_branch_reg): Likewise.
25457         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25458         Likewise.
25459         (rs6000_reassociation_width): Likewise.
25460         (rs6000_invalid_binary_op): Likewise.
25461         (fusion_p9_p): Likewise.
25462         (emit_fusion_p9_load): Likewise.
25463         (emit_fusion_p9_store): Likewise.
25464         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25465         Likewise.
25466         (riscv_hard_regno_mode_ok_p): Likewise.
25467         (riscv_address_insns): Likewise.
25468         (riscv_split_symbol): Likewise.
25469         (riscv_legitimize_move): Likewise.
25470         (riscv_function_value): Likewise.
25471         (riscv_hard_regno_nregs): Likewise.
25472         (riscv_expand_builtin): Likewise.
25473         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25474         (riscv_build_integer): Likewise.
25475         (riscv_split_integer): Likewise.
25476         (riscv_legitimate_constant_p): Likewise.
25477         (riscv_cannot_force_const_mem): Likewise.
25478         (riscv_regno_mode_ok_for_base_p): Likewise.
25479         (riscv_valid_base_register_p): Likewise.
25480         (riscv_valid_offset_p): Likewise.
25481         (riscv_valid_lo_sum_p): Likewise.
25482         (riscv_classify_address): Likewise.
25483         (riscv_legitimate_address_p): Likewise.
25484         (riscv_address_insns): Likewise.
25485         (riscv_load_store_insns): Likewise.
25486         (riscv_force_binary): Likewise.
25487         (riscv_split_symbol): Likewise.
25488         (riscv_force_address): Likewise.
25489         (riscv_legitimize_address): Likewise.
25490         (riscv_move_integer): Likewise.
25491         (riscv_legitimize_const_move): Likewise.
25492         (riscv_legitimize_move): Likewise.
25493         (riscv_address_cost): Likewise.
25494         (riscv_subword): Likewise.
25495         (riscv_output_move): Likewise.
25496         (riscv_canonicalize_int_order_test): Likewise.
25497         (riscv_emit_int_order_test): Likewise.
25498         (riscv_function_arg_boundary): Likewise.
25499         (riscv_pass_mode_in_fpr_p): Likewise.
25500         (riscv_pass_fpr_single): Likewise.
25501         (riscv_pass_fpr_pair): Likewise.
25502         (riscv_get_arg_info): Likewise.
25503         (riscv_function_arg): Likewise.
25504         (riscv_function_arg_advance): Likewise.
25505         (riscv_arg_partial_bytes): Likewise.
25506         (riscv_function_value): Likewise.
25507         (riscv_pass_by_reference): Likewise.
25508         (riscv_setup_incoming_varargs): Likewise.
25509         (riscv_print_operand): Likewise.
25510         (riscv_elf_select_rtx_section): Likewise.
25511         (riscv_save_restore_reg): Likewise.
25512         (riscv_for_each_saved_reg): Likewise.
25513         (riscv_register_move_cost): Likewise.
25514         (riscv_hard_regno_mode_ok_p): Likewise.
25515         (riscv_hard_regno_nregs): Likewise.
25516         (riscv_class_max_nregs): Likewise.
25517         (riscv_memory_move_cost): Likewise.
25518         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25519         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25520         (rl78_addr_space_address_mode): Likewise.
25521         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25522         Likewise.
25523         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25524         (rs6000_reassociation_width): Likewise.
25525         (rs6000_invalid_binary_op): Likewise.
25526         (fusion_p9_p): Likewise.
25527         (emit_fusion_p9_load): Likewise.
25528         (emit_fusion_p9_store): Likewise.
25529         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25530         (ok_for_simple_move_operands): Likewise.
25531         (ok_for_simple_move_strict_operands): Likewise.
25532         (ok_for_simple_arith_logic_operands): Likewise.
25533         (visium_legitimize_reload_address): Likewise.
25534         (visium_select_cc_mode): Likewise.
25535         (output_cbranch): Likewise.
25536         (visium_split_double_move): Likewise.
25537         (visium_expand_copysign): Likewise.
25538         (visium_expand_int_cstore): Likewise.
25539         (visium_expand_fp_cstore): Likewise.
25540         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25541         (visium_function_arg): Likewise.
25542         (visium_function_arg_advance): Likewise.
25543         (visium_libcall_value): Likewise.
25544         (visium_setup_incoming_varargs): Likewise.
25545         (visium_legitimate_constant_p): Likewise.
25546         (visium_legitimate_address_p): Likewise.
25547         (visium_legitimize_address): Likewise.
25548         (visium_secondary_reload): Likewise.
25549         (visium_register_move_cost): Likewise.
25550         (visium_memory_move_cost): Likewise.
25551         (prepare_move_operands): Likewise.
25552         (ok_for_simple_move_operands): Likewise.
25553         (ok_for_simple_move_strict_operands): Likewise.
25554         (ok_for_simple_arith_logic_operands): Likewise.
25555         (visium_function_value_1): Likewise.
25556         (rtx_ok_for_offset_p): Likewise.
25557         (visium_legitimize_reload_address): Likewise.
25558         (visium_split_double_move): Likewise.
25559         (visium_expand_copysign): Likewise.
25560         (visium_expand_int_cstore): Likewise.
25561         (visium_expand_fp_cstore): Likewise.
25562         (visium_split_cstore): Likewise.
25563         (visium_select_cc_mode): Likewise.
25564         (visium_split_cbranch): Likewise.
25565         (output_cbranch): Likewise.
25566         (visium_print_operand_address): Likewise.
25567         * expmed.c (flip_storage_order): Likewise.
25568         * expmed.h (emit_cstore): Likewise.
25569         (flip_storage_order): Likewise.
25570         * genrecog.c (validate_pattern): Likewise.
25571         * hsa-gen.c (gen_hsa_addr): Likewise.
25572         * internal-fn.c (expand_arith_overflow): Likewise.
25573         * ira-color.c (allocno_copy_cost_saving): Likewise.
25574         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25575         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25576         (process_invariant_for_inheritance): Likewise.
25577         * lra-eliminations.c (move_plus_up): Likewise.
25578         * omp-low.c (lower_oacc_reductions): Likewise.
25579         * simplify-rtx.c (simplify_subreg): Likewise.
25580         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25581         (TARGET_CHKP_BOUND_MODE): Likewise..
25582         * targhooks.c (default_chkp_bound_mode): Likewise.
25583         (default_setup_incoming_vararg_bounds): Likewise.
25584         * targhooks.h (default_chkp_bound_mode): Likewise.
25585         (default_setup_incoming_vararg_bounds): Likewise.
25586         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25587         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25588         (have_whole_vector_shift): Likewise.
25589         * tree-vect-stmts.c (vectorizable_load): Likewise.
25590         * doc/tm.texi: Regenerate.
25592 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25594         Graceful degrade if Binutils PR21472 is not available.
25596         PR target/81072
25597         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25598         .rodata in flash test fails.
25599         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25600         * confgure: Regenerate.
25601         * config.in: Regenerate.
25602         * config/avr/avr.c (avr_asm_named_section)
25603         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25604         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25605         (avr_asm_init_sections): Same.
25607 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25609         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25610         (fma<VH:mode>4_intrinsic): Likewise.
25611         (*fmsub<VCVTF:mode>4): Likewise.
25612         (*fmsub<VH:mode>4_intrinsic): Likewise.
25614 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25616         PR target/81305
25617         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25618         Don't depend on "optimize > 0".
25619         (out_movhi_r_mr, out_movqi_mr_r): Same.
25620         (out_movhi_mr_r, out_movqi_r_mr): Same.
25621         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25622         io_address_operand on "optimize > 0".
25624 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25626         * tree-loop-distribution.c: Add general explanantion on the pass.
25627         (generate_loops_for_partition): Mark distributed loop.
25628         (pg_add_dependence_edges): New parameter.  Handle alias data
25629         dependence specially and record it in the parameter if asked.
25630         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25631         (init_partition_graph_vertices, add_partition_graph_edge): New.
25632         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25633         (free_partition_graph_vdata, build_partition_graph): New.
25634         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25635         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25636         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25637         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25638         (version_for_distribution_p, finalize_partitions): New.
25639         (distribute_loop): Handle alias data dependence specially.  Factor
25640         out loop fusion code as functions and call these functions.
25642 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25644         * tree-loop-distribution.c (classify_partition): New parameter and
25645         better handle reduction statement.
25646         (rdg_build_partitions): Revise comment.
25647         (distribute_loop): Compute statements in all partitions and pass it
25648         to classify_partition.
25650 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25652         * tree-loop-distribution.c (enum partition_type): New.
25653         (struct partition): New field type.
25654         (partition_merge_into): Add parameter.  Update partition type.
25655         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25656         (build_rdg_partition_for_vertex): Compute partition type.
25657         (rdg_build_partitions): Dump partition type.
25658         (distribute_loop): Update calls to partition_merge_into.
25660 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25662         * tree-loop-distribution.c (struct ddr_hasher): New.
25663         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25664         (ddrs_table): New.
25665         (classify_partition): Call get_data_dependence.
25666         (pg_add_dependence_edges): Ditto.
25667         (distribute_loop): Release data dependence hash table.
25669 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25671         * tree-loop-distribution.c (ref_base_address): Delete.
25672         (similar_memory_accesses): Rename ...
25673         (share_memory_accesses): ... to this.  Check if partitions access
25674         the same memory reference.
25675         (distribute_loop): Call share_memory_accesses.
25677 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25679         * tree-loop-distribution.c (struct partition): New field recording
25680         its data reference.
25681         (partition_alloc, partition_free): Init and release data refs.
25682         (partition_merge_into): Merge data refs.
25683         (build_rdg_partition_for_vertex): Collect data refs for partition.
25684         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25685         Update uses.
25686         (distribute_loop): Remve data refs from vertice data of partition
25687         graph.
25689 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25691         * tree-loop-distribution.c (params.h): Include header file.
25692         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25693         (datarefs_vec): New global var.
25694         (create_rdg_vertices): Use datarefs_vec directly.
25695         (free_rdg): Don't free data references.
25696         (build_rdg): Update use.  Don't free data references.
25697         (distribute_loop): Compute global variable for data references.
25698         Bail out if there are too many data references.
25700 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25702         * tree-loop-distribution.c (loop_nest): New global var.
25703         (build_rdg): Use loop directly, rather than loop nest.
25704         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25705         variable directly.
25706         (distribute_loop): Compute global variable loop nest.  Update use.
25708 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25710         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25711         (partition_merge_into): New parameter.  Dump reason for fusion.
25712         (distribute_loop): Update use of partition_merge_into.
25714 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25716         * tree-loop-distribution.c (bb_top_order_index): New.
25717         (bb_top_order_index_size, bb_top_order_cmp): New.
25718         (stmts_from_loop): Use topological order.
25719         (pass_loop_distribution::execute): Compute and release topological
25720         order for basic blocks.
25722 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25724         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25725         if no loops.
25727 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25729         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25730         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25731         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25732         * internal-fn.def (LOOP_DIST_ALIAS): New.
25733         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25734         (fold_loop_internal_call): ... this.
25735         (vect_loop_dist_alias_call): New function.
25736         (set_uid_loop_bbs): Call fold_loop_internal_call.
25737         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25738         internal calls.
25740 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25742         PR target/81300
25743         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25744         Require dead FLAGS_REG at the beginning of a peephole.
25746 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25748         PR target/81294
25749         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25750         arguments in the call to __builtin_ia32_sbb_u32.
25751         (_subborrow_u64): Swap _X and _Y arguments in the call to
25752         __builtin_ia32_sbb_u64.
25754 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25756         PR debug/81278
25757         * tree-vrp.c (compare_assert_loc): Turn into a function template
25758         with stable template parameter.  Only test if a->e is NULL,
25759         !a->e == !b->e has been verified already.  Use e == NULL or
25760         e != NULL instead of e or ! e tests.  If stable is true, don't use
25761         iterative_hash_expr, on the other side allow a or b or both NULL
25762         and sort the NULLs last.
25763         (process_assert_insertions): Sort using compare_assert_loc<false>
25764         instead of compare_assert_loc, later sort using
25765         compare_assert_loc<true> before calling process_assert_insertions_for
25766         in a loop.  Use break instead of continue once seen NULL pointer.
25768 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25770         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25771         Cortex-R7 and Cortex-R8 processors.
25773 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25775         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25776         uninitialized while src is not.
25778 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25780         * common/config/arm/arm-common.c: Adjust include path for
25781         arm-cpu-cdata.h
25782         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25783         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25784         (arm-cpu-data.h): Likewise.
25785         (arm-cpu-cdata.h): Likewise.
25786         * config/arm/arm-cpu.h: Delete.
25787         * config/arm/arm-cpu-cdata.h: Delete.
25788         * config/arm/arm-cpu-data.h: Delete.
25790 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25792         * config/arm/arm-cpus.in (cortex-a55): New.
25793         (cortex-a75): Likewise.
25794         (cortex-a75.cortex-a55): Likewise.
25795         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25796         cortex-a75.
25797         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25798         * config/arm/arm-cpu-cdata.h: Regenerate.
25799         * config/arm/arm-cpu-data.h: Regenerate.
25800         * config/arm/arm-cpu.h: Regenerate.
25801         * config/arm/arm-tables.opt: Regenerate.
25802         * config/arm/arm-tune.md: Regenerate.
25804 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25806         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25808 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25810         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25811         probability.
25813 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25815         PR tree-optimization/81292
25816         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25817         full_string_p, also call adjust_related_strinfos if the adjustment
25818         is simple, otherwise invalidate related strinfos.
25820 2017-07-04  Martin Liska  <mliska@suse.cz>
25822         PR sanitizer/81040
25823         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25824         newly created variable as DECL_IGNORED_P.
25826 2017-07-04  Martin Liska  <mliska@suse.cz>
25828         PR ipa/81293
25829         * ipa-inline.c (inline_small_functions):
25830         Use xstrdup_for_dump.
25832 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25834         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25836 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25838         PR target/81033
25839         * config/darwin.c (darwin_function_switched_text_sections):
25840         Fix spaces.
25842 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25844         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25846 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25848         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25850 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25852         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25853         min_profitable_iters, and th as inclusive lower bounds.
25854         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25855         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25856         for min_profitable_iters and min_profitable_estimate.
25857         (vect_transform_loop): Treat th as an inclusive lower bound.
25858         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25860 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25862         PR target/81033
25863         * config/darwin.c (darwin_function_switched_text_sections):
25864         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
25865         in two pieces, and suppress the use of buf.
25867 2017-07-03  Nathan Sidwell  <nathan@acm.org>
25869         * hash-table.h (hash_table_mod1): Fix indentation.
25871 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
25873         PR middle-end/81290
25874         * predict.c (force_edge_cold): Be more careful about propagation
25875         backward.
25876         * profile-count.h (profile_probability::guessed,
25877         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
25878         New.
25879         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
25881 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
25883         * doc/invoke.texi (rcpc architecture extension): Document it.
25885 2017-07-03  Richard Biener  <rguenther@suse.de>
25887         PR tree-optimization/60510
25888         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
25889         the scalar reduction PHI and use it.
25890         (vectorizable_reduction): Properly guard the single_defuse_cycle
25891         path for non-SLP reduction chains where we cannot use it.
25892         Rework reduc_def/index and vector type deduction.  Rework
25893         vector operand gathering during reduction op code-gen.
25894         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
25895         chains dissolve the chain and leave it to non-SLP reduction
25896         handling.
25898 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25900         * tree-data-ref.h (dr_alignment): Declare.
25901         * tree-data-ref.c (dr_alignment): New function.
25902         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
25903         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25904         set it.
25905         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
25907 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25909         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
25910         and base_misalignment fields.
25911         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
25912         * tree-data-ref.c: Include builtins.h.
25913         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
25914         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
25915         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
25916         * tree-vect-data-refs.c: Include tree-cfg.h.
25917         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
25918         fields instead of calculating an alignment here.
25919         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
25920         innermost_loop_behavior fields.
25922 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25924         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
25925         field.
25926         (DR_STEP_ALIGNMENT): New macro.
25927         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
25928         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
25929         (create_data_ref): Print it.
25930         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
25931         to tell whether the step preserves vector (mis)alignment.
25932         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
25933         Move the check for an integer step and generalise to all INTEGER_CST.
25934         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
25935         Print the outer step alignment.
25937 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25939         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
25940         with offset_alignment.
25941         (DR_ALIGNED_TO): Delete.
25942         (DR_OFFSET_ALIGNMENT): New macro.
25943         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
25944         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
25945         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
25946         (create_data_ref): Likewise.
25947         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
25948         (vect_analyze_data_refs): Likewise.
25949         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
25950         creating dummy innermost behavior.
25952 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25954         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
25955         with a "innermost_loop_behavior *" and refeence tree.
25956         * tree-data-ref.c (dr_analyze_innermost): Likewise.
25957         (create_data_ref): Update call accordingly.
25958         * tree-predcom.c (find_looparound_phi): Likewise.
25960 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25962         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
25963         fields with dr_wrt_vec_loop.
25964         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
25965         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
25966         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
25967         (vect_dr_behavior): New function.
25968         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
25969         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
25970         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
25971         track whether the step preserves the misalignment.
25972         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
25973         Use vect_dr_behavior.
25974         (vect_setup_realignment): Update call accordingly.
25975         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
25976         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
25977         call to vect_create_addr_base_for_vector_ref.
25978         (vect_create_cond_for_align_checks): Likewise.
25979         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
25980         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
25981         (vect_recog_mask_conversion_pattern): Likewise.
25982         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
25983         (new_stmt_vec_info): Remove redundant zeroing.
25985 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25987         * common/config/arm/arm-common.c (arm_be8_option): New function.
25988         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
25989         (ISA_ARMv6): Add isa_bit_be8.
25990         * config/arm/arm.h (arm_be8_option): Add prototype.
25991         (BE8_SPEC_FUNCTION): New define.
25992         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
25993         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
25994         (mlittle-endian): Similarly.
25995         (mbe8, mbe32): New options.
25996         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
25997         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
25999 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26001         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26003 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26005         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26006         (cleanup_tree_cfg_bb): Use it.
26007         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26008         New functions.
26009         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26011 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26013         PR bootstrap/81285
26014         * loop-doloop.c (add_test): Update profile.
26016 2017-07-03  Martin Liska  <mliska@suse.cz>
26018         PR sanitize/81040
26019         * sanopt.c (rewrite_usage_of_param): New function.
26020         (sanitize_rewrite_addressable_params): Likewise.
26021         (pass_sanopt::execute): Call rewrite_usage_of_param.
26023 2017-07-03  Richard Biener  <rguenther@suse.de>
26025         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26026         back to using VIEW_CONVERT_EXPR.
26028 2017-07-03  Martin Liska  <mliska@suse.cz>
26030         PR other/78366
26031         * doc/extend.texi: Document when a resolver function is
26032         generated for target_clones.
26034 2017-07-03  Martin Liska  <mliska@suse.cz>
26036         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26037         and shadow memory of auto variables which are subject of
26038         use-after-scope sanitization.
26039         (asan_expand_mark_ifn): Add do set only when is_poison.
26041 2016-07-03  Richard Biener  <rguenther@suse.de>
26043         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26044         reduction PHIs.
26045         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26046         (vectorizable_reduction): Perform reduction PHI creation when
26047         visiting a reduction PHI and adjust and simplify code generation
26048         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26049         (vect_transform_loop): Visit reduction PHIs.
26050         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26051         defs into the SLP tree.
26052         (vect_build_slp_tree): Reduction defs terminate the recursion.
26053         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26054         of reduction defs.
26055         (vect_get_vec_defs_for_stmt_copy): Export.
26056         (vect_get_vec_defs): Likewise.
26057         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26058         purpose.
26059         (vect_get_vec_defs_for_stmt_copy): Declare.
26060         (vect_get_vec_defs): Likewise.
26062 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26064         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26065         parameter with a "loop" parameter and use it instead of the
26066         loop containing DR_STMT.  Don't check simple_iv when doing
26067         BB analysis.  Describe the two analysis modes in the comment.
26069 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26071         PR tree-optimization/69468
26072         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26073         (find_same_succ_bb): Handle ignore_edge_flags.
26075 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26077         PR tree-optimization/81192
26078         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26079         hash.
26080         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26081         differs.
26082         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26084 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26086         PR tree-optimization/81192
26087         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26088         BB_SAME_SUCC (bb) == NULL.
26090 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26092         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26093         consistency.
26095 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26097         * dumpfile.c: Include profile-count.h
26098         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26099         update profile.
26100         (insert_cond_bb): Update profile.
26101         * tree-cfg.h (insert_cond_bb): Update prototype.
26102         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26103         * tree-dump.c: Do not include tree-cfg.
26105 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26107         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26109 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26111         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26112         bb.
26114 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26116         * tree-complex.c (expand_complex_div_wide): update profile.
26118 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26119             Alan Hayward  <alan.hayward@arm.com>
26120             David Sherwood  <david.sherwood@arm.com>
26122         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26123         (CORETYPES_H): New define.
26124         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26125         (insn-modes-inline.h, s-modes-inline-h): New rules.
26126         (generated_files): Add insn-modes-inline.h.
26127         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26128         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26129         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26130         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26131         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26132         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26133         (build/gencodes.o, build/genconditions.o): Likewise.
26134         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26135         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26136         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26137         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26138         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26139         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26140         * coretypes.h: Include everything up to real.h for generators.
26141         Include insn-modes.h first.  Include wide-int-print.h after
26142         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26143         * machmode.h: Don't include insn-modes.h here.
26144         * function-tests.c: Remove includes of signop.h, machmode.h,
26145         double-int.h and wide-int.h.
26146         * rtl.h: Likewise.
26147         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26148         and wide-int.h.
26149         * optc-save-gen.awk: Likewise.
26150         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26151         * godump.c: Remove include of wide-int-print.h.
26152         * pretty-print.h: Likewise.
26153         * wide-int-print.cc: Likewise.
26154         * wide-int.cc: Likewise.
26155         * hash-map-tests.c: Remove include of signop.h.
26156         * hash-set-tests.c: Likewise.
26157         * rtl-tests.c: Likewise.
26158         * mkconfig.sh: Remove include of machmode.h.
26159         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26160         into...
26161         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26162         into an insn-modes-inline.h header file, adding appropriate
26163         include guards and end comments.
26164         (emit_insn_modes_c_header): Remove include of machmode.h.
26165         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26166         machmode.h.
26167         (main): Handle -i flag and call emit_insn_modes_inline_h when
26168         it is passed.
26170 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26172         * tree-ssa-strlen.c (strinfo): Rename the length field to
26173         nonzero_chars.  Add a full_string_p field.
26174         (compare_nonzero_chars, zero_length_string_p): New functions.
26175         (get_addr_stridx): Add an offset_out parameter.
26176         Use compare_nonzero_chars.
26177         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26178         (new_strinfo): Update after above changes to strinfo.
26179         (set_endptr_and_length): Set full_string_p.
26180         (get_string_length): Update after above changes to strinfo.
26181         (unshare_strinfo): Update call to new_strinfo.
26182         (maybe_invalidate): Likewise.
26183         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26184         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26185         as a uhwi instead of an shwi.  Update after above changes to
26186         strinfo and new_strinfo.
26187         (zero_length_string): Assert that chainsi contains full strings.
26188         Use zero_length_string_p.  Update call to new_strinfo.
26189         (adjust_related_strinfos): Update after above changes to strinfo.
26190         Copy full_string_p from origsi.
26191         (adjust_last_stmt): Use zero_length_string_p.
26192         (handle_builtin_strlen): Update after above changes to strinfo and
26193         new_strinfo.  Install the lhs as the string length if the previous
26194         entry didn't describe a full string.
26195         (handle_builtin_strchr): Update after above changes to strinfo
26196         and new_strinfo.
26197         (handle_builtin_strcpy): Likewise.
26198         (handle_builtin_strcat): Likewise.
26199         (handle_builtin_malloc): Likewise.
26200         (handle_pointer_plus): Likewise.
26201         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26202         that aren't necessarily followed by a nul terminator.
26203         (handle_char_store): Likewise.
26205 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26207         PR tree-optimization/80769
26208         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26209         for malloc and calloc.  Document the new invariant that all related
26210         strinfos have delayed lengths or none do.
26211         (verify_related_strinfos): Move earlier in file.
26212         (set_endptr_and_length): New function, split out from...
26213         (get_string_length): ...here.  Also set the lengths of related
26214         strinfos.
26215         (zero_length_string): Assert that chainsi has known (rather than
26216         delayed) lengths.
26217         (adjust_related_strinfos): Likewise.
26219 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26221         PR tree-optimization/81136
26222         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26223         assert that two references with the same misalignment have the same
26224         compile-time misalignment if those compile-time misalignments
26225         are known.
26227 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26229         * print-tree.c (print_node): Print all attributes.
26231 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26233         * cfg.c (scale_bbs_frequencies): New function.
26234         * cfg.h (scale_bbs_frequencies): Declare it.
26235         * cfgloopanal.c (single_likely_exit): Cleanup.
26236         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26237         as parameter.
26238         (scale_loop_profile): Likewise.
26239         (loop_version): Likewise.
26240         (create_empty_loop_on_edge): Update.
26241         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26242         scale_loop_frequencies, scale_loop_profile, loopify,
26243         loop_version): Update prototypes.
26244         * modulo-sched.c (sms_schedule): Update.
26245         * predict.c (unlikely_executed_edge_p): Also check probability.
26246         (probably_never_executed_edge_p): Fix typo.
26247         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26248         * tree-parloops.c (gen_parallel_loop): Update.
26249         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26250         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26251         * tree-ssa-loop-split.c (split_loop): Update.
26252         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26253         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26254         (vect_loop_versioning): Update.
26255         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26257 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26259         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26261 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26263         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26264         to keep profile consistent.
26266 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26268         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26269         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26270         * profile-count.h (max_safe_multiplier): Make unsigned.
26271         (profile_count::guessed_zero): New.
26273 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26275         * bb-reorder.c (fix_up_crossing_landing_pad,
26276         fix_crossing_conditional_branches): Use make_single_succ_edge
26277         to keep profile consistent.
26279 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26281         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26282         to update profile.
26284 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26286         PR sanitizer/81262
26287         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26288         the right scopes, make sure cond_jump isn't preserved between multiple
26289         iterations.  Search for fallthru edge whenever there are 3+ edges and
26290         use find_fallthru_edge for it.
26292 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26294         Patch by Alexander Monakov <amonakov@ispras.ru>
26295         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26296         probabilities consistently.
26298 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26300         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26301         emit_cmp_and_jump_insns.
26303 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26305         PR ipa/81261
26306         * tree-inline.c (expand_call_inline): Combine profile statuses.
26308 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26310         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26311         fold_stmt returned true.
26313 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26315         * ggc.h (empty_string): Delete.
26316         * cfgexpand.c (expand_asm_stmt): Use plain "".
26317         * optabs.c (expand_asm_memory_barrier): Likewise.
26318         * stringpool.c (empty_string): Delete.
26319         (digit_vector, digit_string): Delete.
26320         (ggc_alloc_string): Use plain "", don't optimize single digit
26321         strings.  Use ggc_alloc_atomic.
26323 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26325         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26326         comparison set and one other set, use the cost of the non-comparison
26327         set.
26329 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26331         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26332         some formatting.
26334 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26336         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26337         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26338         just set removed labels to NULL_TREE.
26340 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26342         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26343         (set_range_info): ...here.  Only call set_range_info_raw if domain
26344         is useful.
26345         (set_nonzero_bits): Call set_range_info_raw.
26346         * tree-ssanames.h (set_range_info_raw): New.
26348 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26350         PR target/81225
26351         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26352         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26353         of nonimmediate_operand and <store_mask_constraint> instead of m
26354         for the input operand.  For V8FI iterator, always split if input
26355         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26356         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26357         <store_mask_predicate> instead of register_operand and
26358         <store_mask_constraint> instead of v for the input operand.  Make
26359         sure both operands aren't MEMs for if not <mask_applied>.
26361 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26363         * lto-wrapper.c (copy_file) Close both file descriptors before
26364         exiting normally.
26366 2017-06-30  Martin Liska  <mliska@suse.cz>
26368         PR ipa/81214
26369         * multiple_target.c (create_dispatcher_calls): Make ifunc
26370         also for function that don't have calls or are not referenced.
26372 2017-06-30  Richard Biener  <rguenther@suse.de>
26374         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26375         analyze the first scalar stmt.  Move vector type computation
26376         for the BB case here from ...
26377         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26378         live operation processing in the SLP case properly.
26380 2017-06-30  Richard Biener  <rguenther@suse.de>
26382         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26384 2017-06-30  Martin Liska  <mliska@suse.cz>
26386         PR sanitizer/81021
26387         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26388         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26390 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26392         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26393         ones.
26395 2017-06-30  Richard Biener  <rguenther@suse.de>
26397         PR tree-optimization/81249
26398         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26399         condition reduction result to original scalar type.
26401 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26403         * profile-count.h (enum profile_quality): Fix typos and whitespace
26404         issues.
26406 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26408         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26409         type for branch probabilities.
26411 2017-06-29  Julian Brown  <julian@codesourcery.com>
26412             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26414         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26415         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26416         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26417         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26419 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26421         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26422         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26423         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26424         CC usage from generic code to here.
26425         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26426         CC usage into the target macros.
26428 2017-06-29  Maya Rashish  <coypu@sdf.org>
26430         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26431         objects.
26433 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26435         * arm/arm-builtins.c: Include profile-count.h
26436         * except.c (sjlj_emit_function_enter): Use
26437         profile_probability::unlikely.
26439 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26441         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26442         and tocrel_offset be pointer args rather than implicitly using
26443         static versions.
26444         (legitimate_constant_pool_address_p, rs6000_emit_move,
26445         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26446         tocrel_offset and use in toc_relative_expr_p call.
26447         (print_operand, print_operand_address): Use static tocrel_base_oac
26448         and tocrel_offset_oac.
26449         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26450         tocrel_offset_oac.
26452 2017-06-29  Maya Rashish  <coypu@sdf.org>
26454         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26456 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26458         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26459         objects, take into account only the alignment of 'op0' and 'mode1' if
26460         'op0' is a MEM.
26462 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26464         * ccmp.c (ccmp_tree_comparison_p): New function.
26465         (ccmp_candidate_p): Update to use above function.
26466         (get_compare_parts): New function.
26467         (expand_ccmp_next): Update to use new functions.
26468         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26469         new functions.
26470         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26471         take mode as argument.
26472         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26473         * expr.c (expand_expr_real_1): Pass mode as argument.
26475 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26477         * combine.c (combine_instructions): Print insns to dump_file, together
26478         with their costs.
26480 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26482         * asan.c (asan_emit_stack_protection): Update.
26483         (create_cond_insert_point): Update.
26484         * auto-profile.c (afdo_propagate_circuit): Update.
26485         * basic-block.h (struct edge_def): Turn probability to
26486         profile_probability.
26487         (EDGE_FREQUENCY): Update.
26488         * bb-reorder.c (find_traces_1_round): Update.
26489         (better_edge_p): Update.
26490         (sanitize_hot_paths): Update.
26491         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26492         (make_single_succ_edge): Update.
26493         (check_bb_profile): Update.
26494         (dump_edge_info): Update.
26495         (update_bb_profile_for_threading): Update.
26496         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26497         probabilitycount to 0.
26498         * cfgbuild.c (compute_outgoing_frequencies): Update.
26499         * cfgcleanup.c (try_forward_edges): Update.
26500         (outgoing_edges_match): Update.
26501         (try_crossjump_to_edge): Update.
26502         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26503         (expand_gimple_tailcall): Update.
26504         (construct_init_block): Use make_single_succ_edge.
26505         (construct_exit_block): Use make_single_succ_edge.
26506         * cfghooks.c (verify_flow_info): Update.
26507         (redirect_edge_succ_nodup): Update.
26508         (split_edge): Update.
26509         (account_profile_record): Update.
26510         * cfgloopanal.c (single_likely_exit): Update.
26511         * cfgloopmanip.c (scale_loop_profile): Update.
26512         (set_zero_probability): Remove.
26513         (duplicate_loop_to_header_edge): Update.
26514         * cfgloopmanip.h (loop_version): Update prototype.
26515         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26516         (force_nonfallthru_and_redirect): Update.
26517         (update_br_prob_note): Update.
26518         (rtl_verify_edges): Update.
26519         (purge_dead_edges): Update.
26520         (rtl_lv_add_condition_to_bb): Update.
26521         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26522         * cgraphunit.c (init_lowered_empty_function): Update.
26523         (cgraph_node::expand_thunk): Update.
26524         * cilk-common.c: Include profile-count.h
26525         * dojump.c (inv): Remove.
26526         (jumpifnot): Update.
26527         (jumpifnot_1): Update.
26528         (do_jump_1): Update.
26529         (do_jump): Update.
26530         (do_jump_by_parts_greater_rtx): Update.
26531         (do_compare_rtx_and_jump): Update.
26532         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26533         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26534         * dwarf2cfi.c: Include profile-count.h
26535         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26536         (sjlj_emit_dispatch_table): Likewise.
26537         * explow.c: Include profile-count.h
26538         * expmed.c (emit_store_flag_force): Update.
26539         (do_cmp_and_jump): Update.
26540         * expr.c (compare_by_pieces_d::generate): Update.
26541         (compare_by_pieces_d::finish_mode): Update.
26542         (emit_block_move_via_loop): Update.
26543         (store_expr_with_bounds): Update.
26544         (store_constructor): Update.
26545         (expand_expr_real_2): Update.
26546         (expand_expr_real_1): Update.
26547         * expr.h (try_casesi, try_tablejump): Update prototypes.
26548         * gimple-pretty-print.c (dump_probability): Update.
26549         (dump_profile): New.
26550         (dump_gimple_label): Update.
26551         (dump_gimple_bb_header): Update.
26552         * graph.c (draw_cfg_node_succ_edges): Update.
26553         * hsa-gen.c (convert_switch_statements): Update.
26554         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26555         (find_if_case_1): Update.
26556         (find_if_case_2): Update.
26557         * internal-fn.c (expand_arith_overflow_result_store): Update.
26558         (expand_addsub_overflow): Update.
26559         (expand_neg_overflow): Update.
26560         (expand_mul_overflow): Update.
26561         (expand_vector_ubsan_overflow): Update.
26562         * ipa-cp.c (good_cloning_opportunity_p): Update.
26563         * ipa-split.c (split_function): Use make_single_succ_edge.
26564         * ipa-utils.c (ipa_merge_profiles): Update.
26565         * loop-doloop.c (add_test): Update.
26566         (doloop_modify): Update.
26567         * loop-unroll.c (compare_and_jump_seq): Update.
26568         (unroll_loop_runtime_iterations): Update.
26569         * lra-constraints.c (lra_inheritance): Update.
26570         * lto-streamer-in.c (input_cfg): Update.
26571         * lto-streamer-out.c (output_cfg): Update.
26572         * mcf.c (adjust_cfg_counts): Update.
26573         * modulo-sched.c (sms_schedule): Update.
26574         * omp-expand.c (expand_omp_for_init_counts): Update.
26575         (extract_omp_for_update_vars): Update.
26576         (expand_omp_ordered_sink): Update.
26577         (expand_omp_for_ordered_loops): Update.
26578         (expand_omp_for_generic): Update.
26579         (expand_omp_for_static_nochunk): Update.
26580         (expand_omp_for_static_chunk): Update.
26581         (expand_cilk_for): Update.
26582         (expand_omp_simd): Update.
26583         (expand_omp_taskloop_for_outer): Update.
26584         (expand_omp_taskloop_for_inner): Update.
26585         * omp-simd-clone.c (simd_clone_adjust): Update.
26586         * optabs.c (expand_doubleword_shift): Update.
26587         (expand_abs): Update.
26588         (emit_cmp_and_jump_insn_1): Update.
26589         (expand_compare_and_swap_loop): Update.
26590         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26591         * predict.c (predictable_edge_p): Update.
26592         (edge_probability_reliable_p): Update.
26593         (set_even_probabilities): Update.
26594         (combine_predictions_for_insn): Update.
26595         (combine_predictions_for_bb): Update.
26596         (propagate_freq): Update.
26597         (estimate_bb_frequencies): Update.
26598         (force_edge_cold): Update.
26599         * profile-count.c (profile_count::dump): Add missing space into dump.
26600         (profile_count::debug): Add newline.
26601         (profile_count::differs_from_p): Explicitly convert to unsigned.
26602         (profile_count::stream_in): Update.
26603         (profile_probability::dump): New member function.
26604         (profile_probability::debug): New member function.
26605         (profile_probability::differs_from_p): New member function.
26606         (profile_probability::differs_lot_from_p): New member function.
26607         (profile_probability::stream_in): New member function.
26608         (profile_probability::stream_out): New member function.
26609         * profile-count.h (profile_count_quality): Rename to ...
26610         (profile_quality): ... this one.
26611         (profile_probability): New.
26612         (profile_count): Update.
26613         * profile.c (compute_branch_probabilities): Update.
26614         * recog.c (peep2_attempt): Update.
26615         * sched-ebb.c (schedule_ebbs): Update.
26616         * sched-rgn.c (find_single_block_region): Update.
26617         (compute_dom_prob_ps): Update.
26618         (schedule_region): Update.
26619         * sel-sched-ir.c (compute_succs_info): Update.
26620         * stmt.c (struct case_node): Update.
26621         (do_jump_if_equal): Update.
26622         (get_outgoing_edge_probs): Update.
26623         (conditional_probability): Update.
26624         (emit_case_dispatch_table): Update.
26625         (expand_case): Update.
26626         (expand_sjlj_dispatch_table): Update.
26627         (emit_case_nodes): Update.
26628         * targhooks.c: Update.
26629         * tracer.c (better_p): Update.
26630         (find_best_successor): Update.
26631         * trans-mem.c (expand_transaction): Update.
26632         * tree-call-cdce.c: Update.
26633         * tree-cfg.c (gimple_split_edge): Upate.
26634         (move_sese_region_to_fn): Upate.
26635         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26636         * tree-eh.c (lower_resx): Upate.
26637         (cleanup_empty_eh_move_lp): Upate.
26638         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26639         * tree-inline.c (copy_edges_for_bb): Update.
26640         (copy_cfg_body): Update.
26641         * tree-parloops.c (gen_parallel_loop): Update.
26642         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26643         (gimple_gen_time_profiler): Update.
26644         * tree-ssa-dce.c (remove_dead_stmt): Update.
26645         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26646         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26647         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26648         (unloop_loops): Update.
26649         (try_peel_loop): Update.
26650         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26651         * tree-ssa-loop-split.c (connect_loops): Update.
26652         (split_loop): Update.
26653         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26654         (hoist_guard): Update.
26655         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26656         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26657         (value_replacement): Update.
26658         * tree-ssa-reassoc.c (branch_fixup): Update.
26659         * tree-ssa-tail-merge.c (replace_block_by): Update.
26660         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26661         (create_edge_and_update_destination_phis): Update.
26662         (compute_path_counts): Update.
26663         (recompute_probabilities): Update.
26664         (update_joiner_offpath_counts): Update.
26665         (freqs_to_counts_path): Update.
26666         (duplicate_thread_path): Update.
26667         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26668         (struct switch_conv_info): Update.
26669         (gen_inbound_check): Update.
26670         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26671         (vect_do_peeling): Update.
26672         (vect_loop_versioning): Update.
26673         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26674         (optimize_mask_stores): Update.
26675         * ubsan.c (ubsan_expand_null_ifn): Update.
26676         * value-prof.c (gimple_divmod_fixed_value): Update.
26677         (gimple_divmod_fixed_value_transform): Update.
26678         (gimple_mod_pow2): Update.
26679         (gimple_mod_pow2_value_transform): Update.
26680         (gimple_mod_subtract): Update.
26681         (gimple_mod_subtract_transform): Update.
26682         (gimple_ic): Update.
26683         (gimple_stringop_fixed_value): Update.
26684         (gimple_stringops_transform): Update.
26685         * value-prof.h: Update.
26687 2017-06-29  Carl Love  <cel@us.ibm.com>
26689         * config/rs6000/rs6000-c.c: Add support for built-in functions
26690         vector signed int vec_signed (vector float);
26691         vector signed long long vec_signed (vector double);
26692         vector signed int vec_signed2 (vector double, vector double);
26693         vector signed int vec_signede (vector double);
26694         vector signed int vec_signedo (vector double);
26695         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26696         instruction generator.
26697         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26698         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26699         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26700         Add define_insn.
26701         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26702         vunsignede_v2df): Add define_expands.
26703         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26704         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26705         VEC_UNSIGNEDO): Add definitions.
26706         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26707         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26708         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26709         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26710         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26711         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26712         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26713         vec_unsignede, vec_unsignedo): Add builtin defines.
26714         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26715         declaration.
26716         * doc/extend.texi: Update the built-in documentation file for the
26717         new built-in functions.
26719 2017-06-29  Richard Biener  <rguenther@suse.de>
26721         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26722         reduction chains to LOOP_VINFO_REDUCTIONS.
26723         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26724         SLP reductions after processing reduction chains.
26726 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26728         * builtins.c (fold_builtin_FUNCTION): Use
26729         lang_hooks.decl_printable_name.
26731 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26733         PR middle-end/81194
26734         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26735         with only one label.
26736         * stmt.c (expand_case): Assert NCASES is greater than one.
26738 2017-06-29  Richard Biener  <rguenther@suse.de>
26740         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26741         anything.
26742         (group_case_labels): Likewise.
26743         (find_taken_edge): Push sanity checking on val to workers...
26744         (find_taken_edge_cond_expr): ... here
26745         (find_taken_edge_switch_expr): ... and here, handle cases
26746         with just a default label.
26747         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26748         (group_case_labels): Likewise.
26749         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26750         group_case_labels does anything cleanup the CFG again.
26752 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26754         PR tree-optimization/81196
26755         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26756         exit condition comparing two IVs.
26758 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26760         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26761         profile to the dummy entry at the end of the list of architectures.
26762         * config/arm/arm-cpu-cdata.h: Regenerated.
26764 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26765             Michael Collison <michael.collison@arm.com>
26767         PR target/70119
26768         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26769         New pattern.
26770         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26771         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26772         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26773         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26774         of shift when the shift amount is masked with constant equal to
26775         the size of the mode.
26776         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26777         predicate.
26779 2017-06-29  Martin Liska  <mliska@suse.cz>
26781         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26783 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26785         PR bootstrap/80565
26786         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26787         code.
26788         * ipa-inline.h
26789         (edge_growth_cache_entry::edge_growth_cache_entry): New
26790         function.
26791         (reset_edge_growth_cache): Update to use constructor.
26793 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26795         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26796         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26797         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26799 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26801         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26802         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26804 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26806         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26807         (*-linux-uclibc*): Add t-uclibc tmake_file.
26808         * config/t-musl: New.
26809         * config/t-uclibc: New.
26811 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26813         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26814         context.
26815         (gen_comm_data): Emit architectural setting of arch_prof.
26816         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26817         profile.
26818         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26819         (armv8-m.base, armv8-m.main): Likewise.
26820         * arm-protos.h (arm_build_target): Add profile field.
26821         (arch_option): Likewise.
26822         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26823         the active target.
26824         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26825         arm_active_target.profile.
26827 2017-06-28  Richard Biener  <rguenther@suse.de>
26829         PR middle-end/81227
26830         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26831         TYPE_OVERFLOW_WRAPS.
26832         * match.pd (negate_expr_p): Likewise.
26833         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26834         fold_build2, not fold_binary.
26836 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26838         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26839         Convert memory address to Pmode.
26840         (aarch64_print_operand): Assert MEM operands are always Pmode.
26842 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26844         PR target/79665
26845         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26846         Remove redundant if.
26847         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26848         * config/arm/aarch-common-protos.h
26849         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26850         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26852 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26854         PR ipa/81238
26855         * multiple_target.c (create_dispatcher_calls): Set the default
26856         clone to be static, not public.
26858 2017-06-28  Richard Biener  <rguenther@suse.de>
26860         * tree-vect-loop.c (vectorizable_reduction): Move special
26861         cond reduction IV var creation ...
26862         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
26863         parameter.  Use STMT_VINFO_VECTYPE.
26864         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
26865         constant_p.
26867 2017-06-28  Martin Liska  <mliska@suse.cz>
26869         PR ipa/81128
26870         * ipa-visibility.c (non_local_p): Handle visibility.
26872 2017-06-28  Martin Liska  <mliska@suse.cz>
26874         PR driver/79659
26875         * common.opt: Add IntegerRange to various options.
26876         * opt-functions.awk (integer_range_info): New function.
26877         * optc-gen.awk: Add integer_range_info to cl_options struct.
26878         * opts-common.c (decode_cmdline_option): Handle
26879         CL_ERR_INT_RANGE_ARG.
26880         (cmdline_handle_error): Likewise.
26881         * opts.c (print_filtered_help): Show valid interval in
26882         when --help is provided.
26883         * opts.h (struct cl_option): Add range_min and range_max fields.
26884         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
26886 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
26888         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
26889         (x * C EQ/NE y * C): New transformation.
26891 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
26893         * genmultilib (combination_space): Accept '+' in option names.
26895 2017-06-28  Martin Liska  <mliska@suse.cz>
26897         PR sanitizer/81224
26898         * asan.c (instrument_derefs): Bail out inner references
26899         that are hard register variables.
26901 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
26903         PR target/81175
26904         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
26905         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
26907 2017-06-28  Richard Biener  <rguenther@suse.de>
26909         * tree-vectorizer.h (vect_get_vec_defs): Remove.
26910         (vect_get_slp_defs): Adjust.
26911         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
26912         out from ...
26913         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
26914         simplify.
26915         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
26916         get_initial_defs_for_reduction instead of vect_get_vec_defs.
26917         (vectorizable_reduction): Adjust.
26918         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
26919         handling.
26920         (vect_get_slp_defs): Likewise.
26921         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
26922         (vectorizable_bswap): Adjust.
26923         (vectorizable_call): Likewise.
26924         (vectorizable_conversion): Likewise.
26925         (vectorizable_assignment): Likewise.
26926         (vectorizable_shift): Likewise.
26927         (vectorizable_operation): Likewise.
26928         (vectorizable_store): Likewise.
26929         (vectorizable_condition): Likewise.
26930         (vectorizable_comparison): Likewise.
26932 2017-06-28  Michael Collison  <michael.collison@arm.com>
26934         PR target/68535
26935         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
26936         set of base_reg
26937         (arm_gen_movmemqi): Removed unused variable 'i'.
26938         Convert 'for' loop into 'while' loop.
26939         (arm_expand_prologue): Remove last unnecessary set of insn.
26940         (thumb_pop): Remove unused variable 'pushed_words'.
26941         (thumb_exit): Remove last unnecessary set of regs_to_pop.
26943 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26945         * config/s390/predicates.md: Use s390_rel_address_ok_p.
26946         * config/s390/s390-protos.h: Add prototype of
26947         s390_rel_address_ok_p.
26948         * config/s390/s390.c (s390_got_symbol): New function.
26949         (s390_rel_address_ok_p): New function.
26950         (legitimize_pic_address): Use s390_rel_address_ok_p.
26951         (s390_load_got): Use s390_got_symbol.
26952         (s390_option_override): Issue error if
26953         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
26954         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
26955         New macro.
26956         * config/s390/s390.opt: New option mpic-data-is-text-relative.
26958 2017-06-27  Andrew Pinski  <apinski@cavium.com>
26960         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
26961         (X * copysign (1.0, X)): New pattern.
26962         (X * copysign (1.0, -X)): New pattern.
26963         (copysign (-1.0, CST)): New pattern.
26965 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
26967         * genmultilib (combination_space): Remove variable.
26968         Validate reuse rules against regular expression for any sequence
26969         of multilib options in any order.
26971 2017-06-27  Michael Collison  <michael.collison@arm.com>
26973         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
26974         call aarch64_split_simd_combine.
26975         * (aarch64_combine_internal<mode>): Delete pattern.
26976         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
26977         Allow register and subreg operands.
26979 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
26981         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
26982         specific need, just fallback on defaults.
26983         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
26985 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
26986             Olivier Hainque  <hainque@adacore.com>
26988         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
26989         map for 64bits.
26990         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
26991         targets. Pick a default if no particular attempt applied.
26992         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
26993         larger contexts.
26995 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
26997         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
26998         (x86_64-wrs-vxworks7): Likewise.
27000 2017-06-27  Marek Polacek  <polacek@redhat.com>
27002         PR sanitizer/81223
27003         * ubsan.c (instrument_null): Check get_base_address's result for null.
27005 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27007         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27009 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27011         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27012         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27013         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27014         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27015         New function types.
27016         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27017         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27018         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27019         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27020         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27021         BUILT_IN_FEUPDATEENV): New builtins.
27022         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27023         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27024         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27025         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27026         macros.
27027         (builtin_structptr_types): Adjust size.
27028         * tree.c (builtin_structptr_types): Add four entries.
27030 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27031             Olivier Hainque  <hainque@adacore.com>
27033         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27034         (TLS_SYM): New local macro, forcing reference to __tls__ on
27035         link command lines for VxWorks 7 RTPs, triggering initialization
27036         of tlsLib.
27037         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27038         OS features TLS support, true for RTPs on VxWorks 7.
27039         * config/vxworks.c (vxworks_override_options): Setup emutls
27040         accordingly.
27042 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27044         * predict.c (test_prediction_value_range): Use -1U instead of -1
27045         to avoid narrowing conversion warning.
27046         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27047         to avoid narrowing conversion warning.
27048         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27049         -1.
27050         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27052 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27054         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27055         64bit configurations.
27056         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27057         (SIZE_TYPE): Likewise.
27058         * config/vxworks.c (vxworks_emutls_var_fields): Use
27059         long_unsigned_type_node instead of unsigned_type_node as the offset
27060         field type, which is "pointer" mode in emutls.c.
27062 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27064         PR sanitizer/81209
27065         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27067         PR middle-end/81207
27068         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27069         gimple_vuse copying separately from gimple_vdef copying.
27071 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27073         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27074         condition.
27076 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27077             Olivier Hainque  <hainque@adacore.com>
27079         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27080         for all vxworks7 targets.
27081         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27082         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27083         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27084         variations for VX6/VX7 and 32/64bits later on in ...
27085         (VXWORKS_LIB_SPEC): Leverage new macros.
27086         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27087         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27089 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27091         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27092         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27094 2017-06-26  Carl Love  <cel@us.ibm.com>
27096         * config/rs6000/rs6000-c.c: Add support for built-in functions
27097         vector bool char vec_reve (vector bool char);
27098         vector signed char vec_reve (vector signed char);
27099         vector unsigned char vec_reve (vector unsigned char);
27100         vector bool int vec_reve (vector bool int);
27101         vector signed int vec_reve (vector signed int);
27102         vector unsigned int vec_reve (vector unsigned int);
27103         vector bool long long vec_reve (vector bool long long);
27104         vector signed long long vec_reve (vector signed long long);
27105         vector unsigned long long vec_reve (vector unsigned long long);
27106         vector bool short vec_reve (vector bool short);
27107         vector signed short vec_reve (vector signed short);
27108         vector double vec_reve (vector double);
27109         vector float vec_reve (vector float);
27110         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27111         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27112         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27113         (altivec_vreve): New pattern.
27114         * config/rs6000/altivec.h (vec_reve): New define.
27115         * doc/extend.texi (vec_rev): Update the built-in documentation file
27116         for the new built-in functions.
27118 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27120         PR tree-optimization/71815
27121         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27122         function.
27123         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27124         has_single_use.
27125         (slsr_process_phi): Likewise.
27126         (replace_uncond_cands_and_profitable_phis): Don't replace a
27127         multiply candidate with a stride of 1 (copy or cast).
27128         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27129         has_single_use.
27130         (lowest_cost_path): Likewise.
27131         (total_savings): Likewise.
27133 2017-06-26  Richard Biener  <rguenther@suse.de>
27135         PR target/81175
27136         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27137         Use def_builtin_pure for all gather builtins.
27139 2017-06-26  Richard Biener  <rguenther@suse.de>
27141         PR tree-optimization/81203
27142         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27143         non-dominating BBs.
27145 2017-06-26  Marek Polacek  <polacek@redhat.com>
27147         PR c/80116
27148         * doc/invoke.texi: Document -Wmultistatement-macros.
27150 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27152         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27153         arm_neon_ok_no_float_abi effective target.
27155 2017-06-26  Richard Biener  <rguenther@suse.de>
27157         PR tree-optimization/80928
27158         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27159         (copy_bbs): Set BB_DUPLICATED flag early.
27160         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27161         marked blocks.
27162         (execute_on_shrinking_pred): Likewise.
27163         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27164         BB_DUPLICATED blocks.
27165         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27166         iterate over all PHIs considering removal of *gsi.
27168 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27170         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27171         qdf24xx.
27173 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27175         * config/rs6000/rs6000-string.c: (expand_block_clear,
27176         do_load_for_compare, select_block_compare_mode,
27177         compute_current_alignment, expand_block_compare,
27178         expand_strncmp_align_check, expand_strn_compare,
27179         expand_block_move, rs6000_output_load_multiple)
27180         Move functions related to string/block move/compare
27181         to a separate file.
27182         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27183         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27184         for this function which is now used in two files.
27185         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27186         * config.gcc: Add rs6000-string.o to extra_objs for
27187         targets powerpc*-*-* and rs6000*-*-*.
27189 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27191         PR target/80510
27192         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27193         32-bit, since indexed is not valid for DImode.
27194         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27195         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27196         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27197         (define_peephole2 for Altivec d-form store): Likewise.
27199         PR ipa/81185
27200         * multiple_target.c (create_dispatcher_calls): Only create the
27201         dispatcher call if the function is the default clone of a
27202         versioned function.
27204 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27206         PR middle-end/80902
27207         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27208         a call, force the call to not be a tail call.
27210 2017-06-23  Jeff Law  <law@redhat.com>
27212         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27213         enquire.
27215 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27217         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27218         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27219         handling for early expansion of vector shifts (sl,sr,sra,rl).
27220         (builtin_function_type): Add vector shift right instructions
27221         to the unsigned argument list.
27223 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27225         rtl-optimizatoin/79286
27226         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27227         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27228         trap.  PIC register plus a const unspec without offset can never trap.
27230 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27232         * tree.h (builtin_structptr_type): New type.
27233         (builtin_structptr_types): Declare new array.
27234         * tree.c (builtin_structptr_types): New array.
27235         (free_lang_data, build_common_tree_nodes): Use it.
27237 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27239         PR c++/81187
27240         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27241         -Wnoexcept.
27243 2017-06-22  Matt Turner  <mattst88@gmail.com>
27245         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27246         Lake models to skylake case.  Assume skylake for unknown
27247         models with clflushopt.
27249 2017-06-22  Jeff Law  <law@redhat.com>
27251         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27252         frame sizes that do not satisfy aarch64_uimm12_shift.
27254 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27256         * profile-count.h (apply_probability,
27257         apply_scale, probability_in): Fix checks for zero.
27259 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27261         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27262         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27264 2016-06-22  Richard Biener  <rguenther@suse.de>
27266         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27267         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27268         REDUC_MAX_EXPR support.
27269         (vectorizable_reduction): Likewise.
27270         (vect_create_epilog_for_reduction): Likewise.
27272 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27274         * match.pd (A / (1 << B) -> A >> B): New.
27275         * generic-match-head.c: Include optabs-tree.h.
27276         * gimple-match-head.c: Likewise.
27277         * optabs-tree.h (target_supports_op_p): New.
27278         * optabs-tree.c (target_supports_op_p): New.
27280 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27282         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27283         $gcc_cv_ld --help output.
27284         (gcc_cv_ld_demangle): Likewise.
27285         (gcc_cv_ld_eh_frame_hdr): Likewise.
27286         (gcc_cv_ld_pie): Likewise.
27287         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27288         (gcc_cv_ld_buildid): Likewise.
27289         (gcc_cv_ld_sysroot): Likewise.
27290         (ld_bndplt_support): Likewise.
27291         (ld_pushpopstate_support): Likewise.
27292         * configure: Regenerate.
27293         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27295 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27297         PR target/81151
27298         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27299         operands indexes to avoid gap between operands and match_dups.
27301 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27303         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27304         Increment Arith_shift and Arith_shift_reg by 1.
27305         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27306         New tuning flag.
27307         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27308         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27309         (aarch64_strip_extend): Add new argument and test for it.
27310         (aarch64_cheap_mult_shift_p): New function.
27311         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27312         add a cost if it is true.
27313         Update calls to aarch64_strip_extend.
27314         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27316 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27318         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27319         tunings.
27320         (thunderxt88): Likewise.
27321         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27322         (thunderx_prefetch_tune): New variable.
27323         (thunderx2t99_prefetch_tune): Update for the correct values.
27324         (thunderxt88_tunings): New variable.
27325         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27326         generic_prefetch_tune.
27327         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27329 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27331         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27332         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27333         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27334         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27335         (aarch64_atomic_cas<mode>, GPI): Likewise.
27337 2017-06-21  Martin Liska  <mliska@suse.cz>
27339         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27340         statements on cold and hot labels.
27341         * predict.c (tree_estimate_probability_bb): Remove the
27342         prediction from this place.
27344 2017-06-21  Martin Liska  <mliska@suse.cz>
27346         PR tree-optimization/79489
27347         * gimplify.c (maybe_add_early_return_predict_stmt): New
27348         function.
27349         (gimplify_return_expr): Call the function.
27350         * predict.c (tree_estimate_probability_bb): Remove handling
27351         of early return.
27352         * predict.def: Update comment about early return predictor.
27353         * gimple-predict.h (is_gimple_predict): New function.
27354         * predict.def: Change default value of early return to 66.
27355         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27356         statements.
27357         * passes.def: Put pass_strip_predict_hints to the beginning of
27358         IPA passes.
27360 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27362         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27363         FUNCTION_DECL declarations.
27364         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27365         declarations.
27366         (dwaf2out_decl): Likewise.
27367         * godump.c (go_early_global_decl): Skip call to the real debug hook
27368         for FUNCTION_DECL declarations.
27369         * passes.c (rest_of_decl_compilation): Skip call to the
27370         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27371         -fdump-go-spec is passed.
27373 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27375         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27376         Reorder for compactness.
27377         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27378         (def_builtin_pure, def_builtin_pure2): New functions.
27379         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27381 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27383         * match.pd (nop_convert): New predicate.
27384         ((A +- CST1) +- CST2): Allow some NOP conversions.
27386 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27388         PR c++/81130
27389         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27390         with ctors/dtors if GOVD_SHARED is set.
27392 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27394         * config/aarch64/aarch64.md (movti_aarch64):
27395         Emit mov rather than orr.
27396         (movtf_aarch64): Likewise.
27397         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27398         Emit mov rather than orr.
27400 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27402         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27403         Swap alternatives, make integer dup more expensive.
27405 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27407         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27408         Return true for non-tls symbols.
27410 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27412         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27413         (cortex-a75): Likewise.
27414         (cortex-a75.cortex-a55): Likewise.
27415         * config/aarch64/aarch64-tune.md: Regenerate.
27416         * doc/invoke.texi (-mtune): Document new values for -mtune.
27418 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27420         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27421         stack_size feature.
27422         (Effective-Target Keywords, Other attributes): Suggest using
27423         dg-add-options stack_size feature to get stack limit in stack_size
27424         effective target documentation.
27426 2017-06-21  Julian Brown  <julian@codesourcery.com>
27427             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27429         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27430         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27431         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27432         reservation.
27433         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27434         attribute type list for neon_multiply.
27435         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27436         attribute type list for neon_multiply.
27437         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27438         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27439         attribute type list for neon_multiply.
27440         * config/arm/types.md (crypto_pmull): Add.
27441         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27442         attribute type list.
27444 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27446         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27447         arm1176jzf-s.
27449 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27451         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27452         to make sure not to dereference a NULL cost_classes_ptr pointer.
27454 2017-06-20  Carl Love  <cel@us.ibm.com>
27456         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27457         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27458         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27459         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27460         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27461         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27462         VMULOSW): New enum "unspec" values.
27463         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27464         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27465         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27466         altivec_vmulosw): New patterns.
27467         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27468         VMULOSW): Add definitions.
27470 2017-06-20  Julia Koval  <julia.koval@intel.com>
27472         * config/i386/i386.c: Fix rounding expand for new pattern.
27473         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27475 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27477         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27478         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27480 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27482         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27483         feature string.
27485 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27487         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27488         architecture, then by implementer ID.
27489         * config/aarch64/aarch64-tune.md: Regenerate.
27491 2017-06-20  Richard Biener  <rguenther@suse.de>
27493         PR middle-end/81097
27494         * fold-const.c (split_tree): Fold to type before negating.
27496 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27498         * diagnostic-show-locus.c
27499         (selftest::test_fixit_deletion_affecting_newline): New function.
27500         (selftest::diagnostic_show_locus_c_tests): Call it.
27502 2017-06-20  Andreas Schwab  <schwab@suse.de>
27504         PR target/80970
27505         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27506         instead of "+d".
27508 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27510         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27511         __ARM_FEATURE_COPROC according to support.
27513 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27515         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27516         Rewritten to avoid overflow for > 32-bit pointers.
27518         PR sanitizer/81125
27519         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27520         by removing enum keyword.
27521         (ubsan_type_descriptor): Likewise.  Formatting fix.
27523         PR target/81121
27524         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27525         splitter): Require TARGET_SSE2 in the condition.
27527 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27529         PR target/79799
27530         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27531         for doing vector set of SFmode on ISA 3.0.
27532         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27533         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27534         element.
27535         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27536         SFmode value into a V4SF variable that was extracted from another
27537         V4SF variable without converting the element to double precision
27538         and back to single precision vector format.
27539         (vsx_insert_extract_v4sf_p9_2): Likewise.
27541 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27543         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27544         in UWHI to avoid undefined overflow.
27546         PR sanitizer/81125
27547         * ubsan.h (enum ubsan_encode_value_phase): New.
27548         (ubsan_encode_value): Change second argument to
27549         enum ubsan_encode_value_phase with default value of
27550         UBSAN_ENCODE_VALUE_GENERIC.
27551         * ubsan.c (ubsan_encode_value): Change second argument to
27552         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27553         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27554         create_tmp_var_raw instead of create_tmp_var and use a
27555         TARGET_EXPR.
27556         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27557         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27558         ubsan_encode_value callers.
27560         PR sanitizer/81111
27561         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27562         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27563         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27565 2017-06-19  Richard Biener  <rguenther@suse.de>
27567         PR middle-end/81118
27568         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27569         estimates if we changed anything.
27571 2017-06-19  Richard Biener  <rguenther@suse.de>
27573         PR tree-optimization/80887
27574         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27575         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27576         simplified lookups, then reset mprts_hook.
27577         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27578         simplifying.
27579         (try_to_simplify): Likewise.
27581 2017-06-19  Martin Liska  <mliska@suse.cz>
27583         PR sanitizer/80879
27584         * gimplify.c (gimplify_switch_expr):
27585         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27587 2017-06-19  Martin Liska  <mliska@suse.cz>
27589         * doc/install.texi: Document that PGO runs in 4 stages.
27591 2017-06-19  Martin Liska  <mliska@suse.cz>
27593         PR ipa/80732
27594         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27595         to dispatcher function name.
27596         * multiple_target.c (replace_function_decl): New function.
27597         (create_dispatcher_calls): Redirect both edges and references.
27599 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27601         * profile-count.c (profile_count::dump): Dump quality.
27602         (profile_count::differs_from_p): Update for unsigned val.
27603         * profile-count.h (profile_count_quality): New enum.
27604         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27606 2017-06-19  Richard Biener  <rguenther@suse.de>
27608         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27609         struct function as arg.
27610         (estimate_numbers_of_iterations): Export overload with loop arg.
27611         (free_numbers_of_iterations_estimates_loop): Use an overload of
27612         free_numbers_of_iterations_estimates instead.
27613         * tree-cfg.c (remove_bb): Adjust.
27614         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27615         * tree-parloops.c (gen_parallel_loop): Likewise.
27616         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27617         Likewise.
27618         (tree_unroll_loops_completely): Likewise.
27619         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27620         Use an overload instead and export.
27621         (estimated_loop_iterations): Adjust.
27622         (max_loop_iterations): Likewise.
27623         (likely_max_loop_iterations): Likewise.
27624         (estimate_numbers_of_iterations): Take struct function as arg
27625         and adjust.
27626         (loop_exits_before_overflow): Adjust.
27627         (free_numbers_of_iterations_estimates_loop): Use an overload.
27628         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27629         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27631 2017-06-19  Richard Biener  <rguenther@suse.de>
27633         PR ipa/81112
27634         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27635         RANGE_EXPR conservatively.
27637 2017-06-16  Carl Love  <cel@us.ibm.com>
27639         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27640         definitions for vec_float, vec_float2, vec_floato,
27641         vec_floate built-ins.
27642         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27643         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27644         floate.
27645         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27646         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27647         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27648         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27649         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27650         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27651         vec_floato): Add builtin defines.
27652         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27653         Update the built-in documentation file for the new built-in
27654         functions.
27656 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27658         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27659         (mthumb): Mark as the negative of -marm.
27661 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27663         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27664         extension options.
27665         (ARM Options, -mtune): Document that this accepts the same
27666         extension options as -mcpu.
27667         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27669 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27671         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27672         permitted extensions.
27674 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27676         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27677         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27678         (armv8-m.main): Add option +nodsp.
27679         * config/arm/arm-cpu-cdata.h: Regenerated.
27681 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27683         * config/arm/t-fuchsia: New file.
27684         * config.gcc (arm*-*-fuchsia*): Use it.
27686 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27688         * config/arm/t-symbian: Rewrite for new option infrastructure.
27690 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27692         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27693         (MULTILIB_REQUIRED): Likewise.
27695 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27697         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27698         (MULTILIB_RESUE): Likewise.
27699         (MULTILIB_MATCHES): Likewise.
27700         (MULTLIB_REQUIRED): Likewise.
27702 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27704         * config/arm/t-rtems: Rewrite for new option framework.
27706 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27708         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27709         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27710         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27711         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27712         * config/arm/t-multilib: ... here.
27713         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27714         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27715         armv7-a and armv8*-a when A-profile libraries have not been built.
27716         * config/arm/t-rmprofile: Rewrite.
27718 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27720         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27721         with a backslash.  Remove the backslash after substituting unescaped
27722         periods.
27723         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27725 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27727         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27728         the driver to pass through the default setting of -mfloat-abi.
27729         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27730         rather than NULL.
27731         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27732         (all_feat_combs): New rule.
27733         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27734         default libraries.
27735         * config/arm/t-aprofile: Rewrite.
27737 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27739         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27740         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27741         fpu is not specified by the user/command-line.
27742         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27743         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27744         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27745         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27746         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27747         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27749 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27751         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27752         * config/arm/t-arm-elf: Rewritten.
27754 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27756         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27757         have some floating-point instructions.
27758         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27759         (TARGET_MAYBE_HARD_FLOAT): New macro.
27760         * config/arm/arm-builtins.c (arm_init_builtins): Use
27761         TARGET_MAYBE_HARD_FLOAT.
27762         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27764 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27766         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27767         (configargs.h): Include it.
27768         (arm_print_hint_for_fpu_option): New function.
27769         (arm_parse_fpu_option): New function.
27770         (candidate_extension): New class.
27771         (arm_canon_for_multilib): New function.
27772         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27773         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27774         (ARCH_CANONICAL_SPECS): New macro.
27775         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27777 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27779         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27780         are set after handling multilib fragments.  Set target_cpu_default2
27781         from with_cpu.
27783 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27785         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27786         cpu name.
27787         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27788         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27789         options.
27790         (check_arch): Likewise.
27791         * config/arm/arm.c (arm_configure_build_target): Handle
27792         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27793         options in the default.
27795 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27797         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27798         when an option is an alias of another.
27799         * config/arm/parsecpu.awk (optalias): New parser token.
27800         (gen_comm_data): Mark non-alias options as such.  Emit entries
27801         for extension aliases.
27802         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27803         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27804         (armv6kz, armv6zk, armv6t2): Likewise.
27805         (armv7): Make vfpv3-d16 an alias.
27806         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27807         canonical order.
27808         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27809         Sort in canonical order.
27810         (armv8-a): Sort in canonical order.
27811         (armv8.1-a, armv8.2-a):  Likewise.
27812         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27813         canonical order.
27814         (cortex-a9): Sort in canonical order.
27815         * config/arm/arm.c (selftests.h): Include it.
27816         (arm_test_cpu_arch_data): New function.
27817         (arm_run_self_tests): New function.
27818         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27819         (targetm): Move declaration to the end of the file.
27820         * arm-cpu-cdata.h: Regenerated.
27822 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27824         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27825         call to target_mode_check describing the type of option passed.
27826         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27827         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27828         arm_parse_cpu_option_name to match parameters against list of
27829         available targets.
27830         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27831         arm_arch_core_flags data structure.
27832         * config/arm/arm-cpu_cdata.h: Regenerated.
27834 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27836         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27837         config/arm/arm.c.
27838         (arm_print_hint_for_cpu_option): Likewise.
27839         (arm_print_hint_for_arch_option): Likewise.
27840         (arm_parse_cpu_option_name): Likewise.
27841         (arm_parse_arch_option_name): Likewise.
27842         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27843         of entries in the all_fpus list.
27844         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27845         (arm_parse_cpu_option_name): Declare.
27846         (arm_parse_arch_option_name): Declare.
27847         (arm_parse_option_features): Declare.
27848         (arm_intialize_isa): Declare.
27849         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27850         data tables to ...
27851         (gen_comm_data): ... here.  Make definitions non-static.
27852         * config/arm/arm-cpu-data.h: Regenerated.
27853         * config/arm/arm-cpu-cdata.h: Regenerated.
27855 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27857         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27858         (cpu_arch_extension): New structure.
27859         (cpu_arch_option, arch_option, cpu_option): New structures.
27860         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27861         architecture types.
27862         (gen_data): Generate new format data tables.
27863         * config/arm/arm.c (cpu_tune): New structure.
27864         (cpu_option, processors): Delete.
27865         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
27866         (arm_print_hint_for_cpu_option): ... this and ...
27867         (arm_print_hint_for_arch_option): ... this.
27868         (arm_parse_arch_cpu_name): Delete.  Replace with ...
27869         (arm_parse_cpu_option_name): ... this and ...
27870         (arm_parse_arch_option_name): ... this.
27871         (arm_unrecognized_feature): Change type of target parameter to
27872         cpu_arch_option.
27873         (arm_parse_arch_cpu_features): Delete.  Replace with ...
27874         (arm_parse_option_features): ... this.
27875         (arm_configure_build_target): Rework to use new configuration data
27876         tables.
27877         (arm_print_tune_info): Rework for new configuration data tables.
27878         * config/arm/arm-cpu-data.h: Regenerated.
27879         * config/arm/arm-cpu.h: Regenerated.
27881 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27883         * Makefile.in (OBJS): Move sbitmap.o from here ...
27884         (OBJS-libcommon): ... to here.
27886 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27888         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
27889         (ISA_ALL_CRYPTO): New macro.
27890         (ISA_ALL_SIMD): New macro
27891         (ISA_ALL_FP): New macro.
27892         * config/arm/arm.c (fpu_bitlist): Update initializer.
27893         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
27894         simd or fp.
27895         (arm9e): Add fpu.  Add option for nofp
27896         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
27897         (arm926ej-s, arm1026ej-s): Likewise.
27898         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
27899         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
27900         neon-fp16, neon-vfpv4, nofp and nosimd.
27901         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
27902         (cortex-a8): Add fpu.  Add option for nofp.
27903         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
27904         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
27905         (cortex-r4f): Add fpu.
27906         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
27907         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
27908         for nofp.
27909         (cortex-r8): Likewise.
27910         (cortex-m4): Add fpu.  Add option for nofp.
27911         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
27912         (cortex-a17.cortex-a7): Likewise.
27913         (cortex-a32): Add fpu.  Add options for crypto and nofp.
27914         (cortex-a35, cortex-a53): Likewise.
27915         (cortex-a57): Add fpu.  Add option for crypto.
27916         (cortex-a72, cortex-a73): Likewise.
27917         (exynos-m1): Likewise.
27918         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
27919         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
27920         (cortex-m33): Add fpu.  Add option for nofp.
27921         * config/arm/arm-cpu-cdata.h: Regenerated
27922         * config/arm/arm-cpu-data.h: Regenerated.
27924 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27926         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
27927         (armv5te, armv5tej): Likewise.
27928         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
27929         (armv7): Add options fp and vfpv3-d16.
27930         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
27931         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
27932         nofp and nosimd.
27933         (armv7ve): Likewise.
27934         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
27935         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
27936         (armv8-a): Add nocrypto option.
27937         (armv8.1-a, armv8.2-a): Likewise.
27938         (armv8-m.main): add options fp, fp.dp and nofp.
27940 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27942         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
27943         nofp.
27944         (armv8-a+crc): Delete.
27945         (armv8.1-a): Add options simd, crypto and nofp.
27946         (armv8.2-a): Add options fp16, simd, crypto and nofp.
27947         (armv8.2-a+fp16): Delete.
27948         (armv8-m.main): Add option dsp.
27949         (armv8-m.main+dsp): Delete.
27950         (cortex-a8): Add fpu.  Add nofp option.
27951         (cortex-a9): Add fpu.  Add nofp and nosimd options.
27952         * config/arm/parsecpu.awk (gen_data): Generate option tables and
27953         link to main cpu and architecture data structures.
27954         (gen_comm_data): Only put isa attributes from the main architecture
27955         in common tables.
27956         (option): New statement for architecture and CPU entries.
27957         * arm.c (struct cpu_option): New structure.
27958         (struct processors): Add entry for options.
27959         (arm_unrecognized_feature): New function.
27960         (arm_parse_arch_cpu_name): Ignore any characters after the first
27961         '+' character.
27962         (arm_parse_arch_cpu_feature): New function.
27963         (arm_configure_build_target): Separate out any CPU and architecture
27964         features and parse separately.  Don't error out if -mfpu=auto is
27965         used with only an architecture string.
27966         (arm_print_asm_arch_directives): New function.
27967         (arm_file_start): Call it.
27968         * config/arm/arm-cpu-cdata.h: Regenerated.
27969         * config/arm/arm-cpu-data.h: Likewise.
27970         * config/arm/arm-tables.opt: Likewise.
27972 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27974         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
27975         assembler when it is not -mfpu=auto.
27977 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27979         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
27980         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
27981         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
27982         (ASM_CPU_SPEC): Rewrite.
27983         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
27984         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
27985         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
27986         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
27987         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
27988         copied string is NUL-terminated.  Also strip any characters prefixed
27989         by '+'.
27990         (arm_rewrite_selected_arch): New function.
27991         (arm_rewrite_march): New function.
27993 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27995         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
27996         (x_arm_cpu_string, x_arm_tune_string): Likewise.
27997         (march, mcpu, mtune): Convert to string-based options.
27998         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
27999         (arm_parse_arch_cpu_name): New function.
28000         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28001         identify selected architecture or CPU.
28002         (arm_option_save): New function.
28003         (TARGET_OPTION_SAVE): Redefine.
28004         (arm_option_restore): Restore string options.
28005         (arm_option_print): Print string options.
28007 2017-06-16  Martin Sebor  <msebor@redhat.com>
28009         PR tree-optimization/80933
28010         PR tree-optimization/80934
28011         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28012         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28013         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28014         (gimple_fold_builtin): Call them.
28016 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28018         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28019         as unlikely; update profile.
28021 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28023         * predict.c (force_edge_cold): Handle declaring edges impossible
28024         more aggresively.
28026 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28028         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28029         profile.
28030         (try_unroll_loop_completely): Fix reporting.
28032 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28034         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28036 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28038         PR target/71778
28039         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28040         if given a non-constant argument for an intrinsic which requires a
28041         constant.
28043 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28045         * profile.c (compare_freqs): New function.
28046         (branch_prob): Sort edge list.
28047         (find_spanning_tree): Assume that the list is priority sorted.
28049 2017-06-16  Richard Biener  <rguenther@suse.de>
28051         PR tree-optimization/81090
28052         * passes.def (pass_record_bounds): Remove.
28053         * tree-pass.h (make_pass_record_bounds): Likewise.
28054         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28055         make_pass_record_bounds): Likewise.
28056         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28057         not free niter estimates at the beginning but at the end.
28058         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28060 2017-06-16  Richard Biener  <rguenther@suse.de>
28062         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28063         initializer to workaround ICE in host GCC 4.8.
28065 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28067         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28068         counts.
28069         (clone_inlined_nodes): Update.
28071 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28073         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28074         prefetch settings, and enable prefetching by default at -O3.
28076 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28078         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28079         Set flag_prefetch_loop_arrays according to tuning data.
28081 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28083         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28084         New tune structure.
28085         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28086         [Unrelated to main purpose of the patch] Place the pointer field last
28087         to enable type checking errors when tune structure are wrongly merged.
28088         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28089         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28090         (thunderx2t99_prefetch_tune): New tune constants.
28091         (tune_params *_tunings): Update all tunings (no functional change).
28092         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28093         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28094         from tunings structures.
28096 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28098         PR sanitizer/81094
28099         * ubsan.c (instrument_null): Add T argument, use it instead
28100         of computing it based on IS_LHS.
28101         (instrument_object_size): Likewise.
28102         (pass_ubsan::execute): Adjust instrument_null and
28103         instrument_object_size callers to pass gimple_get_lhs or
28104         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28105         calling get_base_address and instrument_mem_ref.  Handle
28106         aggregate call arguments for object-size sanitization.
28108 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28110         PR tree-optimization/81089
28111         * tree-vrp.c (is_masked_range_test): Validate operands of
28112         subexpression.
28114 2017-06-15  Martin Sebor  <msebor@redhat.com>
28116         PR c++/80560
28117         * dumpfile.c (dump_register): Avoid calling memset to initialize
28118         a class with a default ctor.
28119         * gcc.c (struct compiler): Remove const qualification.
28120         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28121         * hash-table.h: Ditto.
28122         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28123           assignment.
28124         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28125         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28126         default ctor.
28127         * params.h (struct param_info): Make struct members non-const.
28128         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28129         with default initialization.
28130         * vec.h (vec_copy_construct, vec_default_construct): New helper
28131         functions.
28132         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28133         with vec_copy_construct.
28134         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28135         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28136         * doc/invoke.texi (-Wclass-memaccess): Document.
28138 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28140         * emit-rtl.h (is_leaf): Update comment about local
28141         register allocator.
28143 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28145         PR target/78818
28146         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28147         for a variable to have a section before checking if the section has a
28148         name.
28149         Set section to.persistent if persistent attribute is set.
28150         Warn if .persistent attribute is used on an automatic variable.
28152 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28154         PR rtl-optimization/80474
28155         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28157 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28159         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28160         of REGNO.
28162 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28164         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28165         (casesi): Emit bounds checking as RTL.
28166         (casesi_internal_mips16_<mode>): Remove bounds checking.
28168 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28170         * config/xtensa/xtensa.c (xtensa_option_override): Append
28171         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28172         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28173          xtensa_doloop_hooks): Define unconditionally.
28174         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28175         TARGET_LOOPS.
28176         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28177         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28178         for it in xtensa_option_override.
28179         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28180          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28182 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28184         * doc/cppopts.texi: Document '-' special value to -MF.
28186 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28188         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28189         (cortex_a53_fconst): Likewise.
28190         (cortex_a53_fpmul): Likewise.
28191         (cortex_a53_f_load_64): Likewise.
28192         (cortex_a53_f_load_many): Likewise.
28193         (cortex_a53_advsimd_alu): Likewise.
28194         (cortex_a53_advsimd_alu_q): Likewise.
28195         (cortex_a53_advsimd_mul): Likewise.
28196         (cortex_a53_advsimd_mul_q): Likewise.
28197         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28198         Add missing fmul, r2f_cvt and fconst cases.
28200 2017-06-14  Richard Biener  <rguenther@suse.de>
28202         PR middle-end/81088
28203         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28204         literal constants.
28205         (fold_binary_loc): When associating do not treat pre-existing
28206         TREE_OVERFLOW on literal constants as a reason to allow
28207         TREE_OVERFLOW on associated literal constants.
28209 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28211         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28212         (MASK_FEATURES): New macro.
28213         * config/sparc/sparc.c (sparc_option_override): Remove the special
28214         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28216 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28218         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28219         a division of 0 if non-call exceptions are enabled.
28221 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28222             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28224         PR target/71663
28225         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28226         Improve vector initialization code gen for only variable case.
28228 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28230         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28232 2017-06-14  Richard Biener  <rguenther@suse.de>
28234         PR tree-optimization/81083
28235         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28236         as values.
28238 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28240         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28241         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28242         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28243         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28244         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28245         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28247 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28249         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28250         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28252 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28254         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28256 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28258         * config/rs6000/t-rtems: Don't handle SPE.
28260 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28262         * config/rs6000/t-linux: Don't handle SPE.
28264 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28266         * config/rs6000/eabispe.h: Delete file.
28268 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28270         * config/rs6000/t-spe: Delete file.
28272 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28274         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28275         (rs6000_legitimate_offset_address_p): Return false for anything in
28276         V2SImode or V2SFmode.
28278 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28280         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28281         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28282         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28283         and V4HImode.
28284         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28285         (rs6000_legitimate_offset_address_p): Ditto.
28286         (rs6000_emit_move): Ditto.
28287         (rs6000_init_builtins): Remove V4HI_type_node.
28289 2017-06-13  Martin Liska  <mliska@suse.cz>
28291         PR sanitize/78204
28292         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28293         (gate_asan): Likewise.
28294         * asan.h (asan_no_sanitize_address_p): Remove the function.
28295         (sanitize_flags_p): New function.
28296         * builtins.def: Fix coding style.
28297         * common.opt: Use renamed enum value.
28298         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28299         * doc/extend.texi: Document no_sanitize attribute.
28300         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28301         to SANITIZE_UNDEFINED_NONDEFAULT.
28302         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28303         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28304         Use sanitize_flags_p.
28305         * gimplify.c (gimplify_function_tree): Likewise.
28306         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28307         * opts.c (parse_no_sanitize_attribute): New function.
28308         (common_handle_option): Use renamed enum value.
28309         * opts.h (parse_no_sanitize_attribute): Declare.
28310         * tree.c (sanitize_flags_p): New function.
28311         * tree.h: Declared here.
28312         * tsan.c: Use sanitize_flags_p.
28313         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28314         (instrument_mem_ref): Likewise.
28315         (instrument_bool_enum_load): Likewise.
28316         (do_ubsan_in_current_function): Remove the function.
28317         (pass_ubsan::execute): Use sanitize_flags_p.
28318         * ubsan.h: Remove do_ubsan_in_current_function
28319         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28320         (dump_function_to_file): Use it here.
28322 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28324         PR tree-optimization/80803
28325         PR tree-optimization/81063
28326         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28327         (propagate_subaccesses_across_link): Enqueue subtree whenever
28328         necessary instead of relying on the caller.
28330 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28332         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28333         that have a first_link.
28334         (sort_and_splice_var_accesses): Do not check first_link before
28335         enquing.
28336         (subtree_mark_written_and_enqueue): Likewise.
28337         (propagate_all_subaccesses): Likewise and do not stop at first
28338         parent with a first_link.
28340 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28342         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28343         instead of f.
28345 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28347         * match.pd: New pattern.
28349 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28351         * tree-vrp.c (is_masked_range_test): New function.
28352         (register_edge_assert_for): Determine ranges for
28353         some bit tests.
28355 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28357         PR tree-optimization/67328
28358         * fold-const.c (maskable_range_p): New function.
28359         (build_range_check): Generate bittests if possible.
28361 2017-06-13  Martin Liska  <mliska@suse.cz>
28363         * gimple-pretty-print.c (dump_probability): Add new argument.
28364         (dump_edge_probability): Dump both probability and count.
28365         (dump_gimple_label): Likewise.
28366         (dump_gimple_bb_header): Likewise.
28368 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28370         PR target/81072
28371         * config/avr/avr-devices.c: Fix indentation.
28372         * config/avr/gen-avr-mmcu-specs.c: Dito.
28374 2017-06-13  Richard Biener  <rguenther@suse.de>
28376         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28377         instead get vector type from stmt_info.
28378         (vectorizable_reduction): Adjust.  Remove dead code.
28380 2017-06-13  Richard Biener  <rguenther@suse.de>
28382         PR middle-end/81065
28383         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28384         case of C * (x * C2 + C3).
28385         (fold_addr_of_array_ref_difference): Properly fold index difference.
28387 2017-06-12  David S. Miller  <davem@davemloft.net>
28389         PR target/80968
28390         * config/sparc/sparc.md (return expander): Emit frame blockage if
28391         function uses alloca.
28393 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28395         * combine.c (make_field_assignment): Check len rather than the mode
28396         precision when calling force_to_mode.
28398 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28400         Support multilibs and devices that see flash in RAM address range.
28402         PR target/81072
28403         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28404         (avr_mcu_t) <flash_pm_offset>: New field.
28405         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28406         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28407         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28408         (AVR_TINY_PM_OFFSET): Remove macro.
28409         * config/avr/avr.opt (-mshort-calls): New option.
28410         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28411         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28412         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28413         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28414         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28415         instead of avr_arch->have_jmp_call.
28416         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28417         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28418         avr_arch->flash_pm_offset to define.
28419         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28420         new field flash_pm_offset.  Add entry for avrxmega3.
28421         (avr_texinfo): Add entry for avrxmega3.
28422         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28423         attiny212, attiny214,
28424         attiny412, attiny414, attiny416, attiny417,
28425         attiny814, attiny816, attiny817,
28426         attiny1614, attiny1616, attiny1617,
28427         attiny3214, attiny3216, attiny3217.
28428         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28429         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28430         (avr_print_operand_address) [AVR_TINY]: Same.
28431         (avr_asm_init_sections) <readonly_data_section>: Only patch
28432         callback if avr_arch->flash_pm_offset = 0.
28433         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28434         for rodata if avr_arch->flash_pm_offset != 0.
28435         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28436         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28437         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28438         (m_options): Append opt_rcall.
28439         (m_dirnames): Append dir_rcall.
28440         * config/avr/t-multilib: Regenerate.
28442         * configure.ac [target=avr]: Check whether avrxmega3 default
28443         linker description file works as needed.
28444         * configure: Regenerate.
28445         * doc/avr-mmcu.texi: Regenerate.
28446         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28447         <__AVR_ARCH__>: Document avrxmega3 and 103.
28448         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28449         <__AVR_SHORT_CALLS__>: Document it.
28450         <__AVR_PM_BASE_ADDRESS__>: Document it.
28451         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28452         (AVR Variable Attributes) <progmem>: Document this is
28453         not needed for avrxmega3.
28454         (AVR Named Address Spaces) <__flash>: Dito.
28456 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28458         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28460 2017-06-12  Doug Rupp  <rupp@adacore.com>
28462         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28463         Append vxworks-stdint.h to the tm_file list.
28464         * config/vxworks-stdint.h: New file.
28466 2017-06-12  Martin Liska  <mliska@suse.cz>
28468         PR tree-optimization/81041
28469         * tree-profile.c (gimple_gen_ic_func_profiler):
28470         Create an extra BB in profile-generate
28471         (gimple_gen_time_profiler): Likewise.
28473 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28475         PR tree-optimization/81003
28476         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28477         (update_range_test): Use it instead of force_gimple_operand_gsi.
28479 2017-06-12  Richard Biener  <rguenther@suse.de>
28481         PR tree-optimization/81053
28482         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28483         with backedge value not defined in loop.  Simplify def stmt
28484         compute.
28486 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28488         PR target/79939
28489         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28490         Return true.
28491         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28492         nvptx_cannot_force_const_mem.
28494 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28496         * opts.c (finish_options): Move test for flag_split_stack after
28497         it has been initialized.
28499 2017-06-11  Jason Merrill  <jason@redhat.com>
28501         * tree.h (id_equal): New.
28502         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28503         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28504         instead of strcmp of IDENTIFIER_POINTER.
28506 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28508         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28509         (mark_all_inlined_calls_cdtor): Fix formating.
28510         (inline_transform): Rescale profile before inlining.
28512 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28514         * cgraph.h (cgraph_edge::clone): Update prototype.
28515         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28516         (cgraph_node::create_clone): Update.
28517         (cgraph_node::create_version_clone): Update.
28518         * tree-inline.c (copy_bb): Update.
28519         (expand_call_inline): Update.
28521 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28523         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28524         factored out from ...
28525         (rs6000_emit_prologue): ... here.
28527 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28529         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28530         factored out from ...
28531         (rs6000_emit_prologue): ... here.
28533 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28535         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28536         edge counts.
28537         (handle_missing_profiles): Fix computation of tp_first_run.
28538         (counts_to_freqs): Do not touch freqs when count is 0.
28540 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28542         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28543         profile.
28545 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28547         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28548         attributes): Document signal effective target.
28550 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28552         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28553         Document effective target stack_size.
28555 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28557         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28558         to the edit_context if they can be auto-applied.
28560 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28562         * opts.c (finish_options): If -fsplit-stack, disable implicit
28563         -forder-blocks-and-partition.
28564         * doc/invoke.texi (Optimize Options): Document that when using
28565         -fsplit-stack -forder-blocks-and-partition is not implicitly
28566         enabled.
28568 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28570         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28571         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28572         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28573         * builtins.def (abort, trap, unreachable): Declare cold.
28574         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28575         * tree-core.h (ECF_COLD): New.
28576         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28577         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28579 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28581         * predict.c (unlikely_executed_stmt_p): Cleanup.
28583 2017-06-09  Richard Biener  <rguenther@suse.de>
28585         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28586         model if the ref is always written to.
28588 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28590         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28592 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28594         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28595         than udiv.
28597 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28599         PR target/80855
28600         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28601         "target cannot support label values" when encountering LABEL_REF.
28603 2017-06-09  Martin Liska  <mliska@suse.cz>
28605         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28606         (gimple_gen_ic_func_profiler): Emit direct comparison
28607         of __gcov_indirect_call_callee with NULL.
28608         (gimple_gen_time_profiler): Change probability from
28609         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28611 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28613         * profile.c (edge_gcov_counts): Turn to pointer.
28614         (compute_branch_probabilities, compute_branch_probabilities): Update.
28615         (branch_prob): Do not clear edge_gcov_count.
28616         * profile.h (edge_gcov_counts): Turn to pointer.
28617         (edge_gcov_count): Update.
28619 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28621         * gimple.h (gimple_check_failed): Mark cold.
28623 2017-06-09  Richard Biener  <rguenther@suse.de>
28625         PR tree-optimization/66623
28626         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28627         refactor check_reduction into two parts, properly computing
28628         whether we have to check reduction validity for outer loop
28629         vectorization.
28631 2017-06-09  Richard Biener  <rguenther@suse.de>
28633         PR tree-optimization/79483
28634         * graphite-scop-detection.c (order): New global.
28635         (get_order): Compute bb to order mapping that satisfies code
28636         generation constraints.
28637         (cmp_pbbs): New helper.
28638         (build_scops): Start domwalk at entry block, sort generated
28639         pbbs.
28641 2017-06-09  Richard Biener  <rguenther@suse.de>
28643         PR middle-end/81007
28644         * ipa-polymorphic-call.c
28645         (ipa_polymorphic_call_context::restrict_to_inner_class):
28646         Skip FIELD_DECLs with error_mark_node type.
28647         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28648         last again.
28650 2017-06-09  Martin Liska  <mliska@suse.cz>
28652         * predict.c (struct branch_predictor): New struct.
28653         (test_prediction_value_range): New test.
28654         (predict_c_tests): New function.
28655         * selftest-run-tests.c (selftest::run_tests): Run the function.
28656         * selftest.h: Declare new tests.
28658 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28660         PR target/80966
28661         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28662         gen_add3_insn did not fail.
28663         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28664         r0, construct that number in a temporary reg and add that reg to r0.
28665         If asked to put the result in r0 as well, fail.
28667 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28669         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28670         for early expansion of vec_eqv.
28672 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28674         PR middle-end/81005
28675         * ubsan.c (instrument_null): Avoid pointless code temporary.
28676         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28678 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28680         PR target/81015
28681         Revert:
28682         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28684         PR target/59874
28685         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28686         (*clzhi2): Ditto.
28688 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28690         * predict.c (unlikely_executed_edge_p): Move ahead.
28691         (probably_never_executed_edge_p): Use it.
28693 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28695         PR middle-end/79988
28696         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28697         gimple_call_builtin_p call.
28699 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28701         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28702         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28703         rtl_check_failed_type2, rtl_check_failed_code1,
28704         rtl_check_failed_code2, rtl_check_failed_code_mode,
28705         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28706         rtvec_check_failed_bounds, rtl_check_failed_flag,
28707         _fatal_insn_not_found, _fatal_insn): Likewise.
28708         * tree.h (tree_contains_struct_check_failed,
28709         tree_check_failed, tree_not_check_failed,
28710         tree_class_check_failed, tree_range_check_failed,
28711         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28712         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28713         tree_operand_check_failed, omp_clause_check_failed,
28714         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28715         Likewise.
28717 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28719         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28720         flag_branch_probabilities.
28721         * ipa-inline.c (edge_badness): Likewise.
28722         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28723         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28724         * predict.c (maybe_hot_frequency_p): Likewise.
28725         (probably_never_executed): Likewise.
28726         * sched-ebb.c (schedule_ebbs): Likewise.
28727         * sched-rgn.c (find_single_block_region): Likewise.
28728         * tracer.c (tail_duplicate): Likewise.
28730 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28732         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28733         longer requires x_flag_profile_use.
28735 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28737         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28738         instead of flag_reorder_blocks_and_partition.
28739         * dbxout.c (dbxout_function_end): Likewise.
28740         * dwarf2out.c (gen_subprogram_die): Likewise.
28741         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28742         * hw-doloop.c (reorg_loops): Likewise.
28743         * varasm.c (assemble_start_function,
28744         assemble_end_function): Likewise.
28745         (decide_function_section): Do not check for
28746         flag_reorder_blocks_and_partition.
28748 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28750         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28751         New function.
28752         (chkp_get_hard_register_fake_addr_expr): Ditto.
28753         (chkp_build_addr_expr): Add check for hard reg case.
28754         (chkp_parse_array_and_component_ref): Ditto.
28755         (chkp_find_bounds_1): Ditto.
28756         (chkp_process_stmt): Don't generate bounds store for
28757         hard reg case.
28759 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28761         * predict.c (maybe_hot_bb_p): Do not check profile status.
28762         (maybe_hot_edge_p): Likewise.
28763         (probably_never_executed): Check for zero counts even if profile
28764         is not read.
28765         (unlikely_executed_edge_p): New function.
28766         (unlikely_executed_stmt_p): New function.
28767         (unlikely_executed_bb_p): New function.
28768         (set_even_probabilities): Use unlikely predicates.
28769         (combine_predictions_for_bb): Likewise.
28770         (predict_paths_for_bb): Likewise.
28771         (predict_paths_leading_to_edge): Likewise.
28772         (determine_unlikely_bbs): New function.
28773         (estimate_bb_frequencies): Use it.
28774         (compute_function_frequency): Use zero counts even if profile is
28775         not read.
28776         * profile-count.h: Fix typo.
28778 2017-08-08  Julia Koval  <julia.koval@intel.com>
28780         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28781         _mm512_mask_cvtsepi16_storeu_epi8,
28782         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28783         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28784         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28785         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28786         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28787         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28788         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28789         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28790         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28791         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28792         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28793         __builtin_ia32_pmovuswb256mem_mask,
28794         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28795         __builtin_ia32_pmovwb512mem_mask): New builtins.
28797 2017-08-08  Julia Koval  <julia.koval@intel.com>
28799         PR target/73350,80862
28800         * config/i386/subst.md (round): Fix round pattern.
28801         * config/i386/i386.c (ix86_erase_embedded_rounding):
28802         Fix erasing rounding for the fixed pattern.
28804 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28806         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28808 2017-06-08  Martin Liska  <mliska@suse.cz>
28810         PR gcov-profile/80911
28811         * gcov.c (block_info::block_info): New constructor.
28813 2017-06-07  Carl Love  <cel@us.ibm.com>
28815         * config/rs6000/rs6000-c: The return type of the following
28816         built-in functions was implemented as int not long long.  Fix sign
28817         of return value for the unsigned version of vec_mulo and vec_mule.
28818         vector unsigned long long vec_bperm (vector unsigned long long,
28819                                              vector unsigned char)
28820         vector signed long long vec_mule (vector signed int,
28821                                           vector signed int)
28822         vector unsigned long long vec_mule (vector unsigned int,
28823                                             vector unsigned int)
28824         vector signed long long vec_mulo (vector signed int,
28825                                           vector signed int)
28826         vector unsigned long long vec_mulo (vector unsigned int,
28827                                             vector unsigned int)
28828         * doc/extend.texi: Fix the documentation for the built-in
28829         functions.
28831 2017-06-07  Carl Love  <cel@us.ibm.com>
28833         PR target/80982
28834         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28835         for BE.
28837 2017-06-07  Carl Love  <cel@us.ibm.com>
28839         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28840         support, Generate       doublehv for signed int/float for BE case only.
28842 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28844         * doc/invoke.texi (mcx16): Rewrite.
28846 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28848         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28849         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28850         *mov<mode>_softfloat, and an anonymous splitter): Use
28851         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28853 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28855         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28856         SPEFSCR registers.
28857         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28858         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28859         (rs6000_debug_reg_global): Adjust.
28860         (rs6000_init_hard_regno_mode_ok): Adjust.
28861         (rs6000_dbx_register_number): Adjust.
28862         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28863         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28864         Remove SPE_ACC and SPEFSCR.
28865         (REG_ALLOC_ORDER): Ditto.
28866         (FRAME_POINTER_REGNUM): Change to 111.
28867         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
28868         (REG_CLASS_NAMES): Ditto.
28869         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
28870         (REGISTER_NAMES): Ditto.
28871         (ADDITIONAL_REG_NAMES): Ditto.
28872         (rs6000_reg_names): Ditto.
28873         * config/rs6000/rs6000.md: Renumber some register number
28874         define_constants.
28876 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28878         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
28879         registers.
28880         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28881         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
28882         to 117.
28883         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
28884         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28885         Delete the SPE high registers.
28886         (REG_ALLOC_ORDER): Ditto.
28887         (enum reg_class): Remove SPE_HIGH_REGS.
28888         (REG_CLASS_NAMES): Ditto.
28889         (REG_CLASS_CONTENTS): Delete the SPE high registers.
28890         (REGISTER_NAMES): Ditto.
28891         (rs6000_reg_names): Ditto.
28892         * doc/tm.texi.in: Remove SPE as example.
28893         * doc/tm.texi: Regenerate.
28895 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28897         * config/rs6000/8540.md (ppc8540_brinc): Delete.
28898         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
28899         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
28900         * config/rs6000/rs6000.md (type): Remove "brinc".
28902 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28904         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
28905         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
28906         * config/rs6000/linuxspe.h: Delete file.
28907         * config/rs6000/rs6000.md: Don't include spe.md.
28908         * config/rs6000/spe.h: Delete file.
28909         * config/rs6000/spe.md: Delete file.
28910         * config/rs6000/t-rs6000: Remove spe.md.
28912 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28914         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
28915         (reg_or_none500mem_operand): Delete.
28916         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
28917         instead of reg_or_none500mem_operand.
28919 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28921         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28922         handling of SPE flags.
28923         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
28925 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28927         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
28928         SPE ABI handling.
28929         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
28930         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
28931         paired_divv2sf3): Similar.
28932         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
28933         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
28934         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
28935         RS6000_BUILTIN_S.
28936         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
28937         Rename the paired_* instruction patterns.
28938         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
28939         define __SPE__.
28940         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
28941         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
28942         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
28943         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
28944         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
28945         PAIRED_VECTOR_MODE.
28946         (struct machine_function): Delete field spe_insn_chain_scanned_p.
28947         (spe_func_has_64bit_regs_p): Delete.
28948         (spe_expand_predicate_builtin): Delete.
28949         (spe_expand_evsel_builtin): Delete.
28950         (TARGET_DWARF_REGISTER_SPAN): Do not define.
28951         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
28952         (invalid_e500_subreg): Delete.
28953         (rs6000_legitimize_address): Always force_reg op2 as well, for
28954         paired single memory accesses.
28955         (rs6000_member_type_forces_blk): Delete.
28956         (rs6000_spe_function_arg): Delete.
28957         (rs6000_expand_unop_builtin): Delete SPE handling.
28958         (rs6000_expand_binop_builtin): Ditto.
28959         (spe_expand_stv_builtin): Delete.
28960         (bdesc_2arg_spe): Delete.
28961         (spe_expand_builtin): Delete.
28962         (spe_expand_predicate_builtin): Delete.
28963         (spe_expand_evsel_builtin): Delete.
28964         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
28965         (spe_init_builtins): Delete.
28966         (spe_func_has_64bit_regs_p): Delete.
28967         (savres_routine_name): Delete "info" parameter.  Adjust callers.
28968         (rs6000_emit_stack_reset): Ditto.
28969         (rs6000_dwarf_register_span): Delete.
28970         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
28971         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
28972         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
28973         Delete.
28974         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
28975         Delete.
28976         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
28977         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
28978         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
28979         mulv2sf3, divv2sf3): Delete expanders.
28981 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28983         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
28985 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28987         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
28988         * config/rs6000/rs6000.c: Ditto.
28990 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28992         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
28993         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
28994         comparison_operator.
28996 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28998         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
28999         * config/rs6000/rs6000.opt: Ditto.
29000         * config/rs6000/t-rtems: Ditto.
29002 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29004         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29005         TARGET_E500_SINGLE by 0, simplify.
29006         * config/rs6000/rs6000.c: Ditto.
29007         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29008         (spe_build_register_parallel): Delete.
29009         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29010         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29011         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29012         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29013         (E500_CONVERT): Delete.
29014         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29016 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29018         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29019         * config/rs6000/dfp.md: Ditto.
29020         (negdd2, *negdd2_fpr): Merge.
29021         (absdd2, *absdd2_fpr): Merge.
29022         (negtd2, *negtd2_fpr): Merge.
29023         (abstd2, *abstd2_fpr): Merge.
29024         * config/rs6000/e500.h: Delete file.
29025         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29026         TARGET_FPRS by 1 and simplify.
29027         * config/rs6000/rs6000-c.c: Ditto.
29028         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29029         TARGET_DF_SPE by 0.
29030         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29031         TARGET_DF_SPE.
29032         * config/rs6000/rs6000.md: Ditto.
29033         (floatdidf2, *floatdidf2_fpr): Merge.
29034         (move_from_CR_gt_bit): Delete.
29035         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29036         (E500_CR_IOR_COMPARE): Delete.
29037         (All patterns that require !TARGET_FPRS): Delete.
29038         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29040 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29042         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29044 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29046         * graphds.c (add_edge): Intitialize edge's attached data.
29047         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29048         pointer parameter.  Call pointed function on each edge during
29049         graph traversing.  Skip traversing the edge when the function
29050         returns true.
29051         (graphds_dfs, graphds_scc): Ditto.
29052         (for_each_edge): New parameter.  Pass the new parameter to callback
29053         function.
29054         * graphds.h (skip_edge_callback): New function pointer type.
29055         (graphds_dfs, graphds_scc): New function pointer parameter.
29056         (graphds_edge_callback, for_each_edge): New parameter.
29058 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29060         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29061         out code checking if runtime alias check is possible to below ...
29062         Call the new function.
29063         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29064         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29066 2017-06-07  Marek Polacek  <polacek@redhat.com>
29068         PR sanitizer/80932
29069         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29070         TYPE_OVERFLOW_WRAPS check.
29072 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29074         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29075         if versioning is required.
29076         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29077         peeling with the check for versioning.
29079 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29081         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29082         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29083         Set true to new parameter if new ssa variable is defined.
29084         (vect_gen_vector_loop_niters): Refactor.  Set range information
29085         for the new vector loop bound variable.
29086         (vect_do_peeling): Ditto.
29088 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29090         * tree-affine.c (ssa.h): Include header file.
29091         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29092         has wrapping overflow behavior.
29094 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29096         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29098 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29100         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29101         (tree_to_aff_combination): ... here.
29103 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29105         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29106         reg_pressure model function.
29107         (ivopts_global_cost_for_size): Delete.
29108         (determine_set_costs, iv_ca_recount_cost): Call new model function
29109         ivopts_estimate_reg_pressure.
29111 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29113         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29114         expensive than udiv.  Remove floating point cases from mod.
29116 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29118         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29119         Increase idiv cost.
29121 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29123         * config/aarch64/aarch64.md
29124         (copysignsf3): Fix mask generation.
29126 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29128         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29129         TDI_gimple.
29130         (class dump_manager): Add register_dumps method.
29131         * dumpfile.c: Include langhooks.h.
29132         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29133         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29134         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29135         (dump_manager::dump_register): Start with 512 entries instead of 32.
29136         (dump_manager::register_dumps): New method.
29137         * toplev.c (general_init): Instead of invoking register_dumps
29138         langhook, invoke register_dumps method on the dump manager.
29139         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29140         TDI_generic.
29142 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29144         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29145         Say that # requires an associated define_split to exist, and that
29146         the define_split must be suitable for use after register allocation.
29148 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29150         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29151         (compute_outgoing_frequencies): Also initialize zero counts.
29152         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29153         around loops; preserve more of profile when nothing changes.
29155 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29157         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29158         here.
29159         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29160         * config/arm/arm-cpu-cdata.h: Regenerate.
29161         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29162         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29163         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29164         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29165         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29166         support.
29167         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29168         support.
29169         * config/arm/t-rmprofile: Likewise.
29170         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29172 2017-06-06  David S. Miller  <davem@davemloft.net>
29174         PR target/80968
29175         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29176         blockage if function uses alloca.
29178 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29180         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29181         New "uid" fields to hold pretty-print IDs of group and ref.
29182         Memory references are now identified as <group_id>:<ref_id>
29183         instead of using [random] addresses.
29184         (dump_mem_details): Simplify, no functional change.
29185         (dump_mem_ref): Simplify and make output more concise.
29186         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29187         (find_or_create_group): Initialize group uid.
29188         (record_ref): Initialize ref uid.  Improve debug output.
29189         (prune_group_by_reuse, should_issue_prefetch_p,)
29190         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29191         (mark_nontemporal_store, determine_loop_nest_reuse):
29192         Improve debug output.
29194 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29196         * dbgcnt.def (prefetch): New debug counter.
29197         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29198         (schedule_prefetches): Stop issueing prefetches if debug counter
29199         tripped.
29201 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29203         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29204         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29206 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29208         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29209         Use aarch64_reg_or_zero predicate for operand 4.
29210         (aarch64_compare_and_swap<mode> define_insn_and_split):
29211         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29212         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29214 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29216         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29217         (arm_compute_save_core_reg_mask): This.
29218         (thumb1_compute_save_reg_mask): Rename into ...
29219         (thumb1_compute_save_core_reg_mask): This.
29220         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29221         (arm_compute_frame_layout): Likewise.
29223 2017-06-06  Richard Biener  <rguenther@suse.de>
29225         PR tree-optimization/80974
29226         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29227         keep or clear leaders SSA info.
29229 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29231         * config/nvptx/nvptx.c (split_mode_p): New function.
29232         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29234 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29236         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29238 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29240         PR bootstrap/80978
29241         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29242         profile.
29244 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29246         * shrink-wrap.c (handle_simple_exit): Update profile.
29247         (try_shrink_wrapping): Upate profile.
29249 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29251         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29252         (tree_guess_outgoing_edge_probabilities): New.
29253         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29254         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29256 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29258         * ipa-split.c (split_function): Initialize return bb profile.
29260 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29262         * profile.c (compute_branch_probabilities): Also initialize
29263         EXIT_BLOCK profile.
29265 2017-06-06  Richard Biener  <rguenther@suse.de>
29267         PR tree-optimization/80928
29268         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29269         (vect_analyze_loop_operations): Properly guard analysis for
29270         pure SLP case.
29271         (vect_transform_loop): Likewise.
29272         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29273         (vect_model_induction_cost): Do not cost for pure SLP.
29274         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29275         of induction in inner loop vectorization.
29276         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29277         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29278         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29279         recursion.
29280         (vect_analyze_slp_cost_1): Cost induction.
29281         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29282         (vect_get_slp_vect_defs): Likewise.
29283         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29284         (vect_transform_stmt): Handle SLP reductions.
29285         * tree-vectorizer.h (vectorizable_induction): Adjust.
29287 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29289         * config/rs6000/rs6000.c (make_resolver_func): Update
29290         init_lowered_empty_function call.
29292 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29294         * doc/invoke.texi: Document the -fprofile-abs-path option.
29295         * common.opt (fprofile-abs-path): New option.
29296         * gcov-io.h (gcov_write_filename): Declare.
29297         * gcov-io.c (gcov_write_filename): New function.
29298         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29299         * profile.c (output_location): Likewise.
29301 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29303         * shring-wrap.c: Revert accidental commit.
29305 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29307         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29309 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29311         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29312         new edge.
29313         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29314         profile in callgraph edge.
29315         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29316         (apply_scale): Likewise.
29317         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29318         Also scale profile when inlining function with zero profile.
29319         (initialize_cfun): Update exit block profile even when it is zero.
29320         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29321         when profile is read.
29323 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29325         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29326         (CLONE_*): New constants to define the processors we can generate
29327         code for with the target_clone attribute.
29328         (rs6000_clone_map): New array to identify which clone processors
29329         the current program is running on.
29330         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29331         target_clone attribute.
29332         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29333         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29334         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29335         (cpu_expand_builtin): Add support for target_clone attribute.
29336         (rs6000_valid_attribute_p): Allow "default" attribute.
29337         (get_decl_name): New debug function to simplify printing the
29338         current function name in debugging statements.
29339         (rs6000_clone_priority): New functions to support the target_clone
29340         attribute, and be able to generate code to switch between ISA 2.05
29341         through ISA 3.0 (power6 through power9).
29342         (rs6000_compare_version_priority): Likewise.
29343         (rs6000_get_function_versions_dispatcher): Likewise.
29344         (make_resolver_func): Likewise.
29345         (add_condition_to_bb): Likewise.
29346         (dispatch_function_versions): Likewise.
29347         (rs6000_generate_version_dispatcher_body): Likewise.
29348         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29349         (fusion_gpr_load_p): Fix a spacing issue.
29350         * doc/extend.texi (Common Function Attributes): Document that the
29351         PowerPC supports the target_clone attribute.
29353 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29355         * config/arm/arm.h: explain F symbol found in description of ARM
29356         register allocation in its legend.
29358 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29360         * config/mips/frame-header-opt.c: Include profile-count.h.
29361         * config/riscv/riscv.c: Include profile-count.h
29363 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29365         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29366         update profile.
29367         (sm_set_flag_if_changed): Add bbs field.
29368         (execute_sm_if_changed_flag_set): Pass BBS.
29369         (execute_sm): Update.
29371 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29373         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29374         New pattern.
29376 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29378         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29379         (peephole2): New peephole2 to emit the above.
29380         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29382 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29384         * config/aarch64/aarch64.c (define_peephole2 above
29385         *sub_<shift>_<mode>): New peephole.
29387 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29389         * config/i386/i386.c (make_resolver_func): Update.
29390         * Makefile.in: Add profile-count.h and profile-count.o
29391         * auto-profile.c (afdo_indirect_call): Update to new API.
29392         (afdo_set_bb_count): Update.
29393         (afdo_propagate_edge): Update.
29394         (afdo_propagate_circuit): Update.
29395         (afdo_calculate_branch_prob): Update.
29396         (afdo_annotate_cfg): Update.
29397         * basic-block.h: Include profile-count.h
29398         (struct edge_def): Turn count to profile_count.
29399         (struct basic_block_def): Likewie.
29400         (REG_BR_PROB_BASE): Move to profile-count.h
29401         (RDIV): Move to profile-count.h
29402         * bb-reorder.c (max_entry_count): Turn to profile_count.
29403         (find_traces): Update.
29404         (rotate_loop):Update.
29405         (connect_traces):Update.
29406         (sanitize_hot_paths):Update.
29407         * bt-load.c (migrate_btr_defs): Update.
29408         * cfg.c (RDIV): Remove.
29409         (init_flow): Use alloc_block.
29410         (alloc_block): Uninitialize count.
29411         (unchecked_make_edge): Uninitialize count.
29412         (check_bb_profile): Update.
29413         (dump_edge_info): Update.
29414         (dump_bb_info): Update.
29415         (update_bb_profile_for_threading): Update.
29416         (scale_bbs_frequencies_int): Update.
29417         (scale_bbs_frequencies_gcov_type): Update.
29418         (scale_bbs_frequencies_profile_count): New.
29419         * cfg.h (update_bb_profile_for_threading): Update.
29420         (scale_bbs_frequencies_profile_count): Declare.
29421         * cfgbuild.c (compute_outgoing_frequencies): Update.
29422         (find_many_sub_basic_blocks): Update.
29423         * cfgcleanup.c (try_forward_edges): Update.
29424         (try_crossjump_to_edge): Update.
29425         * cfgexpand.c (expand_gimple_tailcall): Update.
29426         (construct_exit_block): Update.
29427         * cfghooks.c (verify_flow_info): Update.
29428         (dump_bb_for_graph): Update.
29429         (split_edge): Update.
29430         (make_forwarder_block): Update.
29431         (duplicate_block): Update.
29432         (account_profile_record): Update.
29433         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29434         (get_estimated_loop_iterations): Update.
29435         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29436         (single_likely_exit): Update.
29437         * cfgloopmanip.c (scale_loop_profile): Update.
29438         (loopify): Update.
29439         (set_zero_probability): Update.
29440         (lv_adjust_loop_entry_edge): Update.
29441         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29442         (purge_dead_edges): Update.
29443         (rtl_account_profile_record): Update.
29444         * cgraph.c (cgraph_node::create): Uninitialize count.
29445         (symbol_table::create_edge): Uninitialize count.
29446         (cgraph_update_edges_for_call_stmt_node): Update.
29447         (cgraph_edge::dump_edge_flags): Update.
29448         (cgraph_node::dump): Update.
29449         (cgraph_edge::maybe_hot_p): Update.
29450         * cgraph.h: Include profile-count.h
29451         (create_clone), create_edge, create_indirect_edge): Update.
29452         (cgraph_node): Turn count to profile_count.
29453         (cgraph_edge0: Likewise.
29454         (make_speculative, clone): Update.
29455         (create_edge): Update.
29456         (init_lowered_empty_function): Update.
29457         * cgraphclones.c (cgraph_edge::clone): Update.
29458         (duplicate_thunk_for_node): Update.
29459         (cgraph_node::create_clone): Update.
29460         * cgraphunit.c (cgraph_node::analyze): Update.
29461         (cgraph_node::expand_thunk): Update.
29462         * final.c (dump_basic_block_info): Update.
29463         * gimple-streamer-in.c (input_bb): Update.
29464         * gimple-streamer-out.c (output_bb): Update.
29465         * graphite.c (print_global_statistics): Update.
29466         (print_graphite_scop_statistics): Update.
29467         * hsa-brig.c: Include basic-block.h.
29468         * hsa-dump.c: Include basic-block.h.
29469         * hsa-gen.c (T sum_slice): Update.
29470         (convert_switch_statements):Update.
29471         * hsa-regalloc.c: Include basic-block.h.
29472         * ipa-chkp.c (chkp_produce_thunks): Update.
29473         * ipa-cp.c (struct caller_statistics): Update.
29474         (init_caller_stats): Update.
29475         (gather_caller_stats): Update.
29476         (ipcp_cloning_candidate_p): Update.
29477         (good_cloning_opportunity_p): Update.
29478         (get_info_about_necessary_edges): Update.
29479         (dump_profile_updates): Update.
29480         (update_profiling_info): Update.
29481         (update_specialized_profile): Update.
29482         (perhaps_add_new_callers): Update.
29483         (decide_about_value): Update.
29484         (ipa_cp_c_finalize): Update.
29485         * ipa-devirt.c (struct odr_type_warn_count): Update.
29486         (struct decl_warn_count): Update.
29487         (struct final_warning_record): Update.
29488         (possible_polymorphic_call_targets): Update.
29489         (ipa_devirt): Update.
29490         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29491         * ipa-icf.c (sem_function::merge): Update.
29492         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29493         * ipa-inline.c (compute_uninlined_call_time): Update.
29494         (compute_inlined_call_time): Update.
29495         (want_inline_small_function_p): Update.
29496         (want_inline_self_recursive_call_p): Update.
29497         (edge_badness): Update.
29498         (lookup_recursive_calls): Update.
29499         (recursive_inlining): Update.
29500         (inline_small_functions): Update.
29501         (dump_overall_stats): Update.
29502         (dump_inline_stats): Update.
29503         * ipa-profile.c (ipa_profile_generate_summary): Update.
29504         (ipa_propagate_frequency): Update.
29505         (ipa_profile): Update.
29506         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29507         * ipa-utils.c (ipa_merge_profiles): Update.
29508         * loop-doloop.c (doloop_modify): Update.
29509         * loop-unroll.c (report_unroll): Update.
29510         (unroll_loop_runtime_iterations): Update.
29511         * lto-cgraph.c (lto_output_edge): Update.
29512         (lto_output_node): Update.
29513         (input_node): Update.
29514         (input_edge): Update.
29515         (merge_profile_summaries): Update.
29516         * lto-streamer-in.c (input_cfg): Update.
29517         * lto-streamer-out.c (output_cfg): Update.
29518         * mcf.c (create_fixup_graph): Update.
29519         (adjust_cfg_counts): Update.
29520         (sum_edge_counts): Update.
29521         * modulo-sched.c (sms_schedule): Update.
29522         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29523         * predict.c (maybe_hot_count_p): Update.
29524         (probably_never_executed): Update.
29525         (dump_prediction): Update.
29526         (combine_predictions_for_bb): Update.
29527         (propagate_freq): Update.
29528         (handle_missing_profiles): Update.
29529         (counts_to_freqs): Update.
29530         (rebuild_frequencies): Update.
29531         (force_edge_cold): Update.
29532         * predict.h: Include profile-count.h
29533         (maybe_hot_count_p, counts_to_freqs): UPdate.
29534         * print-rtl-function.c: Do not include cfg.h
29535         * print-rtl.c: Include basic-block.h
29536         * profile-count.c: New file.
29537         * profile-count.h: New file.
29538         * profile.c (is_edge_inconsistent): Update.
29539         (correct_negative_edge_counts): Update.
29540         (is_inconsistent): Update.
29541         (set_bb_counts): Update.
29542         (read_profile_edge_counts): Update.
29543         (compute_frequency_overlap): Update.
29544         (compute_branch_probabilities): Update; Initialize and deinitialize
29545         gcov_count tables.
29546         (branch_prob): Update.
29547         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29548         (edge_gcov_count): New.
29549         (bb_gcov_count): New.
29550         * shrink-wrap.c (try_shrink_wrapping): Update.
29551         * tracer.c (better_p): Update.
29552         * trans-mem.c (expand_transaction): Update.
29553         (ipa_tm_insert_irr_call): Update.
29554         (ipa_tm_insert_gettmclone_call): Update.
29555         * tree-call-cdce.c: Update.
29556         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29557         (gimple_duplicate_sese_tail): Update.
29558         (gimple_account_profile_record): Update.
29559         (execute_fixup_cfg): Update.
29560         * tree-inline.c (copy_bb): Update.
29561         (copy_edges_for_bb): Update.
29562         (initialize_cfun): Update.
29563         (freqs_to_counts): Update.
29564         (copy_cfg_body): Update.
29565         (expand_call_inline): Update.
29566         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29567         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29568         (try_unroll_loop_completely): Update.
29569         (try_peel_loop): Update.
29570         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29571         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29572         * tree-ssa-loop-split.c (connect_loops): Update.
29573         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29574         * tree-ssa-reassoc.c (branch_fixup): Update.
29575         * tree-ssa-tail-merge.c (replace_block_by): Update.
29576         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29577         (compute_path_counts): Update.
29578         (update_profile): Update.
29579         (recompute_probabilities): Update.
29580         (update_joiner_offpath_counts): Update.
29581         (estimated_freqs_path): Update.
29582         (freqs_to_counts_path): Update.
29583         (clear_counts_path): Update.
29584         (ssa_fix_duplicate_block_edges): Update.
29585         (duplicate_thread_path): Update.
29586         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29587         (struct switch_conv_info): Update.
29588         * tree-tailcall.c (decrease_profile): Update.
29589         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29590         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29591         * value-prof.c (check_counter): Update.
29592         (gimple_divmod_fixed_value): Update.
29593         (gimple_mod_pow2): Update.
29594         (gimple_mod_subtract): Update.
29595         (gimple_ic_transform): Update.
29596         (gimple_stringop_fixed_value): Update.
29597         * value-prof.h (gimple_ic): Update.
29599 2017-06-02  Carl Love  <cel@us.ibm.com>
29601         * config/rs6000/rs6000-c: Add support for built-in functions
29602         vector double vec_doublee (vector signed int);
29603         vector double vec_doublee (vector unsigned int);
29604         vector double vec_doublee (vector float);
29605         vector double vec_doubleh (vector signed int);
29606         vector double vec_doubleh (vector unsigned int);
29607         vector double vec_doubleh (vector float);
29608         vector double vec_doublel (vector signed int);
29609         vector double vec_doublel (vector unsigned int);
29610         vector double vec_doublel (vector float);
29611         vector double vec_doubleo (vector signed int);
29612         vector double vec_doubleo (vector unsigned int);
29613         vector double vec_doubleo (vector float);.
29614         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29615         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29616         UNS_DOUBLEL.
29617         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29618         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29619         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29620         VS_sxwsp.
29621         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29622         vec_doublel, vec_doubleh.
29623         * doc/extend.texi: Update the built-in documentation file for the
29624         new built-in functions.
29626 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29628         PR jit/80954
29629         * ipa-inline-analysis.c (free_growth_caches): Set
29630         edge_removal_hook_holder to NULL after removing it.
29632 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29634         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29635         comparision with zero.
29637 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29638         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29639         for early expansion of vec_min and vec_max builtins.
29640         (builtin_function_type): Add min/max unsigned variants to those
29641         identified as having unsigned arguments.
29643 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29645         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29647 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29649         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29650         Use VALL_F16 iterator rather than VALL.
29652 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29654         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29655         Emit CBNZ inside loop when doing a strong exchange and comparing
29656         against zero.  Generate the CC flags after the loop.
29658 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29660         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29661         (dl_section_ref): New.
29662         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29663         On AIX, append an expression to subtract the size of the
29664         section length to dl_section_ref.
29666 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29668         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29669         for early expansion of vector absolute builtins.
29671 2017-06-02  Richard Biener  <rguenther@suse.de>
29673         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29674         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29676 2017-06-02  Richard Biener  <rguenther@suse.de>
29678         PR tree-optimization/80948
29679         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29680         stmt order as well.
29682 2017-06-02  Richard Biener  <rguenther@suse.de>
29684         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29685         PHIs are ok.
29686         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29687         for inductions as relevant.
29689 2017-06-02  Richard Biener  <rguenther@suse.de>
29691         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29692         (vectorizable_induction): ... this.  Remove dead code.
29694 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29696         * builtins. (expand_builtin_alloca): Remove second parameter and
29697         infer its value from the first parameter instead.
29698         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29700 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29702         PR rtl-optimization/80903
29703         * loop-doloop.c (add_test): Unshare sequence.
29705 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29707         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29709 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29711         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29712         static.
29713         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29714         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29715         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29716         (xlogue_layout::get_stub_rtx): Make static.
29717         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29718         (xlogue_layout::compute_stub_managed_regs): Rename to...
29719         (xlogue_layout::count_stub_managed_regs): ...this.
29720         (xlogue_layout::is_stub_managed_reg): New function.
29721         (xlogue_layout::m_stub_names): Rename to...
29722         (xlogue_layout::s_stub_names): ...this, make static.
29723         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29724         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29725         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29726         xlogue_layout::s_stub_names): Instantiate statics.
29727         (stub_managed_regs): Remove.
29728         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29729         (disable_call_ms2sysv_xlogues): Rename to...
29730         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29731         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29732         warning logic.
29733         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29734         change after reload_completed.
29735         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29736         directly.
29737         (ix86_expand_prologue): Likewise.
29738         (ix86_expand_epilogue): Likewise.
29739         (ix86_expand_split_stack_prologue): Likewise.
29740         (ix86_compute_frame_layout): Remove frame parameter ...
29741         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29742         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29743         only if necessary.
29744         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29745         (ix86_frame): Move from here ...
29746         * config/i386/i386.h (ix86_frame): ... to here.
29747         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29748         complete ix86_frame data structure instead.  Remove some_ld_name.
29750 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29752         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29753         symbols that hold a DECL_VALUE_EXPR.
29755 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29757         PR tree-optimization/80898
29758         * tree-sra.c (process_subtree_disqualification): Removed.
29759         (disqualify_candidate): Do not acll
29760         process_subtree_disqualification.
29761         (subtree_mark_written_and_enqueue): New function.
29762         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29763         RHS has been disqualified and re-queue LHS if necessary.  Apart
29764         from that, ignore disqualified RHS.
29766 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29768         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29769         address fetch for z10 or later.
29771 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29773         * config/arc/arc.md (tst_movb): Add guard when splitting.
29775 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29777         * config/arc/arc.c (arc_can_eliminate): Test against
29778         arc_frame_pointer_needed.
29780 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29782         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29783         to prevent store reordering.
29784         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29785         (type): Add block type.
29786         (stack_tie): Define special instruction to be used in
29787         expand_prologue.
29789 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29791         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29792         constraint. It is not valid for the pattern.
29793         (noncommutative_binary_comparison): Likewise.
29795 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29797         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29798         scaled addresses.
29800 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29802         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29803         be used by the reg-alloc.
29805 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29807         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29808         reg-alloc when having mul64 or mul32x16 instructions.
29809         (mulsidi3): Likewise.
29810         (umulsidi3): Likewise.
29811         (mulsi32x16): New pattern.
29812         (mulsi64): Likewise.
29813         (mulsidi64): Likewise.
29814         (umulsidi64): Likewise.
29815         (MUL32x16_REG): Define.
29816         (mul64_600): Use MUL32x16_REG.
29817         (mac64_600): Likewise.
29818         (umul64_600): Likewise.
29819         (umac64_600): Likewise.
29821 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29823         * config/arc/arc.md (mulsi3_700): Make it commutative.
29825 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29827         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29828         type for movstouw.
29829         (*sign_extendsidi2_insn): Likewise for movstosw.
29831 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29833         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29834         the type of the input discriminant value.  Convert the
29835         discriminant value of signedness vary.
29837 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29839         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29840         Add to -Wall section.
29842 2017-06-01  Richard Biener  <rguenther@suse.de>
29844         PR middle-end/66313
29845         * fold-const.c (fold_plusminus_mult_expr): If the factored
29846         factor may be zero use a wrapping type for the inner operation.
29847         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29848         and handle moved defs.
29849         (process_assignment): Properly guard the unary op case.  Return a
29850         tri-state indicating that moving the stmt before the call may allow
29851         to continue.  Pass through to_move.
29852         (find_tail_calls): Handle moving unrelated defs before
29853         the call.
29855 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
29857         PR target/80618
29858         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29859         splitter result in the canonical way.
29861 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
29863         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29864         also for 32bit target.  Update insn attributes.
29865         (zero-extendsidi2 splitter): Allow all registers for operand 1.
29867 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
29869         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
29870         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
29871         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
29872         (_mm_maskz_min_ss): New intrinsics.
29874 2017-05-31  Martin Liska  <mliska@suse.cz>
29876         * tree-vect-loop.c (vect_create_epilog_for_reduction):
29877         Change comment style to one we normally use.
29878         (vectorizable_reduction): Likewise.
29879         (vectorizable_induction): Likewise.
29880         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
29881         (vectorizable_call): Likewise.
29882         (vectorizable_simd_clone_call): Likewise.
29883         (vectorizable_conversion): Likewise.
29884         (vectorizable_assignment): Likewise.
29885         (vectorizable_shift): Likewise.
29886         (vectorizable_operation): Likewise.
29887         (vectorizable_store): Likewise.
29888         (vectorizable_load): Likewise.
29889         * tree-vectorizer.h: Likewise.
29891 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
29893         * passes.c (emergency_dump_function): New.
29894         * tree-pass.h (emergency_dump_function): Declare.
29895         * plugin.c (plugins_internal_error_function): Remove.
29896         * plugin.h (plugins_internal_error_function): Remove declaration.
29897         * toplev.c (internal_error_function): New static function.  Use it...
29898         (general_init): ...here.
29900 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
29902         * config/arc/arc.c (arc_print_operand): Handle constant operands.
29903         (arc_rtx_costs): Add costs for new patterns.
29904         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
29905         * config/arc/predicates.md: Add _1_2_3_operand predicate.
29907 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29909         * tree-ssa-strlen.c (get_next_strinfo): New function.
29910         (get_stridx_plus_constant): Use it.
29911         (zero_length_string): Likewise.
29912         (adjust_related_strinfos): Likewise.
29913         (adjust_last_stmt): Likewise.
29915 2017-05-31  Richard Biener  <rguenther@suse.de>
29917         PR target/80880
29918         * config/i386/i386.c (ix86_expand_builtin): Remove assert
29919         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
29921 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29923         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
29924         loop_vinfo argument and use of dependence distance vectors.
29925         Check instead whether the two references differ only in their
29926         initial value and assume that they have the same alignment if the
29927         difference is a multiple of the vector alignment.
29928         (vect_analyze_data_refs_alignment): Update call accordingly.
29930 2017-05-31  Martin Liska  <mliska@suse.cz>
29932         PR target/79155
29933         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
29935 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
29937         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
29938         (create_intersect_range_checks): Move from ...
29939         * tree-data-ref.c (create_intersect_range_checks_index)
29940         (create_intersect_range_checks): ... to here.
29941         (create_runtime_alias_checks): New function factored from ...
29942         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
29943         here.  Call above function.
29944         * tree-data-ref.h (create_runtime_alias_checks): New function.
29946 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
29948         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
29949         segment length for dr_b and compute it in wide_int.
29951 2017-05-31  Richard Biener  <rguenther@suse.de>
29953         PR tree-optimization/80906
29954         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
29955         and pass through iv_map.
29956         (copy_bb_and_scalar_dependences): Adjust.
29957         (translate_pending_phi_nodes): Likewise.
29958         (copy_loop_close_phi_args): Handle code-generating IVs instead
29959         of ICEing.
29961 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
29963         * diagnostic-color.c (color_dict): Add "type-diff".
29964         (parse_gcc_colors): Update comment.
29965         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
29966         -fdiagnostics-show-template-tree and -fno-elide-type.
29967         (GCC_COLORS): Add type-diff to example.
29968         (type-diff=): New.
29969         (-fdiagnostics-show-template-tree): New.
29970         (-fno-elide-type): New.
29971         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
29972         the pp_format_decoder callback.  Call any m_format_postprocessor's
29973         "handle" method.
29974         (pretty_printer::pretty_printer): Initialize
29975         m_format_postprocessor.
29976         (pretty_printer::~pretty_printer): Delete any
29977         m_format_postprocessor.
29978         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
29979         (class format_postprocessor): New class.
29980         (struct pretty_printer::format_decoder): Document the new parameters.
29981         (struct pretty_printer::m_format_postprocessor): New field.
29982         * tree-diagnostic.c (default_tree_printer): Update for new
29983         bool and const char ** params.
29984         * tree-diagnostic.h (default_tree_printer): Likewise.
29986 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
29988         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
29989         (lwa_operand): Delete rs6000_gen_cell_microcode test.
29990         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29991         rs6000_gen_cell_microcode code.
29992         (rs6000_final_prescan_insn): Delete.
29993         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
29994         "warn-cell-microcode" entries.
29995         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
29996         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
29997         throughout.  Change cc_reg_not_micro_cr0_operand to
29998         cc_reg_not_cr0_operand throughout.
29999         (*extendhi<mode>2_noload): Delete.
30000         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30001         (mwarn-cell-microcode): Delete.
30002         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30003         -mgen-cell-microcode and -mwarn-cell-microcode.
30005 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30007         PR target/80833
30008         * config/i386/constraints.md (Yd): New constraint.
30009         (Ye): Ditto.
30010         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30011         and (?Yd, r) alternatives.  Update insn attributes.
30012         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30013         and (?*Yd, r) alternatives.  Update insn attributes.
30014         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30016 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30018         * gimplify.c (gimplify_modify_expr): Don't create a
30019         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30020         function.
30022 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30024         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30026 2017-05-30  Richard Biener  <rguenther@suse.de>
30028         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30029         and reduc_def fields.
30030         (STMT_VINFO_REDUC_TYPE): New define.
30031         (STMT_VINFO_REDUC_DEF): Likewise.
30032         (vect_force_simple_reduction): Adjust prototype.
30033         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30034         (vect_is_simple_reduction): Remove check_reduction argument.
30035         (vect_force_simple_reduction): Adjust and set
30036         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30037         (vectorizable_reduction): Do not re-do reduction analysis
30038         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30039         * tree-parloops.c (gather_scalar_reductions): Adjust.
30041 2017-05-30  Richard Biener  <rguenther@suse.de>
30043         PR middle-end/80901
30044         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30045         split_edge code.
30047 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30049         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30050         Introduce unknown_misalignment parameter and remove vf.
30051         (vect_peeling_hash_get_lowest_cost):
30052         Pass unknown_misalignment parameter.
30053         (vect_enhance_data_refs_alignment):
30054         Fix unsupportable data ref treatment.
30056 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30058         * tree-vect-data-refs.c (vect_get_data_access_cost):
30059         Workaround for SLP handling.
30060         (vect_enhance_data_refs_alignment):
30061         Compute costs for doing no peeling at all, compare to the best
30062         peeling costs so far and avoid peeling if cheaper.
30064 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30066         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30067         Return peeling info and set costs to zero for unlimited cost
30068         model.
30069         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30070         with unknown misalignment. Compute and costs for unknown
30071         misalignment, compare them to the costs for known misalignment
30072         and choose the cheapest for peeling.
30074 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30076         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30077         (vect_get_peeling_costs_all_drs): Create function.
30078         (vect_peeling_hash_get_lowest_cost):
30079         Use vect_get_peeling_costs_all_drs.
30080         (vect_peeling_supportable): Create function.
30081         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30083 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30085         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30086         DR_HAS_NEGATIVE_STEP.
30087         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30088         (vect_enhance_data_refs_alignment): Use.
30089         (vect_duplicate_ssa_name_ptr_info): Use.
30090         * tree-vectorizer.h (dr_misalignment): Use.
30091         (known_alignment_for_access_p): Use.
30093 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30095         PR target/78838
30096         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30097         .lowtext.
30098         (has_section_name): New function.
30100 2017-05-30  Martin Liska  <mliska@suse.cz>
30102         PR other/80909
30103         * auto-profile.c (get_function_decl_from_block): Fix
30104         parenthesis.
30106 2017-05-30  Richard Biener  <rguenther@suse.de>
30108         PR middle-end/80876
30109         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30111 2017-05-30  Martin Liska  <mliska@suse.cz>
30113         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30114         * dumpfile.h (struct dump_file_info): Remove ctors.
30116 2017-05-30  Martin Liska  <mliska@suse.cz>
30118         * predict.def: Fix GNU coding style.
30120 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30122         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30123         Mark 'to' argument with ATTRIBUTE_UNUSED.
30125 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30127         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30128         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30129         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30130         format string.
30132 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30134         * doc/install.texi (Options specification): Restore entry of
30135         --enable-sjlj-exceptions.
30137 2017-05-27  Michael Eager  <eager@eagercon.com>
30139         Revert:
30140         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30142         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30144         * config/microblaze/microblaze.h
30145         (FIXED_REGISTERS): Update in macro.
30146         (CALL_USED_REGISTERS): Update in macro.
30148 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30150         * doc/install.texi: Add links to macOS binary distributions.
30152 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30154         PR bootstrap/80887
30155         Revert:
30156         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30158         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30160 2017-05-26  Martin Liska  <mliska@suse.cz>
30162         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30164 2017-05-26  Martin Liska  <mliska@suse.cz>
30166         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30167         always leading ';; '.
30168         (dump_bb_info): Likewise.
30169         (brief_dump_cfg): Likewise.
30170         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30171         * dumpfile.c: Remove usage of TDF_VERBOSE.
30172         * dumpfile.h (enum dump_kind): Likewise.
30173         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30174         * print-tree.c (debug_verbose): Remove.
30175         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30176         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30177         * tree-diagnostic.c (default_tree_printer): Replace
30178         TDF_DIAGNOSTIC with TDF_SLIM.
30180 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30182         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30183         in parameter loop, rather than loop_vinfo.
30184         (create_intersect_range_checks): Ditto.
30185         (vect_create_cond_for_alias_checks): Update call to above functions.
30187 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30189         PR tree-optimization/80815
30190         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30191         for merging runtime alias checks.  Handle negative DR_STEPs.
30193 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30195         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30196         Move from ...
30197         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30198         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30199         out code pruning runtime alias checks.
30200         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30201         factored out from above.
30202         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30203         Move from ...
30204         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30205         ... to here.
30206         (prune_runtime_alias_test_list): New decalaration.
30208 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30210         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30211         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30212         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30213         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30214         (operator==, comp_dr_with_seg_len_pair): Ditto.
30215         (vect_prune_runtime_alias_test_list): Ditto.
30217 2017-05-26  Martin Liska  <mliska@suse.cz>
30219         PR ipa/80663
30220         * params.def: Bound partial-inlining-entry-probability param.
30222 2017-05-26  Marek Polacek  <polacek@redhat.com>
30224         PR sanitizer/80875
30225         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30226         can be negated.
30228 2017-05-26  Richard Biener  <rguenther@suse.de>
30230         PR tree-optimization/80842
30231         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30232         value.
30234 2017-05-26  Richard Biener  <rguenther@suse.de>
30236         PR tree-optimization/80844
30237         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30239 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30241         * doc/md.texi (Machine Constraints): Update x86 family
30242         machine constraints section to match 'config/i386/constraints.md'.
30244 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30246         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30248 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30250         * doc/invoke.texi (--enable-languages): Update documentation.
30252 2017-05-25  Martin Liska  <mliska@suse.cz>
30254         * dumpfile.c: Add TDF_FOLDING.
30255         * dumpfile.h (enum dump_kind): Likewise.
30256         * genmatch.c (dt_simplify::gen_1): Use it.
30258 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30260         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30262 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30264         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30265         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30267 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30269         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30270         * match.pd (X == C): Rewrite it here.
30271         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30272         with_certain_nonzero_bits2): New predicates.
30273         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30275 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30277         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30278         avoid warning.
30280         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30281         warning.
30283 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30285         * config/powerpcspe: New port.  Files are copied from the rs6000
30286         port, with "rs6000" in filenames replaced by "powerpcspe".
30288 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30290         PR rtl-optimization/80754
30291         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30293 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30295         * config/sparc/sparc.md (length): Return the correct value for -mflat
30296         sibcalls to match output_sibcall.
30298 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30300         PR bootstrap/80860
30301         PR bootstrap/80843
30302         * config/rs6000/rs6000.c (struct machine_function): Add new field
30303         n_components.
30304         (rs6000_get_separate_components): Init that field, use it.
30305         (rs6000_components_for_bb): Use the field.
30307 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30309         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30311 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30313         PR middle-end/80823
30314         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30316 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30318         PR target/80725
30319         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30320         address against address_operand predicate.
30321         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30323 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30325         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30326         parameters passed indirectly.
30328 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30330         * config/i386/i386.md (*movdi_internal): Remove SSE4
30331         alternative 18 (?r, *v).  Update insn attributes.
30332         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30333         Update insn attributes.
30334         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30335         Update insn attributes.
30336         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30337         alternative 1 (r, v). Remove isa attribute.
30338         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30339         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30340         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30342 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30344         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30345         dg-line directive.
30347 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30349         * cgraphunit.c (symbol_table::process_new_functions): Update.
30350         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30351         (inline_generate_summary): Rename to ...
30352         (ipa_fn_summary_generate): ... this one.
30353         (inline_read_summary): Rename to ...
30354         (ipa_fn_summary_read): ... this one.
30355         (inline_write_summary): Rename to ...
30356         (ipa_fn_summary_write): ... this one.
30357         (inline_free_summary): Rename to ...
30358         (ipa_free_fn_summary): ... this one.
30359         (pass_data_local_fn_summary, pass_local_fn_summary,
30360         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30361         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30362         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30363         make_pass_ipa_fn_summary): New.
30364         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30365         inline_write_summary, inline_free_summary): Remove.
30366         (ipa_free_fn_summary) : New.
30367         * ipa-inline.c (ipa_inline): Update.
30368         (pass_ipa_inline): Do not generate summaries.
30369         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30370         Remove.
30371         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30372         and add pass_ipa_fn_summary.
30373         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30374         New.
30375         (make_pass_inline_parameters): Remove.
30377 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30379         * omp-low.c (struct omp_context): Remove "default_kind" member.
30380         Adjust all users.
30382         * omp-offload.c (execute_oacc_device_lower): Remove the
30383         parallelism dimensions function attributes for unparallelized
30384         OpenACC kernels constructs.
30386 2017-05-23  Martin Liska  <mliska@suse.cz>
30388         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30389         functions.
30390         (cgraph_edge::make_speculative): Likewise.
30391         (cgraph_edge::resolve_speculation): Likewise.
30392         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30393         (cgraph_node::dump): Likewise.
30394         * cgraph.h: Likewise.
30395         * cgraphunit.c (analyze_functions): Likewise.
30396         (symbol_table::compile): Likewise.
30397         * ipa-cp.c (print_all_lattices): Likewise.
30398         (determine_versionability): Likewise.
30399         (initialize_node_lattices): Likewise.
30400         (ipcp_verify_propagated_values): Likewise.
30401         (estimate_local_effects): Likewise.
30402         (update_profiling_info): Likewise.
30403         (create_specialized_node): Likewise.
30404         (perhaps_add_new_callers): Likewise.
30405         (decide_about_value): Likewise.
30406         (decide_whether_version_node): Likewise.
30407         (identify_dead_nodes): Likewise.
30408         (ipcp_store_bits_results): Likewise.
30409         * ipa-devirt.c (dump_targets): Likewise.
30410         (ipa_devirt): Likewise.
30411         * ipa-icf.c (sem_item::dump): Likewise.
30412         (sem_function::equals): Likewise.
30413         (sem_variable::equals): Likewise.
30414         (sem_item_optimizer::read_section): Likewise.
30415         (sem_item_optimizer::execute): Likewise.
30416         (congruence_class::dump): Likewise.
30417         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30418         (dump_inline_summary): Likewise.
30419         (estimate_node_size_and_time): Likewise.
30420         (inline_analyze_function): Likewise.
30421         * ipa-inline-transform.c (inline_call): Likewise.
30422         * ipa-inline.c (report_inline_failed_reason): Likewise.
30423         (want_early_inline_function_p): Likewise.
30424         (edge_badness): Likewise.
30425         (update_edge_key): Likewise.
30426         (inline_small_functions): Likewise.
30427         * ipa-profile.c (ipa_profile): Likewise.
30428         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30429         (ipa_make_edge_direct_to_target): Likewise.
30430         (remove_described_reference): Likewise.
30431         (ipa_impossible_devirt_target): Likewise.
30432         (propagate_controlled_uses): Likewise.
30433         (ipa_print_node_params): Likewise.
30434         (ipcp_transform_function): Likewise.
30435         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30436         (propagate_pure_const): Likewise.
30437         * ipa-reference.c (generate_summary): Likewise.
30438         (read_write_all_from_decl): Likewise.
30439         (propagate): Likewise.
30440         (ipa_reference_read_optimization_summary): Likewise.
30441         * ipa-utils.c (ipa_merge_profiles): Likewise.
30442         * ipa.c (walk_polymorphic_call_targets): Likewise.
30443         (symbol_table::remove_unreachable_nodes): Likewise.
30444         (ipa_single_use): Likewise.
30445         * passes.c (execute_todo): Likewise.
30446         * predict.c (drop_profile): Likewise.
30447         * symtab.c (symtab_node::get_dump_name): New function.
30448         (symtab_node::dump_name): Likewise.
30449         (symtab_node::dump_asm_name): Likewise.
30450         (symtab_node::dump_references): Likewise.
30451         (symtab_node::dump_referring): Likewise.
30452         (symtab_node::dump_base): Likewise.
30453         (symtab_node::debug_symtab): Likewise.
30454         * tree-sra.c (convert_callers_for_node): Likewise.
30455         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30456         * value-prof.c (init_node_map): Likewise.
30458 2017-05-23  Martin Liska  <mliska@suse.cz>
30460         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30461         and symtab_node::debug_symtab to symbol_table::debug.
30462         * cgraphunit.c (analyze_functions): Use the renamed function.
30463         (symbol_table::compile): Likewise.
30464         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30465         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30466         * passes.c (execute_todo): Likewise.
30467         * symtab.c (symbol_table::dump): New function.
30468         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30470 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30472         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30473         that nonconst implies exec.
30475 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30477         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30478         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30479         (inline_edge_summary_vec): Turn into ...
30480         (ipa_call_summaries): ... this one.
30481         (redirect_to_unreachable, edge_set_predicate,
30482         evaluate_properties_for_edge, inline_summary_alloc,
30483         reset_ipa_call_summary, reset_inline_summary,
30484         inline_summary_t::duplicate): Update.
30485         (inline_edge_duplication_hook): Turn to ...
30486         (ipa_call_summary_t::duplicate): ... this one.
30487         (inline_edge_removal_hook): Turn to ...
30488         (ipa_call_summary_t::remove): ... this one.
30489         (dump_inline_edge_summary): Turn to ...
30490         (dump_ipa_call_summary): ... this one.
30491         (estimate_function_body_sizes): Update.
30492         (inline_update_callee_summaries): Update.
30493         (remap_edge_change_prob): Update.
30494         (remap_edge_summaries): Update.
30495         (inline_merge_summary): Update.
30496         (do_estimate_edge_time): Update.
30497         (inline_generate_summary): Update.
30498         (inline_read_section): Update.
30499         (inline_read_summary): Update.
30500         (inline_free_summary): Update.
30501         * ipa-inline.c (can_inline_edge_p): Update.
30502         (compute_inlined_call_time): Update.
30503         (want_inline_small_function_p): Update.
30504         (edge_badness): Update.
30505         (early_inliner): Update.
30506         * ipa-inline.h (inline_edge_summary): Turn to ...
30507         (ipa_call_summary): ... this one.
30508         (ipa_call_summary_t): New class.
30509         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30510         (ipa_call_summaries): New.
30511         (inline_edge_summary): Remove.
30512         (estimate_edge_growth): Update.
30513         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30514         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30515         * ipa-split.c (execute_split_functions): Update.
30516         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30518 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30520         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30521         attributes): Document rdrand effective target.
30523 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30525         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30526         attributes): Sort alphabetically.
30528 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30530         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30532 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30534         PR target/80718
30535         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30536         V2DF/V2DI splat into two separate patterns, one that handles
30537         registers, and the other that only handles memory.  Drop support
30538         for splatting from a GPR on ISA 2.07 and then splitting the
30539         splat into direct move and splat.
30540         (vsx_splat_<mode>_reg): Likewise.
30541         (vsx_splat_<mode>_mem): Likewise.
30543 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30545         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30547 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30549         PR middle-end/80809
30550         * omp-low.c (finish_taskreg_remap): New function.
30551         (finish_taskreg_scan): If unit size of ctx->record_type
30552         is non-constant, unshare the size expression and replace
30553         decls in it with possible outer var refs.
30555         PR middle-end/80809
30556         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30557         GOVD_SHARED rather than GOVD_PRIVATE with it.
30558         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30559         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30561         PR middle-end/80853
30562         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30563         as last argument to build_outer_var_ref for pointer bases of array
30564         section reductions.
30566 2017-05-19  Martin Sebor  <msebor@redhat.com>
30568         * print-tree.c (print_node): Print DECL_READ_P flag.
30570 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30572         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30573         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30574         * cgraph.c: Likewise.
30575         * cgraphunit.c: Likewise.
30576         * gengtype.c: Likewise.
30577         * ipa-cp.c: Likewise.
30578         * ipa-devirt.c: Likewise.
30579         * ipa-icf.c: Likewise.
30580         * ipa-predicate.c: Likewise.
30581         * ipa-profile.c: Likewise.
30582         * ipa-prop.c: Likewise.
30583         * ipa-split.c: Likewise.
30584         * ipa.c: Likewise.
30585         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30586         edge_predicate_pool, dump_inline_hints,
30587         inline_summary::account_size_time, redirect_to_unreachable,
30588         edge_set_predicate, set_hint_predicate,
30589         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30590         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30591         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30592         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30593         ipa_call_summary_t::remove, initialize_growth_caches,
30594         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30595         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30596         mark_modified, unmodified_parm_1, unmodified_parm,
30597         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30598         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30599         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30600         will_be_nonconstant_predicate, record_modified_bb_info,
30601         get_minimal_bb, record_modified, param_change_prob,
30602         phi_result_unknown_predicate, predicate_for_phi_result,
30603         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30604         estimate_function_body_sizes, compute_inline_parameters,
30605         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30606         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30607         inline_update_callee_summaries, remap_edge_change_prob,
30608         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30609         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30610         inline_analyze_function, inline_summary_t::insert,
30611         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30612         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30613         inline_free_summary): Move to ipa-fnsummary.h
30614         (predicate_t): Remove.
30615         * ipa-fnsummary.c: New file.
30616         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30617         (enum inline_hints_vals, inline_hints, agg_position_info,
30618         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30619         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30620         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30621         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30622         inline_read_summary, inline_write_summary, inline_free_summary,
30623         inline_analyze_function, initialize_inline_failed,
30624         inline_merge_summary, inline_update_overall_summary,
30625         compute_inline_parameters): Move to ipa-fnsummary.h
30626         * ipa-fnsummary.h: New file.
30627         * ipa-inline-transform.h: Include ipa-inline.h.
30628         * ipa-inline.c: LIkewise.
30630 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30632         * ipa-inline.c (edge_badness): Use inlined_time instead of
30633         inline_summaries->get.
30635 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30637         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30639 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30641         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30642         (fdump-lang): Document 'raw' option.
30643         * dumpfile.h (TDI_tu): Delete.
30644         * dumpfile.c (dump_files): Remove translation-unit.
30645         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30647 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30649         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30650         command option from $(AWK) call.
30651         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30652         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30653         [FORMAT]: Remove handling of variable.
30654         * config/avr/t-multilib: Regenerate.
30656 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30658         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30659         self_time.
30660         (dump_inline_summary): Do not print self_time.
30661         (estimate_function_body_sizes): Do not set self_time.
30662         (compute_inline_parameters): Likewise.
30663         (inline_read_section, inline_write_summary): Do not stream self_time.
30664         * ipa-inline.h (inline_summary): Drop self_time.
30666 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30668         * ipa-inline-analysis.c (account_size_time): Rename to ...
30669         (inline_summary::account_size_time): ... this one.
30670         (reset_ipa_call_summary): Turn to ...
30671         (ipa_call_summary::reset): ... this one.
30672         (reset_inline_summary): Turn to ...
30673         (inline_summary::reset): ... this one.
30674         (inline_summary_t::remove): Update.
30675         (inline_summary_t::duplicate): Update.
30676         (ipa_call_summary_t::remove): Update.
30677         (dump_inline_summary): Update.
30678         (estimate_function_body_sizes): Update.
30679         (compute_inline_parameters): Update.
30680         (estimate_node_size_and_time): Update.
30681         (inline_merge_summary): Update.
30682         (inline_update_overall_summary): Update.
30683         (inline_read_section): Update.
30684         (inline_write_summary): Update.
30685         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30686         add account_size_time and reset member functions.
30687         (ipa_call_summary): Add reset function.
30688         * ipa-predicate.h (predicate::operator &): Constify.
30690 2017-05-22  Richard Biener  <rguenther@suse.de>
30692         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30694 2017-05-19  Jason Merrill  <jason@redhat.com>
30696         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30698 2017-05-19  Marek Polacek  <polacek@redhat.com>
30700         PR sanitizer/80800
30701         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30702         TYPE_OVERFLOW_WRAPS checks.
30704 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30706         * tree-core.h (enum omp_clause_default_kind): Add
30707         "OMP_CLAUSE_DEFAULT_PRESENT".
30708         * tree-pretty-print.c (dump_omp_clause): Handle it.
30709         * gimplify.c (enum gimplify_omp_var_data): Add
30710         "GOVD_MAP_FORCE_PRESENT".
30711         (gimplify_adjust_omp_clauses_1): Map it to
30712         "GOMP_MAP_FORCE_PRESENT".
30713         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30715         * gimplify.c (oacc_default_clause): Clarify.
30717 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30719         LANG_HOOK_REGISTER_DUMPS
30720         * toplev.c (general_init): Call register dump lang hook.
30721         * doc/invoke.texi: Document -fdump-lang option family.
30722         * dumpfile.c (dump_files): Remove class dump here.
30723         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30724         * dumpfile.h (tree_dump_index): Remove TDI_class.
30725         * langhooks-def.h (lhd_register_dumps): Declare.
30726         (LANG_HOOKS_REGISTER_DUMPS): Define.
30727         (LANG_HOOKS_INITIALIZER): Add it.
30728         * langhooks.c (lhd_register_dumps): Define.
30729         * langhooks.h (struct lang_hooks): Add register_dumps.
30731 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30733         * context.h (context::set_passes): New.
30734         * context.c (context::context): Do not create pass manager.
30735         * toplev.c (general_init): Create pass manager here.
30737 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30739         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30740         use this splitter if two add or or instructions would also work for
30741         the constant we want to generate.
30743 2017-05-19  Richard Biener  <rguenther@suse.de>
30745         PR build/80821
30746         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30747         predicate evaluation.
30749 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30751         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30752         add ctor.
30753         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30754         unsigned.
30756 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30758         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30759         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30760         (inline_edge_summary_vec): Turn into ...
30761         (ipa_call_summaries): ... this one.
30762         (redirect_to_unreachable, edge_set_predicate,
30763         evaluate_properties_for_edge, inline_summary_alloc,
30764         reset_ipa_call_summary, reset_inline_summary,
30765         inline_summary_t::duplicate): Update.
30766         (inline_edge_duplication_hook): Turn to ...
30767         (ipa_call_summary_t::duplicate): ... this one.
30768         (inline_edge_removal_hook): Turn to ...
30769         (ipa_call_summary_t::remove): ... this one.
30770         (dump_inline_edge_summary): Turn to ...
30771         (dump_ipa_call_summary): ... this one.
30772         (estimate_function_body_sizes): Update.
30773         (inline_update_callee_summaries): Update.
30774         (remap_edge_change_prob): Update.
30775         (remap_edge_summaries): Update.
30776         (inline_merge_summary): Update.
30777         (do_estimate_edge_time): Update.
30778         (inline_generate_summary): Update.
30779         (inline_read_section): Update.
30780         (inline_read_summary): Update.
30781         (inline_free_summary): Update.
30782         * ipa-inline.c (can_inline_edge_p): Update.
30783         (compute_inlined_call_time): Update.
30784         (want_inline_small_function_p): Update.
30785         (edge_badness): Update.
30786         (early_inliner): Update.
30787         * ipa-inline.h (inline_edge_summary): Turn to ...
30788         (ipa_call_summary): ... this one.
30789         (ipa_call_summary_t): New class.
30790         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30791         (ipa_call_summaries): New.
30792         (inline_edge_summary): Remove.
30793         (estimate_edge_growth): Update.
30794         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30795         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30796         * ipa-split.c (execute_split_functions): Update.
30797         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30799 2017-05-19  Richard Biener  <rguenther@suse.de>
30801         PR middle-end/80764
30802         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30804 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30806         * config/rs6000/rs6000.c (struct machine_function): Add field
30807         fpr_is_wrapped_separately.
30808         (rs6000_get_separate_components): Use 64 components.  Handle the
30809         new FPR components.
30810         (rs6000_components_for_bb): Handle the FPR components.
30811         (rs6000_emit_prologue_components): Handle the FPR components.
30812         (rs6000_emit_epilogue_components): Handle the FPR components.
30813         (rs6000_set_handled_components): Handle the FPR components.
30814         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30815         that are already separately shrink-wrapped.
30816         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30817         that are already separately shrink-wrapped.
30819 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30821         PR target/80510
30822         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30823         New predicate.
30825         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30826         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30827         cases where the register allocator uses a move and an offsettable
30828         memory operation to/from a FPR register on ISA 2.06/2.07.
30829         (define_peephole2 for Altivec d-form store): Likewise.
30831 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30833         PR target/80799
30834         * config/i386/mmx.md (*mov<mode>_internal): Enable
30835         alternatives 11, 12, 13 and 14 also for 32bit targets.
30836         Remove alternatives 15, 16, 17 and 18.
30837         * config/i386/sse.md (vec_concatv2di): Change
30838         alternative (!x, *y) to (x, ?!*Yn).
30840 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30842         * dumpfile.h (enum dump_kind): Remove stray comma.
30844 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30846         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30847         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30848         predicate::num_conditions
30849         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30850         (CHANGED): turn into predicate::changed.
30851         (agg_position_info): Move to ipa-predicate.h
30852         (add_condition, predicate::add_clause, predicate::operator &=,
30853         predicate::or_with, predicate::evaluate, predicate::probability,
30854         dump_condition, dump_clause, predicate::dump,
30855         predicate::remap_after_duplication, predicate::remap_after_inlining,
30856         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30857         (evaluate_conditions_for_known_args): Update.
30858         (set_cond_stmt_execution_predicate): Update.
30859         * ipa-inline.h: Include ipa-predicate.h
30860         (condition, inline_param_summary, conditions, agg_position_info,
30861         predicate): Move to ipa-predicate.h
30862         * ipa-predicate.c: New file.
30863         * ipa-predicate.h: New file.
30865 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
30867         * final.c (leaf_function_p): Check we are not in a sequence.
30869 2017-05-18  Martin Liska  <mliska@suse.cz>
30871         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
30872         * dumpfile.c (dump_register): Use new enum dump_kind.
30873         (get_dump_file_name): Likewise.
30874         (dump_enable_all): Likewise.
30875         (dump_switch_p_1): Likewise.
30876         (enable_rtl_dump_file): Remove usage of TDF_RTL.
30877         * dumpfile.h (enum dump_kind): New enum type.
30878         (struct dump_file_info): Create constructor and
30879         format fields and comments.
30880         * passes.c (pass_manager::register_one_dump_file):
30881         Use num dump_kind.
30882         * statistics.c (statistics_early_init): Likewise.
30883         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
30884         TDF_TREE with TDF_SLIM.
30885         (gather_memory_references_ref): Likewise.
30887 2017-05-18  Martin Liska  <mliska@suse.cz>
30889         * vec.h (struct vnull): Use it.
30891 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30893         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
30894         (true_predicate, false_predicate, true_predicate_p,
30895         false_predicate_p): Remove.
30896         (single_cond_predicate, not_inlined_predicate): Turn to member function
30897         in ipa-inline.h
30898         (add_condition): Update.
30899         (add_clause): Turn to...
30900         (predicate::add_clause): ... this one; update; allow passing NULL
30901         as parameter.
30902         (and_predicates): Turn to ...
30903         (predicate::operator &=): ... this one.
30904         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
30905         (or_predicates): Turn to ...
30906         (predicate::or_with): ... this one.
30907         (evaluate_predicate): Turn to ...
30908         (predicate::evaluate): ... this one.
30909         (predicate_probability): Turn to ...
30910         (predicate::probability): ... this one.
30911         (dump_condition): Update.
30912         (dump_predicate): Turn to ...
30913         (predicate::dump): ... this one.
30914         (account_size_time): Update.
30915         (edge_set_predicate): Update.
30916         (set_hint_predicate): UPdate.
30917         (evaluate_conditions_for_known_args): Update.
30918         (evaluate_properties_for_edge): Update.
30919         (remap_predicate_after_duplication): Turn to...
30920         (predicate::remap_after_duplication): ... this one.
30921         (remap_hint_predicate_after_duplication): Update.
30922         (inline_summary_t::duplicate): UPdate.
30923         (dump_inline_edge_summary): Update.
30924         (dump_inline_summary): Update.
30925         (set_cond_stmt_execution_predicate): Update.
30926         (set_switch_stmt_execution_predicate): Update.
30927         (compute_bb_predicates): Update.
30928         (will_be_nonconstant_expr_predicate): Update.
30929         (will_be_nonconstant_predicate): Update.
30930         (phi_result_unknown_predicate): Update.
30931         (predicate_for_phi_result): Update.
30932         (array_index_predicate): Update.
30933         (estimate_function_body_sizes): Update.
30934         (estimate_node_size_and_time): Update.
30935         (estimate_ipcp_clone_size_and_time): Update.
30936         (remap_predicate): Rename to ...
30937         (predicate::remap_after_inlining): ... this one.
30938         (remap_hint_predicate): Update.
30939         (inline_merge_summary): Update.
30940         (inline_update_overall_summary): Update.
30941         (estimate_size_after_inlining): Update.
30942         (read_predicate): Rename to ...
30943         (predicate::stream_in): ... this one.
30944         (read_inline_edge_summary): Update.
30945         (write_predicate): Rename to ...
30946         (predicate::stream_out): ... this one.
30947         (write_inline_edge_summary): Update.
30948         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
30949         (clause_t): Turn to uint32_t
30950         (predicate): Turn to class; implement constructor and operators
30951         ==, !=, &
30952         (size_time_entry): Update.
30953         (inline_summary): Update.
30954         (inline_edge_summary): Update.
30956 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
30958         * fold-const.c (fold_binary_loc): Move transformation...
30959         * match.pd (C - X CMP X): ... here.
30961 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
30963         * config/sparc/sparc.c (sparc_option_override): Set function
30964         alignment for -mcpu=niagara7 to 64 to match the I$ line.
30965         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
30966         latency to 1.
30967         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
30968         latency to 2.
30969         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
30971 2017-05-18  Marek Polacek  <polacek@redhat.com>
30973         PR sanitizer/80797
30974         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
30975         (pass_ubsan::execute): Call gimple_assign_single_p instead of
30976         gimple_assign_load_p.
30978 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
30980         PR middle-end/80692
30981         * real.c (do_compare): Give decimal_do_compare preference over
30982         comparing just the signs.
30984 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
30986         * doc/md.texi (Canonicalization of Instructions): Describe the
30987         canonical form of instructions that inherently set a condition
30988         code register.
30990 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
30992         PR middle-end/80775
30993         * tree-cfg.c: Move deletion of unreachable case statements to after
30994         the merging of consecutive case labels.
30996 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30998         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
30999         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31000         restoring of callee-saved registers.
31002 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31004         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31005         * config/visium/visium.c (single_set_and_flags): Likewise.
31006         * config/visium/visium.md (Substitutions): Likewise.
31008 2017-05-17  Martin Liska  <mliska@suse.cz>
31010         * cfg.c: Introduce dump_flags_t type and
31011         use it instead of int type.
31012         * cfg.h: Likewise.
31013         * cfghooks.c: Likewise.
31014         * cfghooks.h (struct cfg_hooks): Likewise.
31015         * cfgrtl.c: Likewise.
31016         * cfgrtl.h: Likewise.
31017         * cgraph.c (cgraph_node::get_body): Likewise.
31018         * coretypes.h: Likewise.
31019         * domwalk.c: Likewise.
31020         * domwalk.h: Likewise.
31021         * dumpfile.c (struct dump_option_value_info): Likewise.
31022         (dump_enable_all): Likewise.
31023         (dump_switch_p_1): Likewise.
31024         (opt_info_switch_p): Likewise.
31025         * dumpfile.h (enum tree_dump_index): Likewise.
31026         (struct dump_file_info): Likewise.
31027         * genemit.c: Likewise.
31028         * generic-match-head.c: Likewise.
31029         * gengtype.c (open_base_files): Likewise.
31030         * gimple-pretty-print.c: Likewise.
31031         * gimple-pretty-print.h: Likewise.
31032         * graph.c (print_graph_cfg): Likewise.
31033         * graphite-scop-detection.c (dot_all_sese): Likewise.
31034         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31035         * loop-unroll.c (report_unroll): Likewise.
31036         * passes.c (pass_manager::register_one_dump_file): Likewise.
31037         * print-tree.c: Likewise.
31038         * statistics.c: Likewise.
31039         * tree-cfg.c: Likewise.
31040         * tree-cfg.h: Likewise.
31041         * tree-dfa.c: Likewise.
31042         * tree-dfa.h: Likewise.
31043         * tree-dump.c (dump_function): Likewise.
31044         * tree-dump.h (struct dump_info): Likewise.
31045         * tree-pretty-print.c: Likewise.
31046         * tree-pretty-print.h: Likewise.
31047         * tree-ssa-live.c: Likewise.
31048         * tree-ssa-live.h: Likewise.
31049         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31050         * tree-vect-loop.c: Likewise.
31051         * tree-vect-slp.c: Likewise.
31053 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31054             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31056         PR tree-optimization/80457
31057         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31058         of all arguments to a statement as scalar_to_vec operations.
31059         (vectorizable_call): Adjust call to vect_model_simple_cost for
31060         new parameter.
31061         (vectorizable_conversion): Likewise.
31062         (vectorizable_assignment): Likewise.
31063         (vectorizable_shift): Likewise.
31064         (vectorizable_operation): Likewise.
31065         (vectorizable_comparison): Likewise.
31066         (vect_is_simple_cond): Record the def types for operands.
31067         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31068         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31069         for statement argument count.
31071 2017-05-16  Carl Love  <cel@us.ibm.com>
31073         * config/rs6000/rs6000-c: Add support for built-in functions
31074         vector unsigned long long vec_bperm (vector unsigned long long,
31075                                              vector unsigned char)
31076         vector signed long long vec_mule (vector signed int,
31077                                           vector signed int)
31078         vector unsigned long long vec_mule (vector unsigned int,
31079                                             vector unsigned int)
31080         vector signed long long vec_mulo (vector signed int,
31081                                           vector signed int)
31082         vector unsigned long long vec_mulo (vector unsigned int,
31083                                             vector unsigned int)
31084         vector signed char vec_sldw (vector signed char,
31085                                      vector signed char,
31086                                      const int)
31087         vector unsigned char vec_sldw (vector unsigned char,
31088                                        vector unsigned char,
31089                                        const int)
31090         vector signed short vec_sldw (vector signed short,
31091                                       vector signed short,
31092                                       const int)
31093         vector unsigned short vec_sldw (vector unsigned short,
31094                                         vector unsigned short,
31095                                         const int)
31096         vector signed int vec_sldw (vector signed int,
31097                                     vector signed int,
31098                                     const int)
31099         vector unsigned int vec_sldw (vector unsigned int,
31100                                       vector unsigned int,
31101                                       const int)
31102         vector signed long long vec_sldw (vector signed long long,
31103                                           vector signed long long,
31104                                           const int)
31105         vector unsigned long long vec_sldw (vector unsigned long long,
31106                                             vector unsigned long long,
31107                                             const int)
31108         * config/rs6000/rs6000-c: Add support for built-in functions
31109         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31110         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31111         * doc/extend.texi: Update the built-in documentation for the
31112         new built-in functions.
31114 2017-05-16  Marek Polacek  <polacek@redhat.com>
31116         PR sanitizer/80536
31117         PR sanitizer/80386
31118         * tree.c (save_expr): Don't fold the expression.
31120 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31122         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31123         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31124         and (?*y,m).  Update insn attributes.
31126 2017-05-16  Martin Liska  <mliska@suse.cz>
31128         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31129         flags argument of print_gimple_stmt, print_gimple_expr,
31130         print_generic_stmt and print_generic_expr.
31131         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31132         * coretypes.h: Likewise.
31133         * except.c (dump_eh_tree): Likewise.
31134         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31135         * gimple-pretty-print.h: Likewise.
31136         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31137         (backprop::push_to_worklist): Likewise.
31138         (backprop::pop_from_worklist): Likewise.
31139         (backprop::process_use): Likewise.
31140         (backprop::intersect_uses): Likewise.
31141         (note_replacement): Likewise.
31142         * gimple-ssa-store-merging.c
31143         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31144         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31145         (pass_store_merging::execute): Likewise.
31146         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31147         (ssa_base_cand_dump_callback): Likewise.
31148         (dump_incr_vec): Likewise.
31149         (replace_refs): Likewise.
31150         (replace_mult_candidate): Likewise.
31151         (create_add_on_incoming_edge): Likewise.
31152         (create_phi_basis): Likewise.
31153         (insert_initializers): Likewise.
31154         (all_phi_incrs_profitable): Likewise.
31155         (introduce_cast_before_cand): Likewise.
31156         (replace_one_candidate): Likewise.
31157         * gimplify.c (gimplify_expr): Likewise.
31158         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31159         (set_rename): Likewise.
31160         (rename_uses): Likewise.
31161         (copy_loop_phi_nodes): Likewise.
31162         (add_close_phis_to_merge_points): Likewise.
31163         (copy_loop_close_phi_args): Likewise.
31164         (copy_cond_phi_args): Likewise.
31165         (graphite_copy_stmts_from_block): Likewise.
31166         (translate_pending_phi_nodes): Likewise.
31167         * graphite-poly.c (print_pdr): Likewise.
31168         (dump_gbb_cases): Likewise.
31169         (dump_gbb_conditions): Likewise.
31170         (print_scop_params): Likewise.
31171         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31172         (build_cross_bb_scalars_use): Likewise.
31173         (gather_bbs::before_dom_children): Likewise.
31174         * hsa-dump.c (dump_hsa_immed): Likewise.
31175         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31176         (get_replacement_map): Likewise.
31177         * ipa-inline-analysis.c (dump_condition): Likewise.
31178         (estimate_function_body_sizes): Likewise.
31179         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31180         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31181         * ipa-prop.c (ipa_dump_param): Likewise.
31182         (ipa_print_node_jump_functions_for_edge): Likewise.
31183         (ipa_modify_call_arguments): Likewise.
31184         (ipa_modify_expr): Likewise.
31185         (ipa_dump_param_adjustments): Likewise.
31186         (ipa_dump_agg_replacement_values): Likewise.
31187         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31188         * ipa-pure-const.c (check_stmt): Likewise.
31189         (pass_nothrow::execute): Likewise.
31190         * ipa-split.c (execute_split_functions): Likewise.
31191         * omp-offload.c (dump_oacc_loop_part): Likewise.
31192         (dump_oacc_loop): Likewise.
31193         * trans-mem.c (tm_log_emit): Likewise.
31194         (tm_memopt_accumulate_memops): Likewise.
31195         (dump_tm_memopt_set): Likewise.
31196         (dump_tm_memopt_transform): Likewise.
31197         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31198         (print_loop): Likewise.
31199         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31200         (chkp_gather_checks_info): Likewise.
31201         (chkp_get_check_result): Likewise.
31202         (chkp_remove_check_if_pass): Likewise.
31203         (chkp_use_outer_bounds_if_possible): Likewise.
31204         (chkp_reduce_bounds_lifetime): Likewise.
31205         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31206         (chkp_mark_completed_bounds): Likewise.
31207         (chkp_register_incomplete_bounds): Likewise.
31208         (chkp_mark_invalid_bounds): Likewise.
31209         (chkp_maybe_copy_and_register_bounds): Likewise.
31210         (chkp_build_returned_bound): Likewise.
31211         (chkp_get_bound_for_parm): Likewise.
31212         (chkp_build_bndldx): Likewise.
31213         (chkp_get_bounds_by_definition): Likewise.
31214         (chkp_generate_extern_var_bounds): Likewise.
31215         (chkp_get_bounds_for_decl_addr): Likewise.
31216         * tree-chrec.c (chrec_apply): Likewise.
31217         * tree-data-ref.c (dump_data_reference): Likewise.
31218         (dump_subscript): Likewise.
31219         (dump_data_dependence_relation): Likewise.
31220         (analyze_overlapping_iterations): Likewise.
31221         * tree-inline.c (expand_call_inline): Likewise.
31222         (tree_function_versioning): Likewise.
31223         * tree-into-ssa.c (dump_defs_stack): Likewise.
31224         (dump_currdefs): Likewise.
31225         (dump_names_replaced_by): Likewise.
31226         (dump_update_ssa): Likewise.
31227         (update_ssa): Likewise.
31228         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31229         * tree-parloops.c (build_new_reduction): Likewise.
31230         (try_create_reduction_list): Likewise.
31231         (ref_conflicts_with_region): Likewise.
31232         (oacc_entry_exit_ok_1): Likewise.
31233         (oacc_entry_exit_single_gang): Likewise.
31234         * tree-pretty-print.h: Likewise.
31235         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31236         (get_scalar_evolution): Likewise.
31237         (add_to_evolution): Likewise.
31238         (get_loop_exit_condition): Likewise.
31239         (analyze_evolution_in_loop): Likewise.
31240         (analyze_initial_condition): Likewise.
31241         (analyze_scalar_evolution): Likewise.
31242         (instantiate_scev): Likewise.
31243         (number_of_latch_executions): Likewise.
31244         (gather_chrec_stats): Likewise.
31245         (final_value_replacement_loop): Likewise.
31246         (scev_const_prop): Likewise.
31247         * tree-sra.c (dump_access): Likewise.
31248         (disqualify_candidate): Likewise.
31249         (create_access): Likewise.
31250         (reject): Likewise.
31251         (maybe_add_sra_candidate): Likewise.
31252         (create_access_replacement): Likewise.
31253         (analyze_access_subtree): Likewise.
31254         (analyze_all_variable_accesses): Likewise.
31255         (sra_modify_assign): Likewise.
31256         (initialize_constant_pool_replacements): Likewise.
31257         (find_param_candidates): Likewise.
31258         (decide_one_param_reduction): Likewise.
31259         (replace_removed_params_ssa_names): Likewise.
31260         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31261         * tree-ssa-copy.c (dump_copy_of): Likewise.
31262         (copy_prop_visit_cond_stmt): Likewise.
31263         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31264         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31265         (record_equivalences_from_stmt): Likewise.
31266         * tree-ssa-dse.c (compute_trims): Likewise.
31267         (delete_dead_call): Likewise.
31268         (delete_dead_assignment): Likewise.
31269         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31270         (forward_propagate_into_cond): Likewise.
31271         (pass_forwprop::execute): Likewise.
31272         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31273         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31274         Likewise.
31275         (move_computations_worker): Likewise.
31276         (execute_sm): Likewise.
31277         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31278         (remove_exits_and_undefined_stmts): Likewise.
31279         (remove_redundant_iv_tests): Likewise.
31280         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31281         (adjust_iv_update_pos): Likewise.
31282         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31283         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31284         (value_replacement): Likewise.
31285         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31286         * tree-ssa-pre.c (print_pre_expr): Likewise.
31287         (get_representative_for): Likewise.
31288         (create_expression_by_pieces): Likewise.
31289         (insert_into_preds_of_block): Likewise.
31290         (eliminate_insert): Likewise.
31291         (eliminate_dom_walker::before_dom_children): Likewise.
31292         (eliminate): Likewise.
31293         (remove_dead_inserted_code): Likewise.
31294         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31295         * tree-ssa-reassoc.c (get_rank): Likewise.
31296         (eliminate_duplicate_pair): Likewise.
31297         (eliminate_plus_minus_pair): Likewise.
31298         (eliminate_not_pairs): Likewise.
31299         (undistribute_ops_list): Likewise.
31300         (eliminate_redundant_comparison): Likewise.
31301         (update_range_test): Likewise.
31302         (optimize_range_tests_var_bound): Likewise.
31303         (optimize_vec_cond_expr): Likewise.
31304         (rewrite_expr_tree): Likewise.
31305         (rewrite_expr_tree_parallel): Likewise.
31306         (linearize_expr): Likewise.
31307         (break_up_subtract): Likewise.
31308         (linearize_expr_tree): Likewise.
31309         (attempt_builtin_powi): Likewise.
31310         (attempt_builtin_copysign): Likewise.
31311         (transform_stmt_to_copy): Likewise.
31312         (transform_stmt_to_multiply): Likewise.
31313         (dump_ops_vector): Likewise.
31314         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31315         (print_scc): Likewise.
31316         (set_ssa_val_to): Likewise.
31317         (visit_reference_op_store): Likewise.
31318         (visit_use): Likewise.
31319         (sccvn_dom_walker::before_dom_children): Likewise.
31320         (run_scc_vn): Likewise.
31321         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31322         Likewise.
31323         (expr_hash_elt::print): Likewise.
31324         (const_and_copies::pop_to_marker): Likewise.
31325         (const_and_copies::record_const_or_copy_raw): Likewise.
31326         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31327         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31328         (dump_predicates): Likewise.
31329         (find_uninit_use): Likewise.
31330         (warn_uninitialized_phi): Likewise.
31331         (pass_late_warn_uninitialized::execute): Likewise.
31332         * tree-ssa.c (verify_vssa): Likewise.
31333         (verify_ssa): Likewise.
31334         (maybe_optimize_var): Likewise.
31335         * tree-vrp.c (dump_value_range): Likewise.
31336         (dump_all_value_ranges): Likewise.
31337         (dump_asserts_for): Likewise.
31338         (register_edge_assert_for_2): Likewise.
31339         (vrp_visit_cond_stmt): Likewise.
31340         (vrp_visit_switch_stmt): Likewise.
31341         (vrp_visit_stmt): Likewise.
31342         (vrp_visit_phi_node): Likewise.
31343         (simplify_cond_using_ranges_1): Likewise.
31344         (fold_predicate_in): Likewise.
31345         (evrp_dom_walker::before_dom_children): Likewise.
31346         (evrp_dom_walker::push_value_range): Likewise.
31347         (evrp_dom_walker::pop_value_range): Likewise.
31348         (execute_early_vrp): Likewise.
31350 2017-05-16  Richard Biener  <rguenther@suse.de>
31352         * dwarf2out.c (loc_list_from_tree_1): Do not create
31353         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31355 2017-05-16  Richard Biener  <rguenther@suse.de>
31357         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31358         just generated.
31359         (note_variable_value_in_expr): If we resolved the decl ref
31360         do not push to the stack.
31362 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31364         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31365         operations in fast-math mode.
31366         (vaddq_f16): Likewise.
31367         (vmul_f16): Likewise.
31368         (vmulq_f16): Likewise.
31369         (vsub_f16): Likewise.
31370         (vsubq_f16): Likewise.
31371         * config/arm/neon.md (add<mode>3): New.
31372         (sub<mode>3): New.
31373         (fma:<VH:mode>3): New.  Also remove outdated comment.
31374         (mul<mode>3): New.
31376 2017-05-16  Martin Liska  <mliska@suse.cz>
31378         PR ipa/79849.
31379         PR ipa/79850.
31380         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31381         (odr_types_equivalent_p): Likewise.
31383 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31385         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31387 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31389         PR target/80425
31390         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31391         non-interunit SSE move alternatives with '?'.
31392         (zero-extendsidi peephole2): New peephole to skip intermediate
31393         general register in SSE zero-extend sequence.
31395 2017-05-15  Jeff Law  <law@redhat.com>
31397         * reorg.c (relax_delay_slots): Create a new variable to hold
31398         the temporary target rather than clobbering TARGET_LABEL.
31400         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31401         missing argument to extract_bit_field call.
31402         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31404 2017-05-15  Martin Liska  <mliska@suse.cz>
31406         PR driver/31468
31407         * gcc.c (process_command): Do not allow empty argument of -o option.
31409 2017-05-15  Renlin Li  <renlin.li@arm.com>
31411         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31412         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31413         * config/aarch64/constraints.md (Usf): Add long call check.
31414         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31415         (call_value): Likewise.
31416         (sibcall): Likewise.
31417         (sibcall_value): Likewise.
31418         (call_insn): New.
31419         (call_value_insn): New.
31420         (sibcall_insn): Update rtx pattern.
31421         (sibcall_value_insn): Likewise.
31422         (call_internal): Remove.
31423         (call_value_internal): Likewise.
31424         (sibcall_internal): Likewise.
31425         (sibcall_value_internal): Likewise.
31426         (call_reg): Likewise.
31427         (call_symbol): Likewise.
31428         (call_value_reg): Likewise.
31429         (call_value_symbol): Likewise.
31431 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31433         PR target/80600
31434         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31436 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31438         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31439         compatible with CCGOCmode and with CCZmode.
31441 2017-05-14  Martin Sebor  <msebor@redhat.com>
31443         PR middle-end/77671
31444         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31445         (gimple_fold_builtin_snprintf): Same.
31446         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31447         (gimple_fold_builtin_snprintf): Same.
31448         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31449         of character types.
31450         (is_call_safe): New function.
31451         (try_substitute_return_value): Call it.
31452         (try_simplify_call): New function.
31453         (pass_sprintf_length::handle_gimple_call): Call it.
31455 2017-05-14  Martin Sebor  <msebor@redhat.com>
31457         PR middle-end/80669
31458         * builtins.c (expand_builtin_stpncpy): Simplify.
31460 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31462         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31463         * config/i386/i386.h
31464         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31465         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31466         (struct machine_function): Add new members call_ms2sysv,
31467         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31468         (struct machine_frame_state): New fields sp_realigned and
31469         sp_realigned_offset.
31470         * config/i386/i386.c
31471         (enum xlogue_stub): New enum.
31472         (enum xlogue_stub_sets): New enum.
31473         (class xlogue_layout): New class.
31474         (struct ix86_frame): New fields stack_realign_allocate_offset,
31475         stack_realign_offset and outlined_save_offset.  Modify comments to
31476         detail stack layout when using out-of-line stubs.
31477         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31478         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31479         -mcall-ms2sysv-xlogues.
31480         (stub_managed_regs): New static variable.
31481         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31482         registers managed by out-of-line stub.
31483         (disable_call_ms2sysv_xlogues): New function.
31484         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31485         m->call_ms2sysv when appropriate and compute frame layout for
31486         out-of-line stubs.
31487         (sp_valid_at, fp_valid_at): New inline functions.
31488         (choose_basereg): New function.
31489         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31490         all callers.
31491         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31492         Use align parameter of choose_baseaddr to generated aligned SSE movs
31493         when possible.
31494         (pro_epilogue_adjust_stack): Modify to track
31495         machine_frame_state::sp_realigned.
31496         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31497         (ix86_nsaved_sseregs): Likewise.
31498         (ix86_emit_save_regs): Likewise.
31499         (ix86_emit_save_regs_using_mov): Likewise.
31500         (ix86_emit_save_sse_regs_using_mov): Likewise.
31501         (get_scratch_register_on_entry): Likewise.
31502         (gen_frame_set): New function.
31503         (gen_frame_load): Likewise.
31504         (gen_frame_store): Likewise.
31505         (emit_outlined_ms2sysv_save): Likewise.
31506         (emit_outlined_ms2sysv_restore): Likewise.
31507         (ix86_expand_prologue): Modify stack re-alignment code and call
31508         emit_outlined_ms2sysv_save when appropriate.
31509         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31510         parameter rtx_insn *insn, which allows the function to be used to only
31511         generate the notes.
31512         (ix86_expand_epilogue): Modify validity checks of frame and stack
31513         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31514         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31515         * config/i386/predicates.md
31516         (save_multiple): New predicate.
31517         (restore_multiple): Likewise.
31518         * config/i386/sse.md
31519         (save_multiple<mode>): New pattern.
31520         (save_multiple_realign<mode>): Likewise.
31521         (restore_multiple<mode>): Likewise.
31522         (restore_multiple_and_return<mode>): Likewise.
31523         (restore_multiple_leave_return<mode>): Likewise.
31524         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31526 2017-05-14  Julia Koval  <julia.koval@intel.com>
31528         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31529         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31530         (__builtin_ia32_xsetbv): New builtins.
31531         * config/i386/i386.c (ix86_expand_special_args_builtin):
31532         Process new types.
31533         (ix86_expand_builtin): Special expand for new intrinsics.
31534         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31535         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31536         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31538 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31540         * cfganal.c (inverted_post_order_compute): Change argument type
31541         to vec *.
31542         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31543         * df-core.c (rest_of_handle_df_initialize): Adjust.
31544         (rest_of_handle_df_finish): Likewise.
31545         (df_analyze_1): Likewise.
31546         (df_analyze): Likewise.
31547         (loop_inverted_post_order_compute): Change argument to be a vec *.
31548         (df_analyze_loop): Adjust.
31549         (df_get_n_blocks): Likewise.
31550         (df_get_postorder): Likewise.
31551         * df.h (struct df_d): Change field to be a vec.
31552         * lcm.c (compute_laterin): Adjust.
31553         (compute_available): Likewise.
31554         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31555         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31556         * tree-ssa-pre.c (compute_antic): Likewise.
31558 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31560         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31561         (depth_first_search::depth_first_search): Change structure init
31562         function to this constructor.
31563         (depth_first_search::add_bb): Rename function to this member.
31564         (depth_first_search::execute): Likewise.
31565         (flow_dfs_compute_reverse_finish): Adjust.
31567 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31569         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31570         (longest_simple_path): Likewise.
31571         * shrink-wrap.c (spread_components): Likewise.
31572         (disqualify_problematic_components): Likewise.
31573         (emit_common_heads_for_components): Likewise.
31574         (emit_common_tails_for_components): Likewise.
31575         (insert_prologue_epilogue_for_components): Likewise.
31577 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31579         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31580         auto_sbitmap.
31582 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31584         * df-core.c (df_set_blocks): Start using auto_bitmap.
31585         (df_compact_blocks): Likewise.
31586         * df-problems.c (df_rd_confluence_n): Likewise.
31587         * df-scan.c (df_insn_rescan_all): Likewise.
31588         (df_process_deferred_rescans): Likewise.
31589         (df_update_entry_block_defs): Likewise.
31590         (df_update_exit_block_uses): Likewise.
31591         (df_entry_block_bitmap_verify): Likewise.
31592         (df_exit_block_bitmap_verify): Likewise.
31593         (df_scan_verify): Likewise.
31594         * lra-constraints.c (lra_constraints): Likewise.
31595         (undo_optional_reloads): Likewise.
31596         (lra_undo_inheritance): Likewise.
31597         * lra-remat.c (calculate_gen_cands): Likewise.
31598         (do_remat): Likewise.
31599         * lra-spills.c (assign_spill_hard_regs): Likewise.
31600         (spill_pseudos): Likewise.
31601         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31602         (bitmap_set_subtract_values): Likewise.
31604 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31606         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31607         management with auto_bitmap.
31608         (fix_inter_tick): Likewise.
31609         (fix_recovery_deps): Likewise.
31610         * ira.c (add_store_equivs): Likewise.
31611         (find_moveable_pseudos): Likewise.
31612         (split_live_ranges_for_shrink_wrap): Likewise.
31613         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31614         (rtx_reuse_manager::seen_def_p): Likewise.
31615         (rtx_reuse_manager::set_seen_def): Likewise.
31616         * print-rtl.h (class rtx_reuse_manager): Likewise.
31618 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31620         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31621         lifetime.
31622         (migrate_btr_def): Likewise.
31623         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31624         * df-core.c (loop_post_order_compute): Likewise.
31625         (loop_inverted_post_order_compute): Likewise.
31626         * hsa-common.h: Likewise.
31627         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31628         * init-regs.c (initialize_uninitialized_regs): Likewise.
31629         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31630         (inline_small_functions): Likewise.
31631         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31632         * ira.c (combine_and_move_insns): Likewise.
31633         (build_insn_chain): Likewise.
31634         * loop-invariant.c (find_invariants): Likewise.
31635         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31636         * predict.c (tree_predict_by_opcode): Likewise.
31637         (predict_paths_leading_to): Likewise.
31638         (predict_paths_leading_to_edge): Likewise.
31639         (estimate_loops_at_level): Likewise.
31640         (estimate_loops): Likewise.
31641         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31642         (spread_components): Likewise.
31643         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31644         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31645         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31646         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31647         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31648         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31649         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31650         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31651         (mark_threaded_blocks): Likewise.
31652         (thread_through_all_blocks): Likewise.
31653         * tree-ssa.c (verify_ssa): Likewise.
31654         (execute_update_addresses_taken): Likewise.
31655         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31657 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31659         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31660         auto_vec.
31661         (post_order_compute): Likewise.
31662         (inverted_post_order_compute): Likewise.
31663         (pre_and_rev_post_order_compute_fn): Likewise.
31665 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31667         * genrecog.c (int_set::int_set): Explicitly construct our
31668         auto_vec base class.
31669         * vec.h (auto_vec::auto_vec): New constructor.
31671 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31673         * bitmap.h (class auto_bitmap): New constructor taking
31674         bitmap_obstack * argument.
31676 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31678         * bitmap.h (class auto_bitmap): Change type of m_bits to
31679         bitmap_head, and adjust ctor / dtor and member operators.
31681 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31683         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31684         when returned register mode doesn't match original mode.
31686 2017-05-12  Jeff Law  <law@redhat.com>
31687             Jakub Jelinek  <jakub@redhat.com>
31689         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31690         we look for cc setter after the compare-elim changes.
31691         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31692         within the vector to match what compare-elim now expects.
31693         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31694         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31696         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31697         after the compare-elim changes.
31698         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31699         the vector to match what compare-elim now expects.
31700         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31701         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31702         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31703         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31704         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31706         * config/visium/visium.c (single_set_and_flags): Fix where
31707         we look for cc setter after the compare-elim changes.
31708         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31709         with the vector to match what compare-elim now expects.
31710         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31711         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31712         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31713         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31714         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31715         (neg<mode>2_insn_set_overflow): Likewise.
31717 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31719         PR middle-end/79794
31720         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31721         maybe_expand_insn call, set ops[0].target.  If still set after call,
31722         set alt_rtl.  Add extra arg to recursive calls.
31723         (extract_bit_field): Add alt_rtl argument.  Pass to
31724         extract_bit_field.
31725         * expmed.h (extract_bit_field): Fix prototype.
31726         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31727         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31728         to extract_bit_field_calls.
31729         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31730         Pass alt_rtl to extract_bit_field calls.
31731         * calls.c (store_unaligned_arguments_into_psuedos)
31732         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31733         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31734         gen_reg_rtx.
31735         * optabs.h (struct expand_operand): Add target bitfield.
31737 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31739         * compare-elim.c (try_eliminate_compare): Canonicalize
31740         operation with embedded compare to
31741         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31742          (set (reg) (operation)].
31744         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31746 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31748         PR target/80723
31749         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31750         cost of adding a carry flag for ADC instruction.
31751         [case MINUS]: Ignore the cost of subtracting a carry flag
31752         for SBB instruction.
31754 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31756         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31757         and x86intrin.h
31758         * config/rs6000/bmiintrin.h: New file.
31759         * config/rs6000/bmi2intrin.h: New file.
31760         * config/rs6000/x86intrin.h: New file.
31762 2017-05-12  Jeff Law  <law@redhat.com>
31764         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31765         markers.
31767 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31769         PR middle-end/80707
31770         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31772 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31774         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31775         early expansion of vector divide builtins.
31776         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31777         builtins identified as having unsigned arguments.
31779 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31781         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31782         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31783         expansion of vector logical operations (and, andc, or, xor,
31784         nor, orc, nand).
31786 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31788         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31789         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31791 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31793         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31794         early GIMPLE expansion of vector multiplies.
31796 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31798         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31799         TARGET_HAVE_MOVT conditional.
31800         (movt splitter): Likewise.
31802 2017-05-12  Richard Biener  <rguenther@suse.de>
31804         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31805         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31806         Fold all stmts not inplace.
31808 2017-05-12  Richard Biener  <rguenther@suse.de>
31810         PR tree-optimization/80713
31811         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31812         inserted_exprs bit for not removed stmts.
31814 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31816         PR middle-end/69921
31817         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31818         parallelized" attribute for parallelized OpenACC kernels.
31819         * omp-offload.c (execute_oacc_device_lower): Use it.
31821         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31822         Set "oacc kernels" attribute.
31823         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31824         parameter.  Adjust all users.
31825         (oacc_fn_attrib_kernels_p): Remove function.
31826         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31827         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31828         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31829         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31830         assert "oacc kernels" attribute is set.
31832 2017-05-11  Carl Love  <cel@us.ibm.com>
31834         * config/rs6000/rs6000-c: Add support for built-in functions
31835         vector unsigned char vec_popcnt (vector signed char)
31836         vector unsigned char vec_popcnt (vector unsigned char)
31837         vector unsigned short vec_popcnt (vector signed short)
31838         vector unsigned short vec_popcnt (vector unsigned short)
31839         vector unsigned int vec_popcnt (vector signed int)
31840         vector unsigned int vec_popcnt (vector unsigned int)
31841         vector unsigned long long vec_popcnt (vector signed long long)
31842         vector unsigned long long vec_popcnt (vector unsigned long long)
31843         vector signed long long vec_slo (vector signed long long,
31844                                          vector signed char)
31845         vector signed long long vec_slo (vector signed long long,
31846                                          vector unsigned char)
31847         vector unsigned long long vec_slo (vector unsigned long long,
31848                                            vector signed char)
31849         vector unsigned long long vec_slo (vector unsigned long long,
31850                                            vector unsigned char)
31851         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31852         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31853         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31854         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31855         * doc/extend.texi: Update the built-in documentation file for the
31856         new built-in functions.
31858 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31860         * attribs.h (sorted_attr_string): Move machine independent
31861         functions for target clone support from the i386 port to common
31862         code.  Rename ix86_function_versions to common_function_versions.
31863         Rename make_name to make_unique_name.
31864         (common_function_versions): Likewise.
31865         (make_unique_name): Likewise.
31866         (make_dispatcher_decl): Likewise.
31867         (is_function_default_version): Likewise.
31868         * attribs.c (attr_strcmp): Likewise.
31869         (sorted_attr_string): Likewise.
31870         (common_function_versions): Likewise.
31871         (make_unique_name): Likewise.
31872         (make_dispatcher_decl): Likewise.
31873         (is_function_default_version): Likewise.
31874         * config/i386/i386.c (attr_strcmp): Likewise.
31875         (sorted_attr_string): Likewise.
31876         (ix86_function_versions): Likewise.
31877         (make_name): Likewise.
31878         (make_dispatcher_decl): Likewise.
31879         (is_function_default_version): Likewise.
31880         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
31882 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31884         PR target/80695
31885         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
31886         Account for direct move costs for vec_construct of integer
31887         vectors.
31889 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
31891         PR target/80706
31892         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
31893         (UNSPEC_STX_ATOMIC): Ditto.
31894         (loaddi_via_sse): New insn.
31895         (storedi_via_sse): Ditto.
31896         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
31897         Update corresponding peephole2 patterns.
31898         (atomic_storedi_fpu): Ditto.
31900 2017-05-11  Julia Koval  <julia.koval@intel.com>
31902         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
31903         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
31904         New intrinsics.
31905         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
31906         (__builtin_ia32_rsqrt14ss_mask): New builtins.
31907         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
31909 2017-05-11  Nathan Sidwell  <nathan@acm.org>
31911         * graphite-poly.c: Include dumpfile.h.
31913         * dumpfle.h (dump_function): Declare here ...
31914         * tree-dump.h (dump_function): ... not here.
31915         * dumpfile.c: #include tree-cfg.h.
31916         (dump_function): Move here from ...
31917         * tree-dump.c (dump_function): ... here.
31918         * gimplify.c: #include splay-tree.h, not tree-dump.h.
31919         * graphite-poly.c: Don't include tree-dump.h.
31920         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
31921         * print-tree.c: Likewise.
31922         * stor-layout.c: Likewise.
31923         * tree-nested.c: Likewise.
31925         * dumpfile.c (dump_start): Use TDF_FLAGS.
31926         (dump_enable_all): Fix TDF_KIND check thinko.
31928 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
31930         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
31931         array entries to represent two legal parameterizations of the
31932         overloaded __builtin_cmpb function, as represented by the
31933         P6_OV_BUILTIN_CMPB constant.
31934         (altivec_resolve_overloaded_builtin): Add special case handling
31935         for the __builtin_cmpb function, as represented by the
31936         P6_OV_BUILTIN_CMPB constant.
31937         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
31938         (BU_P6_64BIT_2): New macro.
31939         (BU_P6_OVERLOAD_2): New macro
31940         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
31941         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
31942         (CMPB): Add overload support to represent both 32-bit and 64-bit
31943         compare-bytes function.
31944         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
31945         support for TARGET_CMPB.
31946         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
31947         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
31948         documentation of the __builtin_cmpb overloaded built-in function.
31950 2017-05-11  Richard Biener  <rguenther@suse.de>
31952         PR tree-optimization/80705
31953         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
31954         bases are not vectorizable.
31956 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
31958         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
31959         when counting register pressure.
31961 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
31963         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
31964         (iv_ca_more_deps): Renamed to ...
31965         (iv_ca_compare_deps): ... this.
31966         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
31968 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
31970         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
31971         to ...
31972         (determine_group_iv_costs): ... here.
31973         (find_inv_vars_cb): Record inv var if it's not recorded before.
31975 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
31977         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
31978         (get_shiftadd_cost): Ditto.
31980 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
31982         * tree-ssa-address.c: Include header file.
31983         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
31984         address.
31985         (add_to_parts): Refactor.
31986         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
31987         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
31988         in new order.
31990 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
31992         PR tree-optimization/53090
31993         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
31994         COMP_IV_EXPR_2.
31995         (extract_cond_operands): Detect condition with IV on both sides
31996         and return COMP_IV_EXPR_2.
31997         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
31998         (rewrite_use_compare): Simplify by removing call to function
31999         extract_cond_operands.
32001 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32003         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32004         (extract_cond_operands): Detect condition comparing against non-
32005         invariant bound and return appropriate enum value.
32006         (find_interesting_uses_cond): Update use of extract_cond_operands.
32007         Handle its return value accordingly.
32008         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32010 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32012         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32013         nonlinear iv_use computation in loop invariant sensitive way.
32015 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32017         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32018         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32020 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32022         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32023         (dump_cand): Support iv_cand.inv_exprs.
32024         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32025         for candidates.
32026         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32027         iv_cand.inv_exprs.
32029 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32031         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32032         from ...
32033         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32034         as local function.  Include necessary header files.
32035         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32037 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32039         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32041 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32043         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32044         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32045         RSHIFT_EXPR and BIT_NOT_EXPR.
32047 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32049         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32050         (adjust_setup_cost): New parameter supporting round up adjustment.
32051         (struct address_cost_data): Delete.
32052         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32053         (split_address_cost, ptr_difference_cost): Delete.
32054         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32055         (struct ainc_cost_data): New struct.
32056         (get_address_cost_ainc): New function.
32057         (get_address_cost, get_computation_cost): Reimplement.
32058         (determine_group_iv_cost_address): Record inv_expr for all uses of
32059         a group.
32060         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32061         (iv_ca_has_deps): Reimplemented to ...
32062         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32063         than OLD_CP.
32064         (iv_ca_extend): Call iv_ca_more_deps.
32066 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32068         * tree-ssa-address.c (struct mem_address): Move to header file.
32069         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32070         * tree-ssa-address.h (struct mem_address): Move from C file.
32071         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32073 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32075         * tree-affine.h (aff_combination_type): New interface.
32076         (aff_combination_zero_p): Remove static.
32077         (aff_combination_const_p): New interface.
32078         (aff_combination_singleton_var_p): New interfaces.
32080 2017-05-11  Richard Biener  <rguenther@suse.de>
32082         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32083         Skip unreachable blocks and destinations.
32084         (eliminate): Move stmt removal and fixup ...
32085         (fini_eliminate): ... here.  Skip inserted exprs.
32086         (pass_pre::execute): Move fini_pre after fini_eliminate.
32087         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32088         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32089         PRE to get rid of dead code that has invalid SSA form and
32090         split critical edges again.
32092 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32094         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32096 2017-05-11  Richard Biener  <rguenther@suse.de>
32098         * passes.c (execute_function_todo): Verify loops if they are
32099         said to be up-to-date.
32100         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32101         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32103 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32105         PR target/80090
32106         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32107         handle calling assemble_external ourself.
32109         PR target/79027
32110         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32111         modes with zero size.  Enhance comment.
32113 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32115         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32116         built-ins for vec_xl and vec_xst with short and char pointer
32117         arguments.
32119 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32121         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32122         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32123         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32124         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32125         (_mm_maskz_min_round_ss): New intrinsics.
32126         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32127         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32128         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32129         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32130         (__builtin_ia32_minss_mask_round): New builtins.
32131         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32132         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32133         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32134         Rename to ...
32135         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32136         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32137         Change to ...
32138         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32139         ... this.
32141 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32143         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32144         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32145         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32146         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32147         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32148         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32149         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32150         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32151         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32152         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32153         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32154         (__builtin_ia32_mulss_mask_round): New builtins.
32155         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32156         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32157         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32158         Rename to ...
32159         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32160         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32161         Change to ...
32162         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32163         ... this.
32165 2017-05-10  Julia Koval  <julia.koval@intel.com>
32167         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32168         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32169         (_mm256_setr_m128i): New intrinsics.
32171 2017-05-10  Julia Koval  <julia.koval@intel.com>
32173         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32174         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32175         (_mm_maskz_rcp14_ss): New intrinsics.
32176         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32177         (__builtin_ia32_rcp14ss_mask): New builtins.
32178         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32180 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32182         PR tree-optimization/51513
32183         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32184         (assert_unreachable_fallthru_edge_p): Use it.
32185         (group_case_labels_stmt): Likewise.
32186         * tree-cfg.h: Prototype it.
32187         * stmt.c: Include cfghooks.h and tree-cfg.h.
32188         (emit_case_dispatch_table) <gap_label>: New local variable.
32189         Use it to fill dispatch table gaps.
32190         Test for default_label before updating probabilities.
32191         (expand_case) <default_label>: Remove unneeded initialization.
32192         Test for unreachable default case statement and remove its edge.
32193         Set default_label accordingly.
32194         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32196 2017-05-10  Carl Love  <cel@us.ibm.com>
32198         * config/rs6000/rs6000-c: Add support for built-in functions
32199         vector signed char      vec_neg (vector signed char)
32200         vector signed short int vec_neg (vector short int)
32201         vector signed int       vec_neg (vector signed int)
32202         vector signed long long vec_neg (vector signed long long)
32203         vector float            vec_neg (vector float)
32204         vector double           vec_neg (vector double)
32205         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32206         overload.
32207         * config/rs6000/altivec.h: Add define for vec_neg
32208         * doc/extend.texi: Update the built-in documentation for the
32209         new built-in functions.
32211 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32213         PR tree-optimization/77644
32214         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32216 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32218         * dumpfile.h (TDI_lang_all): New.
32219         (TDF_KIND): New. Renumber others
32220         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32221         than bits.
32222         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32223         lang-all.
32224         (get_dump_file_name): Adjust suffix generation.
32225         (dump_enable_all): Use TDF_KIND.
32226         * doc/invoke.texi (-fdump-lang-all): Document.
32228         * dumpfile.h: Tabify.
32230 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32232         PR target/80671
32233         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32234         Move member access before delete.
32236 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32238         * tree-inline.c (expand_call_inline): Split block at stmt
32239         before the call.
32241 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32243         PR target/68163
32244         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32245         are now unused after splitting mov{sf,sd}_hardfloat.
32246         (f32_lr2): Likewise.
32247         (f32_lm): Likewise.
32248         (f32_lm2): Likewise.
32249         (f32_li): Likewise.
32250         (f32_li2): Likewise.
32251         (f32_lv): Likewise.
32252         (f32_sr): Likewise.
32253         (f32_sr2): Likewise.
32254         (f32_sm): Likewise.
32255         (f32_sm2): Likewise.
32256         (f32_si): Likewise.
32257         (f32_si2): Likewise.
32258         (f32_sv): Likewise.
32259         (f32_dm): Likewise.
32260         (f32_vsx): Likewise.
32261         (f32_av): Likewise.
32262         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32263         For movsf, order stores so the VSX stores occur before the GPR
32264         store which encourages the register allocator to use a traditional
32265         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32266         store comes before the VSX stores to allow the power6 to work.
32267         This is due to the power6 not having a 32-bit integer store
32268         instruction from a FPR.
32269         (movsf_hardfloat): Likewise.
32270         (movsd_hardfloat): Likewise.
32272 2017-05-09  Martin Sebor  <msebor@redhat.com>
32274         PR translation/80280
32275         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32276         added in r247778.
32278         PR translation/80280
32279         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32280         data member added in r247778.
32281         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32283 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32285         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32287         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32288         typedefs.
32290 2017-05-09  Marek Polacek  <polacek@redhat.com>
32292         * doc/invoke.texi: Fix typo.
32294 2017-05-09  Richard Biener  <rguenther@suse.de>
32296         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32297         (vrp_val_is_min): Likewise.
32298         (set_value_range_to_value): Likewise.
32299         (set_value_range_to_nonnegative): Likewise.
32300         (gimple_assign_nonzero_p): Likewise.
32301         (gimple_stmt_nonzero_p): Likewise.
32302         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32303         (adjust_range_with_scev): Adjust comments.
32304         (compare_range_with_value): Likewise.
32305         (extract_range_from_phi_node): Likewise.
32306         (test_for_singularity): Likewise.
32308 2017-05-09  Richard Biener  <rguenther@suse.de>
32310         * tree-vrp.c (get_single_symbol): Add assert that we don't
32311         get overflowed constants as invariant part.
32312         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32313         checks.  Use wi::cmp instead of recursing for integer constants.
32314         (compare_values): Just ignore whether we assumed undefined
32315         overflow instead of failing the compare.
32316         (extract_range_for_var_from_comparison_expr): Add comment before the
32317         TREE_NO_WARNING sets.
32318         (test_for_singularity): Likewise.
32319         (extract_range_from_comparison): Do not disable optimization
32320         when we assumed undefined overflow.
32321         (extract_range_basic): Remove init of unused var.
32323 2017-05-09  Richard Biener  <rguenther@suse.de>
32325         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32326         (extract_range_from_multiplicative_op_1): Adjust.
32327         (extract_range_from_binary_expr_1): Use int_const_binop.
32329 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32331         PR target/80101
32332         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32333         rs6000_store_data_bypass_p in seven define_bypass directives and
32334         in several comments.
32335         * config/rs6000/rs6000-protos.h: Add prototype for
32336         rs6000_store_data_bypass_p function.
32337         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32338         function implements slightly different (rs6000-specific) semantics
32339         than store_data_bypass_p, returning false rather than aborting
32340         with assertion error when arguments do not satisfy the
32341         requirements of store data bypass.
32342         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32343         rs6000_store_data_bypass_p.
32345 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32347         * config/xtensa/xtensa-protos.h
32348         (xtensa_initial_elimination_offset): New declaration.
32349         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32350         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32351         macro definition, add case for FRAME_POINTER_REGNUM when
32352         FRAME_GROWS_DOWNWARD.
32353         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32354         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32355         xtensa_initial_elimination_offset.
32357 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32359         * doc/invoke.texi: Alphabetize -fdump options.
32361 2017-05-08  Martin Sebor  <msebor@redhat.com>
32363         PR translation/80280
32364         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32366 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32368         * target.def (compute_frame_layout): New optional target hook.
32369         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32370         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32371         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32372         target hook.
32373         * reload1.c (verify_initial_elim_offsets): Likewise.
32374         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32375         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32376         (arm_get_frame_offsets): Split up into this ...
32377         (arm_compute_frame_layout): ... and this function.
32379 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32381         * config/aarch64/constraints.md (Usa): New constraint.
32382         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32384 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32386         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32387         with_multilib_list after it has been checked.
32389 2017-05-08  Richard Biener  <rguenther@suse.de>
32391         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32392         (bitmap_set_subtract_values): Likewise.
32394 2017-05-08  Richard Biener  <rguenther@suse.de>
32396         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32397         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32398         argument.
32399         (gimple_stmt_nonzero_warnv_p): Rename to ...
32400         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32401         argument.
32402         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32403         (extract_range_basic): Adjust, do not disable propagation on
32404         strict overflow sensitive simplification.
32405         (vrp_visit_cond_stmt): Likewise.
32407 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32409         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32410         body size unconditionally.
32412 2017-05-07  Jeff Law  <law@redhat.com>
32414         Revert:
32415         2017-05-06  Jeff Law  <law@redhat.com>
32416         PR tree-optimization/78496
32417         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32418         code.
32420         PR tree-optimization/78496
32421         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32422         (simplify_stmt_using_ranges): Call it.
32423         (vrp_dom_walker::before_dom_children): Extract equivalences
32424         from an ASSERT_EXPR with an equality comparison against a
32425         constant.
32427 2017-05-06  Jeff Law  <law@redhat.com>
32429         PR tree-optimization/78496
32430         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32431         code.
32433         PR tree-optimization/78496
32434         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32435         (simplify_stmt_using_ranges): Call it.
32436         (vrp_dom_walker::before_dom_children): Extract equivalences
32437         from an ASSERT_EXPR with an equality comparison against a
32438         constant.
32440 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32442         * lra-constraints.c (lra_copy_reg_equiv): New function.
32443         (split_reg): Use it to copy equivalence information from the
32444         original register to the spill register.
32446 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32448         PR rtl-optimization/75964
32449         * simplify-rtx.c (simplify_const_relational_operation): Remove
32450         invalid handling of comparisons of integer ABS.
32452 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32454         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32455         initialize to zero.
32456         (init_regs): Remove declaration.
32457         (function_arg_advance_32): Initialize error_p as boolean variable.
32459 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32461         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32462         lines.  Use for (;;).
32464 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32466         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32467         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32468         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32469         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32470         VF=2 that require versioning.
32472 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32474         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32475         int.
32477 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32479         * diagnostic.h (diagnostic_override_option_index): Convert from
32480         macro to inline function.
32482 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32484         * diagnostic.c (last_module_changed_p): New function.
32485         (set_last_module): New function.
32486         (diagnostic_report_current_module): Convert macro usage to
32487         the above functions.
32488         * diagnostic.h (diagnostic_context::last_module): Strengthen
32489         from const line_map * to const line_map_ordinary *.
32490         (diagnostic_last_module_changed): Delete macro.
32491         (diagnostic_set_last_module): Delete macro.
32493 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32495         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32496         with diagnostic_report_diagnostic.
32497         (diagnostic_n_impl_richloc): Likewise.
32498         * diagnostic.h (report_diagnostic): Delete macro.
32499         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32500         with diagnostic_report_diagnostic.
32501         * substring-locations.c (format_warning_va): Likewise.
32503 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32505         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32506         save/restor of format_spec.  Move option-printing code to...
32507         (print_option_information): ...this new function, and
32508         reimplement by simply printing to the pretty_printer,
32509         rather than appending to the format string.
32511 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32513         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32514         handling logic into...
32515         (update_effective_level_from_pragmas): ...this new function.
32517 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32519         * config/riscv/riscv.opt (mstrict-align): New option.
32520         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32521         (SLOW_UNALIGNED_ACCESS): Define.
32522         (riscv_slow_unaligned_access): Declare.
32523         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32524         field.
32525         (riscv_slow_unaligned_access): New variable.
32526         (rocket_tune_info): Set slow_unaligned_access to true.
32527         (optimize_size_tune_info): Set slow_unaligned_access to false.
32528         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32529         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32530         (riscv_option_override): Set riscv_slow_unaligned_access.
32531         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32533 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32535         * config/riscv/riscv.md: Unify indentation.
32537 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32539         PR target/79038
32540         PR target/79202
32541         PR target/79203
32542         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32543         UNSIGNED_FIX.
32544         (extendsi<mode>2): Add support for doing sign extension via
32545         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32546         don't have ISA 3.0 instructions.
32547         (extendsi<mode>2 splitter): Likewise.
32548         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32549         generate the normal insns since SImode can now go in vector
32550         registers.  Disallow the special UNSPECs needed for previous
32551         machines to hide SImode being used.  Add new insns
32552         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32553         (fix_trunc<mode>si2_stfiwx): Likewise.
32554         (fix_trunc<mode>si2_internal): Likewise.
32555         (fixuns_trunc<mode>si2): Likewise.
32556         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32557         (fctiw<u>z_<mode>_smallint): Likewise.
32558         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32559         of floating point to 32-bit integer from doing a direct move to
32560         the GPR registers to do a store.
32561         (fctiwz_<mode>): Break long line.
32563 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32565         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32566         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32567         (addr_list, addr_offset_valid_p): New.
32568         (split_address_groups): Check offset validity with above function.
32569         (gt-tree-ssa-loop-ivopts.h): Include header file.
32571 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32573         * config.gcc (arm*-*-*): Add missing 'fi'.
32575 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32577         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32578         included in -fopt-info does not matter.
32579         * doc/optinfo.texi (-fopt-info): Fix description of default
32580         behavour. Explicitly say order of options included in -fopt-info
32581         does not matter.
32583 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32585         * config.gcc: Allow combinations of aprofile and rmprofile values for
32586         --with-multilib-list.
32587         * config/arm/t-multilib: New file.
32588         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32589         variables.  Remove setting of ISA and floating-point ABI in
32590         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32591         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32592         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32593         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32594         CPU options.
32595         * config/arm/t-rmprofile: Likewise except for the matches changes.
32596         * doc/install.texi (--with-multilib-list): Document the combination of
32597         aprofile and rmprofile values and warn about pitfalls in doing that.
32599 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32601         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32602         (movdi_aarch64): Likewise.
32604 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32606         PR tree-optimization/80632
32607         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32608         field.
32609         (build_arrays): Initialize it for virtual phis.
32610         (fix_phi_nodes): Use it for virtual phis.
32612         PR tree-optimization/80558
32613         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32614         [x, y] op z into [x op, y op z] for op & or | if conditions
32615         are met.
32617 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32618             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32620         PR target/71607
32621         * config/arm/arm.md (use_literal_pool): Remove.
32622         (64-bit immediate split): No longer takes cost into consideration
32623         if arm_disable_literal_pool is enabled.
32624         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32625         used when arm_disable_literal_pool is enabled.
32626         (arm_max_const_double_inline_cost): Remove use of
32627         arm_disable_literal_pool.
32628         (push_minipool_fix): Add assert.
32629         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32630         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32631         (no_literal_pool_sf_immediate): New.
32633 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32635         PR tree-optimization/80613
32636         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32637         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32639 2017-05-05  Richard Biener  <rguenther@suse.de>
32641         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32643 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32645         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32646         of this flag from insn conditions due to removal from r247495.
32648 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32650         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32651         New function.
32652         (arm_early_store_addr_dep_ptr): Likewise.
32653         * config/arm/aarch-common-protos.h
32654         (arm_early_load_addr_dep_ptr): Add prototype.
32655         (arm_early_store_addr_dep_ptr): Likewise.
32656         * config/arm/cortex-a53.md: Add new bypasses.
32658 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32660         * tree.c (next_type_uid): Change type to unsigned.
32661         (type_hash_canon): Decrement back next_type_uid if
32662         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32663         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32664         if possible.
32666 2017-05-04  Martin Sebor  <msebor@redhat.com>
32668         * builtins.c: Fix a trivial typo in a comment.
32670         PR middle-end/79234
32671         * builtins.c (check_sizes): Adjust to handle reading past the end.
32672         Avoid printing excessive upper bound of ranges.  Use %E to print
32673         tree nodes instead of converting them to %wu.
32674         (expand_builtin_memchr): New function.
32675         (compute_dest_size): Rename...
32676         (compute_objsize): ...to this.
32677         (expand_builtin_memcpy): Adjust.
32678         (expand_builtin_mempcpy): Adjust.
32679         (expand_builtin_strcat): Adjust.
32680         (expand_builtin_strcpy): Adjust.
32681         (check_strncat_sizes): Adjust.
32682         (expand_builtin_strncat): Adjust.
32683         (expand_builtin_strncpy): Adjust and simplify.
32684         (expand_builtin_memset): Adjust.
32685         (expand_builtin_bzero): Adjust.
32686         (expand_builtin_memcmp): Adjust.
32687         (expand_builtin): Handle memcmp.
32688         (maybe_emit_chk_warning): Check strncat just once.
32690 2017-05-04  Martin Sebor  <msebor@redhat.com>
32692         PR preprocessor/79214
32693         PR middle-end/79222
32694         PR middle-end/79223
32695         * builtins.c (check_sizes): Add inlining context and issue
32696         warnings even when -Wno-system-headers is set.
32697         (check_strncat_sizes): Same.
32698         (expand_builtin_strncat): Same.
32699         (expand_builtin_memmove): New function.
32700         (expand_builtin_stpncpy): Same.
32701         (expand_builtin): Handle memmove and stpncpy.
32703 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32705         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32706         which is not used any more.
32708 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32710         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32712 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32714         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32715         (cortexa53_tunings): Likewise.
32716         (cortexa57_tunings): Likewise.
32717         (cortexa72_tunings): Likewise.
32718         (cortexa73_tunings): Likewise.
32720 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32722         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32723         Set loop alignment to 8.
32725 2017-05-04  Martin Sebor  <msebor@redhat.com>
32727         PR translation/80280
32728         * builtins.c (expand_builtin_object_size): Add missing quoting to
32729         %D and like directives.
32730         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32731         (hsa_type_for_tree_type): Same.
32732         (verify_function_arguments): Same.
32733         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32734         * varasm.c (get_section): Same.
32735         (mark_weak): Same.
32737 2017-05-04  Martin Sebor  <msebor@redhat.com>
32739         PR translation/80280
32740         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32742 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32744         * config/aarch64/aarch64.c (generic_addrcost_table):
32745         Change HI/TI mode setting.
32747 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32749         PR tree-optimization/80622
32750         * tree-sra.c (comes_initialized_p): New function.
32751         (build_accesses_from_assign): Only set write lazily when
32752         comes_initialized_p is false.
32753         (analyze_access_subtree): Use comes_initialized_p.
32754         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32755         instead of testing for PARM_DECL.
32757 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32759         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32760         constraint on operand 0 to allow more general addressing modes.
32761         Adjust output template.
32762         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32763         New function.
32764         * config/aarch64/aarch64-protos.h
32765         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32766         * config/aarch64/constraints.md (Dp): New address constraint.
32767         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32768         predicate.
32770 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32772         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32773         update use of estimate_ipcp_clone_size_and_time.
32774         (estimate_local_effects): Update use of
32775         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32776         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32777         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32778         Return nonspecialized time.
32780 2017-05-04  Richard Biener  <rguenther@suse.de>
32782         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32783         for the last VUSE which def dominates the PHI.  Directly call
32784         maybe_skip_until.
32785         (get_continuation_for_phi_1): Remove.
32787 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32789         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32790         to explain the use of truncating division.  Cap the number of
32791         iterations to the maximum given by nb_iterations_upper_bound,
32792         if defined.
32794 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32796         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32797         * configure: Regenerate.
32798         * config.in: Regenerate.
32799         * config/i386/driver-mingw32.c: new file.
32800         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32801         * config.host: Link driver-mingw32.o on MinGW host.
32802         * doc/install.texi: Document new --enable-mingw-wildcard configure
32803         option.
32805 2017-05-04  Marek Polacek  <polacek@redhat.com>
32807         PR tree-optimization/80612
32808         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32810 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32811             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32813         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32814         (movt splitter): Likewise.
32815         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32816         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32817         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32818         block for Thumb-1 with MOVT.
32819         (thumb2_legitimate_address_p): Move code block ...
32820         (can_avoid_literal_pool_for_label_p): ... into this new function.
32821         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32822         literal pool.
32823         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32824         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32825         "M-profile targets with the MOVT instruction".
32827 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32829         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32830         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32831         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32833 2017-05-04  Martin Liska  <mliska@suse.cz>
32835         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32836         variable cond_code.
32838 2017-05-04  Richard Biener  <rguenther@suse.de>
32840         * tree.c (array_at_struct_end_p): Handle arrays at struct
32841         end with flexarrays more conservatively.  Refactor and treat
32842         arrays of arrays or aggregates more strict.  Fix
32843         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32844         * tree.c (array_at_struct_end_p): Adjust prototype.
32845         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32846         * gimple-fold.c (get_range_strlen): Likewise.
32847         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32849 2017-05-04  Richard Biener  <rguenther@suse.de>
32851         PR tree-optimization/31130
32852         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32853         false.
32854         (supports_overflow_infinity): Likewise.
32855         (is_negative_overflow_infinity): Likewise.
32856         (is_positive_overflow_infinity): Likewise.
32857         (is_overflow_infinity): Likewise.
32858         (stmt_overflow_infinity): Likewise.
32859         (overflow_infinity_range_p): Likewise.
32860         (usable_range_p): Remove as always returning true.
32861         (make_overflow_infinity): Remove.
32862         (negative_overflow_infinity): Likewise.
32863         (positive_overflow_infinity): Likewise.
32864         (avoid_overflow_infinity): Likewise.
32865         (set_value_range): Adjust accordingly.
32866         (set_value_range_to_nonnegative): Likewise, remove now unused
32867         overflow_infinity arg.
32868         (vrp_operand_equal_p): Adjust.
32869         (update_value_range): Likewise.
32870         (range_int_cst_singleton_p): Likewise.
32871         (operand_less_p): Likewise.
32872         (compare_values_warnv): Likewise.
32873         (extract_range_for_var_from_comparison_expr): Likewise.
32874         (vrp_int_const_binop): Likewise.
32875         (zero_nonzero_bits_from_vr): Likewise.
32876         (extract_range_from_multiplicative_op_1): Likewise.
32877         (extract_range_from_binary_expr_1): Likewise.
32878         (extract_range_from_unary_expr): Likewise.
32879         (extract_range_from_comparison): Likewise.
32880         (extract_range_basic): Likewise.
32881         (adjust_range_with_scev): Likewise.
32882         (compare_ranges): Likewise.
32883         (compare_range_with_value): Likewise.
32884         (dump_value_range): Likewise.
32885         (test_for_singularity): Likewise, remove strict_overflow_p parameter
32886         never used.
32887         (simplify_cond_using_ranges): Adjust.
32889 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
32891         * brig-builtins.def: Added a builtin for class_f64.
32892         * builtin-types.def: Added a builtin type needed by class_f64.
32894 2017-05-03  Jason Merrill  <jason@redhat.com>
32896         * timevar.def: Add TV_CONSTEXPR.
32898 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
32900         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
32902 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32904         * ipa-prop.c (ipa_update_after_lto_read): Removed.
32905         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
32906         * ipa-cp.c (ipcp_propagate_stage): Do not call
32907         ipa_update_after_lto_read.
32908         * ipa-inline.c (ipa_inline): Likewise.
32910 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32912         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
32913         tag.  Added a default constructor and a destructor.
32914         (ipa_edge_args_sum_t): New class;
32915         (ipa_edge_args_sum): Declare.
32916         (ipa_edge_args_vector): Remove declaration.
32917         (IPA_EDGE_REF): Use ipa_edge_args_sum.
32918         (ipa_free_edge_args_substructures): Remove declaration.
32919         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
32920         (ipa_edge_args_info_available_for_edge_p): Likewise.
32921         * ipa-prop.c (ipa_edge_args_vector): Removed.
32922         (edge_removal_hook_holder): Likewise.
32923         (edge_duplication_hook_holder): Likewise.
32924         (ipa_edge_args_sum): New variable.
32925         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
32926         ipa_edge_args_vector.
32927         (ipa_free_edge_args_substructures): Likewise.
32928         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
32929         ipa_edge_args_vector.
32930         (ipa_edge_removal_hook): Turned into method
32931         ipa_edge_args_sum_t::remove.
32932         (ipa_edge_duplication_hook): Turned into method
32933         ipa_edge_args_sum_t::duplicate.
32934         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
32935         registering edge hooks.
32936         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
32937         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
32938         ipa_edge_args_sum instead of ipa_edge_args_vector.
32939         * ipa-profile.c (ipa_profile): Likewise.
32941 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32943         * symbol-summary.h (function_summary): New method exists.
32944         (function_summary::symtab_removal): Deallocate through release.
32945         (call_summary): New class.
32946         (gt_ggc_mx): New overload.
32947         (gt_pch_nx): Likewise.
32948         (gt_pch_nx): Likewise.
32950 2017-05-03  Jeff Law  <law@redhat.com>
32952         PR tree-optimization/78496
32953         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
32954         from simplify_cond_using_ranges.  Split off code to walk
32955         backwards through casts into ...
32956         (simplify_cond_using_ranges_2): New function.
32957         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
32958         (execute_vrp): After identifying jump threads, call
32959         simplify_cond_using_ranges_2.
32961 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
32963         PR bootstrap/80609
32964         * ipa-inline.h (inline_summary): Add ctor.
32965         (create_ggc): Do not use ggc_cleared_alloc.
32967 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
32968             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
32970         * gcc.c (handle_braces): Support escaping in switch matching
32971         text.
32972         * doc/invoke.texi (Spec Files): Document it.
32973         Remove superfluous @code markup in items.
32975 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
32977         * diagnostic-show-locus.c (struct column_range): New struct.
32978         (get_affected_columns): New function.
32979         (get_printed_columns): New function.
32980         (struct correction): New struct.
32981         (correction::ensure_capacity): New function.
32982         (correction::ensure_terminated): New function.
32983         (struct line_corrections): New struct.
32984         (line_corrections::~line_corrections): New dtor.
32985         (line_corrections::add_hint): New function.
32986         (layout::print_trailing_fixits): Reimplement in terms of the new
32987         classes.
32988         (selftest::test_overlapped_fixit_printing): New function.
32989         (selftest::diagnostic_show_locus_c_tests): Call it.
32991 2017-05-03  Nathan Sidwell  <nathan@acm.org>
32993         Canonicalize canonical type hashing
32994         * tree.h (type_hash_canon_hash): Declare.
32995         * tree.c (type_hash_list, attribute_hash_list): Move into
32996         type_hash_canon_hash.
32997         (build_type_attribute_qual_variant): Break out hash code calc into
32998         type_hash_canon_hash.
32999         (type_hash_canon_hash): New.  Generic type hash computation.
33000         (build_range_type_1, build_array_type_1, build_function_type,
33001         build_method_type_directly, build_offset_type, build_complex_type,
33002         make_vector_type): Call it.
33004 2017-05-03  Richard Biener  <rguenther@suse.de>
33006         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33007         When all DRs have unknown misaligned do not always peel
33008         when there is a store but apply the same costing model as if
33009         there were only loads.
33011 2017-05-03  Richard Biener  <rguenther@suse.de>
33013         Revert
33014         PR tree-optimization/80492
33015         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33016         compare_base_decls returning dont-know properly.
33018 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33020         * config/arm/iterators.md (CCSI): New mode iterator.
33021         (arch): New mode attribute.
33022         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33023         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33024         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33025         code iterator for success result mode.
33026         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33027         the corresponding new insn generators.
33029 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33031         Revert r247509
33032         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33033         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33035 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33037         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33038         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33039         (DDR_A): Wrap DDR argument in brackets.
33040         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33041         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33042         (DDR_REVERSED_P): Likewise.
33044 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33046         PR tree-optimization/79472
33047         * tree-switch-conversion.c (struct switch_conv_info): Add
33048         contiguous_range and default_case_nonstandard fields.
33049         (collect_switch_conv_info): Compute contiguous_range and
33050         default_case_nonstandard fields, don't clear final_bb if
33051         contiguous_range and only the default case doesn't have the required
33052         structure.
33053         (check_all_empty_except_final): Set default_case_nonstandard instead
33054         of failing if contiguous_range and the default case doesn't have empty
33055         block.
33056         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33057         and only the default case doesn't have the required constants.  Skip
33058         virtual phis.
33059         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33060         if default_case_nonstandard.
33061         (build_constructors): Build constant 1 just once.  Assert that default
33062         values aren't inserted in between cases if contiguous_range.  Skip
33063         virtual phis.
33064         (build_arrays): Skip virtual phis.
33065         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33066         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33067         Handle virtual phis.
33068         (gen_inbound_check): Handle default_case_nonstandard case.
33069         (process_switch): Adjust check_final_bb caller.  Call
33070         gather_default_values with the first non-default case instead of
33071         default case if default_case_nonstandard.
33073 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33075         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33076         check.  Fix formatting.
33078 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33080         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33081         errors when comparing specialized and unspecialized times.
33083 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33085         * diagnostic-show-locus.c
33086         (layout::should_print_annotation_line_p): Make private.
33087         (layout::print_annotation_line): Make private.
33088         (layout::annotation_line_showed_range_p): Make private.
33089         (layout::show_ruler): Make private.
33090         (layout::print_source_line): Make private.  Pass in line and
33091         line_width, rather than calling location_get_source_line.  Drop
33092         returned value.
33093         (layout::print_leading_fixits): New method.
33094         (layout::print_any_fixits): Rename to...
33095         (layout::print_trailing_fixits): ...this, and make private.
33096         Don't print newline fixits.
33097         (diagnostic_show_locus): Move logic for printing one row into...
33098         (layout::print_line): ...this new function.  Move the
33099         location_get_source_line call and error-handling from
33100         print_source_line to here.  Call print_leading_fixits, and rename
33101         print_any_fixits to print_trailing_fixits.
33102         (selftest::test_fixit_insert_containing_newline): Update now that
33103         newlines are partially supported.
33104         (selftest::test_fixit_insert_containing_newline_2): New test.
33105         (selftest::test_fixit_replace_containing_newline): Update comments.
33106         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33107         * edit-context.c (class added_line): New class.
33108         (class edited_line): Describe newline handling in comment.
33109         (edited_line::actually_edited_p): New method.
33110         (edited_line::print_content): Delete redundant decl.
33111         (edited_line::m_predecessors): New field.
33112         (edited_file::print_content): Call edited_line::print_content.
33113         (edited_file::print_diff): Update to support newlines.
33114         (edited_file::print_diff_hunk): Likewise.
33115         (edited_file::print_run_of_changed_lines): New function.
33116         (edited_file::print_diff_line): Convert to...
33117         (print_diff_line): ...this.
33118         (edited_file::get_effective_line_count): New function.
33119         (edited_line::edited_line): Initialize new field m_predecessors.
33120         (edited_line::~edited_line): Clean up m_predecessors.
33121         (edited_line::apply_fixit): Handle newlines.
33122         (edited_line::get_effective_line_count): New function.
33123         (edited_line::print_content): New function.
33124         (edited_line::print_diff_lines): New function.
33125         (selftest::test_applying_fixits_insert_containing_newline): New
33126         test.
33127         (selftest::test_applying_fixits_replace_containing_newline): New
33128         test.
33129         (selftest::insert_line): New function.
33130         (selftest::test_applying_fixits_multiple_lines): Add example of
33131         inserting a line.
33132         (selftest::edit_context_c_tests): Call the new tests.
33134 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33136         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33137         parameter cand.  Update dump information.
33138         (get_computation_cost): Update uses.
33140 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33142         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33143         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33144         (get_computation_at, rewrite_use_address): Update use of
33145         get_computation_aff.
33147 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33149         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33150         (get_computation): Delete.
33151         (get_computation_cost): Implement like get_computation_cost_at.
33152         Use get_computation_at.
33153         (get_computation_cost_at): Delete.
33154         (rewrite_use_nonlinear_expr): Use get_computation_at.
33155         (rewrite_use_compare, remove_unused_ivs): Ditto.
33157 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33159         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33161 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33163         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33164         (ivopts_global_cost_for_size): Rename parameter and update uses.
33165         (iv_ca_recount_cost): Update uses.
33166         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33167         candidates seperately in n_invs and n_cands.
33168         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33170 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33172         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33173         (find_inv_vars_cb): New.
33174         (find_depends): Renamed to ...
33175         (find_inv_vars): ... this.
33176         (add_candidate_1, force_var_cost): Call find_inv_vars.
33177         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33179 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33181         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33182         inv_vars.  Add inv_exprs.
33183         (struct iv_cand): Rename depends_on to inv_vars.
33184         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33185         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33186         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33187         (dump_cand): Dump inv_vars.
33188         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33189         (record_invariant, find_depends, add_candidate_1): Ditto.
33190         (set_group_iv_cost, force_var_cost): Ditto.
33191         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33192         (get_computation_cost_at, get_computation_cost): Ditto.
33193         (determine_group_iv_cost_generic): Ditto.
33194         (determine_group_iv_cost_address): Ditto.
33195         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33196         (determine_group_iv_costs): Ditto.
33197         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33198         (iv_ca_set_remove_invariants): Renamed to ...
33199         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33200         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33201         (iv_ca_set_add_invariants):  Renamed to ...
33202         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33203         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33204         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33205         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33206         (create_new_ivs): Remove useless dump.
33208 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33210         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33211         iv_cand code.
33212         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33213         (iv_ca_set_no_cp, create_new_iv): Ditto.
33215 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33217         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33219 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33221         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33222         function tree_check2.
33224 2017-05-02  Martin Liska  <mliska@suse.cz>
33226         * doc/gcov.texi: Add missing preposition.
33227         * gcov.c (function_info::function_info): Properly fill up
33228         all member variables.
33230 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33232         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33234 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33236         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33238 2017-05-02  Martin Liska  <mliska@suse.cz>
33240         PR lto/77954.
33241         * lto-streamer-in.c (lto_read_tree_1): Remove
33242         LTO_STREAMER_DEBUG.
33243         * lto-streamer.c (struct tree_hash_entry): Likewise.
33244         (struct tree_entry_hasher): Likewise.
33245         (tree_entry_hasher::hash): Likewise.
33246         (tree_entry_hasher::equal): Likewise.
33247         (lto_streamer_init): Likewise.
33248         (lto_orig_address_map): Likewise.
33249         (lto_orig_address_get): Likewise.
33250         (lto_orig_address_remove): Likewise.
33251         * lto-streamer.h: Likewise.
33252         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33253         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33255 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33257         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33258         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33259         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33260         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33261         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33262         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33263         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33264         (mm_maskz_sub_ss): New intrinsics.
33265         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33266         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33267         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33268         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33269         (__builtin_ia32_subss_mask_round): New builtins.
33270         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33271         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33272         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33273         Renamed to ...
33274         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33275         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33276         Changed to ...
33277         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33278         ... this.
33280 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33282         PR tree-optimization/78687
33283         * tree-sra.c (access): New field parent.
33284         (process_subtree_disqualification): New function.
33285         (disqualify_candidate): Call it.
33286         (build_accesses_from_assign): Reset write flag if creating an
33287         assighnment link.
33288         (build_access_subtree): Fill in parent field and also prpagate
33289         down grp_write flag.
33290         (create_artificial_child_access): New parameter set_grp_write, set
33291         grp_write to its value.
33292         (propagate_subaccesses_across_link): Also propagate grp_write flag
33293         values.
33294         (propagate_all_subaccesses): Push the closest parent back to work
33295         queue if add_access_to_work_queue returned true.
33297 2017-05-02  Richard Biener  <rguenther@suse.de>
33299         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33300         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33301         -fstrict-overflow documentation.
33302         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33303         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33304         flag_strict_overflow.
33305         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33306         * lto-opts.c (lto_write_options): Do not stream it.
33307         * lto-wrapper.c (merge_and_complain): Do not handle it.
33308         * opts.c (default_options_table): Do not set -fstrict-overflow.
33309         (finish_options): Likewise do not clear it when sanitizing.
33310         * simplify-rtx.c (simplify_const_relational_operation): Do not
33311         test flag_strict_overflow.
33313 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33315         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33316         using enabled attribute.
33317         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33318         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33319         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33320         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33321         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33322         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33323         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33324         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33325         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33326         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33328 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33330         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33332 2017-05-02  Richard Biener  <rguenther@suse.de>
33334         PR tree-optimization/80591
33335         Revert
33336         2017-04-10  Richard Biener  <rguenther@suse.de>
33338         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33339         asm inputs.
33341 2017-05-02  Richard Biener  <rguenther@suse.de>
33343         PR tree-optimization/80549
33344         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33345         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33346         headers if they do not have a preheader.
33348 2017-05-02  Martin Liska  <mliska@suse.cz>
33350         PR other/80589
33351         * common.opt: Fix typo.
33352         * doc/invoke.texi: Likewise.
33354 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33356         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33357         swapping, add (x,x,m,x,n) alternative.
33359 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33361         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33362         unnecessary unadjusted_alignment check.
33364 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33366         PR c++/80038
33367         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33368         operations here.
33369         * gimplify.c (gimplify_cilk_detach): New function.
33370         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33371         * tree-core.h: Document EXPR_CILK_SPAWN.
33372         * tree.h (EXPR_CILK_SPAWN): Define.
33374 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33376         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33377         to use new fixit_hint representation, using the "replace" logic.
33378         (get_line_span_for_fixit_hint): Likewise.
33379         (layout::print_any_fixits): Likewise.
33380         (selftest::test_one_liner_many_fixits): Rename to...
33381         (selftest::test_one_liner_many_fixits_1): ...this, and update
33382         comment and expected output to reflect that the multiple fix-it
33383         hints are now consolidated into one insertion.
33384         (selftest::test_one_liner_many_fixits_2): New test.
33385         (selftest::test_diagnostic_show_locus_one_liner): Update for
33386         above.
33387         (selftest::test_fixit_consolidation): Update for fix-it API
33388         change.
33389         * diagnostic.c (print_parseable_fixits): Likewise.
33390         * edit-context.c (edited_line::m_line_events): Convert from
33391         auto_vec <line_event *> to auto_vec <line_event>.
33392         (class line_event): Convert from abstract base class to a concrete
33393         class, taking over the role of replace_event.
33394         (class insert_event): Delete.
33395         (class replace_event): Rename to class line_event.  Convert to
33396         half-open range.
33397         (edit_context::add_fixits): Reimplement.
33398         (edit_context::apply_insert): Delete.
33399         (edit_context::apply_replace): Rename to...
33400         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33401         (edited_file::apply_insert): Delete.
33402         (edited_file::apply_replace): Rename to...
33403         (edited_file::apply_fixit): ...this.
33404         (edited_line::~edited_line): Drop deletion of events.
33405         (edited_line::apply_insert): Delete.
33406         (edited_line::apply_replace): Rename to...
33407         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33408         Update for change to type of m_line_events.
33409         * edit-context.h (edit_context::apply_insert): Delete.
33410         (edit_context::apply_replace): Rename to...
33411         (edit_context::apply_fixit): ...this.
33413 2017-05-01  Martin Sebor  <msebor@redhat.com>
33415         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33416         known.
33418 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33420         PR target/68491
33421         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33422         __get_cpuid_max returns 0.
33423         (__get_cpuid_count): Ditto.
33425 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33427         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33428         replacement expression is another instance of one of its arguments.
33430 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33432         PR target/79430
33433         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33434         check for stack push/pop autoinc.
33435         * config/i386/i386.c (ix86_agi_dependent): Return false
33436         if the only reason why modified_in_p returned true is that
33437         addr is SP based and set_insn is a push or pop.
33439 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33441         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33442         overflow check.
33444 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33446         PR ipa/79224
33447         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33448         (account_size_time): Use two predicates - exec_pred and
33449         nonconst_pred_ptr.
33450         (evaluate_conditions_for_known_args): Compute both clause and
33451         nonspec_clause.
33452         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33453         (inline_summary_t::duplicate): Update.
33454         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33455         separately.
33456         (compute_inline_parameters): Likewise.
33457         (estimate_edge_size_and_time): Update caluclation of time.
33458         (estimate_node_size_and_time): Compute both time and nonspecialized
33459         time.
33460         (estimate_ipcp_clone_size_and_time): Update.
33461         (inline_merge_summary): Update.
33462         (do_estimate_edge_time): Update.
33463         (do_estimate_edge_size): Update.
33464         (do_estimate_edge_hints): Update.
33465         (inline_read_section, inline_write_summary): Stream both new predicates.
33466         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33467         as argument.
33468         (compute_inlined_call_time): Cleanup.
33469         (big_speedup_p): Update.
33470         (edge_badness): Update.
33471         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33472         (size_time_entry): Replace predicate by exec_predicate and
33473         nonconst_predicate.
33474         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33475         (estimate_edge_time): Return also nonspec_time.
33476         (reset_edge_growth_cache): Update.
33478 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33480         PR rtl-optimization/80491
33481         * ifcvt.c (noce_process_if_block): When looking for x setter
33482         with missing else_bb, don't check only the insn right before
33483         cond_earliest, but look for the last insn that x is modified in
33484         within the same bb.
33486         PR rtl-optimization/80491
33487         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33489 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33491         PR tree-optimization/80487
33492         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33494 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33496         PR tree-optimization/79697
33497         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33498         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33499         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33500         BUILT_IN_STRNDUP.
33501         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33502         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33504 2017-04-28  Martin Sebor  <msebor@redhat.com>
33506         PR tree-optimization/80523
33507         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33508         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33509         functions.
33510         (maybe_warn, format_directive, parse_directive): Use new functions.
33511         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33513 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33515         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33517 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33519         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33520         target_header_dir): Set correctly.
33521         * configure: Regenerated.
33522         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33523         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33524         instead of SYSTEM_HEADER_DIR.
33526 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33528         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33529         (estimate_local_effects): Likewise.
33530         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33531         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33532         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33533         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33534         do_estimate_edge_time, estimate_edge_time): Likewise.
33535         * ipa-inline-analysis.c (estimate_node_size_and_time,
33536         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33537         (estimate_time_after_inlining): Remove.
33539 2017-04-28  Martin Liska  <mliska@suse.cz>
33541         * doc/gcov.texi: Enhance documentation of gcov.
33543 2017-04-28  Martin Liska  <mliska@suse.cz>
33545         * doc/gcov.texi: Sort options in alphabetic order.
33546         * doc/gcov-dump.texi: Likewise.
33547         * doc/gcov-tool.texi: Likewise.
33548         * gcov.c (print_usage): Likewise.
33549         * gcov-dump.c (print_usage): Likewise.
33550         * gcov-tool.c (print_merge_usage_message): Likewise.
33551         (print_rewrite_usage_message): Likewise.
33552         (print_overlap_usage_message): Likewise.
33554 2017-04-28  Martin Liska  <mliska@suse.cz>
33556         PR gcov-profile/53915
33557         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33559 2017-04-28  Martin Liska  <mliska@suse.cz>
33561         PR gcov-profile/79891
33562         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33563         is marked by compiler as living on a line.
33564         (get_cycles_count): Remove usage of the union.
33565         (output_intermediate_file): Likewise.
33566         (find_source): Fix GNU coding style.
33567         (accumulate_line_counts): Remove old non-all block mode.
33568         (output_lines): Remove usage of the union.
33569         * profile.c (output_location): Include all BBs, even if
33570         belonging to a same line (and file) as a previous BB.
33572 2017-04-28  Martin Liska  <mliska@suse.cz>
33574         * gcov.c (process_args): Handle new argument 'w'.
33575         (read_graph_file): Assign ID to BBs.
33576         (output_branch_count): Display BB # if verbose flag is set.
33577         (output_lines): Likewise for arcs.
33578         (print_usage): Add '--verbose' option help.
33579         * doc/gcov.texi: Document --verbose (-w) option.
33581 2017-04-28  Martin Liska  <mliska@suse.cz>
33583         * gcov.c (struct block_location_info): New struct.
33584         (process_file): Fill up the new structure.
33585         (read_graph_file): Replace usage of encoding by the newly added
33586         struct.
33587         (add_line_counts): Likewise.
33588         (accumulate_line_counts): Remove usage of the union.
33589         (function_info::function_info): New function.
33590         (function_info::~function_info): Likewise.
33591         (process_file): Call delete instead of release_function.
33592         (release_function): Release the function.
33593         (release_structures): Call delete instead of release_function.
33594         (solve_flow_graph): Replace usage of num_blocks.
33595         (find_exception_blocks): Likewise.
33596         (output_lines): Fix GNU coding style.
33598 2017-04-28  Martin Liska  <mliska@suse.cz>
33600         PR driver/56469
33601         * coverage.c (coverage_remove_note_file): New function.
33602         * coverage.h: Declare the function.
33603         * toplev.c (finalize): Clean if an error has been seen.
33605 2017-04-28  Martin Liska  <mliska@suse.cz>
33607         PR gcov-profile/80031
33608         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33609         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33610         * gcov.c (read_graph_file): Read just number of blocks.
33611         * profile.c (branch_prob): Do not stream 0 flags per a basic
33612         block.
33614 2017-04-28  Martin Liska  <mliska@suse.cz>
33616         * gcov-dump.c (tag_*): Add new argument to declarations.
33617         (dump_gcov_file): Likewise.
33618         (tag_blocks): Add and use new argument depth.
33619         (tag_arcs): Likewise.
33620         (tag_lines): Likewise.
33621         (tag_counters): Likewise.
33622         (tag_summary): Likewise.
33623         (dump_working_sets): Use depth to do a proper indentation.
33625 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33627         PR bootstrap/80531
33628         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33629         * symtab.c (symtab_node::debug_symtab): Move definition here.
33631 2017-04-28  Richard Biener  <rguenther@suse.de>
33633         * lto-streamer.h (LTO_major_version): Bump to 7.
33635 2017-04-28  Richard Biener  <rguenther@suse.de>
33637         * tree-vrp.c (assert_info): New struct.
33638         (add_assert_info): New helper.
33639         (register_edge_assert_for_2): Refactor to add asserts to a vector
33640         of assert_info.
33641         (register_edge_assert_for_1): Likewise.
33642         (register_edge_assert_for): Likewise.
33643         (finish_register_edge_assert_for): New helper actually registering
33644         asserts where live on edge.
33645         (find_conditional_asserts): Adjust.
33646         (find_switch_asserts): Likewise.
33647         (evrp_dom_walker::try_find_new_range): Generalize.
33648         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33650 2017-04-27  Marek Polacek  <polacek@redhat.com>
33652         PR sanitizer/80349
33653         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33654         arg10 and arg11 to itype.
33656 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33658         * doc/extend.texi (Object Size Checking): Improve grammar.
33660 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33662         PR target/80530
33663         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33664         that the logic for permitting reciprocal estimates matches that
33665         in use_rsqrt_p.
33667 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33669         PR c++/80534
33670         * tree.c (type_cache_hasher::equal): Only compare
33671         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33672         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33673         non-aggregate element types.
33674         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33675         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33677 2017-04-27  Richard Biener  <rguenther@suse.de>
33679         PR middle-end/80533
33680         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33681         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33682         keeping a reference to a trailing array.
33684 2017-04-27  Richard Biener  <rguenther@suse.de>
33686         PR middle-end/80539
33687         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33688         being in loop-closed SSA form conservatively.
33689         (chrec_fold_multiply_poly_poly): Likewise.
33691 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33693         PR middle-end/79665
33694         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33695         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33697 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33699         PR target/77728
33700         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33701         (aarch64_function_arg_alignment): Return unsigned int again, but still
33702         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33703         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33704         Don't emit -Wpsabi note.
33705         (aarch64_function_arg_boundary): Likewise.
33706         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33707         caller.
33709 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33711         * tree.h (crc32_unsigned_n): Declare.
33712         (crc32_unsigned, crc32_unsigned): Make inline.
33713         * tree.c (crc32_unsigned_bits): Replace with ...
33714         (crc32_unsigned_n): ... this.
33715         (crc32_unsigned, crc32_byte): Remove.
33716         (crc32_string): Remove unnecessary braces.
33718 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33720         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33721         * ipa-inline-analysis.c (MAX_TIME): Remove.
33722         (account_size_time): Use sreal for time.
33723         (dump_inline_summary): Update.
33724         (estimate_function_body_sizes): Update.
33725         (estimate_edge_size_and_time): Update.
33726         (estimate_calls_size_and_time): Update.
33727         (estimate_node_size_and_time): Update.
33728         (inline_merge_summary): Update.
33729         (inline_update_overall_summary): Update.
33730         (estimate_time_after_inlining): Update.
33731         (inline_read_section): Update.
33732         (inline_write_summary): Update.
33733         * ipa-inline.c (compute_uninlined_call_time): Update.
33734         (compute_inlined_call_time): Update.
33735         (recursive_inlining): Update.
33736         (inline_small_functions): Update.
33737         (dump_overall_stats): Update.
33738         * ipa-inline.h: Include sreal.h.
33739         (size_time_entry): Turn time to sreal.
33740         (inline_summary): Turn self_time nad time to sreal.
33742 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33744         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33745         data-streamer.h
33746         (sreal::stream_out, sreal::stream_in): New.
33747         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33749 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33751         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33752         environment.
33754 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33756         PR target/70799
33757         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33758         Handle ASHIFTRT.
33759         (dimode_scalar_chain::compute_convert_gain): Ditto.
33760         (dimode_scalar_chain::make_vector_copies): Ditto.
33761         (dimode_scalar_chain::convert_reg): Ditto.
33762         (dimode_scalar_chain::convert_insn): Ditto.
33763         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33764         (VI248_AVX512BW_1): New mode iterator.
33765         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33766         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33767         mode iterator.
33769 2017-04-25  Martin Sebor  <msebor@redhat.com>
33771         PR tree-optimization/80497
33772         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33773         constants are representable in HOST_WIDE_INT.
33774         (parse_directive): Ditto.
33776 2017-04-25  Martin Sebor  <msebor@redhat.com>
33778         PR bootstrap/80486
33779         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33780         (new_zero_array): Adjust signature.
33781         (dom_info::dom_init): Used unsigned rather that size_t.
33782         (dom_info::dom_info): Same.
33784 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33785             Jakub Jelinek  <jakub@redhat.com>
33787         PR target/77728
33788         * config/arm/arm.c: Include gimple.h.
33789         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33790         returns negative, increment ncrn only if it returned positive.
33791         (arm_needs_doubleword_align): Return int instead of bool,
33792         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33793         members, but if there is any such non-FIELD_DECL
33794         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33795         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33796         returns negative, increment nregs only if it returned positive.
33797         (arm_setup_incoming_varargs): Likewise.
33798         (arm_function_arg_boundary): Emit -Wpsabi note if
33799         arm_needs_doubleword_align returns negative, return
33800         DOUBLEWORD_ALIGNMENT only if it returned positive.
33802 2017-04-25  Marek Polacek  <polacek@redhat.com>
33804         PR sanitizer/80349
33805         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33806         first argument to type.
33808 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33810         PR target/80482
33811         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33812         type checks to test for compatibility instead of equality.
33814 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33815             Jakub Jelinek  <jakub@redhat.com>
33817         PR target/77728
33818         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33819         type.
33820         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33821         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33822         the alignment computation, but return their maximum in warn_alignment.
33823         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33824         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33825         is smaller.
33826         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33827         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33828         caller.
33830 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33832         * config/arc/simdext.md (dmpyh): Fix typo.
33834 2017-04-25  Richard Biener  <rguenther@suse.de>
33836         PR tree-optimization/80492
33837         * alias.c (compare_base_decls): Handle registers with asm
33838         specification conservatively.
33839         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33840         compare_base_decls returning dont-know properly.
33842 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33844         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33845         (legitimate_offset_address_p): New function.
33846         (arc_legitimate_address_p): Use above function.
33848 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33850         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33852 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33854         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33855         ACCH registers whenever they are available.
33857 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33859         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33860         double regs fix when not used.
33862 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33864         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
33865         core registers.
33866         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
33867         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
33869 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33871         * config/arc/arc.c (arc_output_addsi): Check for h-register class
33872         when emitting short ADD instructions.
33874 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33876         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
33877         constraint.
33878         (cmpsi_cc_c_insn): Likewise.
33879         (cbranchsi4_scratch): Compute proper instruction length using
33880         compact_hreg_operand.
33881         * config/arc/predicates.md (compact_hreg_operand): New predicate.
33883 2017-04-25  Richard Biener  <rguenther@suse.de>
33885         PR middle-end/80509
33886         * passes.c (pass_manager::pass_manager): Initialize
33887         m_name_to_pass_map.
33889 2017-04-25  Richard Biener  <rguenther@suse.de>
33891         PR tree-optimization/79201
33892         * tree-ssa-sink.c (statement_sink_location): Handle calls.
33894 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33896         PR target/80464
33897         * config/s390/vector.md: Split MEM->GPR vector moves for
33898         non-s_operand addresses.
33900 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33902         PR target/79895
33903         * config/s390/predicates.md (reload_const_wide_int_operand): New
33904         predicate.
33905         * config/s390/s390.md ("movti"): Remove d/P alternative.
33906         ("movti_bigconst"): New pattern definition.
33908 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33910         PR target/80080
33911         * s390-protos.h (s390_expand_cs_hqi): Removed.
33912         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
33913         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
33914         modes as well as CCZ1mode and CCZmode.
33915         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
33916         signature of s390_emit_compare_and_swap.
33917         (s390_expand_cs_hqi): Likewise, make static.
33918         (s390_expand_cs_tdsi): Generate an explicit compare before trying
33919         compare-and-swap, in some cases.
33920         (s390_expand_cs): Wrapper function.
33921         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
33922         atomic_exchange.
33923         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
33924         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
33925         patterns for small and large integers.  Forbid symref memory operands.
33926         Move expander to s390.c.  Require cc register.
33927         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
33928         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
33929         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
33930         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
33931         symref memory operands.  Remove CC mode and call s390_match_ccmode
33932         instead.
33933         ("atomic_exchange<mode>"): Allow and implement all integer modes.
33935 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33937         * config/s390/s390.md (define_peephole2): New peephole to help
33938         combining the load-and-test pattern with volatile memory.
33940 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33942         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
33943         with CCZmode for TARGET_Z196.
33945 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33947         PR rtl-optimization/80501
33948         * combine.c (make_compound_operation_int): Set subreg_code to SET
33949         even for AND with mask of the sign bit of mode.
33951         PR rtl-optimization/80500
33952         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
33953         sum's initial value.
33955 2017-04-25  Julian Brown  <julian@codesourcery.com>
33956             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
33958         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
33960 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
33962         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
33964 2017-04-25  Julian Brown  <julian@codesourcery.com>
33965             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
33967         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
33968         (thunderx2t99_sha): New Reservation.
33970 2017-04-25  Julian Brown  <julian@codesourcery.com>
33971             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
33973         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
33974         type for 1-element load.
33976 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
33978         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
33980 2017-04-24  Martin Jambor  <mjambor@suse.cz>
33982         PR tree-optimization/80293
33983         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
33984         char arrays not totally scalarizable if it is false.
33985         (analyze_all_variable_accesses): Pass correct value in the new
33986         parameter.  Add a statistics counter.
33988 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
33990         PR middle-end/79931
33991         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
33993 2017-04-24  Richard Biener  <rguenther@suse.de>
33995         PR tree-optimization/80494
33996         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
33997         out for complex types.
33999 2017-04-24  Richard Biener  <rguenther@suse.de>
34001         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34002         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34003         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34004         (DFS): Adjust and never fail.
34005         (sccvn_dom_walker::fail): Remove.
34006         (sccvn_dom_walker::before_dom_children): Adjust.
34007         (run_scc_vn): Likewise and never fail.
34008         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34009         (pass_fre::execute): Likewise.
34011 2017-04-24  Richard Biener  <rguenther@suse.de>
34013         PR tree-optimization/79725
34014         * tree-ssa-sink.c (statement_sink_location): Return whether
34015         failure reason was zero uses.  Move that check later.
34016         (sink_code_in_bb): Deal with zero uses by removing the stmt
34017         if possible.
34019 2017-04-24  Richard Biener  <rguenther@suse.de>
34021         PR c++/2972
34022         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34023         pointer-based references.
34025 2017-04-24  Richard Biener  <rguenther@suse.de>
34027         PR bootstrap/79814
34028         * pass_manager.h (pass_manager::operator new): Remove.
34029         (pass_manager::operator delete): Likewise.
34030         * passes.c (pass_manager::operator new): Remove.
34031         (pass_manager::operator delete): Likewise.
34032         (pass_manager::pass_manager): Zero individual pass members.
34034 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34036         PR target/70799
34037         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34038         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34039         Check "XEXP (src, 1)" operand here.
34040         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34041         Check "XEXP (src, 1)" operand here.
34042         (dimode_scalar_chain::make_vector_copies): Detect count register
34043         of a shift instruction.  Zero extend count register from QImode
34044         to DImode to satisfy vector shift pattern count operand predicate.
34045         Substitute vector shift count operand with a DImode copy.
34046         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34047         vector register.
34049 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34051         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34052         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34053         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34054         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34055         (UNSPEC_NOREX_MEM): Remove definition.
34057 2017-04-21  Richard Biener  <rguenther@suse.de>
34059         PR tree-optimization/79547
34060         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34061         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34062         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34063         without any constraints.
34065 2017-04-21  Richard Biener  <rguenther@suse.de>
34067         PR tree-optimization/78847
34068         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34070 2017-04-21  Richard Biener  <rguenther@suse.de>
34072         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34073         (build_distinct_type_copy): Likewise.
34074         (build_variant_type_copy): Likewise.
34075         * tree.c (build_qualified_type): Pass down mem-stat info.
34076         (build_distinct_type_copy): Likewise.
34077         (build_variant_type_copy): Likewise.
34079 2017-04-21  Richard Biener  <rguenther@suse.de>
34081         PR tree-optimization/80237
34082         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34083         defaulted to NULL.
34084         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34085         for a simplified result.
34087 2016-04-21  Richard Biener  <rguenther@suse.de>
34089         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34090         sth as strict as a simple_iv but a chrec without symbols and an
34091         operand defined in the loop we are peeling (and not some subloop).
34092         (propagate_constants_for_unrolling): Propagate all constants.
34094 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34096         PR target/79804
34097         * config/i386/i386.c (print_reg): Remove assert for disalowed
34098         regno values, call output_operand_lossage instead.
34100 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34102         PR target/78090
34103         * config/i386/constraints.md (Yc): New register constraint.
34104         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34105         Use Yc constraint for alternative 2 of operand 0.  Remove
34106         preferred_for_speed attribute.
34108 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34110         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34111         lastprivate clauses in SIMT case.
34113 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34115         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34117 2017-04-20  Richard Biener  <rguenther@suse.de>
34119         PR tree-optimization/57796
34120         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34121         as N scalar stores.
34122         (vect_model_load_cost): Cost gathers as N scalar loads.
34124 2017-04-20  Richard Biener  <rguenther@suse.de>
34126         * ggc-page.c (ggc_allocated_p): Rename to ...
34127         (safe_lookup_page_table_entry): ... this and return the lookup
34128         result.
34129         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34131 2017-04-20  Richard Biener  <rguenther@suse.de>
34133         PR tree-optimization/80453
34134         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34135         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34136         from the conditions.
34137         (vn_phi_eq): Pass them down.
34138         (vn_phi_lookup): Record them.
34139         (vn_phi_insert): Likewise.
34141 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34143         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34144         uninitialized variable warning to avoid buffer overrun.
34146 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34148         PR other/71250
34149         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34150         is suppressed for '{ 0 }' in C.
34152 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34154         * BASE-VER: Set to 8.0.0.
34156 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34158         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34159         priority .init_array and .fini_array section with SECTION_NOTYPE
34160         flag.
34162 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34164         PR middle-end/80423
34165         * tree.h (build_array_type): Add typeless_storage default argument.
34166         * tree.c (type_cache_hasher::equal): Also compare
34167         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34168         (build_array_type): Add typeless_storage argument, set
34169         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34170         recursive call.
34171         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34172         (build_array_type): Likewise.  Add typeless_storage argument.
34174 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34175             Jakub Jelinek  <jakub@redhat.com>
34177         PR tree-optimization/80426
34178         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34179         operation on symbolic operands, also compute the overflow for the
34180         invariant part when the operation degenerates into a negation.
34182 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34184         PR debug/80461
34185         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34186         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34188         PR debug/80436
34189         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34191 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34193         PR target/80462
34194         * config/avr/avr.c (tree.h): Include it.
34195         (cgraph.h): Include it.
34196         (avr_encode_section_info): Don't warn for uninitialized progmem
34197         variable if it's just an alias.
34199 2017-04-19  Richard Biener  <rguenther@suse.de>
34201         PR ipa/65972
34202         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34203         when needed by AutoPGO.
34205 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34207         PR lto/50345
34208         * doc/lto.texi: Remove an extra 'that'.
34210 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34212         PR rtl-optimization/80429
34213         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34214         are only used in debug insns.
34216 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34217             Vladimir Makarov  <vmakarov@redhat.com>
34219         * config/sparc/predicates.md (input_operand): Add comment.  Return
34220         true for any memory operand when LRA is in progress.
34221         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34223 2017-04-18  Jeff Law  <law@redhat.com>
34225         PR target/74563
34226         * mips.md ({return,simple_return}_internal): Do not overwrite
34227         operands[0].
34229 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34231         PR tree-optimization/80443
34232         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34233         instead of adding 1, subtract -1 and similarly instead of subtracting
34234         1 add -1.
34236 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34238         PR rtl-optimization/80357
34239         * haifa-sched.c (tmp_bitmap): New variable.
34240         (model_recompute): Handle duplicate use records.
34241         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34242         (free_global_sched_pressure_data): Free it.
34244 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34246         Revert:
34247         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34248         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34249         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34250         instead of SYSTEM_HEADER_DIR.
34252 2017-04-18  Jeff Law  <law@redhat.com>
34254         PR middle-end/80422
34255         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34256         predecessors after walking up the insn chain.
34258 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34260         PR debug/80263
34261         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34262         sizetype type into debug info.
34264 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34266         PR target/80099
34267         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34268         unneeded test for TARGET_UPPER_REGS_SF.
34269         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34271 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34273         PR sanitizer/80444
34274         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34275         instead of gsi_after_labels.
34277 2017-04-18  Jeff Law  <law@redhat.com>
34279         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34280         stack pointer.
34282         Revert:
34283         2017-04-13  Jeff Law  <law@redhat.com>
34284         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34285         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34287 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34289         PR target/79453
34290         * config/avr/avr.c (intl.h): Include it.
34291         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34293 2017-04-18  Martin Liska  <mliska@suse.cz>
34295         PR gcov-profile/78783
34296         * gcov-tool.c (gcov_output_files): Validate that destination
34297         file is either removed by the tool or by a user.
34299 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34300             Guy Benyei  <guybe@mellanox.com>
34302         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34303         block, and do not negate it, the stored id is already negative.
34305 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34307         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34309 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34311         PR target/80098
34312         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34313         masks of options that should be turned off if the VSX vector
34314         options are turned off.
34315         (OTHER_P8_VECTOR_MASKS): Likewise.
34316         (OTHER_VSX_VECTOR_MASKS): Likewise.
34317         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34318         rs6000_disable_incompatible_switches to validate no type switches
34319         like -mvsx.
34320         (rs6000_incompatible_switch): New function to disallow turning on
34321         other vector options if -mno-vsx, -mno-power8-vector, or
34322         -mno-power9-vector are specified.
34324 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34326         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34328 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34330         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34331         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34332         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34333         (ARG_POINTER_CFA_OFFSET): Likewise.
34335 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34337         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34338         conditions to take advantage of various optimizations.
34340 2017-04-13  Jeff Law  <law@redhat.com>
34342         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34343         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34344         (zero_extendsidi2_dext): Likewise.
34346 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34348         PR sanitizer/80403
34349         * fold-const.c (fold_ternary_loc): Revert
34350         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34351         2017-04-12 change.
34353 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34355         PR rtl-optimization/80343
34356         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34357         new scratch pseudo.
34359 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34361         PR sanitizer/80414
34362         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34363         to ubsan_encode_value.
34365 2017-04-13  Jeff Law  <law@redhat.com>
34367         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34368         appearing in DEBUG_INSNs.
34370 2017-04-13  Martin Liska  <mliska@suse.cz>
34372         PR gcov-profile/80413
34373         * gcov-io.c (gcov_write_string): Copy to buffer just when
34374         allocated size is greater than zero.
34376 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34378         PR debug/80321
34379         * dwarf2out.c (decls_for_scope): Ignore declarations of
34380         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34382 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34384         PR lto/69953
34385         * ipa-visibility.c (non_local_p): Fix typos.
34386         (localize_node): When localizing symbol in same comdat group,
34387         dissolve the group only when we know external symbols are going
34388         to be privatized.
34389         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34391 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34393         PR tree-optimization/79390
34394         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34395         order does not result in usable sequence, retry with reversed operand
34396         order.
34398         PR sanitizer/80403
34399         PR sanitizer/80404
34400         PR sanitizer/80405
34401         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34402         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34403         op0 instead of fold_convert_loc (loc, type, arg0).
34405 2017-04-12  Jeff Law  <law@redhat.com>
34407         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34408         has a delay slot in the generated code.
34410         * config/cris/cris.md (cris_preferred_reload_class): Return
34411         GENNONACR_REGS rather than GENERAL_REGS.
34413 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34415         PR c/80163
34416         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34417         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34418         signedness of the result type.
34420 2017-04-12  Richard Biener  <rguenther@suse.de>
34421             Jeff Law  <law@redhat.com>
34423         PR tree-optimization/80359
34424         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34425         trim stores to TARGET_MEM_REFs.
34427 2017-04-12  Richard Biener  <rguenther@suse.de>
34429         PR tree-optimization/79390
34430         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34431         threading case even more.
34433 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34435         PR target/80382
34436         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34437         for quad_address_p for TImode, instead of just not indexed_address.
34439 2017-04-12  Richard Biener  <rguenther@suse.de>
34440             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34442         PR middle-end/79671
34443         * alias.c (component_uses_parent_alias_set_from): Handle
34444         TYPE_TYPELESS_STORAGE.
34445         (get_alias_set): Likewise.
34446         * tree-core.h (tree_type_common): Add typeless_storage flag.
34447         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34448         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34449         for types containing members with TYPE_TYPELESS_STORAGE.
34450         (place_field): Likewise.
34451         (layout_type): Likewise for ARRAY_TYPE.
34452         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34453         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34454         TYPE_TYPELESS_STORAGE.
34455         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34457 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34459         PR sanitizer/80349
34460         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34461         first argument to type.
34463 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34465         PR target/80376
34466         PR target/80315
34467         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34468         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34469         (rs6000_expand_binop_builtin): Likewise.
34470         (rs6000_expand_ternop_builtin): Likewise; also add missing
34471         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34472         vshasigma built-ins.
34473         * doc/extend.texi: Document that vec_xxpermdi's third argument
34474         must be a constant.
34476 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34478         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34479         Use shift_const cost parameter when calculating gain of STV shifts.
34481 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34483         PR rtl-optimization/70478
34484         * lra-constraints.c (process_alt_operands): Check memory for
34485         disfavoring memory insn operand.
34487 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34489         PR middle-end/80100
34490         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34491         left shift in unsigned HOST_WIDE_INT type.
34493         PR rtl-optimization/80385
34494         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34495         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34497         PR libgomp/80394
34498         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34499         if they have any depend clauses.
34501 2017-04-11  Martin Liska  <mliska@suse.cz>
34503         PR ipa/80212
34504         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34505         * ipa-split.c (split_function): Create a local comdat symbol
34506         if caller is in a comdat group.
34508 2017-04-11  Martin Liska  <mliska@suse.cz>
34510         PR ipa/80212
34511         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34512         flags.
34514 2017-04-11  Martin Sebor  <msebor@redhat.com>
34516         PR middle-end/80364
34517         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34518         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34519         for INTEGER_TYPE.
34520         (directive::set_width, directive::set_precision, format_character):
34521         Adjust.
34522         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34523         INTEGER_TYPE.
34525 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34527         PR target/80389
34528         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34529         conflict, set target->arch_name instead of target->cpu_name.
34531 2017-04-11  Richard Biener  <rguenther@suse.de>
34533         PR tree-optimization/80374
34534         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34535         build_zero_cst, remove fold_convertible_p check again.
34537 2017-04-11  Martin Liska  <mliska@suse.cz>
34539         PR sanitizer/70878
34540         * ubsan.c (instrument_object_size): Do not instrument register
34541         variables.
34543 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34545         PR target/80381
34546         * config/i386/i386-builtin-types.def
34547         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34548         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34549         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34550         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34551         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34552         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34553         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34554         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34555         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34556         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34557         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34558         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34559         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34560         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34561         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34562         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34563         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34564         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34565         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34566         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34567         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34568         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34569         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34570         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34571         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34572         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34573         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34574         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34575         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34576         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34577         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34578         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34579         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34580         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34581         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34582         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34583         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34584         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34585         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34586         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34587         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34588         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34589         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34590         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34591         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34592         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34593         aliases.
34594         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34595         flag to second_arg_count, handle 4 argument function type _COUNT
34596         aliases, handle second_arg_count on second argument rather than last.
34598 2017-04-10  Jeff Law  <law@redhat.com>
34600         PR tree-optimization/80374
34601         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34602         record anything if we can not convert integer_zero_node to the
34603         desired type.
34605 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34607         PR target/80108
34608         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34609         Enhance special handling given to the TARGET_P9_MINMAX option in
34610         relation to certain other options.
34612 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34614         PR tree-optimization/80153
34615         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34616         remove POINTER_PLUS_EXPR's base part directly, rather than through
34617         aff_tree.
34619 2017-04-10  Richard Biener  <rguenther@suse.de>
34620             Bin Cheng  <bin.cheng@arm.com>
34622         PR tree-optimization/80153
34623         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34624         the first element of pointer type aff_tree.  Build result expr in
34625         aff_tree's type.
34626         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34627         fold_convert calls.
34628         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34629         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34631 2017-04-10  Richard Biener  <rguenther@suse.de>
34633         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34634         asm inputs.
34636 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34638         PR rtl-optimization/70478
34639         * lra-constraints.c (curr_small_class_check): New.
34640         (update_and_check_small_class_inputs): New.
34641         (process_alt_operands): Update curr_small_class_check.  Disfavor
34642         alternative insn memory operands.  Check available regs for small
34643         class operands.
34645 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34647         PR target/80057
34648         * config/mips/mips.opt (-mvirt): Update description.
34649         * doc/invoke.texi (-mvirt): Likewise.
34651 2017-04-10  Richard Biener  <rguenther@suse.de>
34653         PR middle-end/80362
34654         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34655         looking for NEGATE_EXPR in -A / -B to A / B folding.
34657 2017-04-10  Martin Liska  <mliska@suse.cz>
34659         PR gcov-profile/80224
34660         * gcov.c (print_usage): Fix usage string.
34661         (get_gcov_intermediate_filename): Remove.
34662         (output_gcov_file): Use both for normal and intermediate format.
34663         (generate_results): Do not initialize special file for
34664         intermediate format.
34666 2017-04-10  Richard Biener  <rguenther@suse.de>
34668         PR tree-optimization/80304
34669         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34670         for safelen.
34672 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34674         PR target/79905
34675         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34676         (rs6000_init_builtins): Use it.
34678 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34680         * config/arm/arm.md (<mrc>): Add mode to SET source.
34681         (<mrrc>): Likewise.
34683 2017-04-10  Richard Biener  <rguenther@suse.de>
34685         PR middle-end/80344
34686         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34688 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34690         PR target/80324
34691         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34692         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34693         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34694         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34695         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34696         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34697         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34698         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34699         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34700         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34701         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34702         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34703         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34704         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34705         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34706         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34707         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34708         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34709         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34710         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34711         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34712         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34713         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34715 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34717         PR rtl-optimization/70478
34718         * lra-constraints.c: Reverse the last patch.
34720 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34722         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34723         Add comment for WCHAR_T.
34725 2017-04-08  Martin Liska  <mliska@suse.cz>
34727         Revert:
34728         2017-04-07  Martin Liska  <mliska@suse.cz>
34730         PR ipa/80212
34731         * ipa-split.c (split_function): Add function part to a same comdat
34732         group.
34734 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34736         PR target/80358
34737         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34739 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34741         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34742         (vec_pairing): ...to this.
34743         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34744         (rs6000_sched_init): Adjust for name change.
34745         (struct rs6000_sched_context): Likewise.
34746         (rs6000_init_sched_context): Likewise.
34747         (rs6000_set_sched_context): Likewise.
34749 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34751         PR target/80322
34752         PR target/80323
34753         PR target/80325
34754         PR target/80326
34755         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34756         intrinsics.
34757         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34758         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34759         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34761 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34763         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34765 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34767         PR rtl-optimization/70703
34768         * ira-color.c (update_conflict_hard_regno_costs): Use
34769         int64_t instead of HOST_WIDE_INT.
34771 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34773         PR rtl-optimization/70478
34774         * lra-constraints.c (process_alt_operands): Disfavor alternative
34775         insn memory operands.
34777 2017-04-07  Jeff Law  <law@redhat.com>
34779         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34780         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34782 2017-04-07  Martin Liska  <mliska@suse.cz>
34784         PR target/79889
34785         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34786         Show error message instead of an ICE.
34788 2017-04-07  Martin Liska  <mliska@suse.cz>
34790         PR ipa/80212
34791         * ipa-split.c (split_function): Add function part to a same comdat
34792         group.
34794 2017-04-07  Richard Biener  <rguenther@suse.de>
34796         PR middle-end/80341
34797         * tree.c (get_unwidened): Also handle ! for_type case for
34798         INTEGER_CSTs.
34799         * convert.c (do_narrow): Split out from ...
34800         (convert_to_integer_1): ... here.  Do not pass final truncation
34801         type to get_unwidened for TRUNC_DIV_EXPR.
34803 2017-04-07  Richard Biener  <rguenther@suse.de>
34805         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34806         than aff_tree.
34807         (aff_combination_const): Adjust.
34808         (aff_combination_scale): Likewise.
34809         (aff_combination_add_elt): Likewise.
34810         (aff_combination_add_cst): Likewise.
34811         (aff_combination_convert): Likewise.
34812         (add_elt_to_tree): Likewise.  Remove unused argument.
34813         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34815 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34817         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34818         definition.
34819         * config/arm/arm.c (arm_default_short_enums): Use
34820         ARM_DEFAULT_SHORT_ENUMS.
34821         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34823 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34825         PR debug/80234
34826         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34827         members with redundant out-of-class redeclaration.
34829 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34831         PR target/80286
34832         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34833         * config/i386/i386.md (*zero_extendsidi2):
34834         Add (?*x,*x) and (?*v,*v) alternatives.
34836 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34838         PR target/79733
34839         * config/i386/i386.c (ix86_expand_builtin)
34840         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34841         mode from insn data. Convert operands to insn operand mode.
34842         Copy operands that don't satisfy insn predicate to a register.
34844 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34846         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34847         Update comments.
34849 2017-04-06  Richard Biener  <rguenther@suse.de>
34851         PR tree-optimization/80334
34852         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34853         preserve alignment of accesses.
34855 2017-04-06  Richard Biener  <rguenther@suse.de>
34857         PR tree-optimization/80262
34858         * tree-sra.c (build_ref_for_offset): Preserve address-space
34859         information.
34860         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34861         Drop useless address-space information on MEM_REF offsets.
34863 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
34865         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
34867 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34869         PR rtl-optimization/70703
34870         * ira-color.c (update_conflict_hard_regno_costs): Use
34871         HOST_WIDE_INT instead of long.
34873 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
34875         PR target/80298
34876         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
34877         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
34878         is not defined.
34879         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
34880         for x86_64 target.  Handle -m3dnowa option.
34882 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34884         PR rtl-optimization/70703
34885         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
34886         (update_conflict_hard_regno_costs): Use long instead of unsigned
34887         arithmetic for cost calculation.
34889 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34890             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34892         PR sanitizer/80308
34893         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
34894         for big endian.
34896 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
34898         PR target/78002
34899         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
34900         ptr_mode with Pmode throughout.
34901         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
34902         into probe_stack_range and use DImode.
34904 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34906         PR target/79890
34907         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
34908         call_eh_return is true.
34910 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34912         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
34913         Initialize last_match_fntype_index.
34915 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34917         PR target/80310
34918         * tree-nvr.c: Include internal-fn.h.
34919         (pass_return_slot::execute): Ignore internal calls without
34920         direct optab.
34922 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34923             Richard Biener  <rguenther@suse.de>
34925         PR c++/80297
34926         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
34927         captures used multiple times, except for the last use.
34928         * generic-match-head.c: Include gimplify.h.
34930 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34932         PR tree-optimization/79390
34933         * target.h (struct noce_if_info): Declare.
34934         * targhooks.h (default_noce_conversion_profitable_p): Declare.
34935         * target.def (noce_conversion_profitable_p): New target hook.
34936         * ifcvt.h (struct noce_if_info): New type, moved from ...
34937         * ifcvt.c (struct noce_if_info): ... here.
34938         (noce_conversion_profitable_p): Renamed to ...
34939         (default_noce_conversion_profitable_p): ... this.  No longer
34940         static nor inline.
34941         (noce_try_store_flag_constants, noce_try_addcc,
34942         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
34943         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
34944         instead of noce_conversion_profitable_p.
34945         * config/i386/i386.c: Include ifcvt.h.
34946         (ix86_option_override_internal): Don't override
34947         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
34948         (ix86_noce_conversion_profitable_p): New function.
34949         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
34950         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
34951         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
34952         * doc/tm.texi: Regenerated.
34954 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34956         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
34957         correction.
34959 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34961         PR target/80307
34962         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
34963         instructions for small multiply cores.
34965 2017-04-04  Jeff Law  <law@redhat.com>
34967         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
34968         added member.
34969         (mips_expand_vec_perm_const): Initialize elements in orig_perm
34970         that are not set by the loop over the elements.
34972 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34974         PR target/80286
34975         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
34976         int mode, convert_modes it to mode as unsigned, otherwise use
34977         lowpart_subreg to mode rather than SImode.
34978         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
34979         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
34980         Use DImode instead of SImode for the shift count operand.
34981         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
34982         Likewise.
34984 2017-04-04  Richard Biener  <rguenther@suse.de>
34986         PR middle-end/80281
34987         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
34988         arithmetic done for the negate or the plus.  Simplify.
34989         (A - (-B) -> A + B): Likewise.
34990         * fold-const.c (split_tree): Make sure to not negate pointers.
34992 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
34994         PR rtl-optimization/60818
34995         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
34996         a compare of comparisons with the thing compared if this results
34997         in a different machine mode.
34999 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35001         * alias.c (base_alias_check): Fix typo in comment.
35002         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35003         * cgraphunit.c (symbol_table::compile): Likewise.
35004         * collect2.c (maybe_run_lto_and_relink): Likewise.
35005         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35006         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35007         * config/avr/avr.c (avr_map_op_t): Likewise.
35008         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35009         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35010         * config/epiphany/epiphany.md (movcc): Likewise.
35011         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35012         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35013         Likewise.
35014         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35015         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35016         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35017         * config/sh/sh.c (sh_rtx_costs): Likewise.
35018         * fold-const.c (fold_truth_andor): Likewise.
35019         * genautomata.c (collapse_flag): Likewise.
35020         * gengtype.h (struct type::u::s): Likewise.
35021         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35022         * input.c (FORMAT_AMOUNT): Likewise.
35023         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35024         (known_aggs_to_agg_replacement_list): Likewise.
35025         * ipa-inline-analysis.c: Likewise.
35026         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35027         * ipa-polymorphic-call.c
35028         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35029         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35030         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35031         Likewise.
35032         * modulo-sched.c (apply_reg_moves): Likewise.
35033         * omp-expand.c (build_omp_regions_1): Likewise.
35034         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35035         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35036         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35037         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35038         * value-prof.c: Likewise.
35039         * var-tracking.c (val_reset): Likewise.
35041 2017-04-03  Richard Biener  <rguenther@suse.de>
35043         PR tree-optimization/80275
35044         * fold-const.c (split_address_to_core_and_offset): Handle
35045         POINTER_PLUS_EXPR.
35047 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35049         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35050         descriptors is at least equal to that of functions.
35052 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35054         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35056 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35058         PR target/80250
35059         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35060         (mov<IMOD4:mode>): New expander.
35061         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35063 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35065         PR rtl-optimization/79405
35066         * fwprop.c (propagations_left): New variable.
35067         (forward_propagate_into): Decrement it.
35068         (fwprop_init): Initialize it.
35069         (fw_prop): If the variable has reached zero, stop propagating.
35070         (fwprop_addr): Ditto.
35072 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35074         PR debug/79255
35075         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35076         a FUNCTION_DECL, pass it as decl instead of origin to
35077         process_scope_var.
35079 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35081         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35082         string.
35084 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35086         PR target/80107
35087         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35088         TARGET_VSX_SMALL_INTEGER.
35090 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35092         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35093         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35095 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35097         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35098         extraction from odd-numbered MSA register.
35100 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35102         PR middle-end/80173
35103         * expmed.c (store_bit_field_1): Don't attempt to create
35104         a word subreg out of hard registers wider than word if they
35105         have HARD_REGNO_NREGS of 1 for their mode.
35107         PR middle-end/80163
35108         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35109         conversions to integer types wider than word and pointer.
35111         PR debug/80025
35112         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35113         (rtx_equal_for_cselib_p): Pass 0 to it.
35114         * cselib.c (cselib_hasher::equal): Likewise.
35115         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35116         is 128, don't look up VALUE locs and punt.  Increment
35117         depth in recursive calls when walking VALUE locs.
35119 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35121         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35122         (make_gcov_file_name): Use the canonical path name for generating
35123         the MD5 value.
35124         (read_line): Fix handling of files with ascii null bytes.
35126 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35128         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35129         to initialise a vector register instead
35130         of using a const_int.
35132 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35134         PR translation/80189
35135         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35136         diagnostic messages.
35138 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35140         PR target/80246
35141         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35142         (dfp_diex_<mode>): Update mode of operand 1.
35143         * doc/extend.texi (dxex, dxexq): Document change to return type.
35144         (diex, diexq): Document change to argument type.
35146 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35148         PR ipa/77333
35149         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35150         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35151         it reflects the signature changes performed at the callee side.
35152         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35153         to cgraph_build_function_type_skip_args.
35154         (build_function_decl_skip_args): Adjust call to the above function.
35156 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35158         PR target/80206
35159         * config/i386/sse.md
35160         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35161         register as dest whenever it is a MEM not rtx_equal_p to the
35162         corresponding dup operand, and when forcing into reg move the
35163         reg into the memory afterwards.
35164         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35165         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35166         for the force_reg mode.
35167         (avx512vl_vextractf128<mode>): Use register as dest either
35168         always when a MEM, or when it is a MEM not rtx_equal_p to the
35169         corresponding dup operand, or even not when it is a CONST_VECTOR
35170         depending on the mode and lo vs. hi.
35171         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35172         parens.
35173         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35174         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35175         Likewise.  Require that operands[2] is even.
35176         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35177         Remove extraneous parens.  Require that operands[2] is a multiple
35178         of 4.
35179         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35180         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35181         disallow memory then.
35183 2017-03-30  Richard Biener  <rguenther@suse.de>
35185         PR tree-optimization/77498
35186         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35187         to non-constants over backedges.
35189 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35191         PR rtl-optimization/80233
35192         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35193         as last_combined_insn.  Do not test for BARRIER_P separately.
35195 2017-03-29  Andreas Schwab  <schwab@suse.de>
35197         PR ada/80146
35198         * calls.c (prepare_call_address): Convert funexp to Pmode before
35199         copying to temp reg.
35201 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35203         PR tree-optimization/80158
35204         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35205         Handle possible future case of more than one alternate
35206         interpretation.
35207         (replace_rhs_if_not_dup): Likewise.
35208         (replace_one_candidate): Likewise.
35210 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35212         PR rtl-optimization/80193
35213         * ira.c (ira): Do not check allocation for LRA.
35215 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35217         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35218         (nvptx_output_simt_exit): Declare.
35219         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35220         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35221         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35222         (nvptx_declare_function_name): ...here.  Emit declaration of local
35223         memory space buffer for omp_simt_enter insn.
35224         (nvptx_output_unisimt_switch): New.
35225         (nvptx_output_softstack_switch): New.
35226         (nvptx_output_simt_enter): New.
35227         (nvptx_output_simt_exit): New.
35228         * config/nvptx/nvptx.h (struct machine_function): New fields
35229         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35230         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35231         (UNSPECV_SIMT_EXIT): Ditto.
35232         (omp_simt_enter_insn): New insn.
35233         (omp_simt_enter): New expansion.
35234         (omp_simt_exit): New insn.
35235         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35237         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35238         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35239         (expand_GOMP_SIMT_EXIT): New.
35240         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35241         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35242         (GOMP_SIMT_EXIT): Ditto.
35243         * target-insns.def (omp_simt_enter): New insn.
35244         (omp_simt_exit): Ditto.
35245         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35246         simt_dlist.
35247         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35248         (lower_rec_input_clauses): Likewise.
35249         (lower_lastprivate_clauses): Handle SIMT privatization.
35251         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35252         (ompdevlow_adjust_simt_enter): New.
35253         (find_simtpriv_var_op): New.
35254         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35255         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35257         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35258         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35259         (copy_decl_for_dup_finish): Ditto.
35261         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35263 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35265         PR target/53383
35266         * config/i386/i386.c (ix86_option_override_internal): Always
35267         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35269 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35271         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35273 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35275         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35276         mark new edge's irreducible flag accordign to it.
35277         (vect_do_peeling): Check loop preheader edge's irreducible flag
35278         and pass it to function slpeel_add_loop_guard.
35280 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35282         PR tree-optimization/80218
35283         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35284         Update block frequencies and counts.
35286 2017-03-28  Richard Biener  <rguenther@suse.de>
35288         PR tree-optimization/78644
35289         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35290         of a simplification result we may not use it at all.
35292 2017-03-28  Richard Biener  <rguenther@suse.de>
35294         PR ipa/80205
35295         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35296         without arguments, generate default definition of a SSA name.
35298 2017-03-28  Richard Biener  <rguenther@suse.de>
35300         PR middle-end/80222
35301         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35302         TYPE_REF_CAN_ALIAS_ALL references.
35303         * fold-const.c (fold_indirect_ref_1): Likewise.
35305 2017-03-28  Martin Liska  <mliska@suse.cz>
35307         PR ipa/80104
35308         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35309         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35311 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35312             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35314         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35315         (EXTRA_SPECS): Define.
35316         (SUBTARGET_EXTRA_SPECS): Likewise.
35317         (SUBTARGET_CPP_SPEC): Likewise.
35318         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35319         SUBTARGET_EXTRA_SPECS.
35320         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35322 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35324         * config/arc/simdext.md (vst64_insn): Update pattern.
35325         (vld32wh_insn): Likewise.
35326         (vld32wl_insn): Likewise.
35327         (vld64_insn): Likewise.
35328         (vld32_insn): Likewise.
35330 2017-03-28  Marek Polacek  <polacek@redhat.com>
35332         PR sanitizer/80067
35333         * fold-const.c (fold_comparison): Use protected_set_expr_location
35334         instead of SET_EXPR_LOCATION.
35336 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35338         * tree.c (add_expr): Avoid name lookup warning.
35340 2017-03-27  Jeff Law  <law@redhat.com>
35342         PR tree-optimization/80216
35343         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35344         function name.  Limit recursion depth.
35345         (record_temporary_equivalences): Corresponding changes.
35347 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35349         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35350         covered first.
35352 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35354         PR target/80102
35355         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35356         notes.
35357         * cfgcleanup.c (reg_note_cfa_p): New array.
35358         (insns_have_identical_cfa_notes): New function.
35359         (old_insns_match_p): Don't cross-jump in between /f
35360         and non-/f instructions.  If both i1 and i2 are frame related,
35361         verify all CFA notes, their order and content.
35363 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35365         PR target/78543
35366         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35367         HImode and SImode with zero extend to DImode to one insn.
35368         (bswap<mode>2_extenddi): Likewise.
35369         (bswapsi2_extenddi): Likewise.
35370         (bswaphi2_extendsi): Likewise.
35371         (bswaphi2): Combine bswap HImode and SImode into one insn.
35372         Separate memory insns from swapping register.
35373         (bswapsi2): Likewise.
35374         (bswap<mode>2): Likewise.
35375         (bswaphi2_internal): Delete, no longer used.
35376         (bswapsi2_internal): Likewise.
35377         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35378         store, and gpr<-gpr swap insns.
35379         (bswap<mode>2_store): Likewise.
35380         (bswaphi2_reg): Register only splitter, combine with the splitter.
35381         (bswaphi2 splitter): Likewise.
35382         (bswapsi2_reg): Likewise.
35383         (bswapsi2 splitter): Likewise.
35384         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35385         the insns into load, store, and register/register insns.
35386         (bswapdi2_ldbrx): Likewise.
35387         (bswapdi2_load): Likewise.
35388         (bswapdi2_store): Likewise.
35389         (bswapdi2_reg): Likewise.
35391 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35393         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35394         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35396 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35398         PR target/80103
35399         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35400         add comments.
35401         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35402         special handling for target option conflicts between dform
35403         options (-mpower9-dform, -mpower9-dform-vector,
35404         -mpower9-dform-scalar) and -mno-direct-move.
35406 2017-03-27  Richard Biener  <rguenther@suse.de>
35408         PR tree-optimization/80181
35409         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35411 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35413         * config/arc/predicates.md (move_double_src_operand): Replace the
35414         call to move_double_src_operand with a call to address_operand.
35416 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35418         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35419         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35420         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35422 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35424         * config/arc/predicates.md (long_immediate_loadstore_operand):
35425         Consider scaled addresses cases.
35427 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35429         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35430         restored when in interrupt.
35431         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35432         doesn't have delay slot.
35434 2017-03-27  Richard Biener  <rguenther@suse.de>
35436         PR ipa/79776
35437         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35438         inlined thunk clones.
35440 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35442         PR sanitizer/80168
35443         * asan.c (instrument_derefs): Copy over last operand from
35444         original COMPONENT_REF to the new COMPONENT_REF with
35445         DECL_BIT_FIELD_REPRESENTATIVE.
35446         * ubsan.c (instrument_object_size): Likewise.
35448 2017-03-27  Richard Biener  <rguenther@suse.de>
35450         PR tree-optimization/80170
35451         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35452         sure DR/SCEV didnt fold in constants we do not see when looking
35453         at the reference base alignment.
35455 2017-03-27  Richard Biener  <rguenther@suse.de>
35457         PR middle-end/80171
35458         * gimple-fold.c (fold_ctor_reference): Properly guard against
35459         NULL return value from canonicalize_constructor_val.
35461 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35463         PR target/80180
35464         * config/i386/i386.c (ix86_expand_builtin)
35465         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35466         flags reg setting and flags reg using instructions.
35467         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35468         clobbering instructions to zero extend op2.
35470 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35472         * doc/install.texi (Configuration) <--with-aix-soname>:
35473         Update link to AIX ld.
35475 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35477         PR rtl-optimization/80160
35478         PR rtl-optimization/80159
35479         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35480         reg_alternate_class into account.
35482 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35484         PR target/80148
35485         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35486         to consider in curr_insn_transform.
35488 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35490         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35491         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35492         and emit_mode_inner.
35494 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35496         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35497         argument to the overloaded builtin variants.  Use the new flag to
35498         deprecate certain builtin variants.
35499         * config/s390/s390-builtin-types.def: Add new builtin types.
35500         * config/s390/s390-builtins.h: Support new flags field for
35501         overloaded builtins.
35502         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35503         (s390_macro_to_expand): Enable vector float data type.
35504         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35505         builtins by incrementing the __VEC__ version number.
35506         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35507         vec_xst.
35508         (s390_resolve_overloaded_builtin): Emit error messages depending
35509         on the builtin flags.
35510         * config/s390/s390.c (s390_expand_builtin): Support additional
35511         flags argument.  Change error message to match the messages
35512         emitted in s390-c.c.
35513         * config/s390/s390.md: New UNSPEC_* constants.
35514         (op_type): Add new instruction types.
35515         * config/s390/vecintrin.h: Add new builtins and test data class
35516         constants.
35517         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35518         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35519         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35520         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35521         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35522         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35524         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35525         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35526         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35527         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35529         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35530         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35531         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35532         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35534         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35535         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35536         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35537         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35538         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35539         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35540         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35542         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35543         ("vec_scatter_element<V_HW_4:mode>_DI")
35544         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35545         ("vec_fpint<mode>", "vflls")
35546         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35547         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35548         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35549         ("*vec_cmphe<mode>_cc"): ... these.
35551         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35552         mode constant instead of magic value.
35554 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35556         * config/s390/s390.c (s390_expand_vec_compare): Support other
35557         vector floating point modes than just V2DF.
35558         (s390_expand_vcond): Likewise.
35559         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35560         (s390_cannot_change_mode_class): Prevent mode changes between TF
35561         and V1TF in vector registers.
35562         * config/s390/s390.md (DF, SF): New mode attributes.
35563         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35564         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35565         SFmode support for VRs.
35566         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35567         vector fp modes.
35568         (VFT, VF_HW): New mode iterators.
35569         (vw, sdx): New mode attributes.
35570         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35571         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35572         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35573         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35574         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35575         also the new vector floating point modes.  Renaming to ...
35577         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35578         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35579         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35580         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35581         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35582         ("vec_unordered<mode>"): ... these.
35584         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35585         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35586         ("*vec_extendv2df"): New insn definitions.
35588 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35590         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35591         ("mulditi3_2", "*muldi3_sign"): New patterns.
35592         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35593         rename the pattern definition.
35595 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35597         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35598         expander.
35599         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35601 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35603         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35604         instruction if possible.
35605         * config/s390/vector.md (vec_halfnumelts): New mode
35606         attribute.
35607         ("*vec_vllezlf<mode>"): New pattern.
35609 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35611         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35612         ("popcountv4si2", "popcountv2di2"): Rename to ...
35613         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35614         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35615         condition.
35616         ("popcount<mode>2_vxe"): New pattern.
35618 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35620         * common/config/s390/s390-common.c (processor_flags_table): Add
35621         arch12.
35622         * config.gcc: Add arch12.
35623         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35624         Default to arch12 for unknown CPU model numbers.
35625         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35626         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35627         PROCESSOR_max sanity check.
35628         * config/s390/s390-opts.h (enum processor_type): Add
35629         PROCESSOR_ARCH12.
35630         * config/s390/s390.c (processor_table): Add arch12.
35631         (s390_expand_builtin): Add check for B_VXE flag.
35632         (s390_issue_rate): Add PROCESSOR_ARCH12.
35633         (s390_get_sched_attrmask): Likewise.
35634         (s390_get_unit_mask): Likewise.
35635         (s390_sched_score): Enable z13 scheduling for arch12.
35636         (s390_sched_reorder): Likewise.
35637         (s390_sched_variable_issue): Likewise.
35638         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35639         PF_VXE.
35640         (s390_tune_attr): Use z13 scheduling also for arch12.
35641         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35642         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35643         (TARGET_VXE_P): New macros.
35644         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35645         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35646         * config/s390/s390.opt: Add arch12 as processor_type.
35648 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35650         * config/s390/s390.md
35651         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35652         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35653         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35655         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35656         Rename expanders to ...
35658         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35659         ("fixuns_truncdddi2_emu"): ... these.
35661         ("fixuns_trunc<mode>si2_emu"): New expander.
35663         ("*fixuns_truncdfdi2_z13"): Rename to ...
35664         ("*fixuns_truncdfdi2_vx"): ... this.
35666 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35668         * config/s390/2964.md: Remove the single element vector compare
35669         instructions which are no longer used.
35670         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35671         vector CCmodes.
35672         (s390_canonicalize_comparison): Remove handling of DFmode
35673         compares.
35674         (s390_expand_vec_compare_scalar): Remove function.
35675         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35676         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35677         pattern.
35678         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35680 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35682         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35683         FP zero.
35684         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35685         will anyway by matched by mov<mode>_64dfp.
35687 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35689         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35690         vlef/vstef.  Add missing operand to vleif.
35692 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35694         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35695         pair for all vector types with 64 bit elements.
35696         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35697         * config/s390/vector.md (V_HW_64): ... here.
35698         (V_128_NOSINGLE): New mode iterator.
35699         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35700         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35701         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35702         ("*vec_load_pairv2di"): Change to ...
35703         ("*vec_load_pair<mode>"): ... this one.
35705 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35707         * config/s390/constraints.md: Add comments.
35708         (jKK): Reject element sizes > 8 bytes.
35709         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35710         s_operands.
35711         * config/s390/s390.md: Add the s_operand checks formerly in
35712         s390_split_ok_p to various splitters where they are still
35713         required.
35714         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35715         for 128 bit vectors.  Plus two splitters.
35717 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35719         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35720         the file.
35722 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35724         PR target/79893
35725         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35726         error if the boundary argument is not constant.
35728 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35730         PR rtl-optimization/80112
35731         * loop-doloop.c (doloop_condition_get): Don't check condition
35732         if cmp isn't SET with IF_THEN_ELSE src.
35734 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35736         PR tree-optimization/80158
35737         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35738         replacing a candidate statement, also replace it for the
35739         candidate's alternate interpretation.
35740         (replace_rhs_if_not_dup): Likewise.
35741         (replace_one_candidate): Likewise.
35743 2017-03-24  Richard Biener  <rguenther@suse.de>
35745         PR tree-optimization/80167
35746         * graphite-isl-ast-to-gimple.c
35747         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35748         properly.
35749         (translate_isl_ast_to_gimple::get_rename): Likewise.
35751 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35753         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35754         handling of certain combinations of target options, including the
35755         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35756         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35758 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35760         PR target/71436
35761         * config/arm/arm.md (*load_multiple): Add reload_completed to
35762         matching condition.
35764 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35765             Richard Biener  <rguenth@suse.de>
35767         PR tree-optimization/79908
35768         PR tree-optimization/80136
35769         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35770         been cast away, gimplify_and_add suffices.
35772 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35774         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35776 2017-03-23  Richard Biener  <rguenther@suse.de>
35778         PR tree-optimization/80032
35779         * gimplify.c (gimple_push_cleanup): Forced unconditional
35780         cleanups still have to go to the conditional_cleanups
35781         sequence.
35783 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35785         PR tree-optimization/80072
35786         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35787         to unsigned int.
35788         (next_operand_entry_id): Change type to unsigned int.
35789         (sort_by_operand_rank): Make sure to return the right return value
35790         even if unsigned fields are bigger than INT_MAX.
35791         (struct oecount): Change cnt and id type to unsigned int.
35792         (oecount_hasher::equal): Formatting fix.
35793         (oecount_cmp): Make sure to return the right return value
35794         even if unsigned fields are bigger than INT_MAX.
35795         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35797         PR c++/80129
35798         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35799         TREE_READONLY on result if writing it more than once.
35801         PR sanitizer/80110
35802         * doc/invoke.texi (-fsanitize=thread): Document that with
35803         -fnon-call-exceptions atomics are not able to throw
35804         exceptions.
35806         PR sanitizer/80110
35807         * tsan.c: Include tree-eh.h.
35808         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35809         maybe_clean_or_replace_eh_stmt where needed.
35810         (instrument_memory_accesses): Add cfg_changed argument.
35811         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35812         if it returned true.
35813         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35815         PR rtl-optimization/63191
35816         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35817         wrapper function, moved the whole old content into ...
35818         (ix86_delegitimize_address_1): ... this.  New inline function.
35819         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35820         true as last argument instead of ix86_delegitimize_address.
35822 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35824         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35825         cortexa57_branch_cost.
35827 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35829         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35831 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35833         PR target/80123
35834         * doc/md.texi (Constraints): Document wA constraint.
35835         * config/rs6000/constraints.md (wA): New.
35836         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35837         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35838         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35839         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35841 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35843         PR c++/80029
35844         * gimplify.c (is_oacc_declared): New function.
35845         (oacc_default_clause): Use it to set default flags for acc declared
35846         variables inside parallel regions.
35847         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35848         declared variables.
35849         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35850         declare attribute to any decl as necessary.
35852 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35854         PR target/80082
35855         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35856         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35857         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35858         (arm_arch_lpae): This.
35859         * config/arm/arm.c (arm_arch7ve): Rename into ...
35860         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
35861         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35862         arm_arch_lpae.
35864 2017-03-22  Martin Liska  <mliska@suse.cz>
35866         PR target/79906
35867         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
35868         error message instead of an ICE.
35870 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35872         * doc/extend.texi (6.11 Additional Floating Types): Revise.
35874 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35876         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
35877         comments.
35878         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35879         comments.
35881 2017-03-21  Martin Sebor  <msebor@redhat.com>
35883         * doc/extend.texi: Use "cannot" instead of "can't."
35884         * doc/hostconfig.texi: Same.
35885         * doc/install.texi: Same.
35886         * doc/invoke.texi: Same.
35887         * doc/loop.texi: Same.
35888         * doc/md.texi: Same.
35889         * doc/objc.texi: Same.
35890         * doc/rtl.texi: Same.
35891         * doc/tm.texi: Same.
35892         * doc/tm.texi.in: Same.
35893         * doc/trouble.texi: Same.
35895 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
35897         PR debug/63238
35898         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
35899         (collect_checksum_attributes): Set it.
35900         (die_checksum_ordered): Use it.
35902 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35904         PR tree-optimization/79908
35905         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
35906         change: For a VA_ARG whose LHS has been cast away, use
35907         force_gimple_operand to construct the side effects.
35909 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
35911         PR translation/80001
35912         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
35913         more amenable to translation.
35914         (oacc_loop_auto_partitions): Likewise.
35916 2017-03-21  Marek Polacek  <polacek@redhat.com>
35917             Martin Sebor  <msebor@redhat.com>
35919         PR tree-optimization/80109
35920         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
35921         on INTEGRAL_TYPE_P.
35923 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
35924             Segher Boessenkool  <segher@kernel.crashing.org>
35926         PR target/80125
35927         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
35928         check reg_used_between_p between insn and one of succ or succ2
35929         depending on if succ is artificial insn not inserted into insn
35930         stream.
35932 2017-03-21  Martin Liska  <mliska@suse.cz>
35934         PR gcov-profile/80081
35935         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
35936         * doc/gcc.texi: Include gcov-dump stuff.
35937         * doc/gcov-dump.texi: New file.
35939 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
35941         PR rtl-optimization/79150
35942         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
35943         conditional jump, if the jump is the last insn of the loop.
35945 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35946             Richard Biener  <rguenth@suse.de>
35948         PR tree-optimization/79908
35949         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35950         been cast away, use force_gimple_operand to construct the side
35951         effects.
35953 2017-03-21  Martin Liska  <mliska@suse.cz>
35955         PR libfortran/79956
35956         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
35957         to NULL.
35959 2017-03-21  Brad Spengler <spender@grsecurity.net>
35961         PR plugins/80094
35962         * plugin.c (htab_hash_plugin): New function.
35963         (add_new_plugin): Use it and adjust.
35964         (parse_plugin_arg_opt): Adjust.
35965         (init_one_plugin): Likewise.
35967 2017-03-21  Richard Biener  <rguenther@suse.de>
35969         PR tree-optimization/80032
35970         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
35971         if set force the cleanup to happen unconditionally.
35972         (gimplify_target_expr): Push inserted clobbers with force_uncond
35973         to avoid them being removed by control-dependent DCE.
35975 2017-03-21  Richard Biener  <rguenther@suse.de>
35977         PR tree-optimization/80122
35978         * tree-inline.c (copy_bb): Do not expans va-arg packs or
35979         va_arg_pack_len when the inlined call stmt requires pack
35980         expansion itself.
35981         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
35983 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
35985         PR sanitizer/78158
35986         * tsan.c (instrument_builtin_call): If the memory model argument
35987         is not a constant, assume it is valid.
35989         PR c/67338
35990         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
35991         avoid UB.
35993 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
35995         PR rtl-optimization/79910
35996         * combine.c (can_combine_p): Do not allow combining an I0 or I1
35997         if its dest is used by an insn before I2 (other than the combined
35998         insns themselves, which are properly handled already).
36000 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36002         Revert:
36003         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36005         * combine.c (record_used_regs): New static function.
36006         (try_combine): Handle situations where there is an additional
36007         instruction between I2 and I3 which needs to have a LOG_LINK
36008         updated.
36010         Revert:
36011         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36013         * combine.c (try_combine): Delete redundant i1 test.  Call
36014         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36016 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36018         PR target/80083
36019         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36020         alternatives 13/14.
36022 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36024         PR tree-optimization/80054
36025         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36026         the optimization if a PHI or any of its arguments is not dominated
36027         by the candidate's basis.  Use gphi* rather than gimple* as
36028         appropriate.
36029         (replace_profitable_candidates): Clean up a gimple* variable that
36030         should be a gphi* variable.
36032 2017-03-20  Martin Sebor  <msebor@redhat.com>
36034         PR c++/52477
36035         * doc/extend.texi (attribute constructor): Document present limitation.
36037 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36039         PR target/79963
36040         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36041         __POWER9_VECTOR__ #ifdef control, change template definition to
36042         use Power9-specific built-in function.
36043         (vec_any_eq): Likewise.
36044         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36045         to control outcomes from this test.
36046         (vector_ae_<mode>p): For VEC_F modes, likewise.
36048 2017-03-20  Ian Lance Taylor  <iant@google.com>
36050         * config/i386/i386.c (ix86_function_regparm): Save an extra
36051         register for -fsplit-stack with DECL_STATIC_CHAIN.
36053 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36055         PR target/79912
36056         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36057         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36059 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36061         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36062         iorw,ow".
36063         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36064         iorw,iorw".
36066 2017-03-20  Marek Polacek  <polacek@redhat.com>
36068         PR sanitizer/80063
36069         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36071 2017-03-20  Richard Biener  <rguenther@suse.de>
36073         PR tree-optimization/80113
36074         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36075         allocate extra SSA name for PHI def.
36076         (add_close_phis_to_outer_loops): Likewise.
36077         (add_close_phis_to_merge_points): Likewise.
36078         (copy_loop_close_phi_args): Likewise.
36079         (copy_cond_phi_nodes): Likewise.
36081 2017-03-20  Martin Liska  <mliska@suse.cz>
36083         PR middle-end/79753
36084         * tree-chkp.c (chkp_build_returned_bound): Do not build
36085         returned bounds for a LHS that's not a BOUNDED_P type.
36087 2017-03-20  Martin Liska  <mliska@suse.cz>
36089         PR target/79769
36090         PR target/79770
36091         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36092         COMPLEX_CST and VECTOR_CST.
36094 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36096         PR target/78857
36097         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36098         target operand.  A new splitter adds the clobber statement in case
36099         the target operand is dead anyway.
36101 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36103         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36104         to age-old versions of binutils and glibc.
36106 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36108         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36110 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36112         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36114 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36116         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36117         requirement for binutils 2.13.
36119 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36121         * combine.c (try_combine): Delete redundant i1 test.  Call
36122         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36124 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36126         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36127         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36128         contents.
36129         <riscv64-*-elf>: Re-arrange section
36130         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36131         <riscv32-*-linux>: Likewise.
36132         <riscv64-*-elf>: Likewise
36133         <riscv64-*-linux>: Likewise.
36135 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36137         PR target/80052
36138         * aarch64.opt(verbose-cost-dump): Fix typo.
36140 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36142         PR target/79951
36143         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36144         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36146 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36148         * reload.c (find_reloads): When reloading a nonoffsettable address,
36149         use RELOAD_OTHER for it and its address reloads.
36151         PR rtl-optimization/79910
36152         * combine.c (record_used_regs): New static function.
36153         (try_combine): Handle situations where there is an additional
36154         instruction between I2 and I3 which needs to have a LOG_LINK
36155         updated.
36157 2017-03-17  Jeff Law  <law@redhat.com>
36159         PR tree-optimization/71437
36160         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36161         conditional in the hash table first.
36162         (vrp_dom_walker::before_dom_children): Extract condition from
36163         ASSERT_EXPR.  Record condition, its inverion and any implied
36164         conditions as well.
36166 2017-03-17  Marek Polacek  <polacek@redhat.com>
36167             Markus Trippelsdorf  <markus@trippelsdorf.de>
36169         PR tree-optimization/80079
36170         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36171         m_stores_head.
36173 2017-03-17  Richard Biener  <rguenther@suse.de>
36175         PR middle-end/80075
36176         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36177         Properly verify the LHS before the RHS possibly claims to be
36178         handled.
36179         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36180         do not throw.
36182 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36184         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36185         (List of -O2 options): Likewise.
36186         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36187         (-fipa-vrp) New.
36189 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36191         * gcov-dump.c (print_usage): Print bug_report_url.
36193 2017-03-17  Richard Biener  <rguenther@suse.de>
36195         PR middle-end/80050
36196         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36197         (parser::peek): Likewise.
36199 2017-03-17  Richard Biener  <rguenther@suse.de>
36201         PR tree-optimization/80048
36202         * sese.c (free_sese_info): Properly release rename_map and
36203         copied_bb_map elements.
36205 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36207         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36208         Add linked-list forward and backlinks.  Insert on
36209         construction, remove on destruction.
36210         (class pass_store_merging): Add m_stores_head field.
36211         (pass_store_merging::terminate_and_process_all_chains):
36212         Iterate over m_stores_head list.
36213         (pass_store_merging::terminate_all_aliasing_chains):
36214         Likewise.
36215         (pass_store_merging::execute): Check for debug stmts first.
36216         Push new chains onto the m_stores_head stack.
36218 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36220         PR target/71294
36221         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36222         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36223         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36225 2017-03-16  Jeff Law  <law@redhat.com>
36227         PR tree-optimization/71437
36228         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36229         member function.  Implementation moved into after_dom_children
36230         member function and into the threader's thread_outgoing_edges
36231         function.
36232         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36233         some code into new thread_outgoing_edges.
36234         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36235         definition.  Simplify marker handling (do it here).   Assume we always
36236         have the available expression and the const/copies tables.
36237         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36238         and tree-vrp.c
36239         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36240         * tree-vrp.c (equiv_stack): No longer file scoped.
36241         (vrp_dom_walker): New class.
36242         (vrp_dom_walker::before_dom_children): New member function.
36243         (vrp_dom_walker::after_dom_children): Likewise.
36244         (identify_jump_threads):  Setup domwalker.  Use it rather than
36245         walking edges in a random order by hand.  Simplify setup/finalization.
36246         (finalize_jump_threads): Remove.
36247         (vrp_finalize): Do not call identify_jump_threads here.
36248         (execute_vrp): Do it here instead and call thread_through_all_blocks
36249         here too.
36251         PR tree-optimization/71437
36252         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36253         callers changed.
36254         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36255         callers changed.
36256         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36257         (dom_opt_dom_walker::thread_across_edge): Remove
36258         handle_dominating_asserts argument.  All callers changed.
36259         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36260         changes.  Remove calls to lhs_of_dominating_assert.  Other
36261         uses of handle_dominating_asserts turn into unconditional code
36262         (simplify_control_stmt_condition_1): Likewise.
36263         (simplify_control_stmt_condition): Likewise.
36264         (thread_through_normal_block, thread_across_edge): Likewise.
36265         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36266         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36267         object if it is not an SSA_NAME.
36268         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36269         before calling into the VRP specific simplifiers.
36270         (identify_jump_threads): Remove handle_dominating_asserts
36271         argument.
36273 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36275         PR fortran/79886
36276         * tree-diagnostic.c (default_tree_printer): No longer static.
36277         * tree-diagnostic.h (default_tree_printer): New prototype.
36279 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36281         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36282         Change ins into fmov.
36284 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36286         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36287         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36288         Use h_con constraint for operand 1.
36289         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36290         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36292 2017-03-15  Jeff Law  <law@redhat.com>
36294         PR tree-optimization/71437
36295         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36296         (record_temporary_equivalences): Use it.
36298         PR tree-optimization/71437
36299         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36300         tree-ssa-scopedtables.
36301         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36302         (record_conditions, record_cond, vuse_eq): Likewise.
36303         (record_edge_info): Adjust to API tweak of record_conditions.
36304         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36305         (record_temporary_equivalences, optimize_stmt): Likewise.
36306         (eliminate_redundant_computations): Likewise.
36307         (record_equivalences_from_stmt): Likewise.
36308         * tree-ssa-scopedtables.c: Include options.h and params.h.
36309         (vuse_eq): New function, moved from tree-ssa-dom.c
36310         (build_and_record_new_cond): Likewise.
36311         (record_conditions): Likewise.  Accept vector of conditions rather
36312         than edge_equivalence structure for first argument.
36313         for the first argument.
36314         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36315         from tree-ssa-dom.c.
36316         (avail_exprs_stack::record_cond): Likewise.
36317         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36318         from tree-ssa-dom.c.
36319         (avail_exprs_stack): Add new member functions lookup_avail_expr
36320         and record_cond.
36321         (record_conditions): Declare.
36323 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36325         PR target/80017
36326         * lra-constraints.c (process_alt_operands): Increase reject for
36327         reloading an input/output operand.
36329 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36331         PR target/79038
36332         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36333         insns to convert from signed/unsigned char/short to IEEE 128-bit
36334         floating point.
36335         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36337 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36339         PR target/80019
36340         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36341         subreg of inner mode for values already in registers.
36343 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36345         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36346         iteration reg is used after the loop.
36348 2017-03-14  Martin Sebor  <msebor@redhat.com>
36350         PR tree-optimization/79800
36351         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36352         precision in negative-positive range.
36353         (format_floating): Call non-const overload with adjusted precision.
36355 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36357         PR target/79947
36358         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36359         -mpowerpc-gfxopt.
36361 2017-03-14  Martin Sebor  <msebor@redhat.com>
36363         PR middle-end/80020
36364         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36365         * builtins.def (aligned_alloc): Use it.
36367         PR c/79936
36368         * Makefile.in (GTFILES): Add calls.c.
36369         * calls.c: Include "gt-calls.h".
36371 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36373         PR rtl-optimization/79728
36374         * regs.h (struct target_regs): New field
36375         x_contains_allocatable_regs_of_mode.
36376         (contains_allocatable_regs_of_mode): New macro.
36377         * reginfo.c (init_reg_sets_1): Initialize it, and change
36378         contains_reg_of_mode so it includes global regs as well.
36379         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36380         rather than contains_regs_of_mode.
36382 2017-03-14  Martin Liska  <mliska@suse.cz>
36384         * doc/invoke.texi: Document options that can't be combined with
36385         -fcheck-pointer-bounds.
36387 2017-03-14  Martin Liska  <mliska@suse.cz>
36389         PR middle-end/79831
36390         * doc/invoke.texi (-Wchkp): Document the option.
36392 2017-03-14  Martin Liska  <mliska@suse.cz>
36394         * Makefile.in: Install gcov-dump.
36396 2017-03-14  Martin Liska  <mliska@suse.cz>
36398         * multiple_target.c (expand_target_clones): Bail out for
36399         an invalid attribute.
36401 2017-03-14  Richard Biener  <rguenther@suse.de>
36403         * alias.c (struct alias_set_entry): Pack properly.
36404         * cfgloop.h (struct loop): Likewise.
36405         * cse.c (struct set): Likewise.
36406         * ipa-utils.c (struct searchc_env): Likewise.
36407         * loop-invariant.c (struct invariant): Likewise.
36408         * lra-remat.c (struct cand): Likewise.
36409         * recog.c (struct change_t): Likewise.
36410         * rtl.h (struct address_info): Likewise.
36411         * symbol-summary.h (function_summary): Likewise.
36412         * tree-loop-distribution.c (struct partition): Likewise.
36413         * tree-object-size.c (struct object_size_info): Likewise.
36414         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36415         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36416         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36417         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36418         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36419         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36420         (struct _stmt_vec_info): Likewise.
36422 2017-03-14  Martin Liska  <mliska@suse.cz>
36424         PR target/79892
36425         * multiple_target.c (create_dispatcher_calls): Check that
36426         a target can create a function dispatcher.
36428 2017-03-14  Martin Liska  <mliska@suse.cz>
36430         PR lto/66295
36431         * multiple_target.c (expand_target_clones): Drop local.local
36432         flag for default implementation.
36434 2017-03-14  Richard Biener  <rguenther@suse.de>
36436         PR tree-optimization/80030
36437         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36439 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36441         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36442         gcc_fallthrough() instead of __attribute__((fallthrough));
36444 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36446         * doc/gcc.texi: Remove "up" link to (DIR).
36447         * doc/gccint.texi: Ditto.
36449 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36451         * doc/install.texi (Specific) <avr>: Remove reference to
36452         binutils 2.13.
36454 2017-03-13  Jeff Law  <law@redhat.com>
36456         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36457         attribute rather than comments.
36459         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36460         match_scratch operand is highest.
36462 2017-03-13  Martin Liska  <mliska@suse.cz>
36464         PR middle-end/78339
36465         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36466         is a CHKP clone, use original declaration.
36468 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36470         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36471         (arc_conditional_register_usage): Use a different allocation order
36472         when optimizing for size.
36473         * common/config/arc/arc-common.c (arc_option_optimization_table):
36474         Section anchors default on when optimizing for size.
36476 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36478         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36480 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36482         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36483         * config/arc/arc.md (cpu_facility): Add cd variant.
36484         (*movqi_insn): Add code density variant.
36485         (*movhi_insn): Likewise.
36486         (*movqi_insn): Likewise.
36487         (*addsi3_mixed): Likewise.
36488         (subsi3_insn): Likewise.
36490 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36492         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36494 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36496         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36497         expressions with MINUS and UNARY ops.
36499 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36501         PR target/79911
36502         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36503         Rename to...
36504         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36505         between vec_select and vector argument.
36506         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36507         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36508         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36509         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36510         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36511         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36513 2017-03-13  Richard Biener  <rguenther@suse.de>
36515         PR other/79991
36516         * params.def (vect-max-peeling-for-alignment): Fix typo.
36518 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36520         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36521         issue that only occurred with binutils below 2.18.
36523 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36525         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36526         refer to binutils 2.11/2.12 minimum.
36528 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36530         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36531         ftp.kernel.org and simplify binutils requirement.
36533 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36535         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36536         optimization.
36537         (Optimize Options): Ditto.  Also remove redundancy.
36539 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36541         PR translation/79848
36542         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36543         "%qs".
36544         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36545         to G_ to avoid double translation.
36547 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36549         PR translation/79923
36550         * auto-profile.c (get_combined_location): Convert leading
36551         character of diagnostics to lower case and remove trailing period.
36552         (read_profile): Likewise for various diagnostics.
36553         * config/arm/arm.c (arm_option_override): Remove trailing period
36554         from various diagnostics.
36555         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36556         (msp430_expand_delay_cycles): Likewise.
36558 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36560         PR target/79925
36561         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36562         full command-line argument, rather than just "str".
36563         (aarch64_validate_march): Likewise.
36564         (aarch64_validate_mtune): Likewise.
36566 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36568         PR rtl-optimization/78911
36569         * lra-assigns.c (must_not_spill_p): New function.
36570         (spill_for): Use it.
36572 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36574         PR tree-optimization/79981
36575         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36576         ATOMIC_COMPARE_EXCHANGE ifn result.
36577         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36578         IFN_ATOMIC_COMPARE_EXCHANGE.
36580 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36582         PR driver/79875
36583         * opts.c (parse_sanitizer_options): Add missing question mark to
36584         "did you mean" message.
36586 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36588         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36589         built-in.
36590         (VMULEUH_UNS): Likewise.
36591         (VMULOUB_UNS): Likewise.
36592         (VMULOUH_UNS): Likewise.
36593         * config/rs6000/rs6000.c (builtin_function_type): Remove
36594         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36596 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36598         PR bootstrap/79952
36599         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36600         x with result of extra_parsing_for_operand_code_0.
36601         (function_reader::extra_parsing_for_operand_code_0): Convert
36602         return type from void to rtx, returning x.  When reading
36603         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36604         larger size containing struct block_symbol.
36606 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36608         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36609         -mfloat128-hardware without -m64.
36611 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36613         PR target/79941
36614         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36615         entries to the case statement that marks unsigned arguments to
36616         overloaded functions.
36618 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36620         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36621         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36623 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36625         PR target/79907
36626         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36627         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36629 2017-03-10  Martin Liska  <mliska@suse.cz>
36631         PR target/65705
36632         PR target/69804
36633         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36634         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36635         FIELD != NULL.
36637 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36639         * tree-switch-conversion (array_value_type): Start by resetting
36640         candidate type to it's main variant.
36642 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36644         PR rtl-optimization/79909
36645         * combine.c (try_combine): Use simplify_replace_rtx on individual
36646         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36647         of the whole CALL_INSN_FUNCTION_USAGE.
36649         PR tree-optimization/79972
36650         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36651         get_range_info on SSA_NAMEs.  Formatting fixes.
36653 2017-03-10  Richard Biener  <rguenther@suse.de>
36654             Jakub Jelinek  <jakub@redhat.com>
36656         PR tree-optimization/77975
36657         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36658         edge to be constant.
36659         (get_val_for): For constant x return it.  Formatting fix.
36660         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36661         would use the same bases as the current one.
36663 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36665         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36666         instead of vec_select for V1TImode.
36667         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36668         longer needed.
36669         (VSX_LE_128): Add V1TI to this mode iterator.
36670         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36671         (*vsx_le_perm_store_<mode>): Likewise.
36672         (pre-reload splitter for VSX stores): Likewise.
36673         (post-reload splitter for VSX stores): Likewise.
36674         (*vsx_xxpermdi2_le_<mode>): Likewise.
36675         (*vsx_lxvd2x2_le_<mode>): Likewise.
36676         (*vsx_stxvd2x2_le_<mode>): Likewise.
36678 2017-03-09  Michael Eager  <eager@eagercon.com>
36680         Correct failures with --enable-checking=yes,rtl.
36682         * config/microblaze/microblaze.c (microblaze_expand_shift):
36683         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36684         test for const0_rtx.
36685         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36686         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36688 2017-03-09  Richard Biener  <rguenther@suse.de>
36690         PR tree-optimization/79977
36691         * graphite-scop-detection.c (scop_detection::merge_sese):
36692         Handle the case of extra exits to blocks dominating the entry.
36694 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36696         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36697         Document rdynamic.
36699 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36701         PR rtl-optimization/79949
36702         * lra-constraints.c (process_alt_operands): Check memory when
36703         trying to predict a cycle.  Print about the overall increase.
36705 2017-03-09  Richard Biener  <rguenther@suse.de>
36707         PR middle-end/79971
36708         * gimple-expr.c (useless_type_conversion_p): Preserve
36709         TYPE_SATURATING for fixed-point types.
36711 2017-03-09  Richard Biener  <rguenther@suse.de>
36713         PR ipa/79970
36714         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36715         alignment of BLKmode params.
36717 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36719         PR target/79913
36720         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36721         (VALL_NO_V2Q): Likewise.
36722         (VDQF_DF): Delete.
36723         * config/aarch64/aarch64-simd.md
36724         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36725         iterator.
36726         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36727         VALL_NO_V2Q mode iterator.
36728         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36730 2017-03-09  Martin Liska  <mliska@suse.cz>
36732         PR tree-optimization/79631
36733         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36734         tree_int_cst_sign_bit just for INTEGER constants.
36736 2017-03-09  Martin Liska  <mliska@suse.cz>
36738         PR target/65705
36739         PR target/69804
36740         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36741         sanitizers.
36743 2017-03-09  Marek Polacek  <polacek@redhat.com>
36745         PR c++/79672
36746         * tree.c (inchash::add_expr): Handle TREE_VEC.
36748 2017-03-09  Martin Liska  <mliska@suse.cz>
36750         PR ipa/79764
36751         (chkp_narrow_size_and_offset): New function.
36752         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36753         (void chkp_parse_bit_field_ref): New function.
36754         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36755         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36757 2017-03-09  Martin Liska  <mliska@suse.cz>
36759         PR ipa/79761
36760         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36761         (chkp_find_bounds_1): Remove gcc_unreachable.
36763 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36765         PR sanitizer/79944
36766         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36767         BUILT_IN_SYNC*, determine the access type from the size suffix and
36768         always build a MEM_REF with that type.  Handle forgotten
36769         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36771         PR target/79932
36772         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36773         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36774         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36775         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36776         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36777         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36778         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36779         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36780         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36781         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36782         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36783         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36784         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36785         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36786         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36787         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36788         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36789         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36790         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36791         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36792         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36793         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36794         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36795         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36796         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36797         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36798         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36799         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36800         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36801         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36802         definitions outside of __OPTIMIZE__ guarded section.
36804         PR target/79932
36805         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36806         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36807         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36808         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36809         guarded section.
36811 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36813         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36814         ("vfenez<mode>"): Add missing constraints.
36816 2017-03-08  Martin Sebor  <msebor@redhat.com>
36818         PR target/79928
36819         * config/nds32/nds32.c (nds32_option_override):
36820         Fix misspelled diagnostic.
36822 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36824         PR c/79940
36825         * gimplify.c (gimplify_omp_for): Replace index var in outer
36826         taskloop statement with an artificial variable and add
36827         OMP_CLAUSE_PRIVATE clause for it.
36829 2017-03-08  Richard Biener  <rguenther@suse.de>
36831         PR tree-optimization/79955
36832         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36833         for accesses that are completely outside of the variable.
36835 2017-03-08  Andrew Haley  <aph@redhat.com>
36837         PR tree-optimization/79943
36838         * tree-ssa-loop-split.c (compute_new_first_bound): When
36839         calculating the new upper bound, (END-BEG) should be added, not
36840         subtracted.
36842 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36844         * config/avr/avr.md (setmemhi): Make sure match_dup
36845         operand number comes before match_scratch.
36847 2017-03-08  Richard Biener  <rguenther@suse.de>
36849         PR tree-optimization/79920
36850         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36851         with ncopies == 1 to ...
36852         (vect_transform_slp_perm_load): ... here.  Properly compute
36853         all element loads by iterating VF times over the group.  Do
36854         not handle ncopies (computed in a broken way) in
36855         vect_create_mask_and_perm.
36857 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36859         PR sanitizer/79904
36860         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36861         is a uniform vector, use uniform_vector_p return value instead of
36862         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36864 2017-03-07  Marek Polacek  <polacek@redhat.com>
36866         PR middle-end/79809
36867         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
36868         (alloca_call_type): Likewise.
36870 2017-03-07  Martin Liska  <mliska@suse.cz>
36872         * gcov.c (process_args): Put comment to correct location.
36874 2017-03-07  Martin Liska  <mliska@suse.cz>
36876         PR middle-end/68270
36877         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
36878         Use array_at_struct_end_p instead of DECL_CHAIN (field).
36879         (chkp_narrow_bounds_for_field): Likewise.
36880         (chkp_parse_array_and_component_ref): Pass one more argument to
36881         call.
36883 2017-03-07  Richard Biener  <rguenther@suse.de>
36885         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
36886         preheaders.
36888 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
36890         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
36891         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
36893 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36895         PR c/79855
36896         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
36897         to end of description.
36898         (PARAM_MAX_STORES_TO_MERGE): Likewise.
36900 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
36902         PR rtl-optimization/79901
36903         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
36904         ...
36905         (*avx512f_<code><mode>3<mask_name>): ... this.
36906         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
36907         iterator instead of VI8_AVX2_AVX512BW.
36909         PR rtl-optimization/79901
36910         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
36911         min/max expander, expand it using expand_vec_cond_expr.
36913         PR sanitizer/79897
36914         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
36915         temporary.
36917 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
36919         PR c++/79821
36920         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
36921         to void * for PCH reasons.
36922         * dwarf2out.c (output_loc_operands, output_die): Cast
36923         v.val_vec.array to unsigned char *.
36925 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
36927         PR target/77850
36928         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
36929         vector types.
36931 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
36933         PR rtl-optimization/79571
36934         * lra-constraints.c (process_alt_operands): Calculate static
36935         reject and subtract it from overall when only addresses will be
36936         reloaded.
36938 2017-03-06  Julia Koval  <julia.koval@intel.com>
36940         PR target/79793
36941         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
36942         incoming stack boundary to 128 for 64-bit targets.
36944 2017-03-06  Richard Biener  <rguenther@suse.de>
36946         PR tree-optimization/79894
36947         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
36948         to NULL after folding it.
36950 2017-03-06  Richard Biener  <rguenther@suse.de>
36952         PR tree-optimization/79824
36953         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
36954         check disabling peeling for gaps.
36956 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
36958         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
36959         attributes): Document gettimeofday.
36961 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
36963         * config/s390/s390.c (s390_option_override_internal): Set
36964         PARAM_MIN_VECT_LOOP_BOUND
36966 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
36968         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
36969         * config/s390/s390.md: Likewise.
36971 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
36973         PR target/79812
36974         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
36975         (<avx2_avx512>_perm<mode>): Rename to ...
36976         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
36977         of VI8F_256_512.
36978         (<avx512>_perm<mode>_mask): Rename to ...
36979         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
36980         of VI8F_256_512.
36981         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
36982         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
36983         instead of VI8F_256_512.
36984         (avx512f_perm<mode>): New define_expand.
36985         (avx512f_perm<mode>_mask): Likewise.
36986         (avx512f_perm<mode>_1<mask_name>): New define_insn.
36987         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
36989 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
36991         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
36992         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
36993         if_then_else.
36994         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
36996 2017-03-06  Martin Liska  <mliska@suse.cz>
36998         PR sanitize/79783
36999         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37000         when having a SSA NAME w/o VAR_DECL assigned to it.
37002 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37004         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37005         msa_dpsub_<su>_d): Fix MODE for vec_select.
37007 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37009         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37010         argument.
37011         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37013 2017-03-06  Richard Biener  <rguenther@suse.de>
37015         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37016         * plugin.c (register_plugin_info): Likewise.
37017         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37019 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37021         * config/i386/sse.md (sse_storehps, sse_storelps,
37022         avx_<castmode><avxsizesuffix>_<castmode>,
37023         avx512f_<castmode><avxsizesuffix>_<castmode>,
37024         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37025         in condition that at least one operand is not a MEM.
37027 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37029         PR middle-end/79805
37030         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37031         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37032         ECF_NOTHROW.
37033         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37034         gimple_call_nothrow_p flag based on whether original builtin can throw.
37035         If it can, emit following stmts on the fallthrough edge.
37036         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37037         don't create new bb if inserting just debug stmts on the edge, try to
37038         insert them on the fallthru bb or just reset debug stmts.
37040 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37042         PR target/43763
37043         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37044         restore recog_data (including the operand rtxes inside it) around
37045         the call to get_insn_template.
37047 2017-03-03  Martin Sebor  <msebor@redhat.com>
37049         PR tree-optimization/79699
37050         * context.c (context::~context): Free MPFR caches to avoid
37051         a memory leak on program exit.
37053 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37055         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37056         Use wide_int::ulow () instead of .elt (0).
37058 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37060         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37061         (*pushxf): Limit oF constraint to 32bit targets and add oC
37062         constraint for 64bit targets.
37063         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37064         (*pushdf): Change rmF constraint to rmC.
37066 2017-03-03  Martin Liska  <mliska@suse.cz>
37068         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37069         Remove unused variable.
37071 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37073         PR target/79807
37074         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37075         is a memory operand, increase num_memory.
37076         (ix86_expand_args_builtin): Likewise.
37078 2017-03-03  Jan Hubicka  <jh@suse.cz>
37080         PR lto/79760
37081         * ipa-devirt.c (maybe_record_node): Properly handle
37082         __cxa_pure_virtual visibility.
37084 2017-03-03  Martin Liska  <mliska@suse.cz>
37086         PR tree-optimization/79803
37087         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37088         assert.
37089         (pass_loop_prefetch::execute): Disabled optimization if an
37090         assumption about L1 cache size is not met.
37092 2017-03-03  Martin Liska  <mliska@suse.cz>
37094         PR rtl-optimization/79574
37095         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37096         (hash_scan_set): Likewise.
37097         (dump_hash_table): Likewise.
37098         (hoist_code): Likewise.
37100 2017-03-03  Richard Biener  <rguenther@suse.de>
37102         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37103         in place of uhigh ().
37104         (fixed_convert_from_real): Likewise.
37106 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37108         PR target/79514
37109         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37111 2017-03-03  Richard Biener  <rguenther@suse.de>
37113         PR middle-end/79818
37114         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37115         TYPE_OVERFLOW_UNDEFINED check.
37117 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37119         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37120         numbers.
37121         (vector_ae_<mode>_p): Likewise.
37122         (vector_nez_<mode>_p): Likewise.
37123         (vector_ne_v2di_p): Likewise.
37124         (vector_ae_v2di_p): Likewise.
37125         (vector_ne_<mode>_p): Likewise.
37126         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37127         numbers.
37128         (vsx_tsqrt<mode>2_fe): Likewise.
37130 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37132         PR target/79514
37133         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37135 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37137         PR rtl-optimization/79780
37138         * cprop.c (one_cprop_pass): When second and further conditional trap
37139         in a single basic block is turned into an unconditional trap, turn it
37140         into a deleted note to avoid RTL verification failures.
37142 2017-03-02  Richard Biener  <rguenther@suse.de>
37144         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37146 2017-03-02  Richard Biener  <rguenther@suse.de>
37148         PR tree-optimization/79345
37149         PR c++/42000
37150         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37151         param and abort the walk, returning -1 if it is hit.
37152         (walk_aliased_vdefs): Take a limit param and pass it on.
37153         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37154         defaulting to 0 and return a signed int.
37155         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37156         (check_defs): New helper.
37157         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37158         about uninitialized memory.
37159         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37160         bogus uninitialized warning.
37161         (fixed_convert_from_real): Likewise.
37163 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37165         PR tree-optimization/66768
37166         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37167         iv_use if base object can't be determined.
37169 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37171         PR tree-optimization/79345
37172         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37173         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37174         (get_pattern_stats): Initialize it.
37175         * genemit.c (gen_expand): Verify match_scratch numbers come after
37176         match_operand/match_dup numbers.
37177         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37178         match_scratch numbers.
37179         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37180         Likewise.
37181         * config/s390/s390.md (trunctdsd2): Likewise.
37183 2017-03-02  Richard Biener  <rguenther@suse.de>
37185         * wide-int.h (wide_int_storage::operator=): Implement in terms
37186         of wi::copy.
37188 2017-03-02  Richard Biener  <rguenther@suse.de>
37190         PR tree-optimization/79777
37191         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37192         the to insert expression to sth existing.
37194 2017-03-01  Martin Sebor  <msebor@redhat.com>
37196         PR middle-end/79692
37197         * gimple-ssa-sprintf.c
37198         (directive::known_width_and_precision): New function.
37199         (format_integer): Use it.
37200         (get_mpfr_format_length): Consider the full range of precision
37201         when computing %g output with the # flag.  Set the likely byte
37202         count to 3 rather than 1 when precision is indeterminate.
37203         (format_floating): Correct the lower bound of precision.
37205 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37207         * doc/invoke.texi: Document default code model for 64-bit Linux.
37209 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37211         PR target/79752
37212         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37213         udiv rather than div since input pattern is unsigned.
37215 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37217         * config/i386/i386.c (print_reg): Warn for values of
37218         unsupported size in integer register.
37220 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37222         PR target/79439
37223         * config/rs6000/predicates.md (current_file_function_operand): Do
37224         not allow self calls to be local if the function is replaceable.
37226 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37228         PR target/79395
37229         * config/rs6000/altivec.h (vec_ctz and others): Change the
37230         preprocessor macro that controls conditional compilation from
37231         _ARCH_PWR9 to __POWER9_VECTOR__.
37232         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37233         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37234         control (instead of _ARCH_PWR9 control) so that template
37235         definition uses power9-specific function.
37236         (vec_any_eq): Likewise.
37237         (vec_all_ne): Change macro definition to use a power9-specific
37238         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37239         _ARCH_PWR9 control).
37240         (vec_any_eq) Likewise.
37241         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37242         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37243         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37244         support for xvcmpnedp instruction.
37245         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37246         macro expansion so that Power9 implementation of vec_all_ne does
37247         not use the AltiVec predicate framework.
37248         (VCMPNEH_P): Likewise.
37249         (VCMPNEW_P): Likewise.
37250         (VCMPNED_P): Likewise.
37251         (VCMPNEFP_P): Likewise.
37252         (VCMPNEDP_P): Likewise.
37253         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37254         implementation of vec_any_eq to not use AltiVec predicate
37255         framework.
37256         (VCMPAEH_P): Likewise.
37257         (VCMPAEW_P): Likewise.
37258         (VCMPAED_P): Likewise.
37259         (VCMPAEFP_P): Likewise.
37260         (VCMPAEDP_P): Likewise.
37261         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37262         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37263         not use the AltiVec predicate framework.
37264         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37265         of vec_any_eq to not use AltiVec predicate framework.
37266         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37267         support for predefined __POWER9_VECTOR__ macro to indicate that
37268         Power9 instruction selection is enabled.
37269         (altivec_overloaded_builtins): Remove extraneous
37270         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37271         function argument types RS6000_BTI_bool_V16QI and
37272         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37273         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37274         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37275         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37276         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37277         Power9 for implementations of vec_cmpne.  Change the signature for
37278         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37279         (representing vec_all_ne) to remove the previously described first
37280         argument of type RS6000_BTI_INTSI, as this was an artifact of
37281         reliance on the AltiVec predicate framework, which is no longer
37282         used in the implementation of these functions.  Add
37283         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37284         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37285         since, unlike the AltiVec predicate framework implementation, we
37286         do not share function descriptors between vec_alle and vec_anyeq.
37287         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37288         set of modes that receive special treatment even when
37289         TARGET_P9_VECTOR is true.  The special treatment emits code that
37290         does not depend on Power9 instructions.
37291         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37292         define_expand to not rely on AltiVec predicate framework.
37293         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37294         function.
37295         (vector_ne_v2di_p): Change this define_expand to not rely on
37296         AltiVec predicate framework.
37297         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37298         function.
37299         (vector_ne_<mode>_p): Change this define_expand to not rely on
37300         AltiVec predicate framework.
37301         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37302         function.
37303         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37304         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37305         define_insn pattern.
37306         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37307         define_insn pattern because the xvcmpne<VSs>. instruction is not
37308         supported.
37309         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37310         instruction is not supported.
37312 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37314         * config/nvptx/nvptx.c: Include intl.h.
37316 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37318         PR lto/78140
37319         * ipa-prop.h (ipa_bits): Removed field known.
37320         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37321         to pointers.  Adjusted their comments to warn about their sharing.
37322         (ipcp_transformation_summary): Change bits to a vector of pointers.
37323         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37324         (ipa_get_ipa_bits_for_value): Declare.
37325         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37326         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37327         (ipa_bits_hash_table): Likewise.
37328         (ipa_vr_ggc_hash_traits): Likewise.
37329         (ipa_vr_hash_table): Likewise.
37330         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37331         being pointers and vr_known being removed.
37332         (ipa_set_jf_unknown): Likewise.
37333         (ipa_get_ipa_bits_for_value): New function.
37334         (ipa_set_jfunc_bits): Likewise.
37335         (ipa_get_value_range): New overloaded functions.
37336         (ipa_set_jfunc_vr): Likewise.
37337         (ipa_compute_jump_functions_for_edge): Use the above functions to
37338         construct bits and vr parts of jump functions.
37339         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37340         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37341         exist.
37342         (ipcp_grow_transformations_if_necessary): Also allocate
37343         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37344         exist.
37345         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37346         them.  Fix too long lines.
37347         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37348         vr_known being removed.
37349         (ipa_read_jump_function): Use new setter functions to construct bits
37350         and vr parts of jump functions or set them to NULL.
37351         (write_ipcp_transformation_info): Adjust for bits being pointers.
37352         (read_ipcp_transformation_info): Likewise.
37353         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37354         space.
37355         Include gt-ipa-prop.h.
37356         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37357         being pointers.
37358         (ipcp_store_bits_results): Likewise.
37359         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37360         Do not write to existing jump functions but use a temporary instead.
37362 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37364         PR c++/79681
37365         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37366         attempt to use its first operand as BIT_FIELD_REF base.
37368 2017-03-01  Richard Biener  <rguenther@suse.de>
37370         PR middle-end/79721
37371         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37372         interpolating formula in wrapping arithmetic.
37373         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37375 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37377         PR tree-optimization/79734
37378         * tree-vect-generic.c (expand_vector_condition): Optimize
37379         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37380         Handle VEC_COND_EXPR where comparison has different inner width from
37381         type's inner width.
37383 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37385         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37386         markup, and similar issues.  Remove @opindex entries for things
37387         that aren't options.  Add missing -mmpy-option entries.
37389 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37391         PR tree-optimization/79737
37392         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37393         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37394         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37395         instead of byte_size.  Formatting fix.
37396         (shift_bytes_in_array_right): Formatting fix.
37398 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37400         PR target/79749
37401         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37402         condition on optimize for the leaf function test.
37404 2017-02-28  Martin Liska  <mliska@suse.cz>
37406         PR lto/79625
37407         * read-rtl-function.c (function_reader::handle_unknown_directive):
37408         Bail out when one uses -flto.
37410 2017-02-28  Martin Liska  <mliska@suse.cz>
37412         * common.opt: Replace space with tabular for options of <number>
37413         type.
37414         * config/i386/i386.opt: Show <number> value for
37415         -mlarge-data-threshold.
37416         * opts.c (print_filtered_help): Do not display number in hexadecimal
37417         format.
37419 2017-02-28  Martin Liska  <mliska@suse.cz>
37421         * common.opt: Fix --help=option -Q for options which are of
37422         an enum type.
37424 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37426         * config/i386/i386.c (print_reg): Error out for values
37427         of 8-bit size in invalid integer register.
37429 2017-02-28  Martin Sebor  <msebor@redhat.com>
37431         PR tree-optimization/79691
37432         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37434 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37436         PR target/79729
37437         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37438         gcc_unreachable with output_operand_lossage.
37440 2017-02-28  Richard Biener  <rguenther@suse.de>
37442         PR tree-optimization/79740
37443         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37444         inserts.
37445         (visit_nary_op): Insert the nary into the hashtable if we
37446         pattern-matched sth.
37447         * tree-ssa-pre.c (eliminate_insert): Robustify.
37449 2017-02-28  Richard Biener  <rguenther@suse.de>
37451         PR middle-end/79731
37452         * fold-const.c (decode_field_reference): Reject out-of-bound
37453         accesses.
37455 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37457         * config/i386/i386.c: Include intl.h.
37458         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37459         instead of just cond ? "..." : "...".
37460         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37461         * coverage.c (read_counts_file): Likewise.
37462         * omp-offload.c: Include intl.h.
37463         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37464         of just cond ? "..." : "...".
37465         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37466         of just cond ? "..." : "...".
37468 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37470         PR target/79742
37471         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37472         entry, if present.
37473         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37474         'tune for' CPU name.
37475         * config/arm/arm-cpu-data.h: Regenerated.
37477 2017-02-28  Richard Biener  <rguenther@suse.de>
37479         PR tree-optimization/79732
37480         * tree-inline.c (expand_call_inline): Do not shadow var.
37482 2017-02-28  Richard Biener  <rguenther@suse.de>
37484         PR tree-optimization/79723
37485         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37486         address-space properly.
37488 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37490         * doc/optinfo.texi (Optimization groups): Fix option used for
37491         OPTGROUP_ALL.
37492         * doc/invoke.texi (-fopt-info): Document "omp".
37493         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37494         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37495         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37496         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37497         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37499         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37500         all users.
37501         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37502         OPTGROUP_OMP with "omp".
37504 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37506         PR target/79544
37507         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37508         for arithmetic shift of unsigned V2DI.
37510 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37512         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37513         arc/linux.h headers.
37514         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37515         (LINK_SPEC): Likewise.
37516         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37517         (EXTRA_SPECS): Likewise.
37518         (STARTFILE_SPEC): Likewise.
37519         (ENDFILE_SPEC): Likewise.
37520         (LIB_SPEC): Likewise.
37521         (TARGET_SDATA_DEFAULT): Likewise.
37522         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37523         (MULTILIB_DEFAULTS): Likewise.
37524         (DWARF2_UNWIND_INFO): Likewise.
37525         * config/arc/big.h: New file.
37526         * config/arc/elf.h: Likewise.
37527         * config/arc/linux.h: Likewise.
37528         * config/arc/t-uClibc: Remove.
37530 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37532         PR tree-optimization/77536
37533         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37534         (tree_transform_and_unroll_loop): Use above function to compute the
37535         estimated niter of unrolled loop and use it when scaling profile.
37536         Also use count info rather than frequency if it's non-zero.
37537         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37538         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37539         (vect_transform_loop): Call above function.
37541 2017-02-27  Richard Biener  <rguenther@suse.de>
37543         PR tree-optimization/45397
37544         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37545         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37546         (visit_nary_op): Add pattern matching for CSEing sign-changed
37547         or truncated operations with wider ones.
37549 2017-02-27  Richard Biener  <rguenther@suse.de>
37551         PR tree-optimization/79690
37552         * tree-vect-stmts.c (vectorizable_store): Use vector type
37553         built from the DR with address-space.
37555 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37557         * doc/invoke.texi (Optimize Options): Refine the description
37558         of asan-use-after-return.
37560 2017-02-25  Alan Modra  <amodra@gmail.com>
37562         PR rtl-optimization/79584
37563         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37564         base, not ad->base_term, the reg within base.  Remove assertion
37565         that ad->base == ad->base_term.  Replace gen_int_mode using
37566         bogus mode with const0_rtx.
37568 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37570         PR middle-end/79396
37571         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37572         FMA_EXPR like tcc_binary or tcc_unary.
37574         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37576         PR debug/77589
37577         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37578         bitfield.
37579         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37580         (output_loc_operands): Handle DW_OP_call_ref and
37581         DW_OP_GNU_variable_value.
37582         (struct variable_value_struct): New type.
37583         (struct variable_value_hasher): Likewise.
37584         (variable_value_hash): New variable.
37585         (string_types): Remove.
37586         (copy_loc_descr): New function.
37587         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37588         (prepend_loc_descr_to_each): New function.
37589         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37590         instead of add_loc_descr_to_each if the first argument is single
37591         location list and the second has multiple.
37592         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37593         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37594         when looking for variable value which doesn't have other location info.
37595         (loc_list_from_tree): Formatting fix.
37596         (gen_array_type_die): Simplify DW_AT_string_length handling.
37597         (adjust_string_types): Remove.
37598         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37599         string_types.  Call resolve_variable_values.
37600         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37601         (resolve_addr_in_expr): Likewise.  Add A argument.
37602         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37603         original expression being DW_OP_GNU_variable_value with optionally
37604         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37605         optionally after it.
37606         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37607         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37608         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37609         (variable_value_hasher::hash, variable_value_hasher::equal): New
37610         methods.
37611         (resolve_variable_value_in_expr, resolve_variable_value,
37612         resolve_variable_values, note_variable_value_in_expr,
37613         note_variable_value): New functions.
37614         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37615         DIEs.
37617 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37619         PR c/79677
37620         * opts.h (handle_generated_option): Add GENERATED_P argument.
37621         * opts-common.c (handle_option): Adjust function comment.
37622         (handle_generated_option): Add GENERATED_P argument, pass it to
37623         handle_option.
37624         (control_warning_option): Pass false to handle_generated_option
37625         GENERATED_P.
37626         * opts.c (maybe_default_option): Pass true to handle_generated_option
37627         GENERATED_P.
37628         * optc-gen.awk: Likewise.
37630 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37632         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37633         a REG, look at the REG it is a SUBREG of.
37634         (splitter for cmpeqsi_t): Ditto.
37636 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37638         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37639         the special USEs with the pattern of the insn, not the insn itself.
37641 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37643         PR target/79473
37644         * doc/invoke.texi: Document -mload-store-pairs.
37646 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37647             Sandra Loosemore  <sandra@codesourcery.com>
37649         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37650         argument isn't a CONST_INT.
37651         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37652         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37653         (nios2_validate_compare): Bypass alternate compare logic if *op2
37654         is not a CONST_INT.
37655         (ldstwm_operation_p): Return false if first_base is not a REG or
37656         if first_offset is not a CONST_INT.
37658 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37660         * config/cris/cris.md: Use correct operand in a define_peephole2.
37662 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37664         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37666 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37668         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37669         this_insn if it is an INSN or JUMP_INSN.
37670         (force_offsettable): Look at base, not at addr.
37671         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37672         on things that aren't necessarily CONST_INTs.
37674 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37676         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37677         -mfpmath=sse is the default also for x86-32 targets with SSE2
37678         instruction set when @option{-ffast-math} is enabled
37680 2017-02-24  Jeff Law  <law@redhat.com>
37682         PR rtl-optimizatoin/79286
37683         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37684         dominance test.
37686 2017-02-24  Richard Biener  <rguenther@suse.de>
37688         PR tree-optimization/79389
37689         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37690         debug insns.
37692 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37694         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37695         function comment to reflect reality.
37696         (loop_exits_before_overflow): Fix typo in function description.
37698 2017-02-24  Richard Biener  <rguenther@suse.de>
37700         PR tree-optimization/79389
37701         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37702         properly that a threading opportunity exists.  Detect conditional
37703         copy/constant propagation opportunities.
37705 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37707         * config/visium/visium.md (type): Add trap.
37708         (b): New mode attribute.
37709         (*btst): Rename into...
37710         (*btst<mode>): ...this and adjust.
37711         (*cbranchsi4_btst_insn): Rename into...
37712         (*cbranch<mode>4_btst_insn): ...this and adjust.
37713         (trap): New define_insn.
37715 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37717         PR tree-optimization/79389
37718         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37719         (noce_reversed_cond_code): New function.
37720         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37721         reversed_comparison_code.  Formatting fix.
37722         (noce_try_store_flag): Test rev_cond != NULL in addition to
37723         reversed_comparison_code.
37724         (noce_try_store_flag_constants): Likewise.
37725         (noce_try_store_flag_mask): Likewise.
37726         (noce_try_addcc): Use rev_cond if non-NULL instead of
37727         reversed_comparison_code.
37728         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37729         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37730         (noce_find_if_block): Initialize rev_cond.
37731         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37732         instead of false as last argument never attempt to reverse it
37733         afterwards.
37735 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37737         PR tree-optimization/79663
37738         * tree-predcom.c (combine_chains): Process refs in reverse order
37739         only for ZERO length chains, and add explaining comment.
37741 2017-02-23  Jeff Law  <law@redhat.com>
37743         PR tree-optimization/79578
37744         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37745         in call to operand_equal_p.
37747 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37749         PR target/71017
37750         * config/i386/cpuid.h: Fix another undefined behavior.
37752 2017-02-23  Richard Biener  <rguenther@suse.de>
37754         PR tree-optimization/79683
37755         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37756         vector types for data-refs.
37758 2017-02-23  Martin Liska  <mliska@suse.cz>
37760         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37762 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37764         PR middle-end/79665
37765         * internal-fn.c (get_range_pos_neg): Moved to ...
37766         * tree.c (get_range_pos_neg): ... here.  No longer static.
37767         * tree.h (get_range_pos_neg): New prototype.
37768         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37769         are known to be in between 0 and signed maximum inclusive, try to
37770         expand both unsigned and signed divmod and use the cheaper one from
37771         those.
37773 2017-02-22  Jeff Law  <law@redhat.com>
37775         PR tree-optimization/79578
37776         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37777         to compare base operands.
37779 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37781         PR target/79211
37782         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37783         gpc_reg_operand instead of fpr_reg_operand.
37785 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37787         * config/mips/mips.c (mips_return_in_memory): Force FP
37788         vector types to be returned in memory for o32 ABI.
37790 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37792         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37793         instead of DW_TAG_member for static data member declarations and don't
37794         set no_linkage_name for static inline data members.
37795         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37796         to DW_TAG_member.
37798 2017-02-22  Martin Liska  <mliska@suse.cz>
37800         * doc/invoke.texi: Replace inequality signs with square brackets
37801         for -Wnormalized.
37803 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37805         PR target/78660
37806         * lra-constraints.c (simplify_operand_subreg): Handle
37807         WORD_REGISTER_OPERATIONS targets.
37809 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37811         PR target/70465
37812         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37813         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37814         elimination by swapping fld*.
37816 2017-02-22  Richard Biener  <rguenther@suse.de>
37818         PR tree-optimization/79673
37819         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37820         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37821         irrelevant address-space qualifiers and avoiding a
37822         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37824 2017-02-22  Richard Biener  <rguenther@suse.de>
37826         PR tree-optimization/79666
37827         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37828         to not symbolically negate if that may introduce undefined
37829         overflow.
37831 2017-02-22  Martin Liska  <mliska@suse.cz>
37833         PR lto/79587
37834         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37835         * data-streamer-out.c (streamer_write_gcov_count_stream):
37836         Likewise.
37837         * value-prof.c (stream_out_histogram_value): Make assert more
37838         precise based on type of counter.
37840 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37842         PR target/79593
37843         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37844         Use nonimmediate_operand instead of memory_operand for operand 1.
37845         (float-extend standard_x87sse_constant_load splitter): Ditto.
37847 2017-02-21  Jeff Law  <law@redhat.com>
37849         PR tree-optimization/79621
37850         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37851         blocks with edges to themselves.
37853 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37855         PR target/79633
37856         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37857         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37858         Use gimple_call_builtin_p.
37860         PR target/79570
37861         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37862         on temporarily removed DEBUG_INSNs.
37864         PR tree-optimization/79649
37865         * tree-loop-distribution.c (classify_partition): Give up on
37866         non-generic address space loads/stores.
37868 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
37870         * doc/loop.texi (Loop manipulation): Remove nonexistent
37871         tree_ssa_loop_version from the documentation.
37872         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
37874 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37876         PR target/79494
37877         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
37878         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
37879         * config/rs6000/rs6000.c: Include except.h.
37880         (rs6000_expand_split_stack_prologue): Call
37881         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
37883 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37885         PR lto/79579
37886         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
37887         have been analyzed.
37889 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37891         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
37892         for backward compatibility only.
37893         * doc/invoke.texi (Option Summary): Remove all references to
37894         -fipa-cp-alignment.
37896 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
37898         PR target/78660
37899         Revert:
37900         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37902         * lra-constraints.c (curr_insn_transform): Handle
37903         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37905 2017-02-21  Martin Liska  <mliska@suse.cz>
37907         * config/i386/i386.opt: Replace -masm-dialect with -masm.
37909 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
37911         PR translation/79638
37912         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
37914 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
37916         PR ada/67205
37917         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
37918         (arm_function_ok_for_sibcall): Return false for an indirect call by
37919         descriptor if all the argument registers are used.
37920         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
37921         alignment of the function.
37923 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37925         PR tree-optimization/61441
37926         * simplify-rtx.c (simplify_const_unary_operation): For
37927         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
37928         the sNaN unmodified.
37930 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
37932         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
37933         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
37934         instead of SYSTEM_HEADER_DIR.
37936 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
37937             Martin LiÅ¡ka  <mliska@suse.cz>
37939         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
37940         Fix typos and grammar, use active voice, and clarify.
37942 2017-02-20  Marek Polacek  <polacek@redhat.com>
37944         PR middle-end/79537
37945         * gimplify.c (gimplify_expr): Handle unused *&&L;.
37947         PR sanitizer/79558
37948         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
37950 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
37952         PR target/79568
37953         * config/i386/i386.c (ix86_expand_builtin): Handle
37954         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
37955         ix86_builtins_isa[fcode].isa as a requirement of those
37956         flags and any other flag in the bitmask.
37957         (ix86_init_mmx_sse_builtins): Use 0 instead of
37958         ~OPTION_MASK_ISA_64BIT as mask.
37959         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
37960         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
37961         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
37962         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
37964 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37966         PR target/78012
37967         * lra-constraints.c (split_reg): Check requested split mode
37968         is supported by the register.
37970 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37972         * lra-constraints.c (simplify_operand_subreg): Remove early
37973         return false.
37975 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37977         PR target/78660
37978         * lra-constraints.c (curr_insn_transform): Tighten condition
37979         for converting SUBREG reloads from OP_OUT to OP_INOUT.
37981 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37983         PR target/78660
37984         * lra-constraints.c (curr_insn_transform): Handle
37985         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37987 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
37989         Revert:
37990         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
37992         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
37994 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
37996         PR c++/69523
37997         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
37998         description.
38000 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38002         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38003         for FMA_EXPR.
38005 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38007         * final.c (last_columnnum, override_columnnum): New variables.
38008         (final_start_function): Set last_columnnum, pass it to begin_prologue
38009         hook and pass 0 to dwarf2out_begin_prologue.
38010         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38011         to source_line debug hook.
38012         (notice_source_line): Compute last_columnnum and for debug_column_info
38013         return true on column changes.
38014         * debug.h (struct gcc_debug_hooks): Add column argument to
38015         source_line and begin_prologue hooks.
38016         (debug_nothing_int_charstar_int_bool): Remove prototype.
38017         (debug_nothing_int_int_charstar,
38018         debug_nothing_int_int_charstar_int_bool): New prototypes.
38019         (dwarf2out_begin_prologue): Add column argument.
38020         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38021         begin_prologue hooks.
38022         (debug_nothing_int_charstar_int_bool): Remove.
38023         (debug_nothing_int_int_charstar,
38024         debug_nothing_int_int_charstar_int_bool): New functions.
38025         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38026         through to dwarf2out_source_line.
38027         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38028         (dwarf2out_source_line): Add column argument, emit it if requested.
38029         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38030         arguments.
38031         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38032         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38033         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38034         through to dwarf2out_begin_prologue.
38035         (vmsdbgout_source_line): Add column argument, pass it through to
38036         dwarf2out_source_line.
38037         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38038         dbxout_source_line caller.
38039         (dbxout_source_line): Add column argument.
38041         * common.opt (gno-column-info, gcolumn-info): New options.
38042         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38043         (check_die): Also test for multiple DW_AT_decl_column attributes.
38044         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38045         DW_AT_decl_column if requested.
38046         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38047         if requested.
38048         (gen_variable_die): Likewise.
38049         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38050         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38052         PR target/79569
38053         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38054         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38055         (ix86_handle_option): Handle OPT_m3dnowa.
38056         * doc/invoke.texi (-m3dnowa): Document.
38057         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38058         -m3dnowa instead of -m3dnow -march=athlon.
38060         PR target/79559
38061         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38062         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38064 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38066         PR target/79261
38067         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38068         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38069         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38070         generator for vsx_xxpermdi_<mode>_be.
38071         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38072         force big-endian semantics.
38073         (vsx_xxpermdi_<mode>_be): New define_expand with same
38074         implementation as previous version of vsx_xxpermdi_<mode>.
38076 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38078         PR tree-optimization/79327
38079         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38080         variable, its initialization and use.
38082 2017-02-17  Julia Koval  <julia.koval@intel.com>
38084         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38085         (OPTION_MASK_ISA_PKU_UNSET): New.
38086         (ix86_handle_option): Handle -mrdpid.
38087         * config/i386/cpuid.h (bit_RDPID): New.
38088         * config/i386/driver-i386.c (host_detect_local_cpu):
38089         Detect RDPID feature.
38090         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38091         * config/i386/i386-c.c (ix86_target_macros_internal):
38092         Handle RDPID flag.
38093         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38094         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38095         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38096         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38097         * config/i386/i386.md (define_insn "rdpid"): New.
38098         * config/i386/i386.opt Add -mrdpid.
38099         * config/i386/immintrin.h (_rdpid_u32): New.
38101 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38103         PR rtl-optimization/79541
38104         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38105         instead of transforming it into USE.
38107 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38109         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38110         If HONOR_SNANS (SFmode) force the input to a register.
38111         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38112         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38113         an frsp or similar insn.
38115 2017-02-17  Martin Liska  <mliska@suse.cz>
38117         PR rtl-optimization/79577
38118         * params.def (selsched-max-sched-times): Increase minimum to 1.
38120 2017-02-17  Martin Liska  <mliska@suse.cz>
38122         PR rtl-optimization/79574
38123         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38125 2017-02-17  Martin Liska  <mliska@suse.cz>
38127         PR tree-optimization/79529
38128         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38129         ssa_defined_default_def_p to handle cases which are implicitly
38130         defined.
38131         * tree-ssa.c (ssa_defined_default_def_p): New function.
38132         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38133         which are implicitly defined.
38134         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38136 2017-02-17  Richard Biener  <rguenther@suse.de>
38138         PR middle-end/79576
38139         * params.def (max-ssa-name-query-depth): Limit to 10.
38141 2017-02-17  Richard Biener  <rguenther@suse.de>
38143         PR tree-optimization/79552
38144         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38145         default defs.
38147 2017-02-17  Richard Biener  <rguenther@suse.de>
38149         PR bootstrap/79567
38150         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38152 2017-02-17  Marek Polacek  <polacek@redhat.com>
38154         PR middle-end/79536
38155         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38156         (fold_negate_expr): New wrapper.
38158 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38160         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38161         Correct terminology and de-emphasize pre-standard behavior.
38163 2017-02-16  Alan Modra  <amodra@gmail.com>
38165         PR rtl-optimization/79286
38166         * ira.c (def_dominates_uses): New function.
38167         (update_equiv_regs): Don't create an equivalence for insns that
38168         may trap where the register def does not dominate the use.
38170 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38172         PR rtl-optimization/78127
38173         * lra.c (lra): Call lra_eliminate before finish the loop after
38174         lra_constraint.
38176 2017-02-16  Richard Biener  <rguenther@suse.de>
38178         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38179         isl/isl_val.h.
38180         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38181         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38182         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38183         (isl_val_int_from_wi): New function.
38184         (extract_affine_gmp): Rename to ...
38185         (extract_affine_wi): ... this, take a widest_int.
38186         (extract_affine_int): Just wrap extract_affine_wi.
38187         (add_param_constraints): Use isl_val_int_from_wi.
38188         (add_loop_constraints): Likewise, and extract_affine_wi.
38190 2017-02-15  Jeff Law  <law@redhat.com>
38192         PR middle-end/79521
38193         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38194         ira_init_register_move_cost_if_necessary.
38196 2017-02-15  Martin Sebor  <msebor@redhat.com>
38198         PR middle-end/32003
38199         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38200         removed in a prior commit.
38202 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38204         PR tree-optimization/79347
38205         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38206         counters during peeling.
38208 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38210         * Makefile.in (site.exp): Remove "set ISLVER".
38212 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38214         PR target/79487
38215         * real.c (real_from_integer): Call real_convert even for decimal.
38217 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38219         PR target/79421
38220         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38222 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38224         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38225         cores and change the partno/implementer to be correct.
38226         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38227         the 'B" as the implementer.
38228         * config/aarch64/aarch64-tune.md: Regenerate.
38230 2017-02-14  Carl Love  <cel@us.ibm.com>
38232         * config/rs6000/rs6000.c: Add case statement entry to make the
38233         xvcvuxdsp built-in argument unsigned.
38234         * config/rs6000/vsx.md: Fix the source and return operand types so they
38235         match the instruction definitions from the ISA document.  Fix typo
38236         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38237         statement.
38239 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38241         PR target/79282
38242         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38243         member early_clobber_alts.
38244         * lra-lives.c (reg_early_clobber_p): New.
38245         (process_bb_lives): Use it.
38246         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38247         (debug_operand_data): Initialize early_clobber_alts.
38248         (setup_operand_alternative): Set up early_clobber_alts.
38249         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38250         alternatives to new_insn_reg.
38251         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38252         it.
38253         (lra_update_insn_regno_info): Pass the new arg.
38255 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38257         PR middle-end/79505
38258         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38259         (new_oacc_loop_raw): Don't clear already cleared fields.
38261         PR target/79481
38262         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38263         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38264         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38266 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38268         PR target/79495
38269         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38271 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38273         PR target/79498
38274         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38275         the extra instruction to the right place to store 128-bit constant
38276         when needed.
38278 2017-02-14  Martin Sebor  <msebor@redhat.com>
38280         PR middle-end/79448
38281         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38282           warning for strings of unknown length.
38284 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38286         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38288 2017-02-14  Jeff Law  <law@redhat.com>
38290         PR target/79404
38291         * ira-costs.c (scan_one_insn): Initialize register move costs
38292         for pseudos seen in USE/CLOBBER insns.
38294         PR tree-optimization/79095
38295         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38296         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38297         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38298         if the operands are known to be not equal, then the resulting range
38299         is ~[0,0].
38300         (intersect_ranges): If the new range is ~[0,0] and the old range is
38301         wide, then prefer ~[0,0].
38302         * tree-vrp.c (overflow_comparison_p_1): New function.
38303         (overflow_comparison_p): New function.
38304         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38305         if NAME is used in an overflow test.
38306         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38307         overflow check that can be expressed as an equality test, then adjust
38308         ops to be that equality test.
38310 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38312         * config/s390/s390-builtin-types.def: Remove flags argument.
38313         * config/s390/s390.c (s390_init_builtins): Likewise.
38315 2017-02-14  Martin Liska  <mliska@suse.cz>
38317         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38318         vector.  Fix trailing white spaces.
38320 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38322         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38323         HFmode.
38325 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38327         PR rtl-optimization/68664
38328         * config/arm/arm.c (arm_sched_can_speculate_insn):
38329         New function.  Declare prototype.
38330         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38332 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38334         PR rtl-optimization/68664
38335         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38336         New function.
38337         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38339 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38341         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38342         max skip bytes for function, loop and jump.
38344 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38346         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38347         ABS_EXPR for gimple dump.
38349 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38351         PR target/79462
38352         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38354         PR tree-optimization/79408
38355         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38356         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38357         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38358         also if rhs1 is INTEGER_CST.
38360 2017-02-14  Richard Biener  <rguenther@suse.de>
38362         PR middle-end/79432
38363         * tree-into-ssa.c (insert_phi_nodes): When the function can
38364         have abnormal edges rewrite SSA names with broken use-def
38365         dominance out of SSA and register them for PHI insertion.
38367 2017-02-13  Martin Sebor  <msebor@redhat.com>
38369         PR middle-end/79496
38370         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38371         clearing info.nowrite flag when snprintf size argument is a range.
38373 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38375         * cprop.c (cprop_jump): Add missing space in string literal.
38376         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38377         (get_constraint_for_component_ref): Likewise.
38378         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38379         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38380         * lra-constraints.c (process_alt_operands): Likewise.
38381         * ipa-inline.c (inline_small_functions): Likewise.
38382         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38383         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38384         * trans-mem.c (diagnose_tm_1_op): Likewise.
38385         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38386         (grid_parallel_clauses_gridifiable): Likewise.
38388         * config/nvptx/mkoffload.c (process): Add space in between
38389         , and %d.
38391         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38392         "MOD4_SSE_REGS" and "ALL_REGS".
38394         * spellcheck.c (test_data): Add , in between "foo" and "food".
38396 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38398         PR target/79449
38399         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38400         boundary crossing check and subsequent code generation agree.
38402 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38404         * config/aarch64/aarch64.c (has_memory_op): Delete.
38405         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38406         has_memory_op.
38408 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38410         PR rtl-optimization/79388
38411         PR rtl-optimization/79450
38412         * combine.c (distribute_notes): When removing TEM_INSN for which
38413         corresponding dest has last value recorded, invalidate that last
38414         value.
38416 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38418         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38419         of explicit '@'.  Add missing assembly comment marker on branch costs
38420         printout.
38422 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38424         * gengtype-lex.l (<in_struct>): Add '/'.
38426 2017-02-13  Martin Liska  <mliska@suse.cz>
38428         PR c/79471
38429         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38431 2017-02-13  Richard Biener  <rguenther@suse.de>
38433         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38434         Remove.
38435         * configure: Re-generate.
38436         * config.in: Likewise.
38437         * graphite-dependences.c: Simplify as if
38438         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38439         * graphite-isl-ast-to-gimple.c: Likewise.
38440         * graphite-optimize-isl.c: Likewise.
38441         * graphite-poly.c: Likewise.
38442         * graphite-sese-to-poly.c: Likewise.
38443         * graphite.h: Likewise.
38444         * toplev.c: Include isl/version.h and use isl_version () for
38445         printing the ISL version.
38446         * doc/install.texi: Update ISL requirement.
38448 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38450         * doc/standards.texi (Standards): Update reference to
38451         Objective-C 2.0.
38453 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38455         * doc/extend.texi (Named Address Spaces): sourceware.org now
38456         defaults to https.
38457         * doc/install.texi (Binaries): Ditto.
38458         (Specific): Ditto.
38460 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38462         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38463         standard terminology "stringize"/"stringizing" throughout.
38464         * doc/cppinternals.texi: Likewise.
38466 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38468         * doc/extend.texi: Fix some spelling mistakes and typos.
38469         * doc/invoke.texi: Likewise.
38471 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38473         PR ipa/79224
38474         * params.def (inline-min-speedup) Change from 10 to 8.
38476 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38478         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38479         4.5.
38481 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38483         PR ipa/79224
38484         * ipa-inline-analysis.c (get_minimal_bb): New function.
38485         (record_modified): Use it.
38486         (remap_edge_change_prob): Handle also ancestor functions.
38488 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38490         * doc/contrib.texi (Contributors): Remove broken link into
38491         the Mauve CVS repository.
38493 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38495         PR middle-end/79454
38496         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38497         result computation whenever lhs doesn't have vector mode, not
38498         just when it has BLKmode.
38500 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38502         * doc/makefile.texi (profiledbootstrap): Refer to the
38503         installation instructions only in textual form.
38505 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38507         PR target/79295
38508         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38510 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38512         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38513         (Specific): Update mingw-w64 reference.
38514         (Binaries): Ditto.
38515         (Specific): Remove broken link to Renesas RX processor.
38517 2017-02-10  Richard Biener  <rguenther@suse.de>
38519         * toplev.c (process_options): Do not mention obsolete graphite
38520         options when printing sorry message about missing graphite support.
38521         Mention -floop-nest-optimize.
38523 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38525         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38526         (vtst_p16): Likewise.
38527         (vtstq_p8): Likewise.
38528         (vtstq_p16): Likewise.
38529         (vtst_p64): New.
38530         (vtstq_p64): Likewise.
38531         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38532         (vset_lane_p64): New.
38533         (vsetq_lane_p64): New.
38535 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38537         PR tree-optimization/79411
38538         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38539         stmt operands are SSA_NAMEs used in abnormal phis.
38540         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38541         phis.
38543 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38545         PR ipa/70795
38546         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38547         flag if needed.
38549 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38551         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38553 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38555         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38556         to avoid warning.
38558         PR c/79413
38559         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38560         not arbitrary TREE_CONSTANT.
38562         PR c/79431
38563         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38564         "omp declare target link" attribute unless is_global_var.
38565         * omp-offload.c (find_link_var_op): Likewise.
38567 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38568             Chung-Lin Tang  <cltang@codesourcery.com>
38570         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38571         OMP_CLAUSE_TILE.
38572         (gimplify_adjust_omp_clauses): Don't delete TILE.
38573         (gimplify_omp_for): Deal with TILE.
38574         * internal-fn.c (expand_GOACC_TILE): New function.
38575         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38576         (GOACC_TILE): New.
38577         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38578         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38579         element fields.
38580         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38581         avoid DIV for outermost collapse var.
38582         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38583         Remove out of date comments, fix whitespace.
38584         * omp-general.c (omp_extract_for_data): Deal with tiling.
38585         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38586         adjust OLF_DIM_BASE value.
38587         (struct omp_for_data): Add tiling field.
38588         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38589         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38590         for auto loops.  Remove default auto determining, moved to
38591         oacc_loop_fixed_partitions.
38592         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38593         stmts, add e_mask field.
38594         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38595         (oacc_thread_numbers): Use oacc_dim_call.
38596         (oacc_xform_tile): New.
38597         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38598         (finish_oacc_loop): Adjust for ifns vector.
38599         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38600         add case for GOACC_TILE fns.
38601         (oacc_loop_xform_loop): Delete.
38602         (oacc_loop_process): Iterate over call list directly, and add
38603         handling for GOACC_TILE fns.
38604         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38605         dump partitioning.
38606         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38607         vector partitioning to outer loops.  Assign 2 partitions to loops
38608         when available. Add TILE handling.
38609         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38610         (execite_oacc_device_lower): Process GOACC_TILE fns,
38611         ignore unknown specs.
38612         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38613         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38614         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38616 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38618         * configure.ac (ACX_BUGURL): Update.
38619         * configure: Regenerate.
38621 2017-02-09  Richard Biener  <rguenther@suse.de>
38623         PR tree-optimization/69823
38624         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38625         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38627 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38629         * config/arc/arc-c.def: Add __NPS400__ definition.
38630         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38631         (TARGET_NPS400): Define.
38633 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38635         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38636         file.
38637         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38638         pointer, arch_info.
38639         (arc_cpu_types): Fill the arch_info field with a pointer into the
38640         arc_arch_types table.
38641         (arc_selected_cpu): Declare.
38642         * config/arc/arc.c (arc_selected_cpu): Make global.
38643         (arc_selected_arch): Delete.
38644         (arc_base_cpu): Delete.
38645         (arc_override_options): Remove references to deleted variables,
38646         update access to arch information.
38647         (ARC_OPT): Update access to arch information.
38648         (ARC_OPTX): Likewise.
38649         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38650         (TARGET_ARC600): Update access to arch information.
38651         (TARGET_ARC601): Likewise.
38652         (TARGET_ARC700): Likewise.
38653         (TARGET_EM): Likewise.
38654         (TARGET_HS): Likewise.
38655         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38656         information.
38658 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38660         PR target/78604
38661         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38662         condition/operands for integer GE/LE/GEU/LEU operations.
38664 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38666         PR translation/79397
38667         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38668         of AltiVec.
38670 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38672         PR ipa/79375
38673         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38674         whether allocation happened.
38675         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38676         nothing was allocated.
38678 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38680         PR tree-optimization/79408
38681         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38682         constant, but SSA_NAME with a known integer range, use the minimum
38683         of that range instead of op1 to determine if modulo can be replaced
38684         with its first operand.
38686 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38688         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38690 2017-02-08  Richard Biener  <rguenther@suse.de>
38692         PR tree-optimization/71824
38693         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38694         Check all loops contained in the merged region.
38696 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38698         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38700 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38702         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38703         (thunderxt88): Likewise.
38704         (thunderxt81): Disable LSE and change v8.1 to v8.
38705         (thunderxt83): Likewise.
38707 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38708             Richard Biener  <rguenther@suse.de>
38710         PR middle-end/79399
38711         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38712         type from int to size_t.
38713         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38715 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38717         PR rtl-optimization/79386
38718         * cprop.c (bypass_conditional_jumps): Initialize
38719         bypass_last_basic_block already before splitting bbs after
38720         unconditional traps...
38721         (bypass_conditional_jumps): ... rather than here.
38723         PR target/79299
38724         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38725         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38726         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38727         fix -masm=intel patterns.
38729 2017-02-07  Richard Biener  <rguenther@suse.de>
38731         PR tree-optimization/79256
38732         PR middle-end/79278
38733         * builtins.c (get_object_alignment_2): Use min_align_of_type
38734         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38735         and ADJUST_FIELD_ALIGN.
38737         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38738         type parameter.
38739         * doc/tm.texi: Regenerate.
38740         * stor-layout.c (layout_decl): Adjust.
38741         (update_alignment_for_field): Likewise.
38742         (place_field): Likewise.
38743         (min_align_of_type): Likewise.
38744         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38745         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38746         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38747         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38748         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38749         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38750         * config/i386/i386.c (x86_field_alignment): Likewise.
38751         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38752         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38753         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38754         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38755         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38756         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38757          Likewise.
38759         Revert
38760         2017-01-30  Richard Biener  <rguenther@suse.de>
38762         PR tree-optimization/79256
38763         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38764         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38765         alignment on TYPE.
38767 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38769         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38770         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38771         builtins to SImode and emit a zero-extend, if necessary.
38773 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38775         * docs/invoke.texi (RISC-V Options): Alphabetize.
38777 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38779         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38780         options.
38782 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38784         * config/riscv/riscv.c: New file.
38785         * common/config/riscv/riscv-common.c: Likewise.
38786         * config.gcc: Likewise.
38787         * config/riscv/constraints.md: Likewise.
38788         * config/riscv/elf.h: Likewise.
38789         * config/riscv/generic.md: Likewise.
38790         * config/riscv/linux.h: Likewise.
38791         * config/riscv/multilib-generator: Likewise.
38792         * config/riscv/peephole.md: Likewise.
38793         * config/riscv/pic.md: Likewise.
38794         * config/riscv/predicates.md: Likewise.
38795         * config/riscv/riscv-builtins.c: Likewise.
38796         * config/riscv/riscv-c.c: Likewise.
38797         * config/riscv/riscv-ftypes.def: Likewise.
38798         * config/riscv/riscv-modes.def: Likewise.
38799         * config/riscv/riscv-opts.h: Likewise.
38800         * config/riscv/riscv-protos.h: Likewise.
38801         * config/riscv/riscv.h: Likewise.
38802         * config/riscv/riscv.md: Likewise.
38803         * config/riscv/riscv.opt: Likewise.
38804         * config/riscv/sync.md: Likewise.
38805         * config/riscv/t-elf-multilib: Likewise.
38806         * config/riscv/t-linux: Likewise.
38807         * config/riscv/t-linux-multilib: Likewise.
38808         * config/riscv/t-riscv: Likewise.
38809         * configure.ac: Likewise.
38810         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38811         Waterman as RISC-V maintainers.
38812         * doc/install.texi: Add RISC-V entries.
38813         * doc/invoke.texi: Add RISC-V options section.
38814         * doc/md.texi: Add RISC-V constraints section.
38815         * configure: Regenerated.
38817 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38819         PR target/66144
38820         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38821         false values to be constant vectors with all 0 or all 1 bits set.
38822         (vcondu<mode><mode>): Likewise.
38823         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38824         predicate.
38825         (fpmask_comparison_operator): Update comment.
38826         (vecint_comparison_operator): New predicate.
38827         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38828         vector conditionals when the true and false values are constant
38829         vectors with all 0 bits or all 1 bits set.
38831 2017-02-06  Martin Sebor  <msebor@redhat.com>
38833         PR  tree-optimization/79376
38834         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38836 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38838         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38839         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38840         to simplify split condition.
38842 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38844         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38845         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38846         false.
38848 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38850         PR rtl-optimization/68664
38851         * target.def (can_speculate_insn): New hook.
38852         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38853         * doc/tm.texi: Regenerate.
38854         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38855         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38856         (rs6000_sched_can_speculate_insn): New function.
38858 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38860         PR tree-optimization/79284
38861         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38862         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38863         vectorizable_mask_load_store, vectorizable_operation,
38864         vect_is_simple_cond, get_same_sized_vectype): Use it instead
38865         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
38866         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
38867         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
38868         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
38869         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
38870         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
38871         is_gimple_assign (stmt).  Replace another such test with
38872         is_gimple_assign (stmt).
38874 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38876         PR target/78883
38877         * config/avr/avr.c (rtl-iter.h): Include it.
38878         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
38879         (avr_legitimate_combined_insn): ...and implementation.
38881 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38883         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
38884         * config/s390/s390.c (s390_const_operand_ok)
38885         (s390_canonicalize_comparison, s390_extract_part)
38886         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
38887         (s390_contiguous_bitmask_p, s390_rtx_costs)
38888         (legitimize_pic_address): Likewise.
38889         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
38890         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
38891         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
38892         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
38893         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
38895 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38897         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
38898         REGNO($0) == REGNO($1).
38900 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38902         * config/s390/linux.h(SIZE_TYPE): Add comment.
38904 2017-02-06  Julian Brown  <julian@codesourcery.com>
38905             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
38906             Virendra Pathak  <virendra.pathak@broadcom.com>
38908         * config/aarch64/aarch64-cores.def: Change the scheduler
38909         to Thunderx2t99.
38910         * config/aarch64/aarch64.md: Include thunderx2t99.md.
38911         * config/aarch64/thunderx2t99.md: New file.
38913 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
38915         * doc/standards.texi (Go Language): Update link to language
38916         standard.
38918 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
38920         * tree-eh.c (lower_resx): Sanitize profile.
38921         (cleanup_empty_eh_move_lp): Likewise.
38923 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
38925         PR tree-ssa/79347
38926         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
38927         ELSE_PROB.
38928         * cfgloopmanip.h (loop_version): Update prototype.
38929         * modulo-sched.c (sms_schedule): Update call of loop_version.
38930         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
38931         * tree-parloops.c (gen_parallel_loop): Likewise.
38932         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
38933         * tree-ssa-loop-split.c (split_loop): Likewise.
38934         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
38935         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
38937 2017-02-05  Martin Liska  <mliska@suse.cz>
38939         PR bootstrap/78985
38940         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
38941         variable to NULL.
38942         (print_operand_address): Initialize a struct to zero.
38944 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
38946         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
38947         garbage collector only in textual form.
38949 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
38951         * doc/extend.texi (x86 specific memory model extensions for
38952         transactional memory): Simplify a phrase.
38954 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
38956         PR target/79353
38957         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
38958         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
38959         (atomic_storedi_1): Likewise.
38961 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
38963         PR tree-optimization/79338
38964         * tree-parloops.c (gather_scalar_reductions): Don't call
38965         vect_analyze_loop_form for loop->inner before destroying loop's
38966         loop_vinfo.
38968 2017-02-03  Martin Sebor  <msebor@redhat.com>
38970         PR tree-optimization/79327
38971         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
38972         when precision has resulted in leading zeros.
38973         (format_integer): Adjust the likely counter to assume an unknown
38974         argument that may be zero is non-zero.
38976 2017-02-03  Jason Merrill  <jason@redhat.com>
38978         PR c++/78689
38979         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
38980         avoid copying non-taken branch.
38982 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
38984         PR tree-optimization/79340
38985         * tree-vect-loop.c (vectorizable_reduction): Release
38986         vec_defs elements after safe_splicing them into other vectors.
38987         Formatting fixes.
38989         PR tree-optimization/79327
38990         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
38991         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
38992         dirtype.
38993         (format_integer): Use wide_int_to_tree instead of build_int_cst
38994         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
38995         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
38996         of shortest and longest sequence.
38998 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39000         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39001         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39003 2017-02-03  Walter Lee  <walt@tilera.com>
39005         PR target/78862
39006         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39007         after initial stackframe link reg save.
39008         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39010 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39012         PR target/79354
39013         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39014         wu for stxssp alternative.
39016 2017-02-03  Martin Sebor  <msebor@redhat.com>
39018         PR tree-optimization/79352
39019         * gimple-fold.c (get_range_strlen): Add argument.
39020         (get_range_strlen): Change return type to bool.
39021         (get_maxval_strlen): Pass in a dummy argument.
39022         * gimple-fold.h (get_range_strlen): Change return type to bool.
39023         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39024         * tree.h (array_at_struct_end_p): Add argument.
39025         * tree.c (array_at_struct_end_p): Handle it.
39027 2017-02-03  Martin Liska  <mliska@suse.cz>
39029         PR lto/66295
39030         * multiple_target.c (create_dispatcher_calls): Redirect edge
39031         from a caller of a dispatcher.
39032         (expand_target_clones): Make the clones local.
39033         (ipa_target_clone): Do both target clones and resolvers.
39034         (ipa_dispatcher_calls): Remove the pass.
39035         (pass_dispatcher_calls::gate): Likewise.
39036         (make_pass_dispatcher_calls): Likewise.
39037         * passes.def (pass_target_clone): Put as very first IPA early
39038         pass.
39040 2017-02-03  Martin Liska  <mliska@suse.cz>
39042         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39043         in case of a function with ifunc attribute.
39045 2017-02-03  Martin Liska  <mliska@suse.cz>
39047         * cgraph.c (cgraph_node::dump): Dump function version info.
39048         * symtab.c (symtab_node::dump_base): Add missing new line.
39050 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39052         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39053         (ifcombine_ifandif): Use it.
39055 2017-02-03  Martin Liska  <mliska@suse.cz>
39057         * doc/invoke.texi: Document default value for
39058         use-after-scope-direct-emission-threshold.
39060 2017-02-03  Martin Liska  <mliska@suse.cz>
39062         PR tree-optimization/79339
39063         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39064         (format_floating): Likewise.
39066 2017-02-03  Martin Liska  <mliska@suse.cz>
39068         PR ipa/79337
39069         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39070         implementation.
39071         (ipa_node_params_t::remove): Likewise.
39072         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39073         initialization from removed ipa_node_params_t::insert.
39074         (ipa_node_params::~ipa_node_params): Move from removed
39075         ipa_node_params_t::release.
39076         * symbol-summary.h (symbol_summary::m_released): New member.
39077         Do not release a summary twice.  Do not allow to call finalizer
39078         for types of a summary that live in GGC memory.
39080 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39082         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39083         cmp_branch fusion.
39085 2017-02-02  Martin Sebor  <msebor@redhat.com>
39087         PR middle-end/79275
39088         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39089         (format_string): Tighten up the range of output for non-constant
39090         strings and correct the expected range for wide non-constant strings.
39092 2017-02-02  Martin Sebor  <msebor@redhat.com>
39094         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39096         PR middle-end/32003
39097         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39098         index.
39099         (-fdump-tree-@var): Add to index and document how to come up
39100         with pass-specific option and dump file names.
39101         (-fdump-passes): Clarify where to look for output.
39103 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39105         PR middle-end/77445
39106         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39107         statistics of the analyzed path; allow threading for speed when
39108         any of BBs along the path are optimized for speed.
39110 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39112         PR middle-end/78468
39113         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39114         settings of the virtual registers.
39116         Revert again
39117         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39119         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39120         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39121         needed.
39123 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39125         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39126         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39128 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39130         * config/s390/s390.md: Add missing comments with the expanded
39131         mnemonics.
39132         * config/s390/vector.md: Likewise.
39133         * config/s390/vx-builtins.md: Likewise.
39135 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39137         PR target/79197
39138         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39139         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39140         conditions on a single line.
39142 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39144         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39145         __S390_VX__ to __VX__.
39147 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39149         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39150         stmt_info to record_stmt_cost.
39151         (vect_get_known_peeling_cost): Pass stmt_info if known to
39152         record_stmt_cost.
39153         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39154         cpu_vector_cost field into
39155         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39156         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39157         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39158         splitting of scalar_stmt_cost and vec_stmt_cost.
39159         (thunderx_vector_cost): Likewise.
39160         (cortexa57_vector_cost): LIkewise.
39161         (exynosm1_vector_cost): Likewise.
39162         (xgene1_vector_cost): Likewise.
39163         (thunderx2t99_vector_cost): Improve after the splitting of the two
39164         fields.
39165         (aarch64_builtin_vectorization_cost): Update for the splitting of
39166         scalar_stmt_cost and vec_stmt_cost.
39168 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39169             Richard Henderson  <rth@redhat.com>
39171         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39172         conditional on existance of a fast atomic load.
39173         * optabs-query.c (can_atomic_load_p): New function.
39174         * optabs-query.h (can_atomic_load_p): Declare it.
39175         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39176         no fast atomic load is available for the particular size of access.
39177         (expand_atomic_compare_and_swap): Likewise.
39178         (expand_atomic_load): Likewise.
39179         (expand_atomic_store): Likewise.
39180         (expand_atomic_fetch_op): Likewise.
39181         * testsuite/lib/target-supports.exp
39182         (check_effective_target_sync_int_128): Remove x86 because it provides
39183         no fast atomic load.
39184         (check_effective_target_sync_int_128_runtime): Likewise.
39186 2017-02-01  Richard Biener  <rguenther@suse.de>
39188         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39189         (graphite_transform_loops): Provide opt-info for optimized nests.
39190         * tree-parloop.c (parallelize_loops): Provide opt-info for
39191         parallelized loops.
39193 2017-02-01  Richard Biener  <rguenther@suse.de>
39195         PR middle-end/79315
39196         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39197         was not set before.
39199 2017-02-01  Richard Biener  <rguenther@suse.de>
39201         PR tree-optimization/71824
39202         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39203         Verify the loops are valid in the merged SESE region.
39204         (scop_detection::can_represent_loop_1): Check analyzing the
39205         evolution of the number of iterations in the region succeeds.
39207 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39209         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39210         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39212 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39214         PR preprocessor/79210
39215         * input.c (get_substring_ranges_for_loc): Replace line_width
39216         assertion with error-handling.
39218 2017-01-31  Richard Biener  <rguenther@suse.de>
39220         PR tree-optimization/77318
39221         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39222         (create_pw_aff_from_tree): Take loop parameter.
39223         (add_condition_to_pbb): Pass loop of the condition to
39224         create_pw_aff_from_tree.
39226 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39228         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39229         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39231 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39233         PR target/78597
39234         PR target/79038
39235         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39236         no longer used.
39237         (convert_int_to_float128): Likewise.
39238         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39239         (convert_int_to_float128): Likewise.
39240         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39241         (UNSPEC_IEEE128_CONVERT): Likewise.
39242         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39243         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39244         Use local variables for IBM extended format.
39245         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39246         (fix_trunc<mode>si2_fprs): Likewise.
39247         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39248         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39249         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39250         to know that we can now have integers of all sizes in vector
39251         registers.
39252         (fix<uns>_<mode>di2_hw): Likewise.
39253         (float<uns>_<mode>si2_hw): Likewise.
39254         (fix_<mode>si2_hw): Likewise.
39255         (fixuns_<mode>si2_hw): Likewise.
39256         (float<uns>_<mode>di2_hw): Likewise.
39257         (float_<mode>di2_hw): Likewise.
39258         (float_<mode>si2_hw): Likewise.
39259         (floatuns_<mode>di2_hw): Likewise.
39260         (floatuns_<mode>si2_hw): Likewise.
39261         (xscvqp<su>wz_<mode>): Delete, no longer used.
39262         (xscvqp<su>dz_<mode>): Likewise.
39263         (xscv<su>dqp_<mode>): Likewise.
39264         (ieee128_mfvsrd_64bit): Likewise.
39265         (ieee128_mfvsrd_32bit): Likewise.
39266         (ieee128_mfvsrwz): Likewise.
39267         (ieee128_mtvsrw): Likewise.
39268         (ieee128_mtvsrd_64bit): Likewise.
39269         (ieee128_mtvsrd_32bit): Likewise.
39271 2017-01-31  Martin Liska  <mliska@suse.cz>
39273         PR ipa/79285
39274         * ipa-prop.c (ipa_free_all_node_params): Call release method
39275         instead of ~sumbol_summary to not to trigger double times
39276         dtor of hash_map.
39278 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39280         PR tree-optimization/71691
39281         * bitmap.h (class auto_bitmap): New.
39282         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39283         is_maybe_undefined instead of ssa_undefined_value_p.
39285 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39287         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39288         __S390_ARCH_LEVEL__ to __ARCH__.
39290 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39292         PR tree-optimization/79267
39293         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39294         if should_remove_lhs_p is true.
39296 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39298         PR debug/63238
39299         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39300         (add_alignment_attribute): New.
39301         (base_type_die): Add alignment attribute.
39302         (subrange_type_die): Likewise.
39303         (modified_type_die): Likewise.
39304         (gen_array_type_die): Likewise.
39305         (gen_descr_array_type_die: Likewise.
39306         (gen_enumeration_type_die): Likewise.
39307         (gen_subprogram_die): Likewise.
39308         (gen_variable_die): Likewise.
39309         (gen_field_die): Likewise.
39310         (gen_ptr_to_mbr_type_die): Likewise.
39311         (gen_struct_or_union_type_die): Likewise.
39312         (gen_subroutine_type_die): Likewise.
39313         (gen_typedef_die): Likewise.
39314         (base_type_cmp): Compare alignment attribute.
39316 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39318         PR target/79170
39319         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39320         (setb_unsigned) New pattern for setb with CCUNS.
39321         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39322         subfc./subfe sequence to avoid overflow problems.  Generate a
39323         shorter sequence with cmpld/setb for power9.
39324         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39325         for generating subfc. instruction.
39326         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39327         now uses this instruction.
39329 2017-01-30  Ian Lance Taylor  <iant@google.com>
39331         PR debug/79289
39332         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39333         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39335 2017-01-30  Martin Sebor  <msebor@redhat.com>
39337         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39338         Move constant to the right of a relational operator.
39339         (get_mpfr_format_length, format_character, format_string): Ditto.
39340         (should_warn_p, maybe_warn): Same.
39342         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39344 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39346         PR lto/79061
39347         * asan.c (get_translation_unit_decl): Remove function.
39348         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39350 2017-01-30  Martin Liska  <mliska@suse.cz>
39352         PR gcov-profile/79259
39353         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39354         -fprofile-generate.
39356 2017-01-30  Martin Liska  <mliska@suse.cz>
39358         PR bootstrap/78985
39359         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39360         Initialize variables with NULL value.
39362 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39364         PR target/79260
39365         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39366         tm_p_file.
39367         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39369 2017-01-30  Richard Biener  <rguenther@suse.de>
39371         PR tree-optimization/79276
39372         * tree-vrp.c (process_assert_insertions): Properly adjust common
39373         when removing a duplicate.
39375 2017-01-30  Richard Biener  <rguenther@suse.de>
39377         PR tree-optimization/79256
39378         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39379         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39380         alignment on TYPE.
39381         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39383 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39385         PR target/79240
39386         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39387         ("*r<noxa>sbg_<mode>_sll_bitmask")
39388         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39389         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39390         Use contiguous_bitmask_nowrap_operand.
39392 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39394         PR target/79268
39395         * config/rs6000/altivec.h (vec_xl): Revise #define.
39396         (vec_xst): Likewise.
39398 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39400         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39402 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39404         PR rtl-optimization/79194
39405         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39406         traps before call to bypass_conditional_jumps.
39408 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39410         PR tree-optimization/71374
39411         * lra-constraints.c (check_conflict_input_operands): New.
39412         (match_reload): Use it.
39414 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39416         PR target/79131
39417         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39418         account to calculate conflict_set.
39420 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39422         PR rtl-optimization/78559
39423         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39424         other_insn in combine.
39426 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39428         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39429         uint16_type_node for BT_UINT16.
39431 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39433         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39434         "RTL Tests" to menu.
39435         (GIMPLE Tests): New node.
39436         (RTL Tests): New node.
39438 2017-01-27  Richard Biener  <rguenther@suse.de>
39440         PR tree-optimization/79245
39441         * tree-loop-distribution.c (distribute_loop): Apply cost
39442         modeling also to detected patterns.
39444 2017-01-27  Richard Biener  <rguenther@suse.de>
39446         PR tree-optimization/71433
39447         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39448         (compare_assert_loc): New function.
39449         (process_assert_insertions): Sort and optimize assert locations
39450         to remove duplicates and push down identical assertions on
39451         edges to their destination block.
39453 2017-01-27  Richard Biener  <rguenther@suse.de>
39455         PR tree-optimization/79244
39456         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39457         out SSA names even if abnormal.
39459 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39461         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39462         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39463         instead of MPFR_RNDN.
39465 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39467         PR target/79239
39468         * arm.c (arm_option_override): Don't call build_target_option_node
39469         until after doing all option overrides.
39470         (arm_valid_target_attribute_tree): Likewise.
39472 2017-01-27  Martin Liska  <mliska@suse.cz>
39474         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39475         for {cd}tors and C++ {cd}tors.
39477 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39479         * config/s390/s390.md ("*setmem_long_and")
39480         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39482 2017-01-26  Martin Sebor  <msebor@redhat.com>
39484         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39485         of precision.
39487 2017-01-26  Martin Sebor  <msebor@redhat.com>
39489         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39490         HAVE_DFmode before using XFmode or DFmode.
39491         (parse_directive): Avoid using the z length modifier to avoid
39492         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39494         PR middle-end/78703
39495         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39496         to accept adjustment as an array.
39497         (get_int_range): New function.
39498         (struct directive): Make width and prec arrays.
39499         (directive::set_width, directive::set_precision): Call get_int_range.
39500         (format_integer, format_floating): Handle width and precision ranges.
39501         (format_string, parse_directive): Same.
39503 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39505         PR debug/79129
39506         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39507         comdat_type_p set, just clone them, but keep the children in the
39508         original DIE.
39510         PR debug/78835
39511         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39512         which have direct callers with -fvar-tracking-assignments enabled
39513         in the current TU.
39514         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39515         inside of type units.
39517 2017-01-26  Martin Sebor  <msebor@redhat.com>
39519         PR middle-end/78703
39520         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39521         unlikely counters.
39522         (struct format_result): Replace number_chars, number_chars_min,
39523         and number_chars_max with a single member of struct result_range.
39524         Remove bounded.
39525         (format_result::operator+=): Adjust.
39526         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39527         counters.
39528         (fmtresult::adjust_for_width_or_precision): New function.
39529         (fmtresult:type_max_digits): New function.
39530         (bytes_remaining): Handle likely and unlikely counters.
39531         (min_bytes_remaining): Remove.
39532         (format_percent): Simplify.
39533         (format_integer, format_floating): Set likely and unlikely counters.
39534         (get_string_length, format_character, format_string): Same.
39535         (format_plain, should_warn_p): New function.
39536         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39537         and handle those for all directives, including plain strings.
39538         (format_directive): Handle likely and unlikely counters.
39539         Remove unnecessary quoting from diagnostics.  Add an informational
39540         note.
39541         (add_bytes): Remove.
39542         (pass_sprintf_length::compute_format_length): Simplify.
39543         (try_substitute_return_value): Handle likely and unlikely counters.
39545 2017-01-26  Carl Love  <cel@us.ibm.com>
39547         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39548         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39550 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39552         PR target/79131
39553         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39554         endianess for subregs into account.
39555         * lra-constraints.c (lra_constraints): Do risky transformations
39556         always on the first iteration.
39557         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39558         last_call_used_reg_set.
39559         (process_bb_lives): Define and use last_call_used_reg_set.
39560         * lra.c (lra): Always continue after lra_constraints on the first
39561         iteration.
39563 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39565         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39566         constant.
39567         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39569 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39571         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39572         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39573         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39574         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39575         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39576         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39577         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39578         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39579         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39581 2017-01-26  Marek Polacek  <polacek@redhat.com>
39583         PR c/79199
39584         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39585         for the third operand.
39587 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39589         PR middle-end/79236
39590         * omp-low.c (struct omp_context): Add simt_stmt field.
39591         (scan_omp_for): Return omp_context *.
39592         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39593         context to the _simt_ SIMD stmt.
39594         (lower_omp_for): For combined SIMD with sibling _simt_
39595         SIMD, make sure to use the same decls in _looptemp_
39596         clauses as in the sibling.
39598 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39600         PR middle-end/79212
39601         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39602         all contexts.
39604 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39606         PR target/70465
39607         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39608         emit fld b; fld a; if possible.
39610         * brig-builtins.def: Update copyright years.
39611         * config/arm/arm_acle_builtins.def: Update copyright years.
39613 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39615         PR target/79179
39616         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39617         constraint instead of o for the stxsd instruction.
39619 2017-01-25  Carl Love  <cel@us.ibm.com>
39621         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39622         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39624 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39626         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39628 2017-01-25  Richard Biener  <rguenther@suse.de>
39630         PR tree-optimization/69264
39631         * target.def (vector_alignment_reachable): Improve documentation.
39632         * doc/tm.texi: Regenerate.
39633         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39634         and add a comment.
39635         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39636         earlier changes with respect to TYPE_USER_ALIGN.
39637         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39639 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39641         PR target/79145
39642         * config/arm/arm.md (xordi3): Force constant operand into a register
39643         for TARGET_IWMMXT.
39645 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39647         * doc/invoke.texi (-fstore-merging): Correct default optimization
39648         levels at which it is enabled.
39649         (-O): Move -fstore-merging from list to...
39650         (-O2): ... Here.
39652 2017-01-25  Richard Biener  <rguenther@suse.de>
39654         PR debug/78363
39655         * omp-expand.c: Include debug.h.
39656         (expand_omp_taskreg): Make sure to generate early debug before
39657         outlining anything from a function.
39658         (expand_omp_target): Likewise.
39659         (grid_expand_target_grid_body): Likewise.
39661 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39663         PR lto/79061
39664         * asan.c (get_translation_unit_decl): New function.
39665         (asan_add_global): Extract modules file name from globals
39666         TRANSLATION_UNIT_DECL name.
39668 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39670         PR target/77439
39671         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39672         for long calls with APCS frame and VFP.
39674 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39676         * cfg.c (original_copy_tables_initialized_p): New function.
39677         * cfg.h (original_copy_tables_initialized_p): New decl.
39678         * cfgrtl.c (relink_block_chain): Guard the call to
39679         free_original_copy_tables with a call to
39680         original_copy_tables_initialized_p.
39681         * cgraph.h (symtab_node::native_rtl_p): New decl.
39682         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39683         (symtab_node::needed_p): Don't assert for early assembly output
39684         for __RTL functions.
39685         (cgraph_node::finalize_function): Set "force_output" for __RTL
39686         functions.
39687         (cgraph_node::analyze): Bail out early for __RTL functions.
39688         (analyze_functions): Update assertion to support __RTL functions.
39689         (cgraph_node::expand): Bail out early for __RTL functions.
39690         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39691         __RTL functions.
39692         * function.h (struct function): Update comment for field
39693         "pass_startwith".
39694         * gimple-expr.c: Include "tree-pass.h".
39695         (gimple_has_body_p): Return false for __RTL functions.
39696         * Makefile.in (OBJS): Add run-rtl-passes.o.
39697         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39698         accessor.
39699         (gcc::pass_manager::get_clean_slate): New accessor.
39700         * passes.c: Include "insn-addr.h".
39701         (should_skip_pass_p): Add logging.  Update logic for running
39702         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39703         property-provider override so it is only done for gimple passes.
39704         Don't skip dfinit.
39705         (skip_pass): New function.
39706         (execute_one_pass): Call skip_pass when skipping passes.
39707         * read-md.c (md_reader::read_char): Support filtering
39708         the input to a subset of line numbers.
39709         (md_reader::md_reader): Initialize fields
39710         m_first_line and m_last_line.
39711         (md_reader::read_file_fragment): New function.
39712         * read-md.h (md_reader::read_file_fragment): New decl.
39713         (md_reader::m_first_line): New field.
39714         (md_reader::m_last_line): New field.
39715         * read-rtl-function.c (function_reader::create_function): Only
39716         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39717         curr_properties.  Set DECL_INITIAL to a dummy block.
39718         (read_rtl_function_body_from_file_range): New function.
39719         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39720         New decl.
39721         * run-rtl-passes.c: New file.
39722         * run-rtl-passes.h: New file.
39724 2017-01-24  Jeff Law  <law@redhat.com>
39726         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39727         buffer size.
39729 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39731         PR tree-optimization/79159
39732         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39733         (record_nonwrapping_iv): Improve boundary using above function if no
39734         value range information.
39736 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39737             Martin Jambor  <mjambor@suse.cz>
39739         * brig-builtins.def: New file.
39740         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39741         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39742         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39743         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39744         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39745         * builtin-types.def (BT_INT8): New.
39746         (BT_INT16): Likewise.
39747         (BT_UINT8): Likewise.
39748         (BT_UINT16): Likewise.
39749         (BT_FN_ULONG): Likewise.
39750         (BT_FN_UINT_INT): Likewise.
39751         (BT_FN_UINT_ULONG): Likewise.
39752         (BT_FN_UINT_LONG): Likewise.
39753         (BT_FN_UINT_PTR): Likewise.
39754         (BT_FN_ULONG_PTR): Likewise.
39755         (BT_FN_INT8_FLOAT): Likewise.
39756         (BT_FN_INT16_FLOAT): Likewise.
39757         (BT_FN_UINT32_FLOAT): Likewise.
39758         (BT_FN_UINT16_FLOAT): Likewise.
39759         (BT_FN_UINT8_FLOAT): Likewise.
39760         (BT_FN_UINT64_FLOAT): Likewise.
39761         (BT_FN_UINT16_UINT32): Likewise.
39762         (BT_FN_UINT32_UINT16): Likewise.
39763         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39764         (BT_FN_INT_PTR_INT): Likewise.
39765         (BT_FN_UINT_PTR_UINT): Likewise.
39766         (BT_FN_LONG_PTR_LONG): Likewise.
39767         (BT_FN_ULONG_PTR_ULONG): Likewise.
39768         (BT_FN_VOID_UINT64_UINT64): Likewise.
39769         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39770         (BT_FN_INT8_INT8_INT8): Likewise.
39771         (BT_FN_INT16_INT16_INT16): Likewise.
39772         (BT_FN_INT_INT_INT): Likewise.
39773         (BT_FN_UINT_FLOAT_UINT): Likewise.
39774         (BT_FN_FLOAT_UINT_UINT): Likewise.
39775         (BT_FN_ULONG_UINT_UINT): Likewise.
39776         (BT_FN_ULONG_UINT_PTR): Likewise.
39777         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39778         (BT_FN_UINT_UINT_UINT): Likewise.
39779         (BT_FN_VOID_UINT_PTR): Likewise.
39780         (BT_FN_UINT_UINT_PTR: Likewise.
39781         (BT_FN_UINT32_UINT64_PTR): Likewise.
39782         (BT_FN_INT_INT_UINT_UINT): Likewise.
39783         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39784         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39785         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39786         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39787         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39788         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39789         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39790         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39791         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39792         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39793         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39794         * doc/frontends.texi: List BRIG FE.
39795         * doc/install.texi (Testing): Add BRIG tesring requirements.
39796         * doc/invoke.texi (Overall Options): Mention BRIG.
39797         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39799 2017-01-24  Richard Biener  <rguenther@suse.de>
39801         PR translation/79208
39802         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39804 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39806         PR bootstrap/79198
39807         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39808         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39809         and known_contexts.
39811 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39813         PR middle-end/79123
39814         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39815         casts from signed to unsigned really don't have a range.
39817 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39819         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39820         GMP_RNDx for compatiblity.
39822 2017-01-24  Martin Liska  <mliska@suse.cz>
39824         PR bootstrap/79132
39825         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39826         that would prevent us to call alloca with -1 as argument.
39828 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39830         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39831         Avoid -Wformat-security warning.
39833 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39835         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39836         cost table.
39838 2017-01-23  Martin Sebor  <msebor@redhat.com>
39840         PR middle-end/78703
39841         * gimple-ssa-sprintf.c (warn_level): New global.
39842         (format_integer): Use it here and throughout the rest of the file.
39843         Use the same switch to compute sign as base.
39844         (maybe_warn): New function.
39845         (format_directive): Factor out warnings into maybe_warn.
39846         Add debugging output.  Use warn_level.
39847         (add_bytes): Use warn_level.
39848         (pass_sprintf_length::compute_format_length): Add debugging output.
39849         (try_substitute_return_value): Same.
39850         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39852         PR middle-end/78703
39853         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39854         (struct fmtresult, format_integer, format_floating): Adjust.
39855         (fmtresult::fmtresult): Set max correctly in two argument ctor.
39856         (get_string_length, format_string,format_directive): Same.
39857         (pass_sprintf_length::compute_format_length): Same.
39858         (try_substitute_return_value): Simplify slightly.
39860         PR middle-end/78703
39861         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39862         (fmtresult::operator+=): Outlined.
39863         (struct fmtresult): Add ctors.
39864         (struct conversion_spec): Rename...
39865         (struct directive): ...to this.  Add and remove data members.
39866         (directive::set_width, directive::set_precision): New functions.
39867         (format_percent): Use fmtresult ctor.
39868         (get_width_and_precision): Remove.
39869         (format_integer): Make naming changes.  Avoid computing width and
39870         precision.
39871         (format_floating): Same.  Adjust indentation.
39872         (format_character, format_none): New functions.
39873         (format_string): Moved character handling to format_character.
39874         (format_directive): Remove arguments, change return type.
39875         (parse_directive): New function.
39876         (pass_sprintf_length::compute_format_length): Move directive
39877         parsing to parse_directive.
39879 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
39881         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
39882         (assign_assembler_name_if_needed): ... this.
39883         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
39884         (assign_assembler_name_if_needed): ... this.
39885         (free_lang_data_in_cgraph): Adjust callers.
39886         * cgraphunit.c (cgraph_node::analyze): Likewise.
39887         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
39888         Likewise.
39890 2017-01-23  Richard Biener  <rguenther@suse.de>
39892         PR tree-optimization/79088
39893         PR tree-optimization/79188
39894         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
39895         resetting loop bounds after last path deletion.  Reset loop
39896         bounds of the target loop, make code match the comments.
39897         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
39898         Make sure loops need no fixups.
39900 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
39902         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
39903         exponent support with double type for first argument.
39904         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
39905         type returned by __builtin_vec_extract_sig,
39906         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
39907         functions from "vector int" to "vector unsigned int" or from
39908         "vector long long int" to "vector unsigned long long int".
39909         Changed type returned by __builtin_vec_extract_exp,
39910         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
39911         functions from "vector int" to "vector unsigned int" or from
39912         "vector long long int" to "vector unsigned long long int".
39913         Changed return type of __builtin_vec_test_data_class,
39914         __builtin_vec_test_data_class_sp, and
39915         __builtin_vec_test_data_class_dp from "vector int" to
39916         "vector bool int" or from "vector long long int" to "vector bool
39917         long long int" and changed second argument type from "unsigned
39918         int" to "int".  Added new overloaded function forms "vector float
39919         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
39920         "vector float __builtin_vec_insert_exp_sp (vector float, vector
39921         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
39922         double, vector unsigned long long int)" and "vector double
39923         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
39924         long int)".  Changed return type of
39925         __builtin_scalar_test_data_class and
39926         __builtin_scalar_test_data_class_sp and
39927         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
39928         int" and changed second argument from "unsigned int" to "int".
39929         Changed type returned by __builtin_scalar_test_neg,
39930         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
39931         from "int" to "bool int".  Added new overloaded function form
39932         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
39933         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
39934         exponent double-precision with floating point first argument.
39935         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
39936         documentation of scalar_test_data_class, scalar_test_neg,
39937         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
39938         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
39939         vec_test_data_class built-in functions to reflect refinements in
39940         their type signatures.
39942 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
39944         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
39945         size of buf.
39946         (aarch64_elf_asm_destructor): Likewise.
39948 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
39950         PR rtl-optimization/78634
39951         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
39952         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
39953         * ifcvt.c (noce_try_cmove): Add missing cost check.
39955         PR rtl-optimization/71724
39956         * combine.c (if_then_else_cond): Look for situations where it is
39957         beneficial to undo the work of one of the recursive calls.
39959 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
39961         PR tree-optimization/70754
39962         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
39963         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
39964         combined stmt before it if not NULL.
39965         (combine_chains): Process refs reversely and compute dominance point
39966         for root ref.
39968 2017-01-23  Martin Liska  <mliska@suse.cz>
39970         PR tree-optimization/79196
39971         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
39972         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
39973         instead of memcmp.
39974         (strlen_optimize_stmt): Call the renamed function.
39976 2017-01-23  Michael Matz  <matz@suse.de>
39978         PR tree-optimization/78384
39979         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
39981 2017-01-23  Richard Biener  <rguenther@suse.de>
39983         PR tree-optimization/79186
39984         * tree-vrp.c (register_new_assert_for): Make sure we've seen
39985         both incoming edges before moving an assert.
39987 2017-01-23  Martin Jambor  <mjambor@suse.cz>
39989         * ipa-prop.c (load_from_param_1): Removed.
39990         (load_from_unmodified_param): Bits from load_from_param_1 put back
39991         here.
39992         (load_from_param): Removed.
39993         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
39994         with stmt.  Reverted back to use of load_from_unmodified_param.
39996 2017-01-23  Martin Jambor  <mjambor@suse.cz>
39998         PR ipa/79108
39999         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40000         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40001         field a pointer to garbage collected vector, mark lattices and
40002         ipcp_orig_node with GTY((skip)).
40003         (ipa_get_param_count): Adjust to descriptors being a pointer.
40004         (ipa_get_param): Likewise.
40005         (ipa_get_type): Likewise.
40006         (ipa_get_param_move_cost): Likewise.
40007         (ipa_set_param_used): Likewise.
40008         (ipa_get_controlled_uses): Likewise.
40009         (ipa_set_controlled_uses): Likewise.
40010         (ipa_is_param_used): Likewise.
40011         (ipa_node_params_t): Move into garbage collector.  New methods insert
40012         and remove.
40013         (ipa_node_params_sum): Annotate wth GTY(()).
40014         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40015         garbage collected.
40016         (ipa_load_from_parm_agg): Adjust declaration.
40017         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40018         * ipa-profile.c (ipa_profile): Likewise.
40019         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40020         (ipa_populate_param_decls): Make descriptors parameter garbage
40021         collected.
40022         (ipa_dump_param): Adjust to descriptors being a pointer.
40023         (ipa_alloc_node_params): Likewise.
40024         (ipa_initialize_node_params): Likewise.
40025         (load_from_param_1): Make descriptors parameter garbage collected.
40026         (load_from_unmodified_param): Likewise.
40027         (load_from_param): Likewise.
40028         (ipa_load_from_parm_agg): Likewise.
40029         (ipa_node_params::~ipa_node_params): Removed.
40030         (ipa_free_all_node_params): Remove call to delete operator.
40031         (ipa_node_params_t::insert): New.
40032         (ipa_node_params_t::remove): Likewise.
40033         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40034         copy known_csts and known_contexts vectors.
40035         (ipa_read_node_info): Adjust to descriptors being a pointer.
40036         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40037         collected.
40038         (ipcp_transform_function): Make descriptors variable garbage
40039         collected.
40041 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40043         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40044         * config/i386/avx512dqintrin.h: Ditto.
40045         * config/i386/avx512fintrin.h: Ditto.
40046         * config/i386/i386.c: Handle new builtins.
40047         * config/i386/i386-builtin.def: Add new builtins.
40048         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40049         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40051 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40052             Martin Liska  <mliska@suse.cz>
40054         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40055         * asan.c (asan_expand_poison_ifn): Support stores and use
40056         appropriate ASAN report function.
40057         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40058         * internal-fn.def (ASAN_POISON_USE): Declare.
40059         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40060         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40061         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40062         ASAN_POISON calls w/o LHS.
40063         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40064         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40065         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40066         * gimplify.c (asan_poison_variables): Add attribute
40067         use_after_scope_memory to variables that really needs to live
40068         in memory.
40069         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40070         having the attribute.
40072 2017-01-23  Martin Liska  <mliska@suse.cz>
40074         * asan.c (create_asan_shadow_var): New function.
40075         (asan_expand_poison_ifn): Likewise.
40076         * asan.h (asan_expand_poison_ifn): New declaration.
40077         * internal-fn.c (expand_ASAN_POISON): Likewise.
40078         * internal-fn.def (ASAN_POISON): New builtin.
40079         * sanopt.c (pass_sanopt::execute): Expand
40080         asan_expand_poison_ifn.
40081         * tree-inline.c (copy_decl_for_dup_finish): Make function
40082         external.
40083         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40084         * tree-ssa.c (is_asan_mark_p): New function.
40085         (execute_update_addresses_taken): Rewrite local variables
40086         (identified just by use-after-scope as addressable) into SSA.
40088 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40090         * doc/install.texi (Specific): opensource.apple.com uses https
40091         now. Remove trailing slash.
40093 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40095         * README.Portability: Remove note on an Irix compatibility issue.
40097 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40099         * gcov.c (INCLUDE_ALGORITHM): Define.
40100         (INCLUDE_VECTOR): Define.
40101         No longer include <vector> and <algorithm> directly.
40103 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40105         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40106         to https.
40107         * doc/invoke.texi (Code Gen Options): Ditto.
40109 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40111         PR lto/78407
40112         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40114 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40116         rtl-optimization/79125
40117         * cprop.c (local_cprop_pass): Handle cases where we make an
40118         unconditional trap.
40120 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40122         PR target/61729
40123         PR target/77850
40124         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40125         read from, for big endian.
40127 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40129         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40130         register pauth builtins for LP64 only.
40132 2017-01-20  Marek Polacek  <polacek@redhat.com>
40134         PR c/79152
40135         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40136         non-case labels.
40138 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40140         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40141         of safelen status.
40142         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40143         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40144         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40146 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40148         PR target/71270
40149         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40150         in big-endian mode when they are not a single duplicated value.
40152 2017-01-20  Richard Biener  <rguenther@suse.de>
40154         * BASE-VER: Bump to 7.0.1.
40156 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40158         * omp-low.c (omplow_simd_context): New struct.  Use it...
40159         (lower_rec_simd_input_clauses): ...here and...
40160         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40161         references to idx, lane, max_vf, is_simt.
40163 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40165         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40166         mcpu=nps400.
40168 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40170         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40171         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40172         gt-hsa-common.h.
40173         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40174         (GTFILES): Rename hsa.c to hsa-common.c.
40175         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40176         * hsa-dump.c: Likewise.
40177         * hsa-gen.c: Likewise.
40178         * hsa-regalloc.c: Likewise.
40179         * ipa-hsa.c: Likewise.
40180         * omp-expand.c: Likewise.
40181         * omp-low.c: Likewise.
40182         * toplev.c: Likewise.
40184 2017-01-20  Marek Polacek  <polacek@redhat.com>
40186         PR c/64279
40187         * doc/invoke.texi: Document -Wduplicated-branches.
40188         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40189         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40190         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40191         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40192         return 0 only when not OEP_LEXICOGRAPHIC.
40193         (fold_build_cleanup_point_expr): Use the expression
40194         location when building CLEANUP_POINT_EXPR.
40195         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40196         * tree.c (add_expr): Handle error_mark_node.
40198 2017-01-20  Martin Liska  <mliska@suse.cz>
40200         PR lto/69188
40201         * tree-profile.c (init_ic_make_global_vars): Do not call
40202         finalize_decl.
40203         (gimple_init_gcov_profiler): Likewise.
40205 2017-01-20  Martin Liska  <mliska@suse.cz>
40207         PR ipa/71190
40208         * cgraph.h (maybe_create_reference): Remove argument and
40209         update comment.
40210         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40211         argument.
40212         * ipa-cp.c (create_specialized_node): Likewise.
40213         * symtab.c (symtab_node::maybe_create_reference): Handle
40214         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40216 2017-01-20  Martin Liska  <mliska@suse.cz>
40218         * read-rtl-function.c (function_reader::create_function): Use
40219         build_decl instread of build_decl_stat.
40221 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40223         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40224         * config/i386/avx512dqintrin.h: Ditto.
40225         * config/i386/avx512fintrin.h: Ditto.
40226         * config/i386/i386-builtin-types.def: Add new types.
40227         * config/i386/i386.c: Handle new types.
40228         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40229         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40230         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40231         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40232         (__builtin_ia32_kshiftridi): New.
40233         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40235 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40237         PR target/78875
40238         PR target/79140
40239         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40240         define to rs6000_init_stack_protect_guard.
40241         (rs6000_init_stack_protect_guard): New function.
40243 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40244             Yunqiang Su  <yunqiang.su@imgtec.com>
40246         * config.gcc (supported_defaults): Add madd4.
40247         (with_madd4): Add validation.
40248         (all_defaults): Add madd4.
40249         * config/mips/mips.opt (mmadd4): New option.
40250         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40251         mmadd4.
40252         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40253         __mips_no_madd4.
40254         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40255         (ISA_HAS_FUSED_MADD4): Likewise.
40256         * doc/invoke.texi (-mmadd4): Document the new option.
40257         * doc/install.texi (--with-madd4): Document the new option.
40259 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40261         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40262         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40263         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40264         (aarch64_init_pauth_hint_builtins): New.
40265         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40266         (aarch64_expand_builtin): Expand new builtins.
40268 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40270         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40271         * combine-stack-adj.c (no_unhandled_cfa): Handle
40272         REG_CFA_TOGGLE_RA_MANGLE.
40273         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40274         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40275         info for return address signing.
40276         (aarch64_expand_epilogue): Likewise.
40278 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40280         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40281         * config/aarch64/aarch64-protos.h
40282         (aarch64_return_address_signing_enabled): New declaration.
40283         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40284         New function.
40285         (aarch64_expand_prologue): Sign return address before it's pushed onto
40286         stack.
40287         (aarch64_expand_epilogue): Authenticate return address fetched from
40288         stack.
40289         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40290         (aarch64_attributes): New function attributes for "sign-return-address".
40291         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40292         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40293         ("*do_return"): Generate combined instructions according to key index.
40294         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40295         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40296         iterators.
40297         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40298         * config/aarch64/aarch64.opt (msign-return-address=): New.
40299         * doc/extend.texi (AArch64 Function Attributes): Documents
40300         "sign-return-address=".
40301         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40303 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40305         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40306         overall option summary.
40308 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40310         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40311         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40312         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40313         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40315 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40317         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40318         -mpower9-minmax by default for -mcpu=power9.
40319         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40320         128-bit floating point.
40322 2017-01-20  Alan Modra  <amodra@gmail.com>
40324         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40325         optimizing for size.
40327 2017-01-20  Alan Modra  <amodra@gmail.com>
40329         PR target/79144
40330         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40331         for strcmp and strncmp from corresponding builtin decl.
40333 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40335         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40336         instead of i386/rtems-64.h.
40337         * config/i386/rtems-64.h: Remove.
40339 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40341         PR target/78478
40342         Revert:
40343         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40345         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40347 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40349         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40350         Change int to HOST_WIDE_INT.
40351         * config/aarch64/aarch64-protos.h
40352         (aarch64_simd_gen_const_vector_dup): Likewise.
40353         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40355 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40357         * langhooks-def.h (lhd_type_for_size): New decl.
40358         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40359         * langhooks.c (lhd_type_for_size): New function, taken from
40360         lto_type_for_size.
40362 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40364         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40365         define_bypass for CR latency.
40366         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40367         (power9-alu2): Add define_bypass for CR latency.
40368         (power9-cmp): New.
40369         (power9-mul): Update insn latency.
40370         (power9-mul-compare): Update insn latency, bypass latency and remove
40371         power9-branch.
40373 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40375         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40376         Delete.
40377         * config/aarch64/aarch64.md
40378         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40379         aarch64_nopcrelative_literal_loads.
40380         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40382 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40384         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40385         TARGET_LOONGSON_3A.
40386         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40388 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40390         PR target/78176
40391         * config.gcc (supported_defaults): Add lxc1-sxc1.
40392         (with_lxc1_sxc1): Add validation.
40393         (all_defaults): Add lxc1-sxc1.
40394         * config/mips/mips.opt (mlxc1-sxc1): New option.
40395         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40396         mlxc1-sxc1.
40397         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40398         __mips_no_lxc1_sxc1.
40399         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40400         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40401         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40403 2017-01-19  Richard Biener  <rguenther@suse.de>
40405         PR tree-optimization/72488
40406         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40407         sure to restore SSA info.
40408         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40410 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40412         PR rtl-optimization/79121
40413         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40414         of the inner type when shifting an extended value.
40416 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40418         PR lto/78407
40419         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40420         interposable aliases.
40422 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40424         PR target/78516
40425         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40426         Use the evmergelohi instruction.
40427         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40428         (mov_sitf_e500_subreg8_2_be): Likewise.
40429         (mov_sitf_e500_subreg12_2_le): Likewise.
40430         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40431         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40432         (mov_sitf_e500_subreg8_2_le): Likewise.
40433         (mov_sitf_e500_subreg12_2_be): Likewise.
40435 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40437         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40438         attribute from vecsimple to vecperm.
40439         (altivec_vbpermq2): Likewise.
40441 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40443         PR target/79040
40444         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40446 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40447         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40448         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40449         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40450         case where N arg is SIZE_MAX.
40451         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40452         (cmpstrsi): Add pattern.
40454 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40456         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40457         __builtin_vec_revb builtins.
40458         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40459         built-in functions to support generation of the ISA 3.0 XXBR<x>
40460         vector byte reverse instructions.
40461         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40462         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40463         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40464         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40465         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40466         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40467         (P9V_BUILTIN_VEC_REVB): Likewise.
40468         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40469         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40470         (p9_xxbrq_v16qi): Likewise.
40471         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40472         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40473         (p9_xxbrh_v8hi): Likewise.
40474         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40475         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40476         vec_revb built-in functions.
40478 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40480         PR rtl-optimization/78952
40481         * config/i386/i386.md (any_extract): New code iterator.
40482         (*insvqi_2): Use any_extract for source operand.
40483         (*insvqi_3): Use any_shiftrt for source operand.
40485 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40487         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40488         New function.
40489         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40491 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40493         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40495 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40497         * config/rs6000/altivec.h (vec_bperm): Change #define.
40498         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40499         (altivec_vbpermq2): New define_insn.
40500         (altivec_vbpermd): Likewise.
40501         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40502         function interface.
40503         (VBPERMD): Likewise.
40504         (VBPERM): New polymorphic function interface.
40505         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40506         Add entries for P9V_BUILTIN_VEC_VBPERM.
40507         * doc/extend.texi: Add interfaces for vec_bperm.
40509 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40511         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40512         first letter of error messages.
40513         (s390_resolve_overloaded_builtin): Likewise.
40514         * config/s390/s390.c (s390_expand_builtin): Likewise.
40515         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40516         (s390_valid_target_attribute_inner_p): Likewise.
40517         * config/s390/s390.md ("tabort"): Likewise.
40519 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40521         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40522         (ISA_AVOID_DIV_HILO): New macro.
40523         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40524         (ISA_HAS_DDIV): Likewise.
40526 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40528         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40530 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40532         * doc/invoke.texi (fabi-version): Spelling fix.
40534 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40536         PR c++/70182
40537         * doc/invoke.texi (fabi-version): Mention mangling fix for
40538         operator names.
40540 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40542         PR c++/77489
40543         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40545 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40547         PR target/78875
40548         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40549         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40550         the new options.
40551         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40552         flexible settings.
40553         (stack_protect_test): Ditto.
40554         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40555         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40556         options.
40557         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40558         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40559         -mstack-protector-guard-offset=.
40560         (RS/6000 and PowerPC Options): Ditto.
40562 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40564         * config/i386/i386.h (MASK_CLASS_P): New define.
40565         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40566         there are no registers from different register sets also when
40567         mask registers are used.  Update function comment.
40568         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40569         to (*k/*r) and (*k/*km) alternatives.
40571 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40573         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40574         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40575         (EH_RETURN_HANDLER_RTX): New define.
40576         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40577         Force frame pointer in EH return functions.
40578         (aarch64_expand_epilogue): Add barrier for eh_return.
40579         (aarch64_final_eh_return_addr): Remove.
40580         (aarch64_eh_return_handler_rtx): New function.
40581         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40582         Remove.
40583         (aarch64_eh_return_handler_rtx): New prototype.
40585 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40587         * config/rs6000/altivec.h (vec_rlmi): New #define.
40588         (vec_vrlnm): Likewise.
40589         (vec_rlnm): Likewise.
40590         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40591         (UNSPEC_VRLNM): Likewise.
40592         (VIlong): New mode iterator.
40593         (altivec_vrl<VI_char>mi): New define_insn.
40594         (altivec_vrl<VI_char>nm): Likewise.
40595         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40596         function entry.
40597         (VRLDNM): Likewise.
40598         (RLNM): New polymorphic function entry.
40599         (VRLWMI): New monomorphic function entry.
40600         (VRLDMI): Likewise.
40601         (RLMI): New polymorphic function entry.
40602         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40603         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40604         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40605         vec_vrlnm.
40607 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40609         PR debug/78839
40610         * dwarf2out.c (field_byte_offset): Restore the
40611         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40612         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40613         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40614         of build2 + fold.
40616 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40618         PR ada/67205
40619         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40621 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40623         PR debug/71669
40624         * dwarf2out.c (add_data_member_location_attribute): For constant
40625         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40626         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40627         DW_AT_byte_size attributes.
40629 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40631         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40632         after forcing to constant memory when the code model is medium.
40634 2017-01-17  Julia Koval  <julia.koval@intel.com>
40636         PR target/76731
40637         * config/i386/avx512fintrin.h
40638         (_mm512_i32gather_ps): Change __addr type to void const*.
40639         (_mm512_mask_i32gather_ps): Ditto.
40640         (_mm512_i32gather_pd): Ditto.
40641         (_mm512_mask_i32gather_pd): Ditto.
40642         (_mm512_i64gather_ps): Ditto.
40643         (_mm512_mask_i64gather_ps): Ditto.
40644         (_mm512_i64gather_pd): Ditto.
40645         (_mm512_mask_i64gather_pd): Ditto.
40646         (_mm512_i32gather_epi32): Ditto.
40647         (_mm512_mask_i32gather_epi32): Ditto.
40648         (_mm512_i32gather_epi64): Ditto.
40649         (_mm512_mask_i32gather_epi64): Ditto.
40650         (_mm512_i64gather_epi32): Ditto.
40651         (_mm512_mask_i64gather_epi32): Ditto.
40652         (_mm512_i64gather_epi64): Ditto.
40653         (_mm512_mask_i64gather_epi64): Ditto.
40654         (_mm512_i32scatter_ps): Change __addr type to void*.
40655         (_mm512_mask_i32scatter_ps): Ditto.
40656         (_mm512_i32scatter_pd): Ditto.
40657         (_mm512_mask_i32scatter_pd): Ditto.
40658         (_mm512_i64scatter_ps): Ditto.
40659         (_mm512_mask_i64scatter_ps): Ditto.
40660         (_mm512_i64scatter_pd): Ditto.
40661         (_mm512_mask_i64scatter_pd): Ditto.
40662         (_mm512_i32scatter_epi32): Ditto.
40663         (_mm512_mask_i32scatter_epi32): Ditto.
40664         (_mm512_i32scatter_epi64): Ditto.
40665         (_mm512_mask_i32scatter_epi64): Ditto.
40666         (_mm512_i64scatter_epi32): Ditto.
40667         (_mm512_mask_i64scatter_epi32): Ditto.
40668         (_mm512_i64scatter_epi64): Ditto.
40669         (_mm512_mask_i64scatter_epi64): Ditto.
40670         * config/i386/avx512pfintrin.h
40671         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40672         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40673         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40674         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40675         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40676         (_mm512_prefetch_i32scatter_ps): Ditto.
40677         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40678         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40679         (_mm512_prefetch_i64scatter_pd): Ditto.
40680         (_mm512_prefetch_i64scatter_ps): Ditto.
40681         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40682         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40683         * config/i386/avx512vlintrin.h
40684         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40685         (_mm_mmask_i32gather_ps): Ditto.
40686         (_mm256_mmask_i32gather_pd): Ditto.
40687         (_mm_mmask_i32gather_pd): Ditto.
40688         (_mm256_mmask_i64gather_ps): Ditto.
40689         (_mm_mmask_i64gather_ps): Ditto.
40690         (_mm256_mmask_i64gather_pd): Ditto.
40691         (_mm_mmask_i64gather_pd): Ditto.
40692         (_mm256_mmask_i32gather_epi32): Ditto.
40693         (_mm_mmask_i32gather_epi32): Ditto.
40694         (_mm256_mmask_i32gather_epi64): Ditto.
40695         (_mm_mmask_i32gather_epi64): Ditto.
40696         (_mm256_mmask_i64gather_epi32): Ditto.
40697         (_mm_mmask_i64gather_epi32): Ditto.
40698         (_mm256_mmask_i64gather_epi64): Ditto.
40699         (_mm_mmask_i64gather_epi64): Ditto.
40700         (_mm256_i32scatter_ps): Change __addr type to void*.
40701         (_mm256_mask_i32scatter_ps): Ditto.
40702         (_mm_i32scatter_ps): Ditto.
40703         (_mm_mask_i32scatter_ps): Ditto.
40704         (_mm256_i32scatter_pd): Ditto.
40705         (_mm256_mask_i32scatter_pd): Ditto.
40706         (_mm_i32scatter_pd): Ditto.
40707         (_mm_mask_i32scatter_pd): Ditto.
40708         (_mm256_i64scatter_ps): Ditto.
40709         (_mm256_mask_i64scatter_ps): Ditto.
40710         (_mm_i64scatter_ps): Ditto.
40711         (_mm_mask_i64scatter_ps): Ditto.
40712         (_mm256_i64scatter_pd): Ditto.
40713         (_mm256_mask_i64scatter_pd): Ditto.
40714         (_mm_i64scatter_pd): Ditto.
40715         (_mm_mask_i64scatter_pd): Ditto.
40716         (_mm256_i32scatter_epi32): Ditto.
40717         (_mm256_mask_i32scatter_epi32): Ditto.
40718         (_mm_i32scatter_epi32): Ditto.
40719         (_mm_mask_i32scatter_epi32): Ditto.
40720         (_mm256_i32scatter_epi64): Ditto.
40721         (_mm256_mask_i32scatter_epi64): Ditto.
40722         (_mm_i32scatter_epi64): Ditto.
40723         (_mm_mask_i32scatter_epi64): Ditto.
40724         (_mm256_i64scatter_epi32): Ditto.
40725         (_mm256_mask_i64scatter_epi32): Ditto.
40726         (_mm_i64scatter_epi32): Ditto.
40727         (_mm_mask_i64scatter_epi32): Ditto.
40728         (_mm256_i64scatter_epi64): Ditto.
40729         (_mm256_mask_i64scatter_epi64): Ditto.
40730         (_mm_i64scatter_epi64): Ditto.
40731         (_mm_mask_i64scatter_epi64): Ditto.
40732         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40733         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40734         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40735         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40736         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40737         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40738         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40739         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40740         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40741         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40742         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40743         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40744         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40745         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40746         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40747         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40748         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40749         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40750         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40751         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40752         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40753         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40754         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40755         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40756         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40757         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40758         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40759         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40760         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40761         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40762         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40763         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40764         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40765         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40766         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40767         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40768         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40769         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40770         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40771         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40772         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40773         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40774         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40775         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40776         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40777         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40778         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40779         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40780         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40781         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40782         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40783         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40784         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40785         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40786         definitions accordingly.
40788 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40789             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40791         PR target/79079
40792         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40793         gen_lowpart.
40795 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40797         PR target/79058
40798         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40799         hard regs for inner regno.
40801 2017-01-17  Martin Liska  <mliska@suse.cz>
40803         PR ipa/71207
40804         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40805         assumption and add comment.
40807 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40809         * ipa-visibility.c (localize_node): New function, broken out of ...
40810         (function_and_variable_visibility): ... here. Call it.
40812 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40814         PR middle-end/77445
40815         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40816         correctly set frequency of oudgoing edge.
40817         (duplicate_thread_path): Fix profile updating.
40819 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40821         PR other/79046
40822         * configure.ac: Add GCC_BASE_VER.
40823         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40824         version from BASE-VER file.
40825         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40826         (gcc.o): Depend on $(BASEVER).
40827         * common.opt (dumpfullversion): New option.
40828         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40829         * doc/invoke.texi: Document -dumpfullversion.
40830         * doc/install.texi: Document --with-gcc-major-version-only.
40831         * configure: Regenerated.
40833 2017-01-17  Richard Biener  <rguenther@suse.de>
40835         PR tree-optimization/71433
40836         * tree-vrp.c (register_new_assert_for): Merge same asserts
40837         on all incoming edges.
40838         (process_assert_insertions_for): Handle insertions at the
40839         beginning of BBs.
40841 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40843         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40844         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40846 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40848         PR target/78633
40849         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40850         RTL sharing.
40852 2017-01-17  Alan Modra  <amodra@gmail.com>
40854         PR target/79066
40855         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40856         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40857         symbolic stack limit when pic.
40859 2017-01-16  Martin Sebor  <msebor@redhat.com>
40861         PR tree-optimization/78608
40862         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40864 2017-01-16  Jeff Law  <law@redhat.com>
40866         Revert:
40867         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
40868         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
40869         for several include directories that may be relative to sysroot.
40870         * config/i386/x-mingw32 (gplus_includedir): Define.
40871         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
40872         (native_system_includedir): Likewise.
40873         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
40874         override if TARGET_SYSTEM_ROOT is defined.
40875         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
40877         PR tree-optimization/79090
40878         PR tree-optimization/33562
40879         PR tree-optimization/61912
40880         PR tree-optimization/77485
40881         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
40882         and computed trims into the dump file.
40884 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40886         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
40888 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
40890         PR c/79089
40891         * gimplify.c (gimplify_init_constructor): If want_value and
40892         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
40893         fix.
40895         PR target/79080
40896         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
40897         sequence.  Formatting fixes.
40898         (doloop_optimize): Formatting fixes.
40900         PR driver/49726
40901         * gcc.c (debug_level_greater_than_spec_func): New function.
40902         (static_spec_functions): Add debug-level-gt spec function.
40903         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
40904         !g0.
40905         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40906         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40907         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
40908         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
40909         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
40910         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
40912 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
40914         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
40915         QImode fixups to general and mask registers only.
40917 2017-01-16  Carl Love  <cel@us.ibm.com>
40919         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
40920         for built-in functions
40921         vector signed char vec_nabs (vector signed char)
40922         vector signed short vec_nabs (vector signed short)
40923         vector signed int vec_nabs (vector signed int)
40924         vector signed long long vec_nabs (vector signed long long)
40925         vector float vec_nabs (vector float)
40926         vector double vec_nabs (vector double)
40927         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
40928         and NABS overload.
40929         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
40930         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
40931         * doc/extend.texi: Update the documentation file for the new built-in
40932         functions.
40934 2017-01-16  Martin Sebor  <msebor@redhat.com>
40936         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
40937         message.
40939 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40941         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
40942         UNSPEC_VSX__XXSPLTD to require special splat handling.
40944 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
40946         PR bootstrap/78616
40947         * system.h: Poison strndup.
40949 2017-01-16  Alan Modra  <amodra@gmail.com>
40951         PR target/79098
40952         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
40953         use a switch.
40955 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
40957         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
40959 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
40961         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
40962         call recog here.  Assert that INSN_CODE (insn) is non-negative.
40964 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
40966         PR target/72749
40967         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
40968         fallthrough.
40969         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
40970         in the currently scheduled RTL fragment.
40972 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
40974         PR rtl-optimization/78751
40975         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
40976         give up.
40978 2017-01-14  Jeff Law  <law@redhat.com>
40980         PR tree-optimization/79090
40981         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
40982         variable length stores.
40983         (compute_trims): Delete dead assignment to *trim_tail.
40984         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
40985         zero length.
40987 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
40989         PR rtl-optimization/78626
40990         PR rtl-optimization/78727
40991         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
40992         of a block, and split such blocks after everything else is finished.
40994 2017-01-14  Alan Modra  <amodra@gmail.com>
40996         PR target/72749
40997         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
40998         target legitimate_combined_insn.
40999         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41000         (rs6000_legitimate_combined_insn): New function.
41001         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41002         all uses.
41003         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41004         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41005         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41007 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41009         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41011 2017-01-13  Jeff Law  <law@redhat.com>
41013         PR tree-optimization/33562
41014         PR tree-optimization/61912
41015         PR tree-optimization/77485
41016         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41017         a statement.
41018         (delete_dead_assignment): Likewise.
41019         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41020         statement to delete_dead_call and delete_dead_assignment.
41022 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41024         PR c/78304
41025         * substring-locations.c (format_warning_va): Strengthen case 1 so
41026         that both endpoints of the substring must be within the format
41027         range for just the substring to be printed.
41029 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41031         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41032         * config/i386/i386.c (ix86_target_string): Add missing options
41033         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41034         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41035         flags_other and ix86_target_other to flags2_other.  Display unknown
41036         isa2 options.
41037         (ix86_valid_target_attribute_inner_p): Add missing options and
41038         reorder options by implied ISAs, as in ix86_target_string.
41040 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41042         * hash-table.h (hash_table::too_empty_p): New function.
41043         (hash_table::expand): Use it.
41044         (hash_table::traverse): Likewise.
41045         (hash_table::empty_slot): Use sizeof (value_type) instead of
41046         sizeof (PTR) to convert bytes to elements.  Shrink the table
41047         if the current size is excessive for the current number of
41048         elements.
41050 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41052         * ira-costs.c (record_reg_classes): Break from the inner loop
41053         early once alt_fail is known to be true.  Update outer loop
41054         handling accordingly.
41056 2017-01-13  Jeff Law  <law@redhat.com>
41058         * tree-ssa-dse.c (decrement_count): New function.
41059         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41060         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41061         when we know the partially dead statement is a mem* function.
41063         PR tree-optimization/61912
41064         PR tree-optimization/77485
41065         * tree-ssa-dse.c: Include expr.h.
41066         (maybe_trim_constructor_store): New function.
41067         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41069         PR tree-optimization/33562
41070         PR tree-optimization/61912
41071         PR tree-optimization/77485
41072         * doc/invoke.texi: Document new dse-max-object-size param.
41073         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41074         * tree-ssa-dse.c: Include params.h.
41075         (dse_store_status): New enum.
41076         (initialize_ao_ref_for_dse): New, partially extracted from
41077         dse_optimize_stmt.
41078         (valid_ao_ref_for_dse, normalize_ref): New.
41079         (setup_live_bytes_from_ref, compute_trims): Likewise.
41080         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41081         (maybe_trim_partially_dead_store): Likewise.
41082         (maybe_trim_complex_store): Likewise.
41083         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41084         Track what bytes live from the original store.  Return tri-state
41085         for dead, partially dead or live.
41086         (dse_dom_walker): Add constructor, destructor and new private members.
41087         (delete_dead_call, delete_dead_assignment): New extracted from
41088         dse_optimize_stmt.
41089         (dse_optimize_stmt): Make a member of dse_dom_walker.
41090         Use initialize_ao_ref_for_dse.
41092         PR tree-optimization/33562
41093         PR tree-optimization/61912
41094         PR tree-optimization/77485
41095         * sbitmap.h (bitmap_count_bits): Prototype.
41096         (bitmap_clear_range, bitmap_set_range): Likewise.
41097         * sbitmap.c (bitmap_clear_range): New function.
41098         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41100 2017-01-13  Martin Liska  <mliska@suse.cz>
41102         PR ipa/79043
41103         * function.c (set_cfun): Add new argument force.
41104         * function.h (set_cfun): Likewise.
41105         * ipa-inline-transform.c (inline_call): Use the function when
41106         strict alising from is dropped for function we inline to.
41108 2017-01-13  Richard Biener  <rguenther@suse.de>
41110         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41111         for dumping GIMPLE INTEGER_CSTs.
41113 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41115         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41116         to 201112L since C++17.
41118 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41120         PR sanitizer/78887
41121         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41122         if -fsanitize=kernel-address is present.
41124 2017-01-13  Richard Biener  <rguenther@suse.de>
41126         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41127         as _Literal ( type ) number in case usual suffixes do not
41128         preserve all information.
41130 2017-01-13  Richard Biener  <rguenther@suse.de>
41132         PR tree-optimization/77283
41133         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41134         and ssa-iterators.h.
41135         (is_feasible_trace): Implement a cost model based on joiner
41136         PHI node uses.
41138 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41140         PR target/79004
41141         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41142         char or short to __float128/_Float128 directly.
41144 2017-01-12  Martin Sebor  <msebor@redhat.com>
41146         to -Wformat-overflow.
41147         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41148         (min_bytes_remaining): Same.
41149         (get_string_length): Same.
41150         (format_string): Same.
41151         (format_directive): Same.
41152         (add_bytes): Same.
41153         (pass_sprintf_length::handle_gimple_call): Same.
41155 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41157         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41158         info.nowrite calls with no lhs that can't throw.  Return bool
41159         whether gsi_remove has been called or not.
41160         (pass_sprintf_length::handle_gimple_call): Return bool whether
41161         try_substitute_return_value called gsi_remove.  Formatting fix.
41162         (pass_sprintf_length::execute): Don't use gsi_remove if
41163         handle_gimple_call returned true.
41165         PR bootstrap/79069
41166         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41167         be removed due to side-effects, don't remove following barrier nor
41168         turn the successor edge into fallthru edge.
41170 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41172         PR target/79044
41173         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41174         element-reversing loads and stores as not swappable.
41176 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41177             Nicolai Stange  <nicstange@gmail.com>
41179         * combine.c (try_combine): Don't ignore result of overlap checking
41180         loop.  Combine overlap & asm check into single loop.
41182 2017-01-12  Richard Biener  <rguenther@suse.de>
41184         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41185         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41187 2017-01-12  Richard Biener  <rguenther@suse.de>
41189         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41190         and TS_TARGET_OPTION directly derive from TS_BASE.
41191         * tree-core.h (tree_optimization_option): Derive from tree_base.
41192         (tree_target_option): Likewise.
41194 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41196         * config/i386/i386.c (memory_address_length): Increase len
41197         only when rip_relative_addr_p returns false.
41199 2017-01-11  Julia Koval  <julia.koval@intel.com>
41201         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41202         (OPTION_MASK_ISA_SGX_SET): New.
41203         (ix86_handle_option): Handle OPT_msgx.
41204         * config.gcc: Added sgxintrin.h.
41205         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41206         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41207         * config/i386/i386.c (ix86_target_string): Add -msgx.
41208         (PTA_SGX): New.
41209         (ix86_option_override_internal): Handle new options.
41210         (ix86_valid_target_attribute_inner_p): Add sgx.
41211         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41212         * config/i386/i386.opt: Add msgx.
41213         * config/i386/sgxintrin.h: New file.
41214         * config/i386/x86intrin.h: Add sgxintrin.h.
41216 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41218         PR c++/71537
41219         * fold-const.c (maybe_nonzero_address): Return 1 for function
41220         local objects.
41221         (tree_single_nonzero_warnv_p): Don't handle function local objects
41222         here.
41224         PR c++/72813
41225         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41226         of c-header.
41228 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41230         PR driver/78877
41231         * opts.c: Include "spellcheck.h"
41232         (struct string_fragment): New struct.
41233         (struct edit_distance_traits<const string_fragment &>): New
41234         struct.
41235         (get_closest_sanitizer_option): New function.
41236         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41238 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41240         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41241         by 12.
41242         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41243         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41244         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41245         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41246         for initial die_offset if dwarf_split_debug_info.
41247         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41248         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41249         fields.
41250         (output_skeleton_debug_sections): Formatting fix.  Use
41251         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41252         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41254 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41256         * config/arm/cortex-a53.md: Add bypasses for
41257         cortex_a53_r2f_cvt.
41258         (cortex_a53_r2f): Only use for transfers.
41259         (cortex_a53_f2r): Likewise.
41260         (cortex_a53_r2f_cvt): Add reservation for conversions.
41261         (cortex_a53_f2r_cvt): Likewise.
41263 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41265         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41266         to all inlined functions, change static to extern.
41268 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41270         PR target/78253
41271         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41272         weak symbol.
41273         (arm_assemble_integer): Likewise.
41275 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41277         * config.gcc: Use new awk script to check CPU, FPU and architecture
41278         parameters for --with-... options.
41279         * config/arm/parsecpu.awk: New file
41280         * config/arm/arm-cpus.in: New file.
41281         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41282         files.
41283         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41284         files.
41285         * config/arm/t-arm: Update dependency rules.
41286         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41287         of processing .def files.
41288         * config/arm/genopt.sh: Deleted.
41289         * config/arm/gentune.sh: Deleted.
41290         * config/arm/arm-cores.def: Deleted.
41291         * config/arm/arm-arches.def: Deleted.
41292         * config/arm/arm-fpus.def: Deleted.
41293         * config/arm/arm-tune.md: Regenerated.
41294         * config/arm/arm-tables.opt: Regenerated.
41295         * config/arm/arm-cpu.h: New generated file.
41296         * config/arm/arm-cpu-data.h: New generated file.
41297         * config/arm/arm-cpu-cdata.h: New generated file.
41299 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41301         PR lto/79042
41302         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41303         bit.
41304         (input_varpool_node): Unpack dynamically_initialized bit.
41306 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41308         PR rtl-optimization/79032
41309         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41310         the alignment of the adjusted memory reference against that of MODE,
41311         instead of the alignment of the original memory reference.
41313 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41315         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41316         test.
41317         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41318         decorated functions.
41320 2017-01-11  Richard Biener  <rguenther@suse.de>
41322         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41323         set range/nonnull info for PHI results.  Do not set it on
41324         stmts marked for removal.
41326 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41328         * expr.c (store_field): In the bitfield case, fetch the return value
41329         from the registers before applying a single big-endian adjustment.
41330         Always do a final load for a BLKmode value not larger than a word.
41332 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41334         PR c++/77949
41335         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41336         that we correctly handle column numbers greater than
41337         LINE_MAP_MAX_COLUMN_NUMBER.
41339 2017-01-10  Martin Sebor  <msebor@redhat.com>
41341         PR middle-end/78245
41342         * gimple-ssa-sprintf.c (get_destination_size): Call
41343         {init,fini}object_sizes.
41344         * tree-object-size.c (addr_object_size): Adjust.
41345         (pass_through_call): Adjust.
41346         (pass_object_sizes::execute): Adjust.
41347         * tree-object-size.h (fini_object_sizes): Declare.
41349 2017-01-10  Martin Sebor  <msebor@redhat.com>
41351         PR tree-optimization/78775
41352         * builtins.c (get_size_range): Move...
41353         * calls.c: ...to here.
41354         (alloc_max_size): Accept zero argument.
41355         (operand_signed_p): Remove.
41356         (maybe_warn_alloc_args_overflow): Call get_size_range.
41357         * calls.h (get_size_range): Declare.
41359 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41361         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41362         from TI's devices.csv file as of September 2016.
41363         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41365 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41367         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41368         * doc/invoke.texi: Likewise.
41369         * doc/md.texi: Likewise.
41370         * doc/objc.texi: Likewise.
41372 2017-01-10  Joshua Conner  <joshconner@google.com>
41374         * config/arm/fuchsia-elf.h: New file.
41375         * config/fuchsia.h: New file.
41376         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41377         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41378         targets.
41379         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41381 2016-01-10  Richard Biener  <rguenther@suse.de>
41383         PR tree-optimization/79034
41384         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41385         Propagate out degenerate PHIs in the joiner.
41387 2017-01-10  Martin Liska  <mliska@suse.cz>
41389         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41390         (sort_congruence_classes_by_decl_uid): Likewise.
41391         (sort_congruence_class_groups_by_decl_uid): Likewise.
41392         (sem_item_optimizer::merge_classes): Sort class, groups in these
41393         classes and members in the groups by DECL_UID of declarations.
41394         This would make merge operations stable.
41396 2017-01-10  Martin Liska  <mliska@suse.cz>
41398         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41399         usage of m_classes_vec.
41400         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41401         (sem_item_optimizer::get_group_by_hash): Likewise.
41402         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41403         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41404         (sem_item_optimizer::verify_classes): Likewise.
41405         (sem_item_optimizer::process_cong_reduction): Likewise.
41406         (sem_item_optimizer::dump_cong_classes): Likewise.
41407         (sem_item_optimizer::merge_classes): Likewise.
41408         * ipa-icf.h (congruence_class_hash): Rename from
41409         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41411 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41413         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41414         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41415         * config.gcc: Add avx512vpopcntdqintrin.h.
41416         * config/i386/avx512vpopcntdqintrin.h: New.
41417         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41418         * config/i386/i386-builtin-types.def: Add new types.
41419         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41420         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41421         __builtin_ia32_vpopcountq_v8di_mask): New.
41422         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41423         __AVX512VPOPCNTDQ__.
41424         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41425         (PTA_AVX512VPOPCNTDQ): Define.
41426         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41427         TARGET_AVX512VPOPCNTDQ_P): Define.
41428         * config/i386/i386.opt: Add mavx512vpopcntdq.
41429         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41430         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41432 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41434         PR middle-end/77484
41435         * predict.def (PRED_CALL): Set to 67.
41437 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41439         * expr.c (store_field): In the bitfield case, if the value comes from
41440         a function call and is of an aggregate type returned in registers, do
41441         not modify the field mode; extract the value in all cases if the mode
41442         is BLKmode and the size is not larger than a word.
41444 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41446         PR target/71017
41447         * config/i386/cpuid.h: Fix undefined behavior.
41449 2017-01-04  Jeff Law  <law@redhat.com>
41451         PR tree-optimization/79007
41452         PR tree-optimization/67955
41453         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41454         conservative for pt.null when flag_non_call_exceptions is on.
41456 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41458         PR translation/79019
41459         PR translation/79020
41460         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41461         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41462         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41463         in descriptions.
41464         * config/avr/avr.opt (maccumulate-args): Likewise.
41465         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41466         * common.opt (freport-bug): Likewise.
41467         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41468         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41469         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41470         translatable string.
41471         * config/i386/i386.c (function_value_32): Likewise.
41472         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41473         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41474         Likewise.
41475         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41476         * common/config/msp430/msp430-common.c (msp430_handle_option):
41477         Likewise.
41478         * symtab.c (symtab_node::verify_base): Likewise.
41479         * opts.c (set_debug_level): Likewise.
41480         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41481         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41482         missing whitespace to translatable strings.
41483         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41484         * config/sh/superh.h: Likewise.
41485         * config/i386/xopintrin.h: Likewise.
41486         * config/i386/znver1.md: Likewise.
41487         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41488         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41489         * double-int.h (struct double_int): Likewise.
41490         * double-int.c (div_and_round_double): Likewise.
41491         * wide-int.cc: Likewise.
41492         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41493         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41494         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41495         (crossjumps_occurred): ... this.
41496         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41497         Adjust all uses.
41499         PR tree-optimization/78899
41500         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41501         returning bool return struct loop *, NULL for failure and the new
41502         loop on success.
41503         (versionable_outer_loop_p): Don't version outer loop if it has
41504         dont_vectorized bit set.
41505         (tree_if_conversion): When versioning outer loop, ensure
41506         tree_if_conversion is performed also on the inner loop of the
41507         non-vectorizable outer loop copy.
41508         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41509         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41510         prevent vectorization of it.
41511         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41512         the outer loop vectorization of the non-scalar version is attempted
41513         before vectorization of the inner loop in scalar version.  If
41514         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41515         vectorization of its inner loop.
41516         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41517         has 2 inner loops, rename also on edges from bb whose single pred
41518         is outer_loop->header.  Fix typo in function comment.
41520 2017-01-09  Martin Sebor  <msebor@redhat.com>
41522         PR bootstrap/79033
41523         * asan.c (asan_emit_stack_protection): Increase local buffer size
41524         to avoid snprintf truncation warning.
41526 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41528         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41529         to reference thunderx2t99 for the tuning structure
41530         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41531         Rename to ...
41532         (thunderx2t99_extra_costs): This.
41533         * config/aarch64/aarch64-tune.md: Regenerate.
41534         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41535         (vulcan_addrcost_table): This.
41536         (vulcan_regmove_cost): Rename to ...
41537         (thunderx2t99_regmove_cost): This.
41538         (vulcan_vector_cost): Rename to ...
41539         (thunderx2t99_vector_cost): this.
41540         (vulcan_branch_cost): Rename to ...
41541         (thunderx2t99_branch_cost): This.
41542         (vulcan_tunings): Rename to ...
41543         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41544         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41546 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41548         PR ipa/78365
41549         PR ipa/78599
41550         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41551         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41552         (propagate_vr_accross_jump_function): Use the above function for all
41553         value range computations for pass-through jump functions and type
41554         converasion from explicit value range values.
41555         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41556         parameters from TYPE_ARG_TYPES.
41557         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41558         (ipa_write_node_info): Stream type of the actual argument.
41559         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41561 2017-01-09  Martin Liska  <mliska@suse.cz>
41563         PR pch/78970
41564         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41565         (lookup_compiler): Do not show error message with have_E.
41567 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41569         PR tree-optimization/78938
41570         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41571         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41572         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41573         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41574         fixes.
41576 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41578         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41579         is const0_rtx.
41581 2017-01-09  Richard Biener  <rguenther@suse.de>
41583         PR tree-optimization/78997
41584         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41585         name condition properly.
41587 2017-01-09  Richard Biener  <rguenther@suse.de>
41589         PR debug/79000
41590         * dwarf2out.c (is_cxx): New overload with context.
41591         (is_naming_typedef_decl): Use it.
41593 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41595         * invoke.texi (Option Summary): Correct spacing in option lists
41596         and add line breaks to fix over-long lines.
41598 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41600         PR middle-end/17660
41602         * extend.texi (Common Variable Attributes): Add xref to GCC
41603         Internals manual to explain mode attribute keywords.
41605 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41607         PR other/16519
41608         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41609         and Preprocessor Options.
41610         (Options for Linking): Document -pthread here....
41611         (RS/6000 and PowerPC Options): ...not here.
41612         (Solaris 2 Options): ...or here.
41613         * doc/cppopts.texi: Document -pthread.
41615 2017-01-08  Martin Sebor  <msebor@redhat.com>
41617         PR middle-end/77708
41618         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41619         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41620         New member functions.
41621         (format_directive): Used them.
41622         (add_bytes): Same.
41623         (pass_sprintf_length::handle_gimple_call): Same.
41624         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41625         to avoid truncation for any argument.
41626         (extract_affine_mul): Same.
41627         * tree.c (get_file_function_name): Same.
41629 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41631         PR middle-end/77484
41632         * predict.def (PRED_INDIR_CALL): Set to 86.
41634 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41636         PR preprocessor/54124
41637         * doc/cppopts.texi: Reformat -d subtable to list the full name
41638         of the options.  Add cross-reference to the docs for the general
41639         compiler -d options.
41640         * doc/invoke.texi (Developer Options): Add cross-reference to the
41641         preprocessor-specific -d option documentation.
41643 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41645         PR preprocessor/13498
41646         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41647         redudant material, and reflect new command-line options.
41648         (System Headers): Likewise.
41650 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41652         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41653         -isystem, and -idirafter.  Copy-edit.
41654         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41655         default for -ftrack-macro-expansion.  Delete obsolete and
41656         badly-formatted implementation details about -fdebug-cpp output.
41657         * doc/cppwarnopts.texi: Copy-edit.
41659 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41661         PR c++/72803
41662         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41663         that the transition from a max line width >= 1<<10 to narrower
41664         lines works correctly.
41666 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41668         * doc/options.texi (PerFunction): New.
41669         * opt-functions.awk (switch_flags): Map both Optimization and
41670         PerFunction to CL_OPTIMIZATION.
41671         * opth-gen.awk: Test for PerFunction flag along with
41672         Optimization.
41673         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41674         it only when the latter is present.  Skip those that don't in
41675         the hash function generator.
41676         * common.opt (fvar-tracking): Mark as PerFunction instead of
41677         Optimization.
41678         (fvar-tracking-assignments): Likewise.
41679         (fvar-tracking-assignments-toggle): Likewise.
41680         (fvar-tracking-uninit): Likewise.
41682 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41684         PR translation/79018
41685         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41686         the and store.
41688 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41690         PR target/57583
41691         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41692         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41693         TARGET_LONG_JUMP_TABLE_OFFSETS.
41694         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41695         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41696         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41697         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41698         * config/m68k/m68k.md (tablejump expander): Likewise.
41699         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41700         TARGET_LONG_JUMP_TABLE_OFFSETS.
41701         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41702         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41704 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41705             David Holsgrove <david.holsgrove@xilinx.com>
41707         * common/config/microblaze/microblaze-common.c
41708         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41709         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41710         New prototype.
41711         * config/microblaze/microblaze.c (microblaze_must_save_register)
41712         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41713         calls_eh_return.
41714         (microblaze_eh_return): New function.
41715         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41716         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41717         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41718         * config/microblaze/microblaze.md (eh_return): New pattern.
41720 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41722         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41723         GCC_DIAGNOSTIC_STRINGIFY): Define.
41725         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41727 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41729         * config/arm/arm.md (<mcrr>): New.
41730         (<mrrc>): New.
41731         * config/arm/arm.c (arm_arch5te): New.
41732         (arm_option_override): Set arm_arch5te.
41733         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41734         and mrrc2.
41735         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41736         (arm_mcrr_qualifiers): ... this. New.
41737         (MRRC_QUALIFIERS): Define to...
41738         (arm_mrrc_qualifiers): ... this. New.
41739         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41740         __arm_mrrc2): New.
41741         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41742         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41743         (MRRCI, mrrc, MRRC): New.
41744         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41745         VUNSPEC_MRRC2): New.
41747 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41749         * config/arm/arm.md (<mcr>): New.
41750         (<mrc>): New.
41751         * config/arm/arm.c (arm_coproc_builtin_available): Add
41752         support for mcr, mrc, mcr2 and mrc2.
41753         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41754         (arm_mcr_qualifiers): ... this. New.
41755         (MRC_QUALIFIERS): Define to ...
41756         (arm_mrc_qualifiers): ... this. New.
41757         (MCR_QUALIFIERS): Define to ...
41758         (arm_mcr_qualifiers): ... this. New.
41759         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41760         __arm_mrc2): New.
41761         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41762         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41763         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41764         VUNSPEC_MRC2): New.
41766 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41768         * config/arm/arm.md (*ldc): New.
41769         (*stc): New.
41770         (<ldc>): New.
41771         (<stc>): New.
41772         * config/arm/arm.c (arm_coproc_builtin_available): Add
41773         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41774         (arm_coproc_ldc_stc_legitimate_address): New.
41775         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41776         'qualifier_const_pointer'.
41777         (LDC_QUALIFIERS): Define to...
41778         (arm_ldc_qualifiers): ... this. New.
41779         (STC_QUALIFIERS): Define to...
41780         (arm_stc_qualifiers): ... this. New.
41781         * config/arm/arm-protos.h
41782         (arm_coproc_ldc_stc_legitimate_address): New.
41783         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41784         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41785         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41786         stc2, stcl, stc2l): New.
41787         * config/arm/constraints.md (Uz): New.
41788         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41789         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41790         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41791         VUNSPEC_STC2L): New.
41793 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41795         * config/arm/arm.md (<cdp>): New.
41796         * config/arm/arm.c (neon_const_bounds): Rename this ...
41797         (arm_const_bounds): ... this.
41798         (arm_coproc_builtin_available): New.
41799         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41800         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41801         (CDP_QUALIFIERS): Define to...
41802         (arm_cdp_qualifiers): ... this. New.
41803         (void_UP): Define.
41804         (arm_expand_builtin_args): Add case for 6 arguments.
41805         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41806         (arm_const_bounds): ... this.
41807         (arm_coproc_builtin_available): New.
41808         * config/arm/arm_acle.h (__arm_cdp): New.
41809         (__arm_cdp2): New.
41810         * config/arm/arm_acle_builtins.def (cdp): New.
41811         (cdp2): New.
41812         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41813         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41814         'arm_const_bounds'.
41815         * config/arm/types.md (coproc): New.
41816         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41817         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41818         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41819         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41821 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41823         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41824         (UBINOP_QUALIFIERS): New.
41825         (si_UP): Define.
41826         (acle_builtin_data): New. Change comment.
41827         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41828         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41829         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41830         arm_acle_builtins.def.
41831         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41832         (arm_init_acle_builtins): New.
41833         (CRC32_BUILTIN): Remove.
41834         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41835         crc32cb, crc32ch and crc32cw.
41836         (arm_init_crc32_builtins): Remove.
41837         (arm_init_builtins): Use arm_init_acle_builtins rather
41838         than arm_init_crc32_builtins.
41839         (arm_expand_acle_builtin): New.
41840         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41841         * config/arm/arm_acle_builtins.def: New.
41843 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41845         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41846         (arm_builtin_datum): ... this.
41847         (arm_init_neon_builtin): Rename to ...
41848         (arm_init_builtin): ... this. Add a new parameters PREFIX
41849         and USE_SIG_IN_NAME.
41850         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41851         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41852         'arm_builtin_datum'.
41853         (arm_init_vfp_builtins): Likewise.
41854         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41855         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41856         (arm_expand_neon_args): Rename to ...
41857         (arm_expand_builtin_args): ... this. Rename builtin_arg
41858         enum values and differentiate between ARG_BUILTIN_MEMORY
41859         and ARG_BUILTIN_NEON_MEMORY.
41860         (arm_expand_neon_builtin_1): Rename to ...
41861         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41862         values, arm_expand_builtin_args and add bool parameter NEON.
41863         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41864         (arm_expand_vfp_builtin): Likewise.
41865         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
41867 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41869         PR middle-end/77484
41870         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
41871         * predict.c (tree_estimate_probability_bb): Reverse direction of
41872         polymorphic call predictor.
41874 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
41876         * passes.c (execute_one_pass): Split out pass-skipping logic into...
41877         (determine_pass_name_match): ...this new function and...
41878         (should_skip_pass_p): ...this new function.
41880 2017-01-06  Nathan Sidwell  <nathan@acm.org>
41882         * ipa-visibility.c (function_and_variable_visibility): Reformat
41883         comments and long lines.  Remove extrneous if.
41884         * symtab.c (symtab_node::make_decl_local): Fix code format.
41885         (symtab_node::set_section_for_node): Fix comment typo.
41887 2017-01-06  Martin Liska  <mliska@suse.cz>
41889         PR bootstrap/79003
41890         * lra-constraints.c: Rename invariant to lra_invariant.
41891         * predict.c (set_even_probabilities): Initialize e to NULL.
41893 2017-01-05  Martin Sebor  <msebor@redhat.com>
41895         PR tree-optimization/78910
41896         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
41897         (format_integer): Correct off-by-one error in the handling
41898         of precision with negative numbers in signed conversions..
41900 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
41902         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
41904 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
41906         PR tree-optimization/71016
41907         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
41908         factor_out_conditional_conversion.  Formatting fix.
41909         (factor_out_conditional_conversion): Add cond_stmt argument.
41910         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
41911         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
41912         Formatting fix.
41914 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
41916         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
41917         read-rtl-function.o, and selftest-rtl.o.
41918         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
41919         (selftest::aarch64_test_loading_full_dump): New function.
41920         (selftest::aarch64_run_selftests): New function.
41921         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
41922         selftest::aarch64_run_selftests.
41923         * config/i386/i386.c
41924         (selftest::ix86_test_loading_dump_fragment_1): New function.
41925         (selftest::ix86_test_loading_call_insn): New function.
41926         (selftest::ix86_test_loading_full_dump): New function.
41927         (selftest::ix86_test_loading_unspec): New function.
41928         (selftest::ix86_run_selftests): Call the new functions.
41929         * emit-rtl.c (maybe_set_max_label_num): New function.
41930         * emit-rtl.h (maybe_set_max_label_num): New decl.
41931         * function.c (instantiate_decls): Guard call to
41932         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
41933         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
41934         "static".
41935         * gensupport.c (gen_reader::gen_reader): Pass "false"
41936         for new "compact" param of rtx_reader.
41937         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
41938         rather than an empty string for NULL strings.
41939         * read-md.c: Potentially include config.h rather than bconfig.h.
41940         Wrap include of errors.h with #ifdef GENERATOR_FILE.
41941         (have_error): New global, copied from errors.c.
41942         (md_reader::read_name): Rename to...
41943         (md_reader::read_name_1): ...this, adding "out_loc" param,
41944         and converting "missing name or number" to returning false, rather
41945         than failing.
41946         (md_reader::read_name): Reimplement in terms of read_name_1.
41947         (md_reader::read_name_or_nil): New function.
41948         (md_reader::read_string): Handle "(nil)" by returning NULL.
41949         (md_reader::md_reader): Add new param "compact".
41950         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
41951         (md_reader::read_file): New method.
41952         * read-md.h (md_reader::md_reader): Add new param "compact".
41953         (md_reader::read_file): New method.
41954         (md_reader::is_compact): New accessor.
41955         (md_reader::read_name): Convert return type from void to file_location.
41956         (md_reader::read_name_or_nil): New decl.
41957         (md_reader::read_name_1): New decl.
41958         (md_reader::m_compact): New field.
41959         (noop_reader::noop_reader): Pass "false" for new "compact" param
41960         of rtx_reader.
41961         (rtx_reader::rtx_reader): Add new "compact" param.
41962         (rtx_reader::read_rtx_operand): Make virtual and convert return
41963         type from void to rtx.
41964         (rtx_reader::read_until): New decl.
41965         (rtx_reader::handle_any_trailing_information): New virtual function.
41966         (rtx_reader::postprocess): New virtual function.
41967         (rtx_reader::finalize_string): New virtual function.
41968         (rtx_reader::m_in_call_function_usage): New field.
41969         (rtx_reader::m_reuse_rtx_by_id): New field.
41970         * read-rtl-function.c: New file.
41971         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
41972         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
41973         (selftest::verify_three_block_rtl_cfg): New decl.
41974         * read-rtl-function.h: New file.
41975         * read-rtl.c: Potentially include config.h rather than bconfig.h.
41976         For host, include function.h, memmodel.h, and emit-rtl.h.
41977         (one_time_initialization): New function.
41978         (struct compact_insn_name): New struct.
41979         (compact_insn_names): New array.
41980         (find_code): Handle insn codes in compact dumps.
41981         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
41982         (bind_subst_iter_and_attr): Likewise.
41983         (add_condition_to_string): Likewise.
41984         (add_condition_to_rtx): Likewise.
41985         (apply_attribute_uses): Likewise.
41986         (add_current_iterators): Likewise.
41987         (apply_iterators): Likewise.
41988         (initialize_iterators): Guard usage of apply_subst_iterator with
41989         #ifdef GENERATOR_FILE.
41990         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
41991         (md_reader::read_mapping): Likewise.
41992         (add_define_attr_for_define_subst): Likewise.
41993         (add_define_subst_attr): Likewise.
41994         (read_subst_mapping): Likewise.
41995         (check_code_iterator): Likewise.
41996         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
41997         logic to...
41998         (one_time_initialization): New function.
41999         (rtx_reader::read_until): New method.
42000         (read_flags): New function.
42001         (parse_reg_note_name): New function.
42002         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42003         Handle reuse_rtx ids.
42004         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42005         Add parsing support for RTL dumps, mirroring the special-cases in
42006         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42007         values, and calling handle_any_trailing_information.
42008         (rtx_reader::read_rtx_operand): Convert return type from void
42009         to rtx, returning return_rtx.  Handle case 'e'.  Call
42010         finalize_string on XSTR and XTMPL fields.
42011         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42012          "(nil)" values were omitted.  Call the postprocess vfunc on the
42013         return_rtx.
42014         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42015         class ctor.  Initialize m_in_call_function_usage.  Call
42016         one_time_initialization.
42017         * rtl-tests.c (selftest::test_uncond_jump): Call
42018         set_new_first_and_last_insn.
42019         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42020         * selftest-rtl.c: New file.
42021         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42022         (selftest::get_insn_by_uid): New decl.
42023         * selftest-run-tests.c (selftest::run_tests): Call
42024         read_rtl_function_c_tests.
42025         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42026         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42027         dumps.
42029 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42031         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42032         operands in a special way.  Assert that pos+len <= mode precision.
42034 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42036         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42037         3 argument Alias with unlimited for the negative form.
42038         (fno-vect-cost-model): Removed.
42040 2017-01-05  Martin Liska  <mliska@suse.cz>
42042         * hsa-gen.c (gen_hsa_divmod): New function.
42043         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42045 2017-01-05  Martin Liska  <mliska@suse.cz>
42047         PR pch/78970
42048         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42049         header.
42051 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42053         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42054         small constant length operands.
42056 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42058         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42059         between loop iterations.
42061 2017-01-05  Martin Liska  <mliska@suse.cz>
42063         PR sanitizer/78815
42064         * gimplify.c (gimplify_decl_expr): Compare to
42065         asan_poisoned_variables instread of checking flags.
42066         (gimplify_target_expr): Likewise.
42067         (gimplify_expr): Likewise.
42068         (gimplify_function_tree): Conditionally initialize
42069         asan_poisoned_variables.
42071 2017-01-04  Jeff Law  <law@redhat.com>
42073         PR tree-optimizatin/78812
42074         * rtl.h (contains_mem_rtx_p): Prototype.
42075         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42076         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42077         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42078         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42079         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42081 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42083         * input.c (assert_char_at_range): Default-initialize actual_range.
42085 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42087         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42088         to match the caller.
42090 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42092         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42093         insns after final jump in test to emit dummy move.
42095 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42097         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42098         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42100 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42102         * multiple_target.c (create_dispatcher_calls): Init e_next.
42103         * tree-ssa-loop-split.c (split_loop): Init border.
42104         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42105         scalar_type.
42107 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42109         PR target/71977
42110         PR target/70568
42111         PR target/78823
42112         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42113         (altivec_register_operand): Do not return true if the operand
42114         contains a SUBREG mixing SImode and SFmode.
42115         (vsx_register_operand): Likewise.
42116         (vsx_reg_sfsubreg_ok): New predicate.
42117         (vfloat_operand): Do not return true if the operand contains a
42118         SUBREG mixing SImode and SFmode.
42119         (vint_operand): Likewise.
42120         (vlogical_operand): Likewise.
42121         (gpc_reg_operand): Likewise.
42122         (int_reg_operand): Likewise.
42123         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42124         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42125         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42126         SImode and SFmode.
42127         (rs6000_emit_move_si_sf_subreg): New helper function.
42128         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42129         fixup SUBREGs involving SImode and SFmode.
42130         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42131         numbers for the new peephole2 optimization.
42132         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42133         the GLIBC math library that do AND/IOR/XOR operations on single
42134         precision floating point.
42135         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42136         target macros to say whether we need to avoid SUBREGs mixing
42137         SImode and SFmode.
42138         (TARGET_ALLOW_SF_SUBREG): Likewise.
42139         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42140         (UNSPEC_SI_FROM_SF): Likewise.
42141         (iorxor): Change spacing.
42142         (and_ior_xor): New iterator for AND, IOR, and XOR.
42143         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42144         (movdi_from_sf_zero_ext): Likewise.
42145         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42146         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42147         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42148         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42149         (fms<mode>4): Likewise.
42150         (fnma<mode>4): Likewise.
42151         (fnms<mode>4): Likewise.
42152         (nfma<mode>4): Likewise.
42153         (nfms<mode>4): Likewise.
42155 2017-01-04  Marek Polacek  <polacek@redhat.com>
42157         PR c++/64767
42158         * doc/invoke.texi: Document -Wpointer-compare.
42160 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42162         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42163         RejectNegative.
42165         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42166         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42167         2-byte data.
42169 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42171         PR target/78056
42172         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42173         documentation of the powerpc_popcntb_ok attribute.
42174         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42175         code to issue warning messages if a requested CPU configuration is
42176         not supported by the binary (assembler and loader) toolchain.
42177         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42178         made to define a built-in function that has been disabled.
42179         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42180         made to define a built-in function that has been disabled.
42181         (altivec_init_builtins): Add comment explaining why definition
42182         of the DST built-in functions is not preceded by an assertion
42183         check.  Add assertions to prevent ICE if attempts are made to
42184         define an altivec predicate or an abs* built-in function that has
42185         been disabled.
42186         (htm_init_builtins): Add comment explaining why definition of the
42187         htm built-in functions is not preceded by an assertion check.
42189 2017-01-04  Jeff Law  <law@redhat.com>
42191         PR tree-optimizatin/67955
42192         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42193         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42194         the points-to solution does not include pt_null.  Use DECL_PT_UID
42195         unconditionally.
42197 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42199         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42200         Use gen_int_mode instead of gen_lopwart for const_int operands.
42202 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42204         PR tree-optimization/71563
42205         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42206         out of range value - has low bits known to be zero.
42208 2017-01-04  Alan Modra  <amodra@gmail.com>
42210         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42211         * configure: Regenerate.
42212         * config.in: Regenerate.
42214 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42216         PR bootstrap/77569
42217         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42218         a substring of the message, but strcmp with the whole message.  Ifdef
42219         ENABLE_NLS, translate the message first using dgettext.
42221 2017-01-03  Jeff Law  <law@redhat.com>
42223         PR tree-optimizatin/78856
42224         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42225         (mark_threaded_blocks): Remove code to truncate thread paths that
42226         cross multiple loop headers.  Instead invalidate the cached loop
42227         iteration information and handle case of a thread path walking
42228         into an irreducible region.
42230 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42232         PR target/78900
42233         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42234         assertions.  Add support for doing the signbit if the IEEE 128-bit
42235         floating point value is in a GPR.
42236         * config/rs6000/rs6000.md (Fsignbit): Delete.
42237         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42238         Update the length attribute if the value is in a GPR.
42239         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42240         the sign or zero extension instruction, since the value is always 0/1.
42241         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42243         PR target/78953
42244         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42245         extracting SImode to a GPR register so that we can generate a
42246         store, limit the vector to be in a traditional Altivec register
42247         for the vextuwrx instruction.
42249 2017-01-03  Ian Lance Taylor  <iant@google.com>
42251         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42253 2017-01-03  Martin Sebor  <msebor@redhat.com>
42255         PR tree-optimization/78696
42256         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42257         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42258         of %g.
42259         (pass_sprintf_length::compute_format_length): Set width and precision
42260         specified by asrerisk to void_node for vararg functions.
42261         (try_substitute_return_value): Adjust dump output.
42263 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42265         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42267 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42269         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42270         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42271         -mlra/-mno-lra was passed to the compiler.
42273 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42275         PR rtl-optimization/65618
42276         * emit-rtl.c (try_split): Move initialization of "before" and
42277         "after" to just before the call to emit_insn_after_setloc.
42279 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42281         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42283 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42285         * dwarf2out.c (gen_enumeration_type_die): When
42286         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42288 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42290         PR tree-optimization/78965
42291         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42292         Change first argument from const call_info & to call_info &.  For %n
42293         set info.nowrite to false.
42295         PR middle-end/78901
42296         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42297         possibly throwing calls.
42299         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42300         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42301         and fns handling, rather than in a separate case SSA_NAME.
42303 2017-01-02  Jeff Law  <law@redhat.com>
42305         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42306         fixes for first_period and second_period variables.
42308 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42310         PR target/78967
42311         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42312         (*insvqi_1): New insn pattern.
42313         (*insvqi_1_mem_rex64): Ditto.
42314         (*insvqi_2): Ditto.
42315         (*insvqi_3): Rename from *insvqi.
42317         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42319 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42321         * doc/cfg.texi (Edges): Remove reference to Java.
42322         (Maintaining the CFG): Ditto.
42324 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42326         PR middle-end/77674
42327         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42328         transparent aliases.
42330 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42332         PR middle-end/77484
42333         * predict.def (PRED_CALL): Update hitrate.
42334         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42335         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42336         into direct/indirect/polymorphic variants.
42338 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42340         Update copyright years.
42342         * gcc.c (process_command): Update copyright notice dates.
42343         * gcov-dump.c (print_version): Ditto.
42344         * gcov.c (print_version): Ditto.
42345         * gcov-tool.c (print_version): Ditto.
42346         * gengtype.c (create_file): Ditto.
42347         * doc/cpp.texi: Bump @copying's copyright year.
42348         * doc/cppinternals.texi: Ditto.
42349         * doc/gcc.texi: Ditto.
42350         * doc/gccint.texi: Ditto.
42351         * doc/gcov.texi: Ditto.
42352         * doc/install.texi: Ditto.
42353         * doc/invoke.texi: Ditto.
42355 Copyright (C) 2017 Free Software Foundation, Inc.
42357 Copying and distribution of this file, with or without modification,
42358 are permitted in any medium without royalty provided the copyright
42359 notice and this notice are preserved.