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 (_slp_instance::unrolling_factor): Change
6 from an unsigned int to a poly_uint64.
7 (_loop_vec_info::slp_unrolling_factor): Likewise.
8 (_loop_vec_info::vectorization_factor): Change from an int
10 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11 (vect_get_num_vectors): New function.
12 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13 (vect_get_num_copies): Use vect_get_num_vectors.
14 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16 (vect_analyze_data_refs): Change min_vf from an int * to a
18 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19 than an unsigned HOST_WIDE_INT.
20 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
21 (vect_analyze_data_ref_dependence): Change max_vf from an int *
23 (vect_analyze_data_ref_dependences): Likewise.
24 (vect_compute_data_ref_alignment): Handle polynomial vf.
25 (vect_enhance_data_refs_alignment): Likewise.
26 (vect_prune_runtime_alias_test_list): Likewise.
27 (vect_shift_permute_load_chain): Likewise.
28 (vect_supportable_dr_alignment): Likewise.
29 (dependence_distance_ge_vf): Take the vectorization factor as a
30 poly_uint64 rather than an unsigned HOST_WIDE_INT.
31 (vect_analyze_data_refs): Change min_vf from an int * to a
33 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
34 vfm1 as a poly_uint64 rather than an int. Make the same change
35 for the returned bound_scalar.
36 (vect_gen_vector_loop_niters): Handle polynomial vf.
37 (vect_do_peeling): Likewise. Update call to
38 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
39 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
41 * tree-vect-loop.c (vect_determine_vectorization_factor)
42 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
43 (vect_get_known_peeling_cost): Likewise.
44 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
45 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
46 (vect_transform_loop): Likewise. Use the lowest possible VF when
47 updating the upper bounds of the loop.
48 (vect_min_worthwhile_factor): Make static. Return an unsigned int
50 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
51 polynomial unroll factors.
52 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
53 (vect_make_slp_decision): Likewise.
54 (vect_supported_load_permutation_p): Likewise, and polynomial
56 (vect_analyze_slp_cost): Handle polynomial vf.
57 (vect_slp_analyze_node_operations): Likewise.
58 (vect_slp_analyze_bb_1): Likewise.
59 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
60 than an unsigned HOST_WIDE_INT.
61 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
62 (vectorizable_load): Handle polynomial vf.
63 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
65 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
67 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
68 Alan Hayward <alan.hayward@arm.com>
69 David Sherwood <david.sherwood@arm.com>
71 * match.pd: Handle bit operations involving three constants
72 and try to fold one pair.
74 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
76 * tree-vect-loop-manip.c: Include gimple-fold.h.
77 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
78 niters_maybe_zero parameters. Handle other cases besides a step of 1.
79 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
80 Add a path that uses a step of VF instead of 1, but disable it
82 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
83 and niters_no_overflow parameters. Update calls to
84 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
85 Create a new SSA name if the latter choses to use a ste other
86 than zero, and return it via niters_vector_mult_vf_var.
87 * tree-vect-loop.c (vect_transform_loop): Update calls to
88 vect_do_peeling, vect_gen_vector_loop_niters and
89 slpeel_make_loop_iterate_ntimes.
90 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
91 (vect_gen_vector_loop_niters): Update declarations after above changes.
93 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
95 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
96 128-bit round to integer instructions.
97 (ceil<mode>2): Likewise.
98 (btrunc<mode>2): Likewise.
99 (round<mode>2): Likewise.
101 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
103 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
104 unaligned VSX load/store on P8/P9.
105 (expand_block_clear): Allow the use of unaligned VSX
108 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
110 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
112 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
113 swap associated with both a load and a store.
115 2018-01-02 Andrew Waterman <andrew@sifive.com>
117 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
118 * config/riscv/riscv.md (clear_cache): Use it.
120 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
122 * web.c: Remove out-of-date comment.
124 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
126 * expr.c (fixup_args_size_notes): Check that any existing
127 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
128 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
129 (emit_single_push_insn): ...here.
131 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
133 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
134 (const_vector_encoded_nelts): New function.
135 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
136 (const_vector_int_elt, const_vector_elt): Declare.
137 * emit-rtl.c (const_vector_int_elt_1): New function.
138 (const_vector_elt): Likewise.
139 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
144 * expr.c: Include rtx-vector-builder.h.
145 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
146 directly on the tree encoding.
147 (const_vector_from_tree): Likewise.
148 * optabs.c: Include rtx-vector-builder.h.
149 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
150 sequence of "u" values.
151 * vec-perm-indices.c: Include rtx-vector-builder.h.
152 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
153 directly on the vec_perm_indices encoding.
155 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
157 * doc/rtl.texi (const_vector): Describe new encoding scheme.
158 * Makefile.in (OBJS): Add rtx-vector-builder.o.
159 * rtx-vector-builder.h: New file.
160 * rtx-vector-builder.c: Likewise.
161 * rtl.h (rtx_def::u2): Add a const_vector field.
162 (CONST_VECTOR_NPATTERNS): New macro.
163 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
164 (CONST_VECTOR_DUPLICATE_P): Likewise.
165 (CONST_VECTOR_STEPPED_P): Likewise.
166 (CONST_VECTOR_ENCODED_ELT): Likewise.
167 (const_vec_duplicate_p): Check for a duplicated vector encoding.
168 (unwrap_const_vec_duplicate): Likewise.
169 (const_vec_series_p): Check for a non-duplicated vector encoding.
170 Say that the function only returns true for integer vectors.
171 * emit-rtl.c: Include rtx-vector-builder.h.
172 (gen_const_vec_duplicate_1): Delete.
173 (gen_const_vector): Call gen_const_vec_duplicate instead of
174 gen_const_vec_duplicate_1.
175 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
176 (gen_const_vec_duplicate): Use rtx_vector_builder.
177 (gen_const_vec_series): Likewise.
178 (gen_rtx_CONST_VECTOR): Likewise.
179 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
180 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
181 Build a new vector rather than modifying a CONST_VECTOR in-place.
182 (handle_special_swappables): Update call accordingly.
183 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
184 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
185 Build a new vector rather than modifying a CONST_VECTOR in-place.
186 (handle_special_swappables): Update call accordingly.
188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
190 * simplify-rtx.c (simplify_const_binary_operation): Use
191 CONST_VECTOR_ELT instead of XVECEXP.
193 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
195 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
196 the selector elements to be different from the data elements
197 if the selector is a VECTOR_CST.
198 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
199 ssizetype for the selector.
201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
203 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
204 before testing each element individually.
205 * tree-vect-generic.c (lower_vec_perm): Likewise.
207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
209 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
210 * selftest-run-tests.c (selftest::run_tests): Call it.
211 * vector-builder.h (vector_builder::operator ==): New function.
212 (vector_builder::operator !=): Likewise.
213 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
214 (vec_perm_indices::all_from_input_p): New function.
215 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
216 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
217 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
218 instead of reading the VECTOR_CST directly. Detect whether both
219 vector inputs are the same before constructing the vec_perm_indices,
220 and update the number of inputs argument accordingly. Use the
221 utility functions added above. Only construct sel2 if we need to.
223 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
225 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
226 the broadcast of the low byte.
227 (expand_mult_highpart): Use an explicit encoding for the permutes.
228 * optabs-query.c (can_mult_highpart_p): Likewise.
229 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
230 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
231 (vectorizable_bswap): Likewise.
232 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
233 explicit encoding for the power-of-2 permutes.
234 (vect_permute_store_chain): Likewise.
235 (vect_grouped_load_supported): Likewise.
236 (vect_permute_load_chain): Likewise.
238 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
240 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
241 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
242 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
243 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
244 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
245 (vect_gen_perm_mask_any): Likewise.
247 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
249 * int-vector-builder.h: New file.
250 * vec-perm-indices.h: Include int-vector-builder.h.
251 (vec_perm_indices): Redefine as an int_vector_builder.
252 (auto_vec_perm_indices): Delete.
253 (vec_perm_builder): Redefine as a stand-alone class.
254 (vec_perm_indices::vec_perm_indices): New function.
255 (vec_perm_indices::clamp): Likewise.
256 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
257 (vec_perm_indices::new_vector): New function.
258 (vec_perm_indices::new_expanded_vector): Update for new
259 vec_perm_indices class.
260 (vec_perm_indices::rotate_inputs): New function.
261 (vec_perm_indices::all_in_range_p): Operate directly on the
262 encoded form, without computing elided elements.
263 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
264 encoding. Update for new vec_perm_indices class.
265 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
266 the given vec_perm_builder.
267 (expand_vec_perm_var): Update vec_perm_builder constructor.
268 (expand_mult_highpart): Use vec_perm_builder instead of
269 auto_vec_perm_indices.
270 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
271 vec_perm_indices instead of auto_vec_perm_indices. Use a single
272 or double series encoding as appropriate.
273 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
274 vec_perm_indices instead of auto_vec_perm_indices.
275 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
276 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
277 (vect_permute_store_chain): Likewise.
278 (vect_grouped_load_supported): Likewise.
279 (vect_permute_load_chain): Likewise.
280 (vect_shift_permute_load_chain): Likewise.
281 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
282 (vect_transform_slp_perm_load): Likewise.
283 (vect_schedule_slp_instance): Likewise.
284 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
285 (vectorizable_mask_load_store): Likewise.
286 (vectorizable_bswap): Likewise.
287 (vectorizable_store): Likewise.
288 (vectorizable_load): Likewise.
289 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
290 vec_perm_indices instead of auto_vec_perm_indices. Use
291 tree_to_vec_perm_builder to read the vector from a tree.
292 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
293 vec_perm_builder instead of a vec_perm_indices.
294 (have_whole_vector_shift): Use vec_perm_builder and
295 vec_perm_indices instead of auto_vec_perm_indices. Leave the
296 truncation to calc_vec_perm_mask_for_shift.
297 (vect_create_epilog_for_reduction): Likewise.
298 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
299 from auto_vec_perm_indices to vec_perm_indices.
300 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
301 instead of changing individual elements.
302 (aarch64_vectorize_vec_perm_const): Use new_vector to install
303 the vector in d.perm.
304 * config/arm/arm.c (expand_vec_perm_d::perm): Change
305 from auto_vec_perm_indices to vec_perm_indices.
306 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
307 instead of changing individual elements.
308 (arm_vectorize_vec_perm_const): Use new_vector to install
309 the vector in d.perm.
310 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
311 Update vec_perm_builder constructor.
312 (rs6000_expand_interleave): Likewise.
313 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
314 (rs6000_expand_interleave): Likewise.
316 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
318 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
319 to qimode could truncate the indices.
320 * optabs.c (expand_vec_perm_var): Likewise.
322 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
324 * Makefile.in (OBJS): Add vec-perm-indices.o.
325 * vec-perm-indices.h: New file.
326 * vec-perm-indices.c: Likewise.
327 * target.h (vec_perm_indices): Replace with a forward class
329 (auto_vec_perm_indices): Move to vec-perm-indices.h.
330 * optabs.h: Include vec-perm-indices.h.
331 (expand_vec_perm): Delete.
332 (selector_fits_mode_p, expand_vec_perm_var): Declare.
333 (expand_vec_perm_const): Declare.
334 * target.def (vec_perm_const_ok): Replace with...
335 (vec_perm_const): ...this new hook.
336 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
337 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
338 * doc/tm.texi: Regenerate.
339 * optabs.def (vec_perm_const): Delete.
340 * doc/md.texi (vec_perm_const): Likewise.
341 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
342 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
343 expand_vec_perm for constant permutation vectors. Assert that
344 the mode of variable permutation vectors is the integer equivalent
345 of the mode that is being permuted.
346 * optabs-query.h (selector_fits_mode_p): Declare.
347 * optabs-query.c: Include vec-perm-indices.h.
348 (selector_fits_mode_p): New function.
349 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
350 is defined, instead of checking whether the vec_perm_const_optab
351 exists. Use targetm.vectorize.vec_perm_const instead of
352 targetm.vectorize.vec_perm_const_ok. Check whether the indices
353 fit in the vector mode before using a variable permute.
354 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
355 vec_perm_indices instead of an rtx.
356 (expand_vec_perm): Replace with...
357 (expand_vec_perm_const): ...this new function. Take the selector
358 as a vec_perm_indices rather than an rtx. Also take the mode of
359 the selector. Update call to shift_amt_for_vec_perm_mask.
360 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
361 Use vec_perm_indices::new_expanded_vector to expand the original
362 selector into bytes. Check whether the indices fit in the vector
363 mode before using a variable permute.
364 (expand_vec_perm_var): Make global.
365 (expand_mult_highpart): Use expand_vec_perm_const.
366 * fold-const.c: Includes vec-perm-indices.h.
367 * tree-ssa-forwprop.c: Likewise.
368 * tree-vect-data-refs.c: Likewise.
369 * tree-vect-generic.c: Likewise.
370 * tree-vect-loop.c: Likewise.
371 * tree-vect-slp.c: Likewise.
372 * tree-vect-stmts.c: Likewise.
373 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
375 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
376 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
377 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
378 (aarch64_vectorize_vec_perm_const): ...this new function.
379 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
380 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
381 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
382 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
383 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
384 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
385 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
387 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
388 check for NEON modes.
389 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
390 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
391 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
392 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
394 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
395 the old VEC_PERM_CONST conditions.
396 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
397 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
398 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
399 (ia64_vectorize_vec_perm_const_ok): Merge into...
400 (ia64_vectorize_vec_perm_const): ...this new function.
401 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
402 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
403 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
404 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
405 * config/mips/mips.c (mips_expand_vec_perm_const)
406 (mips_vectorize_vec_perm_const_ok): Merge into...
407 (mips_vectorize_vec_perm_const): ...this new function.
408 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
409 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
410 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
411 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
412 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
413 (rs6000_expand_vec_perm_const): Delete.
414 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
416 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
417 (altivec_expand_vec_perm_const_le): Take each operand individually.
418 Operate on constant selectors rather than rtxes.
419 (altivec_expand_vec_perm_const): Likewise. Update call to
420 altivec_expand_vec_perm_const_le.
421 (rs6000_expand_vec_perm_const): Delete.
422 (rs6000_vectorize_vec_perm_const_ok): Delete.
423 (rs6000_vectorize_vec_perm_const): New function.
424 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
425 an element count and rtx array.
426 (rs6000_expand_extract_even): Update call accordingly.
427 (rs6000_expand_interleave): Likewise.
428 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
429 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
430 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
431 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
432 (rs6000_expand_vec_perm_const): Delete.
433 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
434 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
435 (altivec_expand_vec_perm_const_le): Take each operand individually.
436 Operate on constant selectors rather than rtxes.
437 (altivec_expand_vec_perm_const): Likewise. Update call to
438 altivec_expand_vec_perm_const_le.
439 (rs6000_expand_vec_perm_const): Delete.
440 (rs6000_vectorize_vec_perm_const_ok): Delete.
441 (rs6000_vectorize_vec_perm_const): New function. Remove stray
442 reference to the SPE evmerge intructions.
443 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
444 an element count and rtx array.
445 (rs6000_expand_extract_even): Update call accordingly.
446 (rs6000_expand_interleave): Likewise.
447 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
448 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
450 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
452 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
454 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
455 vector mode and that that mode matches the mode of the data
457 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
458 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
459 directly using expand_vec_perm_1 when forcing selectors into
461 (expand_vec_perm_var): New function, split out from expand_vec_perm.
463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
465 * optabs-query.h (can_vec_perm_p): Delete.
466 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
467 * optabs-query.c (can_vec_perm_p): Split into...
468 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
469 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
470 particular selector is valid.
471 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
472 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
473 (vect_grouped_load_supported): Likewise.
474 (vect_shift_permute_load_chain): Likewise.
475 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
476 (vect_transform_slp_perm_load): Likewise.
477 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
478 (vectorizable_bswap): Likewise.
479 (vect_gen_perm_mask_checked): Likewise.
480 * fold-const.c (fold_ternary_loc): Likewise. Don't take
481 implementations of variable permutation vectors into account
482 when deciding which selector to use.
483 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
484 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
485 with a false third argument.
486 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
487 to test whether the constant selector is valid and can_vec_perm_var_p
488 to test whether a variable selector is valid.
490 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
492 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
493 * optabs-query.c (can_vec_perm_p): Likewise.
494 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
495 instead of vec_perm_indices.
496 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
497 (vect_gen_perm_mask_checked): Likewise,
498 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
499 (vect_gen_perm_mask_checked): Likewise,
501 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
503 * optabs-query.h (qimode_for_vec_perm): Declare.
504 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
505 (qimode_for_vec_perm): ...this new function.
506 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
508 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
510 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
511 does not have a conditional at the top.
513 2018-01-02 Richard Biener <rguenther@suse.de>
515 * ipa-inline.c (big_speedup_p): Fix expression.
517 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
520 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
523 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
527 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
528 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
529 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
530 cond_taken_branch_cost 3->4.
532 2018-01-01 Jakub Jelinek <jakub@redhat.com>
534 PR tree-optimization/83581
535 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
536 TODO_cleanup_cfg if any changes have been made.
539 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
540 convert_modes if target mode has the right side, but different mode
544 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
545 last argument when extracting from CONCAT. If either from_real or
546 from_imag is NULL, use expansion through memory. If result is not
547 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
548 the parts directly to inner mode, if even that fails, use expansion
552 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
553 check for bswap in mode rather than HImode and use that in expand_unop
556 2017-12-31 Jakub Jelinek <jakub@redhat.com>
559 * config/i386/i386.c (ix86_attribute_table): Remove excess
560 initializer for "shared" attribute.
562 2017-12-30 Tom de Vries <tom@codesourcery.com>
565 * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
567 * lto-streamer-out.c (prune_offload_funcs): New function. Remove
568 offload_funcs entries that no longer have a corresponding cgraph_node.
569 Mark the remaining ones as DECL_PRESERVE_P.
570 (output_lto): Call prune_offload_funcs.
572 2017-12-30 Jakub Jelinek <jakub@redhat.com>
574 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
575 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
576 vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
577 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
578 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
579 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
580 vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
581 vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
582 vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
583 vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
584 vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
585 vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
586 vaesenc_<mode>, vpclmulqdq_<mode>,
587 avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
589 2017-12-28 Michael Meissner <meissner@linux.vnet.ibm.com>
591 * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
592 _Float<N> and _Float<N>X variants for rounding built-in
594 (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
595 (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
596 (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
597 (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
598 (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
599 * builtins.c (mathfn_built_in_2): Likewise.
600 * internal-fn.def (CEIL): Likewise.
602 (NEARBYINT): Likewise.
606 * convert.c (convert_to_integer_1): Likewise.
607 * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
608 (integer_valued_real_call_p): Likewise.
609 * fold-const-call.c (fold_const_call_ss): Likewise.
610 * gencfn-macros.c (print_case_cfn): Change CFN and operator
611 printers to take a const char * suffix instead of a bool.
612 (print_define_operator_list): Likewise.
613 (fltall_suffixes): New list of suffixes, that include the
614 traditional suffixes as well as all of the _Float<N> and
616 (main): For _Float<N> and _Float<N>X functions, emit both
617 <name>_FN and <name>_ALL variants. The <macro>_FN variant only
618 has the _Float<N> and _Float<N>X case names or operators. The
619 <name>_ALL variant has both the traditional and the
620 _Float<N>/_Float<N>X case names or operators.
621 * match.pd (COPYSIGN optimizations): Provide optimizations for
622 _Float<N> and _Float<N>X types where possible.
623 (MIN/MAX optimizations): Likewise.
624 (sqrt optimizations): Likewise.
625 (rounding optimizations): Likewise.
627 2017-12-28 Richard Sandiford <richard.sandiford@linaro.org>
629 * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
630 gen_int_mode rather than GEN_INT.
631 * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
633 2017-12-28 Richard Sandiford <richard.sandiford@linaro.org>
635 * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
636 instead of CONSTANT_P.
637 (gen_vec_series): Likewise.
638 * simplify-rtx.c (simplify_binary_operation_1): Likewise.
640 2017-12-28 Andreas Schwab <schwab@linux-m68k.org>
642 * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
645 2017-12-27 Kugan Vivekanandarajah <kuganv@linaro.org>
647 * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
648 (aarch64_ld1x2<VDC:mode>): Likewise.
649 (aarch64_simd_ld1<mode>_x2): Likewise.
650 (aarch64_simd_ld1<mode>_x2): Likewise.
651 * config/aarch64/arm_neon.h (vld1_u8_x2): New.
652 (vld1_s8_x2): Likewise.
653 (vld1_u16_x2): Likewise.
654 (vld1_s16_x2): Likewise.
655 (vld1_u32_x2): Likewise.
656 (vld1_s32_x2): Likewise.
657 (vld1_u64_x2): Likewise.
658 (vld1_s64_x2): Likewise.
659 (vld1_f16_x2): Likewise.
660 (vld1_f32_x2): Likewise.
661 (vld1_f64_x2): Likewise.
662 (vld1_p8_x2): Likewise.
663 (vld1_p16_x2): Likewise.
664 (vld1_p64_x2): Likewise.
665 (vld1q_u8_x2): Likewise.
666 (vld1q_s8_x2): Likewise.
667 (vld1q_u16_x2): Likewise.
668 (vld1q_s16_x2): Likewise.
669 (vld1q_u32_x2): Likewise.
670 (vld1q_s32_x2): Likewise.
671 (vld1q_u64_x2): Likewise.
672 (vld1q_s64_x2): Likewise.
673 (vld1q_f16_x2): Likewise.
674 (vld1q_f32_x2): Likewise.
675 (vld1q_f64_x2): Likewise.
676 (vld1q_p8_x2): Likewise.
677 (vld1q_p16_x2): Likewise.
678 (vld1q_p64_x2): Likewise.
680 2017-12-27 Martin Liska <mliska@suse.cz>
682 PR tree-optimization/83552
683 * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
684 of get_string_lenth to a SSA_NAME if not a GIMPLE value.
686 2017-12-27 Tom de Vries <tom@codesourcery.com>
688 * config/nvptx/nvptx.c (nvptx_option_override): Disable
689 -gstatement-frontiers.
691 2017-12-26 Alexander Monakov <amonakov@ispras.ru>
693 PR rtl-optimization/83513
694 * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
695 before priority comparison.
697 2017-12-25 Jakub Jelinek <jakub@redhat.com>
700 * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
701 ix86_isa_flags2 to ix86_isa_flags.
702 * config/i386/i386-c.c (ix86_target_macros_internal): Test
703 OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
704 isa_flags rather than isa_flags2.
705 * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
706 and -mavx512bitalg from isa2_opts to isa_opts.
707 (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
708 in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
709 and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
711 (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
712 OPTION_MASK_ISA_AVX512F ored with another option similarly to
713 OPTION_MASK_ISA_AVX512VL. Even for OPTION_MASK_ISA_AVX512VL don't
714 clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
715 (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
716 OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
717 OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
718 * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
719 builtins from bdesc_args2 to bdesc_args section.
720 (__builtin_ia32_compressstoreuqi512_mask,
721 __builtin_ia32_compressstoreuhi512_mask,
722 __builtin_ia32_compressstoreuqi256_mask,
723 __builtin_ia32_expandloadqi512_mask,
724 __builtin_ia32_expandloadqi512_maskz,
725 __builtin_ia32_expandloadhi512_mask,
726 __builtin_ia32_expandloadhi512_maskz,
727 __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
728 __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
729 __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
730 __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
731 __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
732 __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
733 __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
734 __builtin_ia32_vpshldv_v32hi_maskz,
735 __builtin_ia32_vpopcountb_v64qi_mask,
736 __builtin_ia32_vpopcountw_v32hi_mask,
737 __builtin_ia32_vpshufbitqmb512_mask,
738 __builtin_ia32_vpshufbitqmb256_mask): Add
739 " | OPTION_MASK_ISA_AVX512BW".
740 (__builtin_ia32_expandloadqi256_mask,
741 __builtin_ia32_expandloadqi256_maskz,
742 __builtin_ia32_vpopcountb_v32qi_mask): Add
743 " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
744 (__builtin_ia32_expandloadhi256_mask,
745 __builtin_ia32_expandloadhi256_maskz,
746 __builtin_ia32_expandloadqi128_mask,
747 __builtin_ia32_expandloadqi128_maskz,
748 __builtin_ia32_expandloadhi128_mask,
749 __builtin_ia32_expandloadhi128_maskz,
750 __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
751 __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
752 __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
753 __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
754 __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
755 __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
756 __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
757 __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
758 __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
759 __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
760 __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
761 __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
762 __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
763 __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
764 __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
765 __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
766 __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
767 __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
768 __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
769 __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
770 __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
771 __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
772 __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
773 __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
774 __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
775 __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
776 __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
777 __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
778 __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
779 __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
780 __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
781 __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
782 __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
783 __builtin_ia32_vpopcountw_v8hi_mask): Add
784 " | OPTION_MASK_ISA_AVX512VL".
785 * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
786 _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
787 _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
788 _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
789 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
790 _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
791 _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
792 _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
793 _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
794 _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
795 _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
796 * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
797 _mm_mask_bitshuffle_epi64_mask): Likewise.
798 * common/config/i386/i386-common.c
799 (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
800 OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
801 (OPTION_MASK_ISA_AVX512F_UNSET): Or in
802 OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
803 OPTION_MASK_ISA_AVX512BITALG_UNSET.
804 (OPTION_MASK_ISA2_AVX512F_UNSET,
805 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
806 (ix86_handle_option): For -mno-general-regs-only, clear from
807 ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
808 just OPTION_MASK_ISA_MPX. For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
809 -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
810 ix86_isa_flags2. For -mno-avx512f likewise, instead of masking
811 individually listed ISAs. For -m{,no-}avx512{vpopcntdq,bitalg} adjust
812 for moving from ix86_isa_flags2 to ix86_isa_flags.
814 2017-12-23 Jakub Jelinek <jakub@redhat.com>
817 * fold-const.c (struct contains_label_data): New type.
818 (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
819 inside of a SWITCH_BODY seen during the walk.
820 (contains_label_p): Use walk_tree instead of
821 walk_tree_without_duplicates, prepare data for contains_label_1 and
824 2017-12-22 Jakub Jelinek <jakub@redhat.com>
827 * tree-iterator.c (alloc_stmt_list): Start with cleared
828 TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
832 * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
833 enable avx512f explicitly in #pragma GCC target.
834 * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
835 __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
836 __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
837 __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
838 __builtin_ia32_vpdpbusds_v8si_mask,
839 __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
840 __builtin_ia32_vpdpbusds_v4si_mask,
841 __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
842 __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
843 __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
844 __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
845 __builtin_ia32_vpdpwssds_v8si_mask,
846 __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
847 __builtin_ia32_vpdpwssds_v4si_mask,
848 __builtin_ia32_vpdpwssds_v4si_maskz): Use
849 OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
850 just OPTION_MASK_ISA_AVX512VNNI.
852 2017-12-22 Martin Jambor <mjambor@suse.cz>
855 * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
858 2017-12-22 Julia Koval <julia.koval@intel.com>
859 Sebastian Peryt <sebastian.peryt@intel.com>
861 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
862 OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
863 (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
864 * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
865 * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
866 _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
867 _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
868 _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
869 _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
870 _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
871 _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
872 _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
873 _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
874 * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
875 _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
876 _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
877 _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
878 _mm256_maskz_popcnt_epi64): New intrinsics.
879 * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
880 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
881 * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
882 V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
883 V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
884 * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
885 __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
886 __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
887 __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
888 __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
889 __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
890 __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
891 __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
892 __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
893 __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
894 __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
895 __builtin_ia32_vpshufbitqmb256_mask,
896 __builtin_ia32_vpshufbitqmb512_mask): New builtins.
897 * config/i386/i386-c.c (__AVX512BITALG__): New.
898 * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
899 (ix86_valid_target_attribute_inner_p): Ditto.
900 (ix86_expand_args_builtin): Handle new types.
901 * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
902 * config/i386/i386.opt: Add -mavx512bitalg.
903 * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
904 avx512bitalgintrin.h.
905 * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
906 (vpopcount<mode><mask_name>): Add more types.
907 (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
908 * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
910 2017-12-22 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
912 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
913 Or in OPTION_MASK_ISA_AVX512F_SET.
914 (OPTION_MASK_ISA_AVX512F_UNSET): Or in
915 OPTION_MASK_ISA_AVX512VNNI_UNSET.
916 (ix86_handle_option): Adjust for
917 OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
918 * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
920 * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
921 isa_flag instead of isa_flag2.
922 * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
923 isa_opts2 to isa_opts.
924 * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
927 2017-12-22 Mike Stump <mikestump@comcast.net>
928 Eric Botcazou <ebotcazou@adacore.com>
930 * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
932 2017-12-21 Jakub Jelinek <jakub@redhat.com>
935 * config/i386/i386.c (ix86_function_arg_boundary): Return
936 PARM_BOUNDARY for TYPE_EMPTY_P types.
939 * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
940 if navail is >= dir.len.
942 2017-12-21 Steve Ellcey <sellcey@cavium.com>
944 * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
947 2017-12-21 Jakub Jelinek <jakub@redhat.com>
949 PR rtl-optimization/80747
950 PR rtl-optimization/83512
951 * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
952 succ edge from ENTRY, copy partition from e->dest to the newly
954 * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
955 ENTRY, use BB_PARTITION of its successor block as current_partition.
956 Don't copy partition when splitting succ edge from ENTRY.
958 PR tree-optimization/83523
959 * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
960 for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
961 (convert_mult_to_fma): Likewise.
963 PR tree-optimization/83521
964 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
965 gimple_build_assign without code on result of
966 fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
969 2017-12-21 Andrew Pinski <apinski@cavium.com>
970 Steve Ellcey <sellcey@cavium.com>
972 * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
973 multi-arch for ilp32.
975 2017-12-21 Uros Bizjak <ubizjak@gmail.com>
978 * config/i386/i386.md (*ashl<mode>3_mask): Add operand
979 constraints to operand 2.
980 (*ashl<mode>3_mask_1): Ditto.
981 (*<shift_insn><mode>3_mask): Ditto.
982 (*<shift_insn><mode>3_mask_1): Ditto.
983 (*<rotate_insn><mode>3_mask): Ditto.
984 (*<rotate_insn><mode>3_mask_1): Ditto.
986 2017-12-21 Alexandre Oliva <aoliva@redhat.com>
988 * reorg.c (make_return_insns): Reemit each insn with its own location.
990 2017-12-21 James Greenhalgh <james.greenhalgh@arm.com>
992 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
993 generation for cases where splatting a value is not useful.
994 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
995 across a vec_duplicate and a paradoxical subreg forming a vector
996 mode to a vec_concat.
998 2017-12-21 James Greenhalgh <james.greenhalgh@arm.com>
1000 * combine.c (simplify_set): Do not transform subregs to zero_extends
1001 if the destination is not a scalar int mode.
1003 2017-12-21 Eric Botcazou <ebotcazou@adacore.com>
1006 * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1007 shared zero if the input has overflowed.
1009 2017-12-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1011 * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1012 support for Cortex-A55 and Cortex-A75.
1014 2017-12-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1016 * common/config/arm/arm-common.c (compare_opt_names): Add function
1017 comment. Use strcmp instead of manual loop.
1019 2017-12-21 Martin Liska <mliska@suse.cz>
1021 PR gcov-profile/83509
1022 * gcov-dump.c (dump_gcov_file): Do not read info about
1023 support_unexecuted_blocks for gcda files.
1025 2017-12-21 Jakub Jelinek <jakub@redhat.com>
1027 PR rtl-optimization/82973
1028 * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1029 (valid_for_const_vector_p): ... this.
1030 * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1031 (valid_for_const_vector_p): ... this. Adjust function comment.
1032 (gen_vec_duplicate): Adjust caller.
1033 * optabs.c (expand_vector_broadcast): Likewise.
1034 * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1035 CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1037 (simplify_const_binary_operation): Likewise. Use CONST_FIXED_P macro
1038 instead of GET_CODE == CONST_FIXED.
1039 (simplify_subreg): Use CONST_FIXED_P macro instead of
1040 GET_CODE == CONST_FIXED.
1043 * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1044 -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1045 -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1046 (ix86_option_override_internal): Adjust for
1047 OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1048 and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1049 (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1050 (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1051 Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1052 and OPTION_MASK_ISA_CLZERO builtins. Use def_builtin instead of
1053 def_builtin2 for CET builtins.
1054 (ix86_expand_builtin): Remove bdesc_special_args2 handling. Fix
1055 up formatting in IX86_BUILTIN_RDPID code.
1056 * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1057 section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1058 * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1059 ix86_isa_flags2 to ix86_isa_flags.
1060 (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1062 * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1063 OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1064 Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1066 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1067 Or in OPTION_MASK_ISA_AVX512F_SET.
1068 (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1069 OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1070 (ix86_handle_option): Adjust for
1071 OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1072 and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1074 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1075 Alan Hayward <alan.hayward@arm.com>
1076 David Sherwood <david.sherwood@arm.com>
1078 * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1079 factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1080 * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1081 Track polynomial offsets.
1083 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1084 Alan Hayward <alan.hayward@arm.com>
1085 David Sherwood <david.sherwood@arm.com>
1087 * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1088 Treat drb->init as a poly_int. Fail if its misalignment wrt
1089 vector_alignment isn't known.
1091 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1092 Alan Hayward <alan.hayward@arm.com>
1093 David Sherwood <david.sherwood@arm.com>
1095 * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1097 (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1098 (vect_loop_versioning): Take the loop versioning threshold as a
1100 * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1101 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1102 versioning_threshold.
1103 (vect_analyze_loop_2): Compute the loop versioning threshold
1104 whenever loop versioning is needed, and store it in the new
1105 field rather than combining it with the cost model threshold.
1106 (vect_transform_loop): Update call to vect_loop_versioning.
1107 Try to combine the loop versioning and cost thresholds here.
1109 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1110 Alan Hayward <alan.hayward@arm.com>
1111 David Sherwood <david.sherwood@arm.com>
1113 * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1114 poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1115 * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1116 * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1117 an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1118 (group_compare_offset): Update accordingly.
1119 (split_small_address_groups_p): Likewise.
1120 (record_use): Take addr_offset as a poly_uint64 rather than
1121 an unsigned HOST_WIDE_INT.
1122 (strip_offset): Return the offset as a poly_uint64 rather than
1123 an unsigned HOST_WIDE_INT.
1124 (record_group_use, split_address_groups): Track polynomial offsets.
1125 (add_iv_candidate_for_use): Likewise.
1126 (addr_offset_valid_p): Take the offset as a poly_int64 rather
1127 than a HOST_WIDE_INT.
1128 (strip_offset_1): Return the offset as a poly_int64 rather than
1131 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1132 Alan Hayward <alan.hayward@arm.com>
1133 David Sherwood <david.sherwood@arm.com>
1135 * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1136 rather than a HOST_WIDE_INT.
1137 * tree.c (get_binfo_at_offset): Likewise.
1139 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1140 Alan Hayward <alan.hayward@arm.com>
1141 David Sherwood <david.sherwood@arm.com>
1143 * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1144 rather than a HOST_WIDE_INT.
1145 * tree-sra.c (build_ref_for_offset): Likewise.
1147 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1148 Alan Hayward <alan.hayward@arm.com>
1149 David Sherwood <david.sherwood@arm.com>
1151 * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1153 * tree.c (mem_ref_offset): Likewise.
1154 (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1155 * builtins.c (get_object_alignment_2): Likewise.
1156 * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1157 * gimple-fold.c (get_base_constructor): Likewise.
1158 * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1159 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1161 * ipa-polymorphic-call.c
1162 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1163 * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1164 (get_ancestor_addr_info): Likewise.
1165 * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1166 * match.pd: Likewise.
1167 * tree-data-ref.c (dr_analyze_innermost): Likewise.
1168 * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1169 * tree-eh.c (tree_could_trap_p): Likewise.
1170 * tree-object-size.c (addr_object_size): Likewise.
1171 * tree-ssa-address.c (copy_ref_info): Likewise.
1172 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1173 (indirect_refs_may_alias_p): Likewise.
1174 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1175 * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1176 (non_rewritable_mem_ref_base): Likewise.
1177 * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1178 * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1179 * varasm.c (decode_addr_const): Likewise.
1181 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1182 Alan Hayward <alan.hayward@arm.com>
1183 David Sherwood <david.sherwood@arm.com>
1185 * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1186 offsets for MEM_REFs.
1188 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1189 Alan Hayward <alan.hayward@arm.com>
1190 David Sherwood <david.sherwood@arm.com>
1192 * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1193 as a poly_uint64 rather than an unsigned int.
1194 * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1196 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1197 Alan Hayward <alan.hayward@arm.com>
1198 David Sherwood <david.sherwood@arm.com>
1200 * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1202 (decode_addr_const): Update accordingly.
1204 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1205 Alan Hayward <alan.hayward@arm.com>
1206 David Sherwood <david.sherwood@arm.com>
1208 * tree.h (bit_field_size, bit_field_offset): New functions.
1209 * hsa-gen.c (gen_hsa_addr): Use them.
1210 * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1211 (simplify_vector_constructor): Likewise.
1212 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1213 * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1214 BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1215 * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1218 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1219 Alan Hayward <alan.hayward@arm.com>
1220 David Sherwood <david.sherwood@arm.com>
1222 * expr.h (emit_group_load, emit_group_load_into_temps)
1223 (emit_group_store): Take the size as a poly_int64 rather than an int.
1224 * expr.c (emit_group_load_1, emit_group_load): Likewise.
1225 (emit_group_load_into_temp, emit_group_store): Likewise.
1227 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1228 Alan Hayward <alan.hayward@arm.com>
1229 David Sherwood <david.sherwood@arm.com>
1231 * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1232 an unsigned int to a poly_uint64.
1233 * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1234 sizes as poly_uint64s rather than unsigned ints.
1235 * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1238 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1239 Alan Hayward <alan.hayward@arm.com>
1240 David Sherwood <david.sherwood@arm.com>
1242 * lra-constraints.c (emit_inc): Change inc_amount from an int
1245 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1246 Alan Hayward <alan.hayward@arm.com>
1247 David Sherwood <david.sherwood@arm.com>
1249 * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1251 (add_stack_var, stack_var_cmp, partition_stack_vars)
1252 (dump_stack_var_partition): Update accordingly.
1253 (alloc_stack_frame_space): Take the size as a poly_int64 rather
1254 than a HOST_WIDE_INT.
1255 (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1256 (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1257 (account_stack_vars, expand_one_var): Likewise. Return a poly_uint64
1258 rather than a HOST_WIDE_INT.
1260 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1261 Alan Hayward <alan.hayward@arm.com>
1262 David Sherwood <david.sherwood@arm.com>
1264 * target.def (return_pops_args): Treat both the input and output
1265 sizes as poly_int64s rather than HOST_WIDE_INTS.
1266 * targhooks.h (default_return_pops_args): Update accordingly.
1267 * targhooks.c (default_return_pops_args): Likewise.
1268 * doc/tm.texi: Regenerate.
1269 * emit-rtl.h (incoming_args): Change pops_args, size and
1270 outgoing_args_size from int to poly_int64_pod.
1271 * function.h (expr_status): Change x_pending_stack_adjust and
1272 x_stack_pointer_delta from int to poly_int64.
1273 (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1274 (ARGS_SIZE_RTX): Update accordingly.
1275 * calls.c (highest_outgoing_arg_in_use): Change from int to
1277 (stack_usage_watermark, stored_args_watermark): New variables.
1278 (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1279 (emit_call_1): Change the stack_size and rounded_stack_size
1280 parameters from HOST_WIDE_INT to poly_int64. Track n_popped
1282 (save_fixed_argument_area): Check stack_usage_watermark.
1283 (initialize_argument_information): Change old_pending_adj from
1284 a HOST_WIDE_INT * to a poly_int64_pod *.
1285 (compute_argument_block_size): Return the size as a poly_int64
1287 (finalize_must_preallocate): Track polynomial argument sizes.
1288 (compute_argument_addresses): Likewise.
1289 (internal_arg_pointer_based_exp): Track polynomial offsets.
1290 (mem_overlaps_already_clobbered_arg_p): Rename to...
1291 (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1292 size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1293 Check stored_args_used_watermark.
1294 (load_register_parameters): Update accordingly.
1295 (check_sibcall_argument_overlap_1): Likewise.
1296 (combine_pending_stack_adjustment_and_call): Take the unadjusted
1297 args size as a poly_int64 rather than an int. Return a bool
1298 indicating whether the optimization was possible and return
1299 the new adjustment by reference.
1300 (check_sibcall_argument_overlap): Track polynomail argument sizes.
1301 Update stored_args_watermark.
1302 (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1303 (expand_call): Likewise. Maintain stack_usage_watermark and
1304 stored_args_watermark. Update calls to
1305 combine_pending_stack_adjustment_and_call.
1306 (emit_library_call_value_1): Handle polynomial argument sizes.
1307 Call stack_region_maybe_used_p and mark_stack_region_used.
1308 Maintain stack_usage_watermark.
1309 (store_one_arg): Likewise. Update call to
1310 mem_overlaps_already_clobbered_arg_p.
1311 * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1313 * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1314 * config/microblaze/microblaze.c (microblaze_function_prologue):
1316 * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1317 TARGET_RETURN_POPS_ARGS interface.
1318 (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1320 * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1321 * config/i386/i386.c (ix86_return_pops_args): Update for new
1322 TARGET_RETURN_POPS_ARGS interface.
1323 (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1324 * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1325 * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1326 TARGET_RETURN_POPS_ARGS interface.
1327 * config/vax/vax.c (vax_return_pops_args): Likewise.
1328 * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1329 (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1330 * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1331 * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1332 * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1333 * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1334 * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1335 * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1336 * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1337 * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1338 and x_stack_pointer_delta from int to poly_int64.
1339 * dojump.c (do_pending_stack_adjust): Update accordingly.
1340 * explow.c (allocate_dynamic_stack_space): Handle polynomial
1341 stack_pointer_deltas.
1342 * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1343 (pad_to_arg_alignment): Track polynomial offsets.
1344 (assign_parm_find_stack_rtl): Likewise.
1345 (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1346 * toplev.c (output_stack_usage): Update reference to
1347 current_function_pushed_stack_size.
1349 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1350 Alan Hayward <alan.hayward@arm.com>
1351 David Sherwood <david.sherwood@arm.com>
1353 * function.c (in_arg_offset, var_offset, dynamic_offset)
1354 (out_arg_offset, cfa_offset): Change from int to poly_int64.
1355 (instantiate_new_reg): Return the new offset as a poly_int64_pod
1356 rather than a HOST_WIDE_INT.
1357 (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1358 (instantiate_virtual_regs_in_insn): Likewise.
1360 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1361 Alan Hayward <alan.hayward@arm.com>
1362 David Sherwood <david.sherwood@arm.com>
1364 * rtl.h (get_args_size, add_args_size_note): New functions.
1365 (find_args_size_adjust): Return a poly_int64 rather than a
1367 (fixup_args_size_notes): Likewise. Make the same change to the
1368 end_args_size parameter.
1369 * rtlanal.c (get_args_size, add_args_size_note): New functions.
1370 * builtins.c (expand_builtin_trap): Use add_args_size_note.
1371 * calls.c (emit_call_1): Likewise.
1372 * explow.c (adjust_stack_1): Likewise.
1373 * cfgcleanup.c (old_insns_match_p): Update use of
1374 find_args_size_adjust.
1375 * combine.c (distribute_notes): Track polynomial arg sizes.
1376 * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1377 end_true_args_size, beg_delay_args_size and end_delay_args_size
1378 from HOST_WIDE_INT to poly_int64.
1379 (add_cfi_args_size): Take the args_size as a poly_int64 rather
1380 than a HOST_WIDE_INT.
1381 (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1382 (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1383 polynomial arg sizes.
1384 * emit-rtl.c (try_split): Use get_args_size.
1385 * recog.c (peep2_attempt): Likewise.
1386 * reload1.c (reload_as_needed): Likewise.
1387 * expr.c (find_args_size_adjust): Return the adjustment as a
1388 poly_int64 rather than a HOST_WIDE_INT.
1389 (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1390 to a poly_int64 and change the return type in the same way.
1391 (emit_single_push_insn): Track polynomial arg sizes.
1393 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1394 Alan Hayward <alan.hayward@arm.com>
1395 David Sherwood <david.sherwood@arm.com>
1397 * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1398 from HOST_WIDE_INT to poly_int64.
1399 * expr.c (push_block, emit_push_insn): Likewise.
1401 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1402 Alan Hayward <alan.hayward@arm.com>
1403 David Sherwood <david.sherwood@arm.com>
1405 * function.h (frame_space): Change start and length from HOST_WIDE_INT
1407 (get_frame_size): Return the size as a poly_int64 rather than a
1409 (frame_offset_overflow): Take the offset as a poly_int64 rather
1410 than a HOST_WIDE_INT.
1411 (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1412 (assign_stack_temp): Likewise for the size.
1413 * function.c (get_frame_size): Return a poly_int64 rather than
1415 (frame_offset_overflow): Take the offset as a poly_int64 rather
1416 than a HOST_WIDE_INT.
1417 (try_fit_stack_local): Take the start, length and size as poly_int64s
1418 rather than HOST_WIDE_INTs. Return the offset as a poly_int64_pod
1419 rather than a HOST_WIDE_INT.
1420 (add_frame_space): Take the start and end as poly_int64s rather than
1422 (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1423 (assign_stack_temp): Likewise for the size.
1424 (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1426 (find_temp_slot_from_address): Handle polynomial offsets.
1427 (combine_temp_slots): Likewise.
1428 * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1430 * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1431 poly_int64 rather than a HOST_WIDE_INT.
1432 (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1433 than a HOST_WIDE_INT.
1434 (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1435 polynomial frame offsets.
1436 * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1437 as a poly_int64 rather than an int.
1438 * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1439 * config/v850/v850-protos.h (compute_frame_size): Likewise.
1440 * config/v850/v850.c (compute_frame_size): Likewise.
1441 * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1442 * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1443 * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1444 * config/pa/pa.c (pa_compute_frame_size): Likewise.
1445 * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1446 rather than a HOST_WIDE_INT.
1447 * explow.c (get_dynamic_stack_base): Likewise.
1448 * final.c (final_start_function): Use the constant lower bound
1449 of the frame size for -Wframe-larger-than.
1450 * ira.c (do_reload): Adjust for new get_frame_size return type.
1451 * lra.c (lra): Likewise.
1452 * reload1.c (reload): Likewise.
1453 * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1454 * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1455 * rtlanal.c (get_initial_register_offset): Return the offset as
1456 a poly_int64 rather than a HOST_WIDE_INT.
1458 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1459 Alan Hayward <alan.hayward@arm.com>
1460 David Sherwood <david.sherwood@arm.com>
1462 * reload1.c (elim_table): Change initial_offset, offset and
1463 previous_offset from HOST_WIDE_INT to poly_int64_pod.
1464 (offsets_at): Change the target array's element type from
1465 HOST_WIDE_INT to poly_int64_pod.
1466 (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1467 (elimination_costs_in_insn, update_eliminable_offsets)
1468 (verify_initial_elim_offsets, set_offsets_for_label)
1469 (init_eliminable_invariants): Update after above changes.
1471 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1472 Alan Hayward <alan.hayward@arm.com>
1473 David Sherwood <david.sherwood@arm.com>
1475 * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1476 * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1477 (decomposition): Change start and end from HOST_WIDE_INT
1479 (decompose, immune_p): Update accordingly.
1480 (find_inc_amount): Return a poly_int64 rather than an int.
1481 * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1484 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1485 Alan Hayward <alan.hayward@arm.com>
1486 David Sherwood <david.sherwood@arm.com>
1488 * tree.h (get_inner_reference): Return the bitsize and bitpos
1489 as poly_int64_pods rather than HOST_WIDE_INT.
1490 * fold-const.h (ptr_difference_const): Return the pointer difference
1491 as a poly_int64_pod rather than a HOST_WIDE_INT.
1492 * expr.c (get_inner_reference): Return the bitsize and bitpos
1493 as poly_int64_pods rather than HOST_WIDE_INT.
1494 (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1496 * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1497 rather than a HOST_WIDE_INT. Update call to get_inner_reference.
1498 (optimize_bit_field_compare): Update call to get_inner_reference.
1499 (decode_field_reference): Likewise.
1500 (fold_unary_loc): Track polynomial offsets and sizes.
1501 (split_address_to_core_and_offset): Return the bitpos as a
1502 poly_int64_pod rather than a HOST_WIDE_INT.
1503 (ptr_difference_const): Likewise for the pointer difference.
1504 * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1505 * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1506 * dbxout.c (dbxout_expand_expr): Likewise.
1507 * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1508 (loc_list_from_tree_1, fortran_common): Likewise.
1509 * gimple-laddress.c (pass_laddress::execute): Likewise.
1510 * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1511 * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1512 * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1513 * tree-affine.c (tree_to_aff_combination): Likewise.
1514 (get_inner_reference_aff): Likewise.
1515 * tree-data-ref.c (split_constant_offset_1): Likewise.
1516 (dr_analyze_innermost): Likewise.
1517 * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1518 * tree-sra.c (ipa_sra_check_caller): Likewise.
1519 * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1520 * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1521 (instrument_bool_enum_load, instrument_object_size): Likewise.
1522 * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1523 to get_inner_reference.
1524 * hsa-gen.c (gen_hsa_addr): Likewise.
1525 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1526 * tsan.c (instrument_expr): Likewise.
1527 * match.pd: Update call to ptr_difference_const.
1529 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1530 Alan Hayward <alan.hayward@arm.com>
1531 David Sherwood <david.sherwood@arm.com>
1533 * fold-const.c (fold_comparison): Track sizes and offsets as
1534 poly_int64s rather than HOST_WIDE_INTs when folding address
1537 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1538 Alan Hayward <alan.hayward@arm.com>
1539 David Sherwood <david.sherwood@arm.com>
1541 * expr.h (get_bit_range): Return the bitstart and bitend as
1542 poly_uint64s rather than unsigned HOST_WIDE_INTs. Return the bitpos
1543 as a poly_int64 rather than a HOST_WIDE_INT.
1544 * expr.c (get_bit_range): Likewise.
1545 (expand_assignment): Update call accordingly.
1546 * fold-const.c (optimize_bit_field_compare): Likewise.
1548 2017-12-21 Richard Sandiford <richard.sandiford@linaro.org>
1549 Alan Hayward <alan.hayward@arm.com>
1550 David Sherwood <david.sherwood@arm.com>
1552 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1553 (aarch64_legitimate_address_p): Use it instead of an rtx code,
1554 as an optional final parameter.
1555 * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1556 (aarch64_legitimate_address_p): Likewise.
1557 (aarch64_print_address_internal): Take an aarch64_addr_query_type
1558 instead of an rtx code.
1559 (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1560 (aarch64_legitimate_address_hook_p): Likewise.
1561 (aarch64_print_ldpstp_address): Likewise.
1562 (aarch64_print_operand_address): Likewise.
1563 (aarch64_address_cost): Likewise.
1564 * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1565 * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1566 (aarch64_mem_pair_lanes_operand): Likewise.
1568 2017-12-20 Richard Biener <rguenther@suse.de>
1570 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1571 update_stmt_if_modified.
1573 2017-12-20 Wilco Dijkstra <wdijkstr@arm.com>
1575 PR tree-optimization/83491
1576 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1577 before walking uses. Improve coding style and comments.
1579 2017-12-20 Tom de Vries <tom@codesourcery.com>
1581 * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1583 2017-12-20 Jakub Jelinek <jakub@redhat.com>
1586 * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1588 (pass_ipa_free_fn_summary): Add small_p private data member,
1589 initialize to false in the ctor.
1590 (pass_ipa_free_fn_summary::clone,
1591 pass_ipa_free_fn_summary::set_pass_param,
1592 pass_ipa_free_fn_summary::gate): New methods.
1593 (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1594 | TODO_dump_symtab if small_p.
1595 * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1596 entry and add another instance of the pass with false parm after
1598 * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1599 ipa_free_fn_summary here.
1601 2017-12-20 Paolo Carlini <paolo.carlini@oracle.com>
1603 * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1605 2017-12-20 Martin Sebor <msebor@redhat.com>
1608 * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1609 (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1610 (expand_builtin_strncmp): Same.
1612 2017-12-20 Alexandre Oliva <aoliva@redhat.com>
1615 * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1616 allowed debug stmts before labels.
1617 (expand_gimple_basic_block): Likewise.
1618 * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1619 * gimple-iterator.h (gsi_after_labels): Likewise.
1620 * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1621 rename reused variable, and simplify using gsi_move_before.
1622 * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1623 * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1624 (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1625 (gimple_verify_flow_info, gimple_block_label): Likewise.
1626 (make_blocks): Move debug markers after adjacent labels.
1627 * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1628 allowed debug insns outside blocks.
1629 * df-scan.c (df_insn_delete): Likewise.
1630 * lra-constraints.c (update_ebb_live_info): Likewise.
1631 * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1632 (vt_initialize, delete_vta_debug_insns): Likewise.
1633 (reemit_marker_as_note): Drop BB parm. Adjust callers.
1635 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1636 Alan Hayward <alan.hayward@arm.com>
1637 David Sherwood <david.sherwood@arm.com>
1639 * poly-int-types.h (round_down_to_byte_boundary): New macro.
1640 (round_up_to_byte_boundary): Likewise.
1641 * expr.h (get_bit_range): Add temporary shim.
1642 * gimple-ssa-store-merging.c (store_operand_info): Change the
1643 bitsize, bitpos, bitregion_start and bitregion_end fields from
1644 unsigned HOST_WIDE_INT to poly_uint64.
1645 (merged_store_group): Likewise load_align_base.
1646 (compatible_load_p, compatible_load_p): Update accordingly.
1647 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1648 (split_group, imm_store_chain_info::output_merged_store): Likewise.
1649 (mem_valid_for_store_merging): Return the bitsize, bitpos,
1650 bitregion_start and bitregion_end as poly_uint64s rather than
1651 unsigned HOST_WIDE_INTs. Track polynomial offsets internally.
1652 (handled_load): Take the bitsize, bitpos,
1653 bitregion_start and bitregion_end as poly_uint64s rather than
1654 unsigned HOST_WIDE_INTs.
1655 (pass_store_merging::process_store): Update call to
1656 mem_valid_for_store_merging.
1658 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1659 Alan Hayward <alan.hayward@arm.com>
1660 David Sherwood <david.sherwood@arm.com>
1662 * builtins.c (get_object_alignment_2): Track polynomial offsets
1663 and sizes. Update the alignment handling.
1665 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1666 Alan Hayward <alan.hayward@arm.com>
1667 David Sherwood <david.sherwood@arm.com>
1669 * tree.h (get_inner_reference): Add a version that returns the
1670 offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1671 * cfgexpand.c (expand_debug_expr): Track polynomial offsets. Simply
1672 the case in which bitpos is not associated with the first byte.
1674 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1675 Alan Hayward <alan.hayward@arm.com>
1676 David Sherwood <david.sherwood@arm.com>
1678 * tree-affine.h (get_inner_reference_aff): Return the size as a
1680 * tree-affine.c (get_inner_reference_aff): Likewise.
1681 * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1682 * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1684 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1685 Alan Hayward <alan.hayward@arm.com>
1686 David Sherwood <david.sherwood@arm.com>
1688 * fold-const.c (pointer_may_wrap_p): Take the offset as a
1689 HOST_WIDE_INT rather than a poly_int64.
1691 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1692 Alan Hayward <alan.hayward@arm.com>
1693 David Sherwood <david.sherwood@arm.com>
1695 * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1696 HOST_WIDE_INT to poly_int64_pod.
1697 (perform_symbolic_merge): Update accordingly.
1698 (bswap_replace): Likewise.
1700 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1701 Alan Hayward <alan.hayward@arm.com>
1702 David Sherwood <david.sherwood@arm.com>
1704 * tree-affine.h (aff_tree::offset): Change from widest_int
1706 (wide_int_ext_for_comb): Delete.
1707 (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1708 constants as poly_widest_int rather than widest_int.
1709 (aff_combination_constant_multiple_p): Return the multiplier
1710 as a poly_widest_int.
1711 (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1713 * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1714 version static and add an overload for poly_widest_int.
1715 (aff_combination_const, aff_combination_add_cst)
1716 (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1717 the constants as poly_widest_int rather than widest_int.
1718 (tree_to_aff_combination): Generalize INTEGER_CST case to
1720 (aff_combination_to_tree): Track offsets as poly_widest_ints.
1721 (aff_combination_add_product, aff_combination_mult): Handle
1723 (aff_combination_constant_multiple_p): Return the multiplier
1724 as a poly_widest_int.
1725 * tree-predcom.c (determine_offset): Return the offset as a
1727 (split_data_refs_to_components, suitable_component_p): Update
1729 (valid_initializer_p): Update call to
1730 aff_combination_constant_multiple_p.
1731 * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1732 * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1733 as a poly_int64 rather than a HOST_WIDE_INT.
1734 (get_address_cost): Handle polynomial offsets.
1735 (iv_elimination_compare_lt): Likewise.
1736 (rewrite_use_nonlinear_expr): Likewise.
1738 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1739 Alan Hayward <alan.hayward@arm.com>
1740 David Sherwood <david.sherwood@arm.com>
1742 * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1743 as a poly_int64_pod rather than a HOST_WIDE_INT.
1744 (get_addr_base_and_unit_offset): Likewise.
1745 * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1746 (get_addr_base_and_unit_offset): Likewise.
1747 * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1748 to poly_int64 in example.
1749 * fold-const.c (fold_binary_loc): Update call to
1750 get_addr_base_and_unit_offset.
1751 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1752 (maybe_canonicalize_mem_ref_addr): Likewise.
1753 (gimple_fold_stmt_to_constant_1): Likewise.
1754 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1756 * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1757 * match.pd: Likewise.
1758 * omp-low.c (lower_omp_target): Likewise.
1759 * tree-sra.c (build_ref_for_offset): Likewise.
1760 (build_debug_ref_for_model): Likewise.
1761 * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1762 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1763 * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1764 * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1765 (constant_pointer_difference): Likewise.
1766 * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1767 * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1768 * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1769 * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1770 (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1771 (set_ssa_val_to): Likewise.
1772 * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1773 (maybe_diag_stxncpy_trunc): Likewise.
1774 * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1775 * tree.c (build_simple_mem_ref_loc): Likewise.
1776 (array_at_struct_end_p): Likewise.
1778 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1779 Alan Hayward <alan.hayward@arm.com>
1780 David Sherwood <david.sherwood@arm.com>
1782 * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1783 max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1784 (get_ref_base_and_extent_hwi): Declare.
1785 * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1786 max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1787 (get_ref_base_and_extent_hwi): New function.
1788 * cfgexpand.c (expand_debug_expr): Update call to
1789 get_ref_base_and_extent.
1790 * dwarf2out.c (add_var_loc_to_decl): Likewise.
1791 * gimple-fold.c (get_base_constructor): Return the offset as a
1792 poly_int64_pod rather than a HOST_WIDE_INT.
1793 (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1794 * ipa-polymorphic-call.c
1795 (ipa_polymorphic_call_context::set_by_invariant)
1796 (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1797 (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1798 (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1799 rather than get_ref_base_and_extent.
1800 (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1801 * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1802 (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1804 * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1805 call to get_ref_base_and_extent.
1806 * tree-sra.c (create_access, get_access_for_expr): Likewise.
1807 * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1808 (stmt_kills_ref_p): Likewise.
1809 * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1810 * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1812 * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1813 Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1814 when calling native_encode_expr.
1815 * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1816 call to get_ref_base_and_extent.
1817 (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1818 get_ref_base_and_extent.
1819 * var-tracking.c (track_expr_p): Likewise.
1821 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1822 Alan Hayward <alan.hayward@arm.com>
1823 David Sherwood <david.sherwood@arm.com>
1825 * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1826 HOST_WIDE_INT to poly_int64_pod.
1827 * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1828 polynomail parameter offsets.
1830 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1831 Alan Hayward <alan.hayward@arm.com>
1832 David Sherwood <david.sherwood@arm.com>
1834 * gengtype.c (main): Handle poly_int64_pod.
1835 * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1836 (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1837 (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1838 from HOST_WIDE_INT to poly_int64_pod.
1839 * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1840 (copy_cfa): New function.
1841 (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1842 (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1843 (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
1844 (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
1845 (initial_return_save): Treat offsets as poly_ints.
1846 (def_cfa_0): Likewise. Cache the CFA in dw_cfi_cfa_loc if either
1847 offset is nonconstant.
1848 (reg_save): Take the offset as a poly_int64. Fall back to
1849 DW_CFA_expression for nonconstant offsets.
1850 (queue_reg_save): Take the offset as a poly_int64.
1851 * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
1853 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1854 Alan Hayward <alan.hayward@arm.com>
1855 David Sherwood <david.sherwood@arm.com>
1857 * rtl.h (operand_subword, operand_subword_force): Take the offset
1858 as a poly_uint64 an unsigned int.
1859 * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
1861 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1862 Alan Hayward <alan.hayward@arm.com>
1863 David Sherwood <david.sherwood@arm.com>
1865 * doc/rtl.texi: Update documentation of SUBREG_BYTE. Document the
1866 'p' format code. Use INT_LIST rather than SUBREG as the example of
1867 a code with an XINT and an XEXP. Remove the implication that
1868 accessing an rtx field using XINT is expected to work.
1869 * rtl.def (SUBREG): Change format from "ei" to "ep".
1870 * rtl.h (rtunion::rt_subreg): New field.
1871 (XCSUBREG): New macro.
1872 (SUBREG_BYTE): Use it.
1873 (subreg_shape): Change offset from an unsigned int to a poly_uint16.
1874 Update constructor accordingly.
1875 (subreg_shape::operator ==): Update accordingly.
1876 (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
1877 than an unsigned int.
1878 (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
1879 a poly_uint64 rather than an unsigned int.
1880 (subreg_lsb_1): Likewise. Take the offset as a poly_uint64 rather
1881 than an unsigned int.
1882 (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
1883 (subreg_size_highpart_offset): Return a poly_uint64 rather than
1884 an unsigned int. Take the sizes as poly_uint64s.
1885 (subreg_offset_from_lsb): Return a poly_uint64 rather than
1886 an unsigned int. Take the shift as a poly_uint64 rather than
1888 (subreg_regno_offset, subreg_offset_representable_p): Take the offset
1889 as a poly_uint64 rather than an unsigned int.
1890 (simplify_subreg_regno): Likewise.
1891 (byte_lowpart_offset): Return the memory offset as a poly_int64
1893 (subreg_memory_offset): Likewise. Take the subreg offset as a
1894 poly_uint64 rather than an unsigned int.
1895 (simplify_subreg, simplify_gen_subreg, subreg_get_info)
1896 (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
1897 poly_uint64 rather than an unsigned int.
1898 * rtl.c (rtx_format): Describe 'p' in comment.
1899 (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
1900 * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
1901 offset as a poly_uint64 rather than an unsigned int.
1902 (byte_lowpart_offset): Return the memory offset as a poly_int64
1904 (subreg_memory_offset): Likewise. Take the subreg offset as a
1905 poly_uint64 rather than an unsigned int.
1906 (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
1907 mode sizes as poly_uint64s rather than unsigned ints. Return a
1908 poly_uint64 rather than an unsigned int.
1909 (subreg_lowpart_p): Treat subreg offsets as poly_ints.
1910 (copy_insn_1): Handle 'p'.
1911 * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
1912 (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
1913 an unsigned int. Return the shift in the same way.
1914 (subreg_lsb): Return the shift as a poly_uint64 rather than an
1916 (subreg_size_offset_from_lsb): Take the sizes and shift as
1917 poly_uint64s rather than unsigned ints. Return the offset as
1919 (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
1920 (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
1922 * rtlhash.c (add_rtx): Handle 'p'.
1923 * genemit.c (gen_exp): Likewise.
1924 * gengenrtl.c (type_from_format, gendef): Likewise.
1925 * gensupport.c (subst_pattern_match, get_alternatives_number)
1926 (collect_insn_data, alter_predicate_for_insn, alter_constraints)
1927 (subst_dup): Likewise.
1928 * gengtype.c (adjust_field_rtx_def): Likewise.
1929 * genrecog.c (find_operand, find_matching_operand, validate_pattern)
1930 (match_pattern_2): Likewise.
1931 (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
1932 (rtx_test::subreg_field): New function.
1933 (operator ==, safe_to_hoist_p, transition_parameter_type)
1934 (print_nonbool_test, print_test): Handle SUBREG_FIELD.
1935 * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
1936 * genpeep.c (match_rtx): Likewise.
1937 * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
1938 (rtx_writer::print_rtx_operand): Handle 'p'.
1939 (print_value): Handle SUBREG.
1940 * read-rtl.c (apply_int_iterator): Likewise.
1941 (rtx_reader::read_rtx_operand): Handle 'p'.
1942 * alias.c (rtx_equal_for_memref_p): Likewise.
1943 * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
1944 * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
1946 * calls.c (expand_call): Likewise.
1947 * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
1948 (make_extraction, gen_lowpart_for_combine): Likewise.
1949 * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
1951 * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
1952 rather than an unsigned int. Treat subreg offsets as poly_ints.
1953 (exp_equiv_p): Handle 'p'.
1954 (hash_rtx_cb): Likewise. Treat subreg offsets as poly_ints.
1955 (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
1956 * dse.c (find_shift_sequence): Likewise.
1957 * dwarf2out.c (rtl_for_decl_location): Likewise.
1958 * expmed.c (extract_low_bits): Likewise.
1959 * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
1960 (expand_expr_real_2): Likewise.
1961 * final.c (alter_subreg): Likewise.
1962 (leaf_renumber_regs_insn): Handle 'p'.
1963 * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
1964 Treat subreg offsets as poly_ints.
1965 * fwprop.c (forward_propagate_and_simplify): Likewise.
1966 * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
1967 * ira.c (get_subreg_tracking_sizes): Likewise.
1968 * ira-conflicts.c (go_through_subreg): Likewise.
1969 * ira-lives.c (process_single_reg_class_operands): Likewise.
1970 * jump.c (rtx_renumbered_equal_p): Likewise. Handle 'p'.
1971 * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
1972 as a poly_uint64 rather than an unsigned int.
1973 (simplify_gen_subreg_concatn, resolve_simple_move): Treat
1974 subreg offsets as poly_ints.
1975 * lra-constraints.c (operands_match_p): Handle 'p'.
1976 (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
1977 * lra-spills.c (assign_mem_slot): Likewise.
1978 * postreload.c (move2add_valid_value_p): Likewise.
1979 * recog.c (general_operand, indirect_operand): Likewise.
1980 * regcprop.c (copy_value, maybe_mode_change): Likewise.
1981 (copyprop_hardreg_forward_1): Likewise.
1982 * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
1983 (record_subregs_of_mode): Likewise.
1984 * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
1985 * reload.c (operands_match_p): Handle 'p'.
1986 (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
1987 * reload1.c (alter_reg, choose_reload_regs): Likewise.
1988 (compute_reload_subreg_offset): Likewise, and return an poly_int64.
1989 * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
1990 (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
1991 (simplify_const_poly_int_tests<N>::run): Likewise.
1992 (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
1993 a poly_uint64 rather than an unsigned int.
1994 * valtrack.c (debug_lowpart_subreg): Likewise.
1995 * var-tracking.c (var_lowpart): Likewise.
1996 (loc_cmp): Handle 'p'.
1998 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
1999 Alan Hayward <alan.hayward@arm.com>
2000 David Sherwood <david.sherwood@arm.com>
2002 * ira.c (get_subreg_tracking_sizes): New function.
2003 (init_live_subregs): Take an integer size rather than a register.
2004 (build_insn_chain): Use get_subreg_tracking_sizes. Update calls
2005 to init_live_subregs.
2007 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2008 Alan Hayward <alan.hayward@arm.com>
2009 David Sherwood <david.sherwood@arm.com>
2011 * expr.c (store_constructor_field): Change bitsize from a
2012 unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2013 HOST_WIDE_INT to a poly_int64.
2014 (store_constructor): Change size from a HOST_WIDE_INT to
2016 (store_field): Likewise bitsize and bitpos.
2018 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2019 Alan Hayward <alan.hayward@arm.com>
2020 David Sherwood <david.sherwood@arm.com>
2022 * expmed.h (store_bit_field): Change bitregion_start and
2023 bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2024 * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2025 (store_bit_field_1, store_integral_bit_field, store_bit_field)
2026 (store_fixed_bit_field, store_split_bit_field): Likewise.
2027 * expr.c (store_constructor_field, store_field): Likewise.
2028 (optimize_bitfield_assignment_op): Likewise. Make the same change
2029 to bitsize and bitpos.
2030 * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2031 and m_bitregion_end from HOST_WIDE_INT to poly_int64. Make the
2032 same change in the constructor arguments.
2033 (get_best_mode): Change bitregion_start and bitregion_end from
2034 unsigned HOST_WIDE_INT to poly_uint64.
2035 * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2036 Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2038 (bit_field_mode_iterator::next_mode): Update for new types
2039 of m_bitregion_start and m_bitregion_end.
2040 (get_best_mode): Change bitregion_start and bitregion_end from
2041 unsigned HOST_WIDE_INT to poly_uint64.
2043 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2044 Alan Hayward <alan.hayward@arm.com>
2045 David Sherwood <david.sherwood@arm.com>
2047 * rtl.h (simplify_gen_subreg): Add a temporary overload that
2048 accepts poly_uint64 offsets.
2049 * expmed.h (extract_bit_field): Take bitsize and bitnum as
2050 poly_uint64s rather than unsigned HOST_WIDE_INTs.
2051 * expmed.c (lowpart_bit_field_p): Likewise.
2052 (extract_bit_field_as_subreg): New function, split out from...
2053 (extract_bit_field_1): ...here. Take bitsize and bitnum as
2054 poly_uint64s rather than unsigned HOST_WIDE_INTs. For vector
2055 extractions, check that BITSIZE matches the size of the extracted
2056 value and that BITNUM is an exact multiple of that size.
2057 If all else fails, try forcing the value into memory if
2058 BITNUM is variable, and adjusting the address so that the
2059 offset is constant. Split the part that can only handle constant
2060 bitsize and bitnum out into...
2061 (extract_integral_bit_field): ...this new function.
2062 (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2063 rather than unsigned HOST_WIDE_INTs.
2065 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2066 Alan Hayward <alan.hayward@arm.com>
2067 David Sherwood <david.sherwood@arm.com>
2069 * expmed.h (store_bit_field): Take bitsize and bitnum as
2070 poly_uint64s rather than unsigned HOST_WIDE_INTs.
2071 * expmed.c (simple_mem_bitfield_p): Likewise. Add a parameter
2072 that returns the byte size.
2073 (store_bit_field_1): Take bitsize and bitnum as
2074 poly_uint64s rather than unsigned HOST_WIDE_INTs. Update call
2075 to simple_mem_bitfield_p. Split the part that can only handle
2076 constant bitsize and bitnum out into...
2077 (store_integral_bit_field): ...this new function.
2078 (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2079 than unsigned HOST_WIDE_INTs.
2080 (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2082 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2083 Alan Hayward <alan.hayward@arm.com>
2084 David Sherwood <david.sherwood@arm.com>
2086 * lra-int.h (lra_reg): Change offset from int to poly_int64.
2087 (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2089 (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2090 update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2091 (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2092 offset as a poly_int64 rather than an int.
2093 * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2094 (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2095 * lra-constraints.c (equiv_address_substitution): Track offsets
2097 (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2098 (curr_insn_transform): Handle the new form of sp_offset.
2099 * lra-eliminations.c (lra_elim_table): Change previous_offset
2100 and offset from HOST_WIDE_INT to poly_int64.
2101 (print_elim_table, update_reg_eliminate): Update accordingly.
2102 (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2103 (get_elimination): Update accordingly.
2104 (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2105 (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2106 update_sp_offset from a HOST_WIDE_INT to a poly_int64. Handle
2107 poly_int64 offsets generally.
2108 (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2109 (mark_not_eliminable, init_elimination): Update accordingly.
2110 (remove_reg_equal_offset_note): Return a bool and pass the new
2111 offset back by pointer as a poly_int64.
2112 * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2113 rather than a HOST_WIDE_INT.
2114 (do_remat): Track offsets poly_int64s.
2115 * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2117 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2118 Alan Hayward <alan.hayward@arm.com>
2119 David Sherwood <david.sherwood@arm.com>
2121 * rtl.h (mem_attrs): Add a default constructor. Change size and
2122 offset from HOST_WIDE_INT to poly_int64.
2123 * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2124 (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2125 (widen_memory_access): Take the sizes and offsets as poly_int64s
2126 rather than HOST_WIDE_INTs.
2127 * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2128 (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2130 (adjust_offset_for_component_ref): Change the offset from a
2131 HOST_WIDE_INT to a poly_int64.
2132 (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2133 * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2134 * dce.c (find_call_stack_args): Likewise.
2135 * dse.c (record_store): Likewise.
2136 * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2137 * print-rtl.c (rtx_writer::print_rtx): Likewise.
2138 * read-rtl-function.c (test_loading_mem): Likewise.
2139 * rtlanal.c (may_trap_p_1): Likewise.
2140 * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2141 * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2142 * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2143 (mem_attrs::mem_attrs): New function.
2144 (set_mem_attributes_minus_bitpos): Change bitpos from a
2145 HOST_WIDE_INT to poly_int64.
2146 (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2147 (clear_mem_offset, clear_mem_size, change_address)
2148 (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2149 initialize mem_attrs.
2150 (set_mem_offset, set_mem_size, adjust_address_1)
2151 (adjust_automodify_address_1, offset_address, widen_memory_access):
2152 Likewise. Take poly_int64s rather than HOST_WIDE_INT.
2154 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2155 Alan Hayward <alan.hayward@arm.com>
2156 David Sherwood <david.sherwood@arm.com>
2158 * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2159 as poly_int64s rather than HOST_WIDE_INTs. Use a size of -1
2160 rather than 0 to represent an unknown size. Assert that the size
2161 is known when the mode isn't BLKmode.
2162 (may_trap_p_1): Use -1 for unknown sizes.
2163 (rtx_addr_can_trap_p): Likewise. Pass BLKmode rather than VOIDmode.
2165 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2166 Alan Hayward <alan.hayward@arm.com>
2167 David Sherwood <david.sherwood@arm.com>
2169 * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2170 to poly_int64. Update commentary for positions_needed.large.
2171 (read_info_type): Change offset and width from HOST_WIDE_INT
2173 (set_usage_bits): Likewise.
2174 (canon_address): Return the offset as a poly_int64 rather than
2175 a HOST_WIDE_INT. Use strip_offset_and_add.
2176 (set_all_positions_unneeded, any_positions_needed_p): Use
2177 positions_needed.large to track stores with non-constant widths.
2178 (all_positions_needed_p): Likewise. Take the offset and width
2179 as poly_int64s rather than ints. Assert that rhs is nonnull.
2180 (record_store): Cope with non-constant offsets and widths.
2181 Nullify the rhs of an earlier store if we can't tell which bytes
2183 (find_shift_sequence): Take the access_size and shift as poly_int64s
2185 (get_stored_val): Take the read_offset and read_width as poly_int64s
2186 rather than HOST_WIDE_INTs.
2187 (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2188 non-constant offsets and widths.
2190 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2191 Alan Hayward <alan.hayward@arm.com>
2192 David Sherwood <david.sherwood@arm.com>
2194 * inchash.h (inchash::hash::add_poly_int): New function.
2195 * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2196 Use poly_int64 rather than HOST_WIDE_INT.
2197 (ao_ref::max_size_known_p): New function.
2198 * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2199 rather than HOST_WIDE_INT.
2200 * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2201 to temporaries until its interface is adjusted to match.
2202 (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2203 (aliasing_component_refs_p, decl_refs_may_alias_p)
2204 (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2205 the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2206 (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2208 * alias.c (ao_ref_from_mem): Likewise.
2209 * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2210 * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2211 (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2212 (maybe_trim_complex_store, maybe_trim_constructor_store)
2213 (live_bytes_read, dse_classify_store): Likewise.
2214 * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2215 (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2216 (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2217 (vn_reference_lookup_3): Likewise.
2218 * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2220 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2221 Alan Hayward <alan.hayward@arm.com>
2222 David Sherwood <david.sherwood@arm.com>
2224 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2225 (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2226 instead of ranges_overlap_p.
2228 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2229 Alan Hayward <alan.hayward@arm.com>
2230 David Sherwood <david.sherwood@arm.com>
2232 * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2233 sizes as poly_int64s rather than HOST_WIDE_INTs.
2235 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2236 Alan Hayward <alan.hayward@arm.com>
2237 David Sherwood <david.sherwood@arm.com>
2239 * gimple-fold.h (fold_ctor_reference): Take the offset and size
2240 as poly_uint64 rather than unsigned HOST_WIDE_INT.
2241 * gimple-fold.c (fold_ctor_reference): Likewise.
2243 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2244 Alan Hayward <alan.hayward@arm.com>
2245 David Sherwood <david.sherwood@arm.com>
2247 * target.def (dwarf_poly_indeterminate_value): New hook.
2248 * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2249 * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2250 * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2251 * doc/tm.texi: Regenerate.
2252 * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2253 offset as a poly_int64.
2254 * dwarf2out.c (new_reg_loc_descr): Move later in file. Take the
2255 offset as a poly_int64.
2256 (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2257 Take the offset as a poly_int64.
2258 (build_cfa_loc): Likewise. Use loc_descr_plus_const.
2259 (frame_pointer_fb_offset): Change to a poly_int64.
2260 (int_loc_descriptor): Take the offset as a poly_int64. Use
2261 targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2262 (based_loc_descr): Take the offset as a poly_int64.
2263 Use strip_offset_and_add to handle (plus X (const)).
2264 Use new_reg_loc_descr instead of an open-coded version of the
2265 previous implementation.
2266 (mem_loc_descriptor): Handle CONST_POLY_INT.
2267 (compute_frame_pointer_to_fb_displacement): Take the offset as a
2268 poly_int64. Use strip_offset_and_add to handle (plus X (const)).
2270 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2271 Alan Hayward <alan.hayward@arm.com>
2272 David Sherwood <david.sherwood@arm.com>
2274 * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2275 (gen_rtx_REG_offset): Take the offset as a poly_int64.
2276 * inchash.h (inchash::hash::add_poly_hwi): New function.
2277 * gengtype.c (main): Register poly_int64.
2278 * emit-rtl.c (reg_attr_hasher::hash): Use inchash. Treat the
2279 offset as a poly_int.
2280 (reg_attr_hasher::equal): Use must_eq to compare offsets.
2281 (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2282 offset as a poly_int64.
2283 (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2284 * print-rtl.c (print_poly_int): New function.
2285 (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2287 * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2289 (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2290 (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2291 (vt_get_decl_and_offset): Return the offset as a poly_int64.
2292 Enforce track_offset_p for parts of a PARALLEL.
2293 (vt_add_function_parameter): Use const_offset for the final
2294 offset to track. Use get_tracked_reg_offset for the parts
2297 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2298 Alan Hayward <alan.hayward@arm.com>
2299 David Sherwood <david.sherwood@arm.com>
2301 * target.def (truly_noop_truncation): Take poly_uint64s instead of
2302 unsigned ints. Change default to hook_bool_puint64_puint64_true.
2303 * doc/tm.texi: Regenerate.
2304 * hooks.h (hook_bool_uint_uint_true): Delete.
2305 (hook_bool_puint64_puint64_true): Declare.
2306 * hooks.c (hook_bool_uint_uint_true): Delete.
2307 (hook_bool_puint64_puint64_true): New function.
2308 * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2309 instead of unsigned ints.
2310 * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2311 * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2313 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2314 Alan Hayward <alan.hayward@arm.com>
2315 David Sherwood <david.sherwood@arm.com>
2317 * optabs.h (expand_operand): Add an int_value field.
2318 (create_expand_operand): Add an int_value parameter and use it
2319 to initialize the new expand_operand field.
2320 (create_integer_operand): Replace with a declaration of a function
2321 that accepts poly_int64s. Move the implementation to...
2322 * optabs.c (create_integer_operand): ...here.
2323 (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2324 the mode preserves the value of int_value, instead of calling
2325 const_int_operand on the rtx. Use gen_int_mode to generate
2328 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2329 Alan Hayward <alan.hayward@arm.com>
2330 David Sherwood <david.sherwood@arm.com>
2332 * dumpfile.h (dump_dec): Declare.
2333 * dumpfile.c (dump_dec): New function.
2334 * pretty-print.h (pp_wide_integer): Turn into a function and
2335 declare a poly_int version.
2336 * pretty-print.c (pp_wide_integer): New function for poly_ints.
2338 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2339 Alan Hayward <alan.hayward@arm.com>
2340 David Sherwood <david.sherwood@arm.com>
2342 * doc/generic.texi (POLY_INT_CST): Document.
2343 * tree.def (POLY_INT_CST): New tree code.
2344 * treestruct.def (TS_POLY_INT_CST): New tree layout.
2345 * tree-core.h (tree_poly_int_cst): New struct.
2346 (tree_node): Add a poly_int_cst field.
2347 * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2348 (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2349 instead of a wide_int_ref.
2350 (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2352 (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2353 instead of an unsigned HOST_WIDE_INT.
2354 (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2355 (ptrdiff_tree_p): Declare.
2356 (tree_to_poly_int64, tree_to_poly_uint64): Likewise. Provide
2357 extern inline implementations if the target doesn't use POLY_INT_CST.
2358 (poly_int_tree_p): New function.
2359 (wi::unextended_tree): New class.
2360 (wi::int_traits <unextended_tree>): New override.
2361 (wi::extended_tree): Add a default constructor.
2362 (wi::extended_tree::get_tree): New function.
2363 (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2364 (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2365 (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2366 (wi::tree_to_poly_wide_ref): New typedefs.
2367 (wi::ints_for): Provide overloads for extended_tree and
2369 (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2370 (wi::to_wide): New functions.
2371 (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2372 * tree.c (poly_int_cst_hasher): New struct.
2373 (poly_int_cst_hash_table): New variable.
2374 (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2375 (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2377 (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2378 (init_ttree): Initialize poly_int_cst_hash_table.
2379 (build_int_cst, build_int_cst_type, build_invariant_address): Take
2380 a poly_int64 instead of a HOST_WIDE_INT.
2381 (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2382 instead of an unsigned HOST_WIDE_INT.
2383 (wide_int_to_tree): Rename to...
2384 (wide_int_to_tree_1): ...this.
2385 (build_new_poly_int_cst, build_poly_int_cst): New functions.
2386 (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2387 (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2388 (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2389 (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2390 * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2392 * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2393 (streamer_read_tree_body): Likewise.
2394 * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2395 (streamer_write_tree_body): Likewise.
2396 * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2397 * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2398 * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2399 * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2400 * gimple-expr.h (is_gimple_constant): Likewise.
2401 * gimplify.c (maybe_with_size_expr): Likewise.
2402 * print-tree.c (print_node): Likewise.
2403 * tree-data-ref.c (data_ref_compare_tree): Likewise.
2404 * tree-pretty-print.c (dump_generic_node): Likewise.
2405 * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2406 * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2407 * tree-vrp.c (compare_values_warnv): Likewise.
2408 * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2409 (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2410 (force_expr_to_var_cost): Likewise.
2411 * tree-ssa-loop.c (for_each_index): Likewise.
2412 * fold-const.h (build_invariant_address, size_int_kind): Take a
2413 poly_int64 instead of a HOST_WIDE_INT.
2414 * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2415 (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2417 (size_binop_loc): Likewise. Allow int_const_binop_1 to fail.
2418 (int_const_binop_2): New function, split out from...
2419 (int_const_binop_1): ...here. Handle POLY_INT_CST.
2420 (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2421 * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2422 * gimple-ssa-strength-reduction.c (slsr_process_add)
2423 (slsr_process_mul): Check for INTEGER_CSTs before using them
2425 * stor-layout.c (bits_from_bytes): New function.
2426 (bit_from_pos): Use it.
2427 (layout_type): Likewise. For vectors, multiply the TYPE_SIZE_UNIT
2428 by BITS_PER_UNIT to get the TYPE_SIZE.
2429 * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2430 MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2432 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2433 Alan Hayward <alan.hayward@arm.com>
2434 David Sherwood <david.sherwood@arm.com>
2436 * doc/rtl.texi (const_poly_int): Document. Also document the
2437 rtl sharing behavior.
2438 * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2439 * rtl.h (const_poly_int_def): New struct.
2440 (rtx_def::u): Add a cpi field.
2441 (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2442 (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2443 (wi::rtx_to_poly_wide_ref): New typedef
2444 (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2445 (poly_int_rtx_p): New functions.
2446 (trunc_int_for_mode): Declare a poly_int64 version.
2447 (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2448 (immed_wide_int_const): Take a poly_wide_int_ref rather than
2450 (strip_offset): Declare.
2451 (strip_offset_and_add): New function.
2452 * rtl.def (CONST_POLY_INT): New rtx code.
2453 * rtl.c (rtx_size): Handle CONST_POLY_INT.
2454 (shared_const_p): Use poly_int_rtx_p.
2455 * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2457 (gen_int_shift_amount): Likewise.
2458 * emit-rtl.c (const_poly_int_hasher): New class.
2459 (const_poly_int_htab): New variable.
2460 (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2461 (const_poly_int_hasher::hash): New function.
2462 (const_poly_int_hasher::equal): Likewise.
2463 (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2464 (immed_wide_int_const): Rename to...
2465 (immed_wide_int_const_1): ...this and make static.
2466 (immed_wide_int_const): New function, taking a poly_wide_int_ref
2467 instead of a wide_int_ref.
2468 (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2469 (gen_lowpart_common): Handle CONST_POLY_INT.
2470 * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2471 * cselib.c (cselib_hash_rtx): Likewise.
2472 * dwarf2out.c (const_ok_for_output_1): Likewise.
2473 * expr.c (convert_modes): Likewise.
2474 * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2475 * rtlhash.c (add_rtx): Likewise.
2476 * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2477 (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2478 Handle existing CONST_POLY_INT rtxes.
2479 * expmed.h (expand_shift): Take a poly_int64 instead of a
2481 * expmed.c (expand_shift): Likewise.
2482 * rtlanal.c (strip_offset): New function.
2483 (commutative_operand_precedence): Give CONST_POLY_INT the same
2484 precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2486 * rtl-tests.c (const_poly_int_tests): New struct.
2487 (rtl_tests_c_tests): Use it.
2488 * simplify-rtx.c (simplify_const_unary_operation): Handle
2490 (simplify_const_binary_operation): Likewise.
2491 (simplify_binary_operation_1): Fold additions of symbolic constants
2492 and CONST_POLY_INTs.
2493 (simplify_subreg): Handle extensions and truncations of
2495 (simplify_const_poly_int_tests): New struct.
2496 (simplify_rtx_c_tests): Use it.
2497 * wide-int.h (storage_ref): Add default constructor.
2498 (wide_int_ref_storage): Likewise.
2499 (trailing_wide_ints): Use GTY((user)).
2500 (trailing_wide_ints::operator[]): Add a const version.
2501 (trailing_wide_ints::get_precision): New function.
2502 (trailing_wide_ints::extra_size): Likewise.
2504 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2505 Alan Hayward <alan.hayward@arm.com>
2506 David Sherwood <david.sherwood@arm.com>
2508 * emit-rtl.h (gen_int_shift_amount): Declare.
2509 * emit-rtl.c (gen_int_shift_amount): New function.
2510 * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2512 * calls.c (shift_return_value): Likewise.
2513 * cse.c (fold_rtx): Likewise.
2514 * dse.c (find_shift_sequence): Likewise.
2515 * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2516 (expand_shift, expand_smod_pow2): Likewise.
2517 * lower-subreg.c (shift_cost): Likewise.
2518 * optabs.c (expand_superword_shift, expand_doubleword_mult)
2519 (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2520 (expand_vec_perm_var): Likewise.
2521 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2522 (simplify_binary_operation_1): Likewise.
2523 * combine.c (try_combine, find_split_point, force_int_to_mode)
2524 (simplify_shift_const_1, simplify_shift_const): Likewise.
2525 (change_zero_ext): Likewise. Use simplify_gen_binary.
2527 2017-12-20 Richard Sandiford <richard.sandiford@linaro.org>
2529 * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2531 2017-12-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2533 * doc/invoke.texi (ARM Options): Document accepted extension options
2534 for -march=armv8.3-a.
2536 2017-12-20 Richard Earnshaw <rearnsha@arm.com>
2539 * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2540 or --with-float=softfp, set the default CPU to arm10e.
2542 2017-12-20 Eric Botcazou <ebotcazou@adacore.com>
2544 * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2545 * config/visium/predicates.md (const_shift_operand): Likewise.
2546 * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2547 (visium_legitimize_reload_address): Likewise.
2549 2017-12-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2552 * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2554 2017-12-20 Jakub Jelinek <jakub@redhat.com>
2557 * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2559 2017-12-20 Julia Koval <julia.koval@intel.com>
2561 * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2562 OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2563 (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2564 * config.gcc: Include vpclmulqdqintrin.h.
2565 * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2566 * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2568 * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2569 __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2570 * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2571 * config/i386/i386.c (isa2_opts): Add -mcx16.
2572 (isa_opts): Add -mpclmulqdq, remove -mcx16.
2573 (ix86_option_override_internal): Move mcx16 to flags2.
2574 (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2575 (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2576 * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2577 * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2578 * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2579 * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2580 * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2581 _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2582 * doc/invoke.texi: Add -mvpclmulqdq.
2584 2017-12-20 Tom de Vries <tom@codesourcery.com>
2587 * config/i386/i386.c (ix86_static_chain): Move
2588 DECL_STATIC_CHAIN test ...
2589 * calls.c (rtx_for_static_chain): ... here. New function.
2590 * calls.h (rtx_for_static_chain): Declare.
2591 * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2592 instead of targetm.calls.static_chain.
2593 * df-scan.c (df_get_entry_block_def_set): Same.
2595 2017-12-19 Tom de Vries <tom@codesourcery.com>
2597 PR tree-optimization/83493
2598 * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2601 2017-12-19 Jakub Jelinek <jakub@redhat.com>
2603 * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2604 inform with hardcoded english plural handling.
2606 2017-12-18 Jeff Law <law@redhat.com>
2608 PR tree-optimization/83477
2609 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2610 a non-virtual PHI, always push a new range.
2612 2017-12-19 Martin Sebor <msebor@redhat.com>
2615 * builtins.c (compute_objsize): Handle non-constant offsets.
2617 2017-12-19 Jakub Jelinek <jakub@redhat.com>
2619 PR tree-optimization/83444
2620 * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2621 character load case, if get_stridx on MEM_REF's operand doesn't
2622 look usable, retry with get_addr_stridx.
2624 2017-12-19 Alexandre Oliva <aoliva@redhat.com>
2627 * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2628 (variable_tracking_main_1): Keep markers even when VTA fails.
2631 * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2632 even if there are markers before it.
2633 * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2635 2017-12-19 Jakub Jelinek <jakub@redhat.com>
2637 * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2638 typical order conditions.
2639 * sel-sched.c (extract_new_fences_from): Likewise.
2640 * config/visium/constraints.md (J, K, L): Likewise.
2641 * config/visium/predicates.md (const_shift_operand): Likewise.
2642 * config/visium/visium.c (visium_legitimize_address,
2643 visium_legitimize_reload_address): Likewise.
2644 * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2645 * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2646 * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2647 * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2648 SET_DUMP_DETAIL): Likewise.
2649 * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2650 * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2651 avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2652 avr_set_core_architecture, avr_set_current_function,
2653 avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2654 output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2655 avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2656 avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2657 avr_map_decompose, avr_fold_builtin): Likewise.
2658 * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2659 * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2660 * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2661 * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2662 * config/m32c/m32c.c (m32c_conditional_register_usage,
2663 m32c_address_cost): Likewise.
2664 * config/m32c/predicates.md (shiftcount_operand,
2665 longshiftcount_operand): Likewise.
2666 * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2667 * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2668 can_use_cdx_ldstw): Likewise.
2669 * config/nios2/nios2.h (CDX_REG_P): Likewise.
2670 * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2672 * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2673 * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2674 * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2675 * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2676 * config/vax/vax.c (adjacent_operands_p): Likewise.
2677 * config/ft32/constraints.md (L, b, KA): Likewise.
2678 * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2680 * cfgexpand.c (expand_stack_alignment): Likewise.
2681 * gcse.c (insert_expr_in_table): Likewise.
2682 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2683 * cgraphunit.c (cgraph_node::expand): Likewise.
2684 * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2685 * emit-rtl.c (add_insn): Likewise.
2686 * input.c (dump_location_info): Likewise.
2687 * passes.c (NEXT_PASS): Likewise.
2688 * read-rtl-function.c (parse_note_insn_name,
2689 function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2691 * sched-rgn.c (sched_rgn_init): Likewise.
2692 * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2693 * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2694 if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2696 * explow.c (eliminate_constant_term): Likewise.
2697 * final.c (leaf_renumber_regs_insn): Likewise.
2698 * cfgrtl.c (print_rtl_with_bb): Likewise.
2699 * genhooks.c (emit_init_macros): Likewise.
2700 * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2701 * tree-data-ref.c (conflict_fn): Likewise.
2702 * selftest.c (assert_streq): Likewise.
2703 * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2704 * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2705 fold_binary_loc, multiple_of_p): Likewise.
2706 * reload.c (push_reload, find_equiv_reg): Likewise.
2707 * et-forest.c (et_nca, et_below): Likewise.
2708 * dbxout.c (dbxout_symbol_location): Likewise.
2709 * reorg.c (relax_delay_slots): Likewise.
2710 * dojump.c (do_compare_rtx_and_jump): Likewise.
2711 * gengtype-parse.c (type): Likewise.
2712 * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2713 simplify_const_relational_operation): Likewise.
2714 * reload1.c (do_output_reload): Likewise.
2715 * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2716 * gengtype.c (type_for_name): Likewise.
2717 * gimple-ssa-sprintf.c (format_directive): Likewise.
2719 2017-12-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2722 * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2723 accessing it. Adjust comment.
2725 2017-12-19 Jakub Jelinek <jakub@redhat.com>
2728 * predict.c (zero_one_minusone): New function.
2729 (apply_return_prediction): Avoid return prediction for functions
2730 returning only -1, 0 and 1 values, unless they only return -1 and 0
2733 2017-12-19 Claudiu Zissulescu <claziss@synopsys.com>
2735 * config/arc/arc.c (legitimate_scaled_address_p): Clean
2736 fall-through warning.
2737 (arc_compute_frame_size): Remove unused variables.
2738 (arc_print_operand): Fix fprintif format.
2739 (arc_can_follow_jump): Clean fall-through warning.
2741 2017-12-19 Marek Polacek <polacek@redhat.com>
2744 * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2747 2017-12-19 Claudiu Zissulescu <claziss@synopsys.com>
2749 * config/arc/arc.c (overriderregs): New variable.
2750 (arc_override_options): Track fixed/call saved/call options.
2751 (arc_conditional_register_usage): Check against overrideregs
2752 variable whenever we change register properties.
2754 2017-12-19 Nathan Sidwell <nathan@acm.org>
2756 * opts.c (finish_options): Don't prefix dump_base_name if it
2757 already contains directories.
2759 2017-12-19 Martin Liska <mliska@suse.cz>
2761 PR rtl-optimization/82675
2762 * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2763 more element in sbitmap.
2765 2017-12-19 Paolo Carlini <paolo.carlini@oracle.com>
2767 * gimplify.c (gimplify_expr): Use error_operand_p.
2769 2017-12-19 Sebastian Huber <sebastian.huber@embedded-brains.de>
2772 * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2775 2017-12-19 Daniel Cederman <cederman@gaisler.com>
2777 * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2778 the jump is to a label.
2780 2017-12-19 Jakub Jelinek <jakub@redhat.com>
2782 PR tree-optimization/83444
2783 * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2788 * ipa-inline.c (flatten_remove_node_hook): New function.
2789 (ipa_inline): Keep only nodes with flatten attribute at the end of
2790 the array in the order from ipa_reverse_postorder, only walk that
2791 portion of array for flattening, if there is more than one such
2792 node, temporarily register a removal hook and ignore removed nodes.
2794 PR tree-optimization/80631
2795 * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2796 induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2797 IFN_REDUC_MAX or IFN_REDUC_MIN.
2799 2017-12-18 Jeff Law <law@redhat.com>
2801 * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2802 record symbolic equivalences from backedges in the CFG.
2805 2017-11-19 Jeff Law <law@redhat.com>
2807 * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2808 of degenerates resulting from ignoring an edge.
2810 2017-12-18 Martin Sebor <msebor@redhat.com>
2813 PR tree-optimization/78450
2814 * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2815 (handle_builtin_strlen): Call it.
2817 2017-12-18 Segher Boessenkool <segher@kernel.crashing.org>
2819 PR rtl-optimization/83424
2820 * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2822 2017-12-18 Marek Polacek <polacek@redhat.com>
2825 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2826 Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2829 2017-12-18 Claudiu Zissulescu <claziss@synopsys.com>
2831 * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2832 (umaddsidi4, umaddsidi_split): Likewise.
2834 2017-12-18 Claudiu Zissulescu <claziss@synopsys.com>
2836 * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2839 2017-12-18 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2842 * config/s390/s390.c (s390_option_override): Avoid strncpy.
2844 2017-12-18 Richard Biener <rguenther@suse.de>
2846 PR tree-optimization/81877
2847 * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
2848 (outermost_indep_loop): Adjust.
2849 (ref_indep_loop_p_1): Likewise. Remove safelen handling again.
2850 (can_sm_ref_p): Adjust.
2852 2017-12-18 Richard Biener <rguenther@suse.de>
2855 * tree.c (array_at_struct_end_p): Return true if the underlying
2856 object has space for at least one element in excess of what
2857 the array domain specifies.
2859 2017-12-17 Sandra Loosemore <sandra@codesourcery.com>
2861 * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
2862 example to avoid overfull hbox.
2863 * doc/invoke.texi (Option Summary): Add missing @gol.
2864 (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
2867 2017-12-17 Sandra Loosemore <sandra@codesourcery.com>
2868 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2870 * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
2871 to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
2872 (AArch64 Options): Document -mverbose-cost-dump.
2873 (ARM Options): Likewise, plus -mflip-thumb.
2875 2017-12-17 Martin Sebor <msebor@redhat.com>
2878 * gimple-ssa-warn-restrict.c
2879 (builtin_memref::offset_out_of_bounds): Correct the handling of
2882 2017-12-17 Sandra Loosemore <sandra@codesourcery.com>
2884 * doc/invoke.texi (ARC Options): Add missing -mlra entry.
2886 2017-12-17 John David Anglin <danglin@gcc.gnu.org>
2888 * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
2890 2017-12-17 Markus Trippelsdorf <markus@trippelsdorf.de>
2892 * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
2895 2017-12-16 Sandra Loosemore <sandra@codesourcery.com>
2897 * doc/invoke.texi: Fix some typos.
2899 2017-12-16 Martin Sebor <msebor@redhat.com>
2901 PR tree-optimization/78918
2902 * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
2903 * builtins.c (check_sizes): Rename...
2904 (check_access): ...to this. Rename function arguments for clarity.
2905 (check_memop_sizes): Adjust names.
2906 (expand_builtin_memchr, expand_builtin_memcpy): Same.
2907 (expand_builtin_memmove, expand_builtin_mempcpy): Same.
2908 (expand_builtin_strcat, expand_builtin_stpncpy): Same.
2909 (check_strncat_sizes, expand_builtin_strncat): Same.
2910 (expand_builtin_strncpy, expand_builtin_memset): Same.
2911 (expand_builtin_bzero, expand_builtin_memcmp): Same.
2912 (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
2913 (maybe_emit_sprintf_chk_warning): Same.
2914 (expand_builtin_strcpy): Adjust.
2915 (expand_builtin_stpcpy): Same.
2916 (expand_builtin_with_bounds): Detect out-of-bounds accesses
2917 in pointer-checking forms of memcpy, memmove, and mempcpy.
2918 (gcall_to_tree_minimal, max_object_size): Define new functions.
2919 * builtins.h (max_object_size): Declare.
2920 * calls.c (alloc_max_size): Call max_object_size instead of
2921 hardcoding ssizetype limit.
2922 (get_size_range): Handle new argument.
2923 * calls.h (get_size_range): Add a new argument.
2924 * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
2925 * doc/invoke.texi (-Wrestrict): Adjust, add example.
2926 * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
2928 (gimple_fold_builtin_memory_chk): Same.
2929 (gimple_fold_builtin_stxcpy_chk): New function.
2930 * gimple-ssa-warn-restrict.c: New source.
2931 * gimple-ssa-warn-restrict.h: New header.
2932 * gimple.c (gimple_build_call_from_tree): Propagate location.
2933 * passes.def (pass_warn_restrict): Add new pass.
2934 * tree-pass.h (make_pass_warn_restrict): Declare.
2935 * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
2937 (handle_builtin_strcat): Same.
2938 (strlen_optimize_stmt): Rename...
2939 (strlen_check_and_optimize_stmt): ...to this. Handle strncat,
2940 stpncpy, strncpy, and their checking forms.
2942 2017-12-16 Jan Hubicka <hubicka@ucw.cz>
2944 PR rtl-optimization/82849
2945 * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
2946 and get_max_loop_iterations_int.
2948 2017-12-16 Richard Sandiford <richard.sandiford@linaro.org>
2949 Alan Hayward <alan.hayward@arm.com>
2950 David Sherwood <david.sherwood@arm.com>
2952 * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
2953 (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
2954 size as a poly_uint64.
2955 (mode_for_vector, mode_for_int_vector): Take the number of vector
2956 elements as a poly_uint64.
2957 * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
2958 size as a poly_uint64.
2959 (mode_for_vector, mode_for_int_vector): Take the number of vector
2960 elements as a poly_uint64.
2962 2017-12-16 Richard Sandiford <richard.sandiford@linaro.org>
2963 Alan Hayward <alan.hayward@arm.com>
2964 David Sherwood <david.sherwood@arm.com>
2966 * machmode.h (MACRO_MODE): New macro.
2967 * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
2968 * targhooks.c (default_libcall_value, default_secondary_reload)
2969 (default_memory_move_cost, default_register_move_cost)
2970 (default_class_max_nregs): Likewise.
2972 2017-12-16 Richard Sandiford <richard.sandiford@linaro.org>
2973 Alan Hayward <alan.hayward@arm.com>
2974 David Sherwood <david.sherwood@arm.com>
2976 * doc/sourcebuild.texi: Document IN_TARGET_CODE.
2977 * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
2979 * genautomata.c (main): Likewise.
2980 * genconditions.c (write_header): Likewise.
2981 * genemit.c (main): Likewise.
2982 * genextract.c (print_header): Likewise.
2983 * genopinit.c (main): Likewise.
2984 * genoutput.c (output_prologue): Likewise.
2985 * genpeep.c (main): Likewise.
2986 * genpreds.c (write_insn_preds_c): Likewise.
2987 * genrecog.c (writer_header): Likewise.
2988 * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
2989 * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
2990 * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
2991 * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
2992 * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
2993 * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
2994 * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
2995 * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
2996 * config/arc/arc.c (IN_TARGET_CODE): Likewise.
2997 * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
2998 * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
2999 * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3000 * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3001 * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3002 * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3003 * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3004 * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3005 * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3006 * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3007 * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3008 * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3009 * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3010 * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3011 * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3012 * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3013 * config/darwin.c (IN_TARGET_CODE): Likewise.
3014 * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3015 * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3016 * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3017 * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3018 * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3019 * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3020 * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3021 * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3022 * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3023 * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3024 * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3025 * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3026 * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3027 * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3028 * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3029 * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3030 * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3031 * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3032 * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3033 * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3034 * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3035 * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3036 * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3037 * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3038 * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3039 * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3040 * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3041 * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3042 * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3043 * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3044 * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3045 * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3046 * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3047 * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3048 * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3049 * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3050 * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3051 * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3052 * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3053 * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3054 * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3055 * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3056 * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3057 * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3058 * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3059 * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3060 * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3061 * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3062 * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3063 * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3064 * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3065 * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3066 * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3067 * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3068 * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3069 * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3070 * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3071 * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3072 * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3073 * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3074 * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3075 * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3076 * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3077 * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3078 * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3079 * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3080 * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3081 * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3082 * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3083 * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3084 * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3085 * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3086 * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3087 * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3088 * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3089 * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3090 * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3091 * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3092 * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3093 * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3094 * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3095 * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3096 * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3097 * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3098 * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3099 * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3100 * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3101 * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3102 * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3103 * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3104 * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3105 * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3106 * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3107 * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3108 * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3109 * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3110 * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3111 * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3112 * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3113 * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3114 * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3115 * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3116 * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3117 * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3118 * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3119 * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3121 2017-12-16 Richard Sandiford <richard.sandiford@linaro.org>
3122 Alan Hayward <alan.hayward@arm.com>
3123 David Sherwood <david.sherwood@arm.com>
3125 * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3128 2017-12-16 Richard Sandiford <richard.sandiford@linaro.org>
3129 Alan Hayward <alan.hayward@arm.com>
3130 David Sherwood <david.sherwood@arm.com>
3132 * doc/generic.texi (VEC_SERIES_EXPR): Document.
3133 * doc/md.texi (vec_series@var{m}): Document.
3134 * tree.def (VEC_SERIES_EXPR): New tree code.
3135 * tree.h (build_vec_series): Declare.
3136 * tree.c (build_vec_series): New function.
3137 * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3138 * tree-pretty-print.c (dump_generic_node): Likewise.
3139 * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3140 * tree-inline.c (estimate_operator_cost): Likewise.
3141 * expr.c (expand_expr_real_2): Likewise.
3142 * optabs-tree.c (optab_for_tree_code): Likewise.
3143 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3144 * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3145 * expmed.c (make_tree): Handle VEC_SERIES.
3146 * optabs.def (vec_series_optab): New optab.
3147 * optabs.h (expand_vec_series_expr): Declare.
3148 * optabs.c (expand_vec_series_expr): New function.
3149 * tree-vect-generic.c (expand_vector_operations_1): Check that
3150 the operands also have vector type.
3152 2017-12-16 Richard Sandiford <richard.sandiford@linaro.org>
3153 Alan Hayward <alan.hawyard@arm.com>
3154 David Sherwood <david.sherwood@arm.com>
3156 * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3157 (VEC_COND_EXPR): Add missing @tindex.
3158 * doc/md.texi (vec_duplicate@var{m}): Document.
3159 * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3160 * tree.c (build_vector_from_val): Add stubbed-out handling of
3161 variable-length vectors, using VEC_DUPLICATE_EXPR.
3162 (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3163 * cfgexpand.c (expand_debug_expr): Likewise.
3164 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3165 * tree-inline.c (estimate_operator_cost): Likewise.
3166 * tree-pretty-print.c (dump_generic_node): Likewise.
3167 * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3168 * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3169 (test_vec_duplicate_folding): New function.
3170 (fold_const_c_tests): Call it.
3171 * optabs.def (vec_duplicate_optab): New optab.
3172 * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3173 * optabs.h (expand_vector_broadcast): Declare.
3174 * optabs.c (expand_vector_broadcast): Make non-static. Try using
3175 vec_duplicate_optab.
3176 * expr.c (store_constructor): Try using vec_duplicate_optab for
3178 (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3180 2017-12-15 Markus Trippelsdorf <markus@trippelsdorf.de>
3183 * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3184 div/mod latencies a bit.
3186 2017-12-15 Jeff Law <law@redhat.com>
3188 PR tree-optimization/36550
3189 * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3190 (mark_threaded_blocks): Rewrite code to avoid block copying when
3191 optimizing for size. Don't pessimize blocks which will be
3192 copied, but all the statements will be dead.
3194 2017-12-15 Alexandre Oliva <aoliva@redhat.com>
3196 PR tree-optimization/81165
3197 * tree-ssa-threadupdate.c (uses_in_bb): New.
3198 (estimate_threading_killed_stmts): New.
3199 * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3200 * tree-ssa-threadedge.c
3201 (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3204 2017-12-15 Jeff Law <law@redhat.com>
3206 PR tree-optimization/83410
3207 * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3208 threads when parallelizing loops.
3210 2017-12-15 Jakub Jelinek <jakub@redhat.com>
3212 * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3214 * config/alpha/alpha.c (vms_attribute_table): Swap
3215 affects_type_identity and handler fields, adjust comments.
3216 * config/mips/mips.c (mips_attribute_table): Likewise.
3217 * config/visium/visium.c (visium_attribute_table): Likewise.
3218 * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3219 * config/microblaze/microblaze.c (microblaze_attribute_table):
3221 * config/spu/spu.c (spu_attribute_table): Likewise.
3222 * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3223 * config/arc/arc.c (arc_attribute_table): Likewise.
3224 * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3225 * config/v850/v850.c (v850_handle_interrupt_attribute,
3226 v850_handle_data_area_attribute): Formatting fixes.
3227 (v850_attribute_table): Swap affects_type_identity and handler
3228 fields, adjust comments.
3229 * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3230 * config/arm/arm.c (arm_attribute_table): Likewise.
3231 * config/avr/avr.c (avr_attribute_table): Likewise.
3232 * config/s390/s390.c (s390_attribute_table): Likewise.
3233 * config/sh/sh.c (sh_attribute_table): Likewise.
3234 * config/i386/i386.c (ix86_handle_cconv_attribute,
3235 ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3236 (ix86_attribute_table): Swap affects_type_identity and handler
3237 fields, adjust comments.
3238 * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3239 * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3240 * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3241 * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3242 * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3243 * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3244 * config/rx/rx.c (rx_attribute_table): Likewise.
3245 * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3246 * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3247 * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3248 * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3249 * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3250 * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3251 * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3252 * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3253 * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3254 * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3255 * doc/plugins.texi (user_attr): Likewise. Add NULL for
3257 * attribs.c (empty_attribute_table): Swap affects_type_identity and
3259 (register_scoped_attributes, decl_attributes): Formatting fixes.
3261 PR tree-optimization/83269
3262 * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3263 subtraction in arg0's type if type is signed and arg0 is unsigned.
3267 * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3268 convert? on @0 instead of convert. Check type of @1, not @0.
3269 ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3270 convert? on @0 instead of convert. Check type of @1, not @0.
3271 ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3272 only optimize if either both @1 and @2 types are narrower
3273 precision, or both are wider or equal precision, and in the former
3274 case only if both have undefined overflow.
3276 2017-12-15 Richard Biener <rguenther@suse.de>
3279 * internal-fn.def (IFN_NOP): Add.
3280 * internal-fn.c (expand_NOP): Do nothing.
3281 * lto-streamer-in.c (input_function): Instead of removing
3282 sanitizer calls replace them with IFN_NOP calls.
3284 2017-12-15 Richard Sandiford <richard.sandiford@linaro.org>
3285 Alan Hayward <alan.hayward@arm.com>
3286 David Sherwood <david.sherwood@arm.com>
3288 * dse.c (store_info, read_info_type): Replace begin and end with
3290 (print_range): New function.
3291 (set_all_positions_unneeded, any_positions_needed_p)
3292 (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3294 (record_store): Likewise. Optimize the case in which all positions
3296 (get_stored_val): Replace read_begin and read_end with read_offset
3298 (replace_read): Update call accordingly.
3300 2017-12-15 Bin Cheng <bin.cheng@arm.com>
3302 * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3303 (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3304 (loop_cand::loop_cand): Initialize above members.
3305 (loop_cand::supported_operations): Delete.
3306 (loop_cand::can_interchange_p): Inline above function.
3307 (loop_cand::classify_simple_reduction): Record number of constant
3308 initialized simple reductions.
3309 (should_interchange_loops): New parameters. Check stmt cost of loops
3311 (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3312 Update call to should_interchange_loops.
3313 (should_interchange_loop_nest): Update call to
3314 should_interchange_loops.
3316 2017-12-15 Eric Botcazou <ebotcazou@adacore.com>
3319 * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3320 * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3321 * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3323 2017-12-15 Julia Koval <julia.koval@intel.com>
3325 * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3326 __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3328 * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3329 * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3330 _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3332 2017-12-15 Julia Koval <julia.koval@intel.com>
3334 * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3335 __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3336 * config/i386/sse.md (vaesenc_<mode>): New pattern.
3337 * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3338 _mm_aesenc_epi128): New intrinsics.
3340 2017-12-15 Julia Koval <julia.koval@intel.com>
3342 * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3343 __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3345 * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3346 * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3347 _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3349 2017-12-14 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3351 * gimple-ssa-strength-reduction.c (analyze_increments):
3352 Distinguish replacement costs for constant strides from those for
3355 2017-12-14 Jakub Jelinek <jakub@redhat.com>
3357 * var-tracking.c (variable_tracking_main_1): Formatting fix.
3359 2017-12-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
3361 * doc/invoke.texi: Document -Wcast-function-type.
3362 * recog.h (stored_funcptr): Change signature.
3363 * tree-dump.c (dump_node): Avoid warning.
3364 * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3366 2017-12-14 Qing Zhao <qing.zhao@oracle.com>
3369 * gimple-fold.c (get_range_strlen): Add the handling of non-member
3372 2017-12-14 David Malcolm <dmalcolm@redhat.com>
3374 PR tree-optimization/83312
3375 * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3376 * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3377 "val" param, and to cope with arbitrary basic blocks.
3378 (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3379 handle NULL_TREE for "val", dropping "bb" param.
3380 (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3381 drop "bb" param. Handle NULL_TREE for "val".
3382 (find_case_label_for_value): Make "switch_stmt" param const.
3383 * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3385 (vrp_prop::check_all_array_refs): Reimplement as...
3386 (check_array_bounds_dom_walker::before_dom_children): ...this new
3387 vfunc. Replace linear search through BB block list, excluding
3388 those with non-executable in-edges via dominator walk.
3390 2017-12-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3392 * config/arm/arm.opt (mverbose-cost-dump): New option.
3393 * config/arm/arm.c (arm_rtx_costs): Use it.
3395 2017-12-14 Andreas Schwab <schwab@linux-m68k.org>
3398 * reload1.c (emit_input_reload_insns): Skip debug markers.
3400 2017-12-14 Alexandre Oliva <aoliva@redhat.com>
3402 * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3406 * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3407 * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3408 * config/sh/sh.c: Skip debug insns besides notes.
3409 * config/sh/sh.md: Likewise.
3410 * config/sh/sh_treg_combine.cc: Likewise.
3411 * config/sh/sync.md: Likewise.
3413 2017-12-14 Tom de Vries <tom@codesourcery.com>
3415 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3416 Add item for weak_undefined.
3418 2017-12-14 Richard Biener <rguenther@suse.de>
3420 PR tree-optimization/67842
3421 * sese.h (bb_in_region): Remove #if 0'ed code.
3423 2017-12-14 Richard Biener <rguenther@suse.de>
3425 PR tree-optimization/83326
3426 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3427 may_be_zero parameter and handle it by not marking the first
3428 peeled copy as not exiting the loop.
3429 (try_peel_loop): Likewise.
3430 (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3431 to handle the case of constant or zero iterations and perform
3432 loop header copying on-the-fly.
3434 2017-12-14 Richard Biener <rguenther@suse.de>
3436 PR tree-optimization/83418
3437 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3438 Instead of asserting we don't get unfolded comparisons deal with them.
3440 2017-12-14 Jakub Jelinek <jakub@redhat.com>
3443 * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3444 basic blocks. Assert debug bind insns don't appear outside of bbs,
3445 don't reset them. Assert insns without BLOCK_FOR_INSN are outside of
3448 PR tree-optimization/83198
3449 * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3450 dir.modifier, regardless of TREE_TYPE (arg). Assume non-REAL_CST
3451 value if arg is a REAL_CST with incompatible type.
3453 2017-12-14 Sudakshina Das <sudi.das@arm.com>
3454 Bin Cheng <bin.cheng@arm.com>
3457 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3459 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3462 2017-12-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3464 * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3465 cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3467 2017-12-14 Richard Sandiford <richard.sandiford@linaro.org>
3468 Alan Hayward <alan.hayward@arm.com>
3469 David Sherwood <david.sherwood@arm.com>
3471 * poly-int.h: New file.
3472 * poly-int-types.h: Likewise.
3473 * coretypes.h: Include them.
3474 (POLY_INT_CONVERSION): Define.
3475 * target.def (estimated_poly_value): New hook.
3476 * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3477 * doc/tm.texi: Regenerate.
3478 * doc/poly-int.texi: New file.
3479 * doc/gccint.texi: Include it.
3480 * doc/rtl.texi: Describe restrictions on subreg modes.
3481 * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3482 * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3483 (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3484 * targhooks.h (default_estimated_poly_value): Declare.
3485 * targhooks.c (default_estimated_poly_value): New function.
3486 * target.h (estimated_poly_value): Likewise.
3487 * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3488 (wi::unary_traits): Delete.
3489 (wi::binary_traits::signed_shift_result_type): Define for
3490 offset_int << HOST_WIDE_INT, etc.
3491 (generic_wide_int::operator <<=): Define for all types and use
3492 wi::lshift instead of <<.
3493 (wi::hwi_with_prec): Add a default constructor.
3494 (wi::ints_for): New class.
3495 (operator <<): Define for all wide-int types.
3496 (operator /): New function.
3497 (operator %): Likewise.
3498 * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3499 (ASSERT_MAYBE_NE_AT): New macros.
3501 2017-12-13 Eric Botcazou <ebotcazou@adacore.com>
3502 Dominik Vogt <vogt@linux.vnet.ibm.com>
3505 * emit-rtl.c (init_emit): Remove ??? comment.
3506 * explow.c (get_dynamic_stack_size): Take known alignment of stack
3507 pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3508 * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3509 alignment of 3 virtual registers to BITS_PER_WORD.
3511 * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3513 2017-12-13 Peter Bergner <bergner@vnet.ibm.com>
3515 * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3516 * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3518 2017-12-13 Alexandre Oliva <aoliva@redhat.com>
3521 * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3523 2017-12-13 Jakub Jelinek <jakub@redhat.com>
3525 * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3526 with the exception of debug begin stmt markers appear before
3530 * final.c (rest_of_handle_final): Call variable_tracking_main only
3531 if !flag_var_tracking.
3533 2017-12-13 Alexandre Oliva <aoliva@redhat.com>
3534 Jakub Jelinek <jakub@redhat.com>
3538 * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3539 labels debug stmts that can only appear after labels.
3541 2017-12-13 Alexander Monakov <amonakov@ispras.ru>
3543 PR rtl-optimization/82398
3544 * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3545 EXPR_USEFULNESS in priority comparison.
3547 2017-12-13 Segher Boessenkool <segher@kernel.crashing.org>
3549 PR rtl-optimization/83393
3550 * combine.c (move_deaths): If reg_stat points to a too new insn in
3551 last_death, do not use it: find the proper insn instead.
3553 2017-12-12 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3556 * config/rs6000/rs6000-string.c (expand_block_compare,
3557 expand_strn_compare): Fix set_mem_size() calls.
3559 2017-12-12 Jeff Law <law@redhat.com>
3561 PR tree-optimization/83298
3562 PR tree-optimization/83362
3563 PR tree-optimization/83383
3564 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3565 push_value_range a public interface. Add new argument to
3566 record_ranges_from_stmt.
3567 * gimple-ssa-evrp-analyze.c
3568 (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3569 Update comments. Handle recording temporary equivalences.
3570 * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3571 new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3572 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3573 * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3574 gimple-ssa-evrp-analyze.h.
3575 (record_temporary_equivalences_from_phis): Add new argument. When
3576 the PHI arg is an SSA_NAME, set the result's range to the range
3578 (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3579 from statements too.
3580 (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3581 Pass it down to children as needed.
3582 (thread_outgoing_edges): Likewise.
3583 (thread_across_edge): Likewise. Push/pop range state as needed.
3584 * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3586 2017-12-12 Julia Koval <julia.koval@intel.com>
3588 * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3589 (PTA_CANNONLAKE): Remove PTA_CLWB.
3591 2017-12-12 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3594 * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3595 (vcondv2div2df): Likewise.
3596 (vconduv2dfv2di): Likewise.
3598 2017-12-12 Jakub Jelinek <jakub@redhat.com>
3600 * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3601 BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3602 BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3603 ATTR_MATHFN_FPROUNDING_ERRNO.
3605 2017-12-12 Richard Biener <rguenther@suse.de>
3607 PR tree-optimization/83385
3608 * graphite-scop-detection.c (get_order, order): Remove.
3609 (bb_to_rpo): New global.
3611 (build_scops): Sort pbbs in RPO order.
3613 2017-12-12 James Greenhalgh <james.greenhalgh@arm.com>
3615 * combine.c (simplify_set): Do not transform subregs to zero_extends
3616 if the destination mode is a vector mode.
3618 2017-12-12 Jakub Jelinek <jakub@redhat.com>
3620 PR tree-optimization/83359
3621 * tree-cfg.h (fold_loop_internal_call): Declare.
3622 * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3623 * tree-cfg.c (fold_loop_internal_call): ... here. No longer static.
3624 (find_loop_dist_alias): New function.
3625 (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3626 updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3629 PR tree-optimization/80631
3630 * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3631 (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3632 arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3633 hardcoding zero as the value if COND_EXPR is never true. For
3634 INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3635 INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3636 hardcoding MAX_EXPR as the reduction operation.
3637 (is_nonwrapping_integer_induction): Allow negative step.
3638 (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3639 vect_create_epilog_for_reduction, if no value is suitable, don't
3640 use INTEGER_INDUC_COND_REDUCTION for now. Formatting fixes.
3642 2017-12-12 Richard Biener <rguenther@suse.de>
3644 PR tree-optimization/81889
3645 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3646 range info from the non-wrapping IV instead of just the range
3649 2017-12-12 Julia Koval <julia.koval@intel.com>
3651 * config.gcc: Add vaesintrin.h.
3652 * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3654 * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3655 __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3657 * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3658 * config/i386/immintrin.h: Include vaesintrin.h.
3659 * config/i386/sse.md (vaesdec_<mode>): New pattern.
3660 * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3661 _mm_aesdec_epi128): New intrinsics.
3663 2017-12-12 Julia Koval <julia.koval@intel.com>
3665 * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3666 OPTION_MASK_ISA_VAES_UNSET): New.
3667 (ix86_handle_option): Handle -mvaes.
3668 * config/i386/cpuid.h: Define bit_VAES.
3669 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3670 * config/i386/i386-c.c (__VAES__): New.
3671 * config/i386/i386.c (ix86_target_string): Add -mvaes.
3672 (ix86_valid_target_attribute_inner_p): Ditto.
3673 * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3674 * config/i386/i386.opt: Add -mvaes.
3675 * doc/invoke.texi: Ditto.
3677 2017-12-12 Alexandre Oliva <aoliva@redhat.com>
3679 * debug.h (gcc_debug_hooks): Add inline_entry.
3680 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3681 * debug.c (do_nothing_debug_hooks): Likewise.
3682 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3683 * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3684 (dwarf2_lineno_debug_hooks): Likewise.
3686 * common.opt (gstatement-frontiers): New, setting
3687 debug_nonbind_markers_p.
3688 * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3689 * toplev.c (process_options): Autodetect value for debug statement
3691 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3692 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3694 * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3695 markers. Integrate source bind into debug stmt expand loop.
3696 (pass_expand::execute): Check debug marker limit. Avoid deep
3697 TER and expand debug locations for debug bind insns only.
3698 * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3700 * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3701 * final.c (reemit_insn_block_notes): Take current block from
3702 nonbind markers. Declare note where it's first set.
3703 (final_scan_insn): Handle begin stmt notes. Emit is_stmt according to
3704 begin stmt markers if enabled.
3705 (notice_source_line): Handle nonbind markers. Fail if their
3706 location is unknown or that of builtins.
3707 (rest_of_handle_final): Convert begin stmt markers to notes if
3708 var-tracking didn't run.
3709 (rest_of_clean_state): Skip begin stmt markers.
3710 * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3712 * function.c (allocate_struct_function): Set begin_stmt_markers.
3713 * function.h (struct function): Add debug_marker_count counter
3714 and debug_nonbind_markers flag.
3715 * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3716 * gimple-low.c (lower_function_body): Adjust
3717 debug_nonbind_markers.
3718 (lower_stmt): Drop or skip gimple debug stmts.
3719 (lower_try_catch): Skip debug stmts.
3720 * gimple.c (gimple_build_debug_begin_stmt): New.
3721 (gimple_copy): Increment debug_marker_count if copying one.
3722 * gimple.h (gimple_build_debug_begin_stmt): Declare.
3723 * gimplify.c (rexpr_location): New.
3724 (rexpr_has_location): New.
3725 (warn_switch_unreachable_r): Handle gimple debug stmts.
3726 (shortcut_cond_r): Call expr_location.
3728 (find_goto_label): New.
3729 (shortcut_cond_expr): Call expr_has_location, expr_location, and
3731 (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3733 (gimplify_expr): Handle begin stmt markers. Reject debug expr decls.
3734 * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New. Add to...
3735 (LANG_HOOKS_INITIALIZER): ... this.
3736 * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3737 * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3739 (update_ebb_live_info): Skip debug insn markers.
3740 * lra.c (debug_insn_static_data): Rename to...
3741 (debug_bind_static_data): ... this.
3742 (debug_marker_static_data): New.
3743 (lra_set_insn_recog_data): Select one of the above depending
3745 (lra_update_isn_regno_info): Don't assume debug insns have
3747 (push_insns): Skip debug insns.
3748 * lto-streamer-in.c (input_function): Drop debug stmts
3749 depending on active options. Adjust debug_nonbind_markers.
3750 * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3751 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3752 begin stmt marker notes.
3753 (print_insn): Likewise.
3754 * recog.c (extract_insn): Recognize rtl for debug markers.
3755 * rtl.def (DEBUG_MARKER): New.
3756 * tree-inline.c: Include params.h.
3757 (remap_gimple_stmt): Handle nonbind markers.
3758 (maybe_move_debug_stmts_to_successors): Likewise.
3759 (copy_debug_stmt): Likewise.
3760 * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3761 markers regardless of no side effects.
3762 (tsi_link_before): Don't update container's side effects when adding
3763 a begin stmt marker.
3764 (tsi_link_after): Likewise.
3765 (expr_first): Skip begin stmt markers.
3766 (expr_last): Likewise.
3767 * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3768 * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3769 Disregard nonbind markers.
3770 * tree.c (make_node_stat): Don't set side effects for begin stmt
3772 (build1_stat): Likewise.
3773 * tree.def (DEBUG_BEGIN_STMT): New.
3774 * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3775 * var-tracking.c (delete_debug_insns): Renamed to...
3776 (delete_vta_debug_insns): ... this.
3777 (reemit_marker_as_note): New.
3778 (vt_initialize): Reemit markers.
3779 (delete_vta_debug_insns): Likewise.
3780 (vt_debug_insns_local): Reemit or delete markers.
3781 (variable_tracking_main_1): Likewise.
3782 * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3783 * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3784 (gimple_debug_nonbind_marker_p): New.
3785 (gimple_build_debug_bind): Adjust.
3786 (gimple_build_debug_begin_stmt): New.
3787 * doc/invoke.texi (max-debug-marker-count): New param.
3788 * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3789 (debug_parameter_ref, debug_marker): New.
3790 (NOTE_INSN_BEGIN_STMT): New.
3791 (DEBUG_INSN): Describe begin stmt markers.
3793 * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3794 without debug insns, we wouldn't, but clean up debug insns
3795 after a control flow insn nevertheless.
3797 * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3798 block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3799 * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3800 * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3801 (rtl_tidy_fallthru_edge): Likewise.
3802 (rtl_verify_fallthru): Likewise.
3803 (rtl_verify_bb_layout): Likewise.
3804 (skip_insns_after_block): Likewise.
3805 (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3806 * dwarf2out.c: Include print-rtl.h.
3807 (dwarf2out_next_real_insn): New.
3808 (dwarf2out_var_location): Call it. Disregard begin stmt markers.
3809 Dump debug binds in asm comments.
3810 * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3811 * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3812 callers to use gsi_start_nondebug_bb instead.
3813 (gsi_after_labels): Skip gimple debug stmts.
3814 (gsi_start_nondebug): New.
3815 * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3816 (proper_loop_form_for_interchange): Adjust.
3817 * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3818 * gimple.h (gimple_seq_last_nondebug_stmt): New.
3819 * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3820 (collect_fallthrough_labels): Likewise.
3821 (should_warn_for_implicit_fallthrough): Likewise.
3822 (warn_implicit_fallthrough_r): Likewise.
3823 (expand_FALLTHROUGH_r): Likewise.
3824 * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3825 (graphite_copy_stmts_from_block): Skip nonbind markers.
3826 * haifa-sched.c (sched_extend_bb): Skip debug insns.
3827 * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3828 * jump.c (clean_barriers): Skip debug insns.
3829 * omp-expand.c (expand_parallel_call): Skip debug insns.
3830 (expand_task_call): Likewise.
3831 (remove_exit_barrier): Likewise.
3832 (expand_omp_taskreg): Likewise.
3833 (expand_omp_for_init_counts): Likewise.
3834 (expand_omp_for_generic): Likewise.
3835 (expand_omp_for_static_nochunk): Likewise.
3836 (expand_omp_for_static_chunk): Likewise.
3837 (expand_omp_simd): Likewise.
3838 (expand_omp_taskloop_for_outer): Likewise.
3839 (expand_omp_taskloop_for_inner): Likewise.
3840 (expand_oacc_for): Likewise.
3841 (expand_omp_sections): Likewise.
3842 (expand_omp_single): Likewise.
3843 (expand_omp_synch): Likewise.
3844 (expand_omp_atomic_load): Likewise.
3845 (expand_omp_atomic_store): Likewise.
3846 (expand_omp_atomic_fetch_op): Likewise.
3847 (expand_omp_atomic_pipeline): Likewise.
3848 (expand_omp_atomic_mutex): Likewise.
3849 (expand_omp_target): Likewise.
3850 (grid_expand_omp_for_loop): Likewise.
3851 (grid_expand_target_grid_body): Likewise.
3852 (build_omp_regions_1): Likewise.
3853 * omp-low.c (check_combined_parallel): Skip debug stmts.
3854 * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
3855 * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
3856 * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
3858 * tree-cfg.c (make_blobs_1): Skip debug stmts.
3859 (make_edges): Likewise.
3860 (cleanup_dead_labels): Likewise.
3861 (gimple_can_merge_blocks_p): Likewise.
3862 (stmt_starts_bb_p): Likewise.
3863 (gimple_block_label): Likewise.
3864 (gimple_redirect_edge_and_branch): Likewise.
3865 * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
3867 (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
3869 * tree-pretty-print (print_declaration): Omit initializer in slim
3871 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
3873 (eliminate_unnecessary_stmts): Stabilize block removal order.
3874 * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
3875 * var-tracking.c (get_first_insn): New.
3876 (vt_emit_notes): Call it.
3877 (vt_initialize): Walk any insns before the first BB.
3878 (delete_debug_insns): Likewise.
3880 * gimple.h (enum gimple_debug_subcode): Add
3881 GIMPLE_DEBUG_BEGIN_STMT.
3882 (gimple_debug_begin_stmt_p): New.
3883 (gimple_debug_nonbind_marker_p): New.
3884 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
3885 (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
3886 (MAY_HAVE_DEBUG_STMTS): ... this. Check both.
3887 * insn-notes.def (BEGIN_STMT): New.
3888 * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
3889 (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
3890 (MAY_HAVE_DEBUG_INSNS): ... this. Check both.
3891 (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
3892 (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
3893 (INSN_DEBUG_MARKER_KIND): New.
3894 (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
3895 (INSN_VAR_LOCATION): Check for VAR_LOCATION.
3896 (INSN_VAR_LOCATION_PTR): New.
3897 * cfgexpand.c (expand_debug_locations): Handle debug bind insns
3899 (expand_gimple_basic_block): Likewise. Emit debug temps for TER
3900 deps only if debug bind insns are enabled.
3901 (pass_expand::execute): Avoid deep TER and expand
3902 debug locations for debug bind insns only.
3903 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
3904 debug stmts special handling down to debug bind stmts.
3905 * combine.c (try_combine): Narrow debug insns special handling
3906 down to debug bind insns.
3907 * cse.c (delete_trivially_dead_insns): Handle debug bindings.
3908 Narrow debug insns preexisting special handling down to debug
3910 * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
3911 handling down to debug bind insns.
3912 * function.c (instantiate_virtual_regs): Skip debug markers,
3913 adjust handling of debug binds.
3914 * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
3915 temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
3916 * haifa-sched.c (schedule_insn): Narrow special handling of debug
3917 insns to debug bind insns.
3918 * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
3919 special handling of debug stmts to debug bind stmts.
3920 * ipa-split.c (split_function): Likewise.
3921 * ira.c (combine_and_move_insns): Adjust debug bind insns only.
3922 * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
3924 * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
3925 * regrename.c (build_def_use): Likewise.
3926 * regcprop.c (copyprop_hardreg_forward_1): Likewise.
3927 (pass_cprop_hardreg): Narrow special casing of debug insns to
3929 * regstat.c (regstat_init_n_sets_and_refs): Likewise.
3930 * reload1.c (reload): Likewise.
3931 * sese.c (sese_insert_phis_for_liveouts): Narrow special
3932 casing of debug stmts to debug bind stmts.
3933 * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
3934 * ssa-iterators.h (num_imm_uses): Likewise.
3935 * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
3936 debug stmts to debug bind stmts.
3937 * tree-inline.c (tree_function_versioning): Narrow special casing
3938 of debug stmts to debug bind stmts.
3939 * tree-loop-distribution.c (generate_loops_for_partition):
3940 Narrow special casing of debug stmts to debug bind stmts.
3941 * tree-sra.c (analyze_access_subtree): Narrow special casing
3942 of debug stmts to debug bind stmts.
3943 * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
3944 stmts to debug bind stmts.
3945 * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
3946 casing of debug stmts to debug bind stmts.
3947 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3948 * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
3949 casing of debug stmts to debug bind stmts.
3950 * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3952 * tree-ssa.c (flush_pending_stmts): Narrow special casing of
3953 debug stmts to debug bind stmts.
3954 (gimple_replace_ssa_lhs): Likewise.
3955 (insert_debug_temp_for_var_def): Likewise.
3956 (insert_debug_temps_for_defs): Likewise.
3957 (reset_debug_uses): Likewise.
3958 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3959 * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
3960 (adjust_debug_stmts): Likewise.
3961 (adjust_phi_and_debug_stmts): Likewise.
3962 (vect_do_peeling): Likewise.
3963 * tree-vect-loop.c (vect_transform_loop): Likewise.
3964 * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
3965 * var-tracking.c (adjust_mems): Narrow special casing of debug
3966 insns to debug bind insns.
3967 (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
3968 (compute_bb_dataflow, vt_find_locations): Likewise.
3969 (vt_expand_loc, emit_notes_for_changes): Likewise.
3970 (vt_init_cfa_base): Likewise.
3971 (vt_emit_notes): Likewise.
3972 (vt_initialize): Likewise.
3973 (vt_finalize): Likewise.
3975 * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
3976 (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
3977 (next_nonnote_nondebug_insn_bb): New.
3978 (prev_nonnote_nondebug_insn_bb): New.
3979 (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
3980 * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
3981 (prev_nonnote_nondebug_insn_bb): Declare.
3982 (next_nonnote_nondebug_insn_bb): Declare.
3983 * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
3984 * cfgrtl.c (get_last_bb_insn): Likewise.
3985 * lra.c (push_insns): Likewise.
3987 2017-12-11 David Malcolm <dmalcolm@redhat.com>
3990 * selftest-run-tests.c (selftest::run_tests): Move start/finish code
3992 * selftest.c (selftest::test_runner::test_runner): New ctor.
3993 (selftest::test_runner::~test_runner): New dtor.
3994 * selftest.h (class selftest::test_runner): New class.
3996 2017-12-11 Carl Love <cel@us.ibm.com>
3998 * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
3999 vec_extract_fp32_from_shortl]): Add #defines.
4000 * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4001 * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4002 ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4003 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4004 ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4005 ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4006 * doc/extend.texi: Add documentation for the added builtins.
4008 2017-12-11 Alexandre Oliva <aoliva@redhat.com>
4010 PR rtl-optimization/80693
4011 PR rtl-optimization/81019
4012 PR rtl-optimization/81020
4013 * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4014 are not mentioned in i3. Place the REG_UNUSED note on i2,
4015 possibly modified to REG_DEAD, if it did not originate in i3.
4017 2017-12-11 Jakub Jelinek <jakub@redhat.com>
4019 * recog.c (store_data_bypass_p_1): New function.
4020 (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER. Use
4021 store_data_bypass_p_1 to avoid code duplication. Formatting fixes.
4023 2017-12-11 Segher Boessenkool <segher@kernel.crashing.org>
4025 PR rtl-optimization/83361
4026 * ifcvt.c (if_convert): Call fixup_partitions.
4028 2017-12-11 Will Schmidt <will_schmidt@vnet.ibm.com>
4030 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4031 early folding of splat_u{8,16,32}.
4033 2017-12-11 Jakub Jelinek <jakub@redhat.com>
4035 * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4036 output_operand_lossage first argument with capital letter.
4037 (aarch64_override_options): Don't start error and sorry first argument
4038 with capital letter.
4040 2017-12-11 Andi Kleen <ak@linux.intel.com>
4042 PR gcov-profile/83355
4043 * auto-profile.c (string_table::get_index_by_decl): Don't
4044 recurse when abstract origin points to itself.
4046 2017-12-11 Bin Cheng <bin.cheng@arm.com>
4048 PR tree-optimization/83320
4049 * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4050 (prune_datarefs_not_in_loop): Ditto.
4052 2017-12-10 Gerald Pfeifer <gerald@pfeifer.com>
4054 * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4056 2017-12-10 Jakub Jelinek <jakub@redhat.com>
4058 PR tree-optimization/83337
4059 * gimple-loop-interchange.cc (compute_access_stride): Handle
4060 bitfield DRs properly.
4062 2017-12-09 Jakub Jelinek <jakub@redhat.com>
4064 PR tree-optimization/83338
4065 * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4066 vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4067 vector of unsigned integers to vector of signed integers.
4069 2017-12-08 Vladimir Makarov <vmakarov@redhat.com>
4071 PR rtl-optimization/83317
4072 * lra-constraints.c (process_address_1): Add insn code check.
4074 2017-12-08 Michael Matz <matz@suse.de>
4076 Fix PR tree-optimization/83323
4077 * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4078 head-controlled loops and loop BBs.
4079 * common.opt (funroll-and-jam): Remove, instead ...
4080 (floop-unroll-and-jam): ... reuse this option.
4081 * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4082 * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4083 (-floop-unroll-and-jam): ... this option.
4085 2017-12-08 Jakub Jelinek <jakub@redhat.com>
4087 * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4088 a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4089 * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4091 2017-12-08 Richard Biener <rguenther@suse.de>
4094 * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4095 handle accesses outside of zero-sized vars.
4097 2017-12-08 Martin Jambor <mjambor@suse.cz>
4099 PR tree-optimization/83141
4100 * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4101 test for MEM_REFs implicitely changing types with padding. Remove
4103 (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4105 2017-12-08 Jakub Jelinek <jakub@redhat.com>
4107 * config/arc/arc.c (arc_attribute_table): Add exclusions to
4109 * config/avr/avr.c (avr_attribute_table): Likewise.
4110 * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4111 * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4112 * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4113 * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4114 of struct attribute_spec.
4115 * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4117 2017-12-08 Ulrich Weigand <uweigand@de.ibm.com>
4120 * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4122 2017-12-08 Jan Hubicka <hubicka@ucw.cz>
4125 * profile-count.c (profile_count::from_gcov_type): Move from
4126 profile-count.h; handle overflow.
4127 * profile-count.h (profile_count::from_gcov_type): Move offline.
4129 2017-12-08 Segher Boessenkool <segher@kernel.crashing.org>
4131 PR rtl-optimization/83304
4132 * combine.c (move_deaths): If we do not know where a register died,
4135 2017-12-08 Richard Biener <rguenther@suse.de>
4137 * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4138 Provide -fopt-info-loop feedback when we interchange in a nest.
4140 2017-06-08 Richard Earnshaw <rearnsha@arm.com>
4142 * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4143 for armv6 ARM CPU IDs.
4145 2017-06-08 Richard Earnshaw <rearnsha@arm.com>
4147 * common/config/arm/arm-common.c: Include <algorithm>.
4148 (INCLUDE_VECTOR): Define.
4149 (compare_opt_names): New function.
4150 (arm_rewrite_selected_arch): Only strip out extensions that can be
4151 expressed through -mfpu. Sort the remaining extensions
4154 2017-06-08 Richard Earnshaw <rearnsha@arm.com>
4156 * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4157 (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4158 (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4159 * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4161 2017-06-08 Tristan Gingold <gindold@adacore.com>
4164 * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4165 .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4167 2017-12-08 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4169 * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4170 of struct attribute_spec.
4172 2017-12-08 Julia Koval <julia.koval@intel.com>
4174 * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4175 _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4176 * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4177 _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4178 _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4179 _mm_maskz_dpwssds_epi32): Ditto.
4181 2017-12-08 Richard Biener <rguenther@suse.de>
4183 PR tree-optimization/81303
4184 * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4185 conditions try to create a comparison vector type matching
4186 the data vector type.
4187 (vectorizable_condition): Adjust.
4188 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4189 Leave invariant conditions alone in case we can vectorize those.
4191 2017-12-08 Julia Koval <julia.koval@intel.com>
4193 * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4194 _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4195 * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4196 _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4197 _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4199 2017-12-08 Julia Koval <julia.koval@intel.com>
4201 * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4202 _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4203 * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4204 _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4205 _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4206 _mm_maskz_dpbusds_epi32): New intrinsics.
4208 2017-12-07 Sandra Loosemore <sandra@codesourcery.com>
4210 * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4212 (ld<bh>io_signed): Likewise.
4213 (st<bhw>io): Likewise.
4214 * config/nios2/predicates.md (ldstio_memory_operand): Allow
4215 SMALL_INT12 constant integer operand.
4217 2017-12-07 Sandra Loosemore <sandra@codesourcery.com>
4219 * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4221 (nios2_large_constant_allowed): ...this. Adjust uses.
4222 (nios2_plus_symbolic_constant_p): Rename to...
4223 (nios2_plus_large_constant_p): ...this. Adjust uses.
4224 (nios2_legitimate_address_p): Correct CONST_INT handling.
4225 (nios2_symbolic_memory_operand_p): Rename to...
4226 (nios2_large_constant_memory_operand_p): ...this. Adjust uses.
4227 (nios2_large_constant_p): Check for large constant integers too.
4228 (nios2_split_large_constant): Handle constant integers.
4229 (nios2_split_symbolic_memory_operand): Rename to...
4230 (nios2_split_large_constant_memory_operand): ...this. Adjust uses.
4231 (nios2_legitimize_constant_address): Handle constant integers.
4232 (r0rel_constant_p): Handle small constant integers.
4233 (nios2_print_operand_address): Handle r0-relative integer addresses.
4234 * config/nios2/nios2-protos.h: Adjust for renamed functions.
4235 * config/nios2/nios2.md: Adjust for renamed functions.
4237 2017-12-07 Andrew Waterman <andrew@sifive.com>
4239 * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4240 (TARGET_HAVE_SRODATA_SECTION): New define.
4241 (riscv_select_section): New function.
4243 2017-12-08 Joseph Myers <joseph@codesourcery.com>
4244 Alexander Monakov <amonakov@ispras.ru>
4245 Jakub Jelinek <jakub@redhat.com>
4248 * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4250 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4252 * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4253 a bool success value. Don't call output_operand_lossage here.
4254 (aarch64_print_ldpstp_address): Return a bool success value.
4255 (aarch64_print_operand_address): Call output_addr_const if
4256 aarch64_print_address_internal fails.
4257 (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4258 'y'; call output_operand_lossage instead. Call output_operand_lossage
4259 if aarch64_print_ldpstp_address fails.
4261 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4263 * tree-vector-builder.h
4264 (tree_vector_builder::binary_encoded_nelts): Declare.
4265 * tree-vector-builder.c
4266 (tree_vector_builder::binary_encoded_nelts): New function.
4267 * fold-const.c (negate_expr_p): Likewise.
4268 (operand_equal_p, fold_checksum_tree): Likewise.
4269 * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4270 * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4271 (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4272 (uniform_vector_p): Likewise.
4273 * varasm.c (const_hash_1, compare_constant): Likewise.
4274 * tree-ssa-ccp.c: Include tree-vector-builder.h.
4275 (valid_lattice_transition): Operate directly on the VECTOR_CST
4277 * ipa-icf.c: Include tree-vector-builder.h.
4278 (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4279 * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4281 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4283 * tree.c (build_vector): Delete.
4284 * tree.h (build_vector): Make static and move into the self-testing
4287 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4289 * vector-builder.h (vector_builder::derived): New const overload.
4290 (vector_builder::elt): New function.
4291 * tree-vector-builder.h (tree_vector_builder::type): New function.
4292 (tree_vector_builder::apply_step): Declare.
4293 * tree-vector-builder.c (tree_vector_builder::apply_step): New
4295 * gimple-fold.h (tree_vector_builder): Declare.
4296 (gimple_build_vector): Take a tree_vector_builder instead of a
4297 type and vector of elements.
4298 * gimple-fold.c (gimple_build_vector): Likewise.
4299 * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4301 (get_initial_defs_for_reduction): Likewise.
4302 (vectorizable_induction): Likewise.
4304 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4306 * tree-vector-builder.h
4307 (tree_vector_builder::new_binary_operation): Declare.
4308 * tree-vector-builder.c
4309 (tree_vector_builder::new_binary_operation): New function.
4310 * fold-const.c (fold_relational_const): Use it.
4311 (const_binop): Likewise. Check that both input vectors have
4312 the same number of elements, thus excluding things like WIDEN_SUM.
4313 Check whether it is possible to operate directly on the encodings
4316 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4318 * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4319 new_unary_operation, operating only on the encoded elements.
4320 (const_unop): Likewise.
4321 (exact_inverse): Likewise.
4322 (distributes_over_addition_p): New function.
4323 (const_binop): Use tree_vector_builder and new_unary_operation
4324 for combinations of VECTOR_CST and INTEGER_CST. Operate only
4325 on the encoded elements unless the encoding is strided and the
4326 operation does not distribute over addition.
4327 (fold_convert_const): Use tree_vector_builder and
4328 new_unary_operation. Operate only on the encoded elements
4329 for truncating integer conversions, or for non-stepped encodings.
4331 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4333 * config/sparc/sparc.c: Include tree-vector-builder.h.
4334 (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4335 * expmed.c: Include tree-vector-builder.h.
4336 (make_tree): Use tree_vector_builder instead of build_vector.
4337 * fold-const.c: Include tree-vector-builder.h.
4338 (const_binop): Use tree_vector_builder instead of build_vector.
4339 (const_unop): Likewise.
4340 (native_interpret_vector): Likewise.
4341 (fold_vec_perm): Likewise.
4342 (fold_ternary_loc): Likewise.
4343 * gimple-fold.c: Include tree-vector-builder.h.
4344 (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4346 * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4347 (simplify_vector_constructor): Use tree_vector_builder instead
4349 * tree-vect-generic.c: Include tree-vector-builder.h.
4350 (add_rshift): Use tree_vector_builder instead of build_vector.
4351 (expand_vector_divmod): Likewise.
4352 (optimize_vector_constructor): Likewise.
4353 * tree-vect-loop.c: Include tree-vector-builder.h.
4354 (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4355 of build_vector. Explicitly use a stepped encoding for
4357 * tree-vect-slp.c: Include tree-vector-builder.h.
4358 (vect_get_constant_vectors): Use tree_vector_builder instead
4360 (vect_transform_slp_perm_load): Likewise.
4361 (vect_schedule_slp_instance): Likewise.
4362 * tree-vect-stmts.c: Include tree-vector-builder.h.
4363 (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4364 (vect_gen_perm_mask_any): Likewise.
4365 (vectorizable_call): Likewise. Explicitly use a stepped encoding.
4366 * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4368 (build_vector_from_val): Likewise. Explicitly use a duplicate
4371 2017-12-07 Richard Sandiford <richard.sandiford@arm.com>
4373 * doc/generic.texi (VECTOR_CST): Describe new representation of
4375 * vector-builder.h: New file.
4376 * tree-vector-builder.h: Likewise.
4377 * tree-vector-builder.c: Likewise.
4378 * Makefile.in (OBJS): Add tree-vector-builder.o.
4379 * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4380 * tree-core.h (tree_base): Add a vector_cst field to the u union.
4381 (tree_vector): Change the number of elements to
4382 vector_cst_encoded_nelts.
4383 * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4384 (VECTOR_CST_ELTS): Delete.
4385 (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4386 (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4387 (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4388 (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4389 (VECTOR_CST_ENCODED_ELT): Likewise.
4390 (vector_cst_encoded_nelts): New function.
4391 (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4392 VECTOR_CST_NELTS_PER_PATTERN as arguments.
4393 (vector_cst_int_elt, vector_cst_elt): Declare.
4394 * tree.c: Include tree-vector-builder.h.
4395 (tree_code_size): Abort if passed VECTOR_CST.
4396 (tree_size): Update for new VECTOR_CST layout.
4397 (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4398 VECTOR_CST_NELTS_PER_PATTERN as arguments.
4399 (build_vector): Use tree_vector_builder.
4400 (vector_cst_int_elt, vector_cst_elt): New functions.
4401 (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4402 encoded elements and then create the vector in the canonical form.
4403 (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4404 (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4405 (tree_c_tests): Call test_vector_cst_patterns.
4406 * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4408 (hash_tree): Likewise.
4409 * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4410 (streamer_write_tree_header): Likewise.
4411 * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4412 (streamer_alloc_tree): Likewise. Update call to make_vector.
4413 * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4415 2017-12-07 Richard Sandiford <richard.sandiford@linaro.org>
4417 * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4418 (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4419 * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4421 2017-12-07 Bin Cheng <bin.cheng@arm.com>
4422 Richard Biener <rguenther@suse.de>
4424 PR tree-optimization/81303
4425 * Makefile.in (gimple-loop-interchange.o): New object file.
4426 * common.opt (floop-interchange): Reuse the option from graphite.
4427 * doc/invoke.texi (-floop-interchange): Ditto. New document for
4428 -floop-interchange and mention it for -O3.
4429 * opts.c (default_options_table): Enable -floop-interchange at -O3.
4430 * gimple-loop-interchange.cc: New file.
4431 * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4432 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4433 * passes.def (pass_linterchange): New pass.
4434 * timevar.def (TV_LINTERCHANGE): New time var.
4435 * tree-pass.h (make_pass_linterchange): New declaration.
4436 * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4437 interchange. Record IV before/after increment in new parameters.
4438 * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4439 * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4441 (check_reduction_path): ...New function here.
4442 * tree-vectorizer.h (check_reduction_path): New declaration.
4444 2017-12-07 Vladimir Makarov <vmakarov@redhat.com>
4447 PR rtl-optimization/80818
4448 * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4449 always early clobbered.
4450 * lra-lives.c (process_bb_lives): Check input hard regs for early
4451 clobbered non-operand hard reg.
4453 2017-12-07 Jakub Jelinek <jakub@redhat.com>
4456 * tree-cfg.c (verify_gimple_assign_binary): Don't require
4457 types_compatible_p, just that TYPE_MODE is the same.
4459 2017-12-07 Martin Sebor <msebor@redhat.com>
4462 * attribs.c (empty_attribute_table): Initialize new member of
4463 struct attribute_spec.
4464 (decl_attributes): Add argument. Handle mutually exclusive
4465 combinations of attributes.
4466 (selftests::test_attribute_exclusions): New function.
4467 (selftests::attribute_c_tests): Ditto.
4468 * attribs.h (decl_attributes): Add default argument.
4469 * selftest.h (attribute_c_tests): Declare.
4470 * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4471 * tree-core.h (attribute_spec::exclusions, exclude): New type and
4473 * doc/extend.texi (Common Function Attributes): Update const and pure.
4474 * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4475 of struct attribute_spec.
4476 * config/arc/arc.c (arc_attribute_table): Same.
4477 * config/arm/arm.c (arm_attribute_table): Same.
4478 * config/avr/avr.c ( avr_attribute_table): Same.
4479 * config/bfin/bfin.c (bfin_attribute_table): Same.
4480 * config/cr16/cr16.c (cr16_attribute_table): Same.
4481 * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4482 * config/h8300/h8300.c (h8300_attribute_table): Same.
4483 * config/i386/i386.c (ix86_attribute_table): Same.
4484 * config/ia64/ia64.c (ia64_attribute_table): Same.
4485 * config/m32c/m32c.c (m32c_attribute_table): Same.
4486 * config/m32r/m32r.c (m32r_attribute_table): Same.
4487 * config/m68k/m68k.c (m68k_attribute_table): Same.
4488 * config/mcore/mcore.c (mcore_attribute_table): Same.
4489 * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4490 * config/mips/mips.c (mips_attribute_table): Same.
4491 * config/msp430/msp430.c (msp430_attribute_table): Same.
4492 * config/nds32/nds32.c (nds32_attribute_table): Same.
4493 * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4494 * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4495 * config/rl78/rl78.c (rl78__attribute_table): Same.
4496 * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4497 * onfig/rx/rx.c (rx_attribute_table): Same.
4498 * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4499 * config/sh/sh.c (sh_attribute_table): Same.
4500 * config/sparc/sparc.c (sparc_attribute_table): Same.
4501 * config/spu/spu.c (spu_attribute_table): Same.
4502 * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4503 * config/v850/v850.c (v850_attribute_table): Same.
4504 * config/visium/visium.c (visium_attribute_table): Same.
4506 2017-12-07 Tamar Christina <tamar.christina@arm.com>
4509 * config/arm/arm.c (INCLUDE_STRING): Define.
4510 (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4511 (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4514 2017-12-07 Michael Matz <matz@suse.de>
4516 Add unroll and jam pass
4518 * gimple-loop-jam.c: New file.
4519 * Makefile.in (OBJS): Add gimple-loop-jam.o.
4520 * common.opt (funroll-and-jam): New option.
4521 * opts.c (default_options_table): Add unroll-and-jam at -O3.
4522 * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4523 (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4524 * passes.def: Add pass_loop_jam.
4525 * timevar.def (TV_LOOP_JAM): Add.
4526 * tree-pass.h (make_pass_loop_jam): Declare.
4527 * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4528 * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4529 * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4530 to flow_loop_tree_node_add.
4531 (duplicate_subloops, copy_loops_to): Append to sibling list.
4532 * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4533 * doc/invoke.texi (-funroll-and-jam): Document new option.
4534 (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4536 2017-12-07 Richard Biener <rguenther@suse.de>
4538 PR tree-optimization/83296
4539 PR tree-optimization/67769
4540 * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4541 flow sensitive info in an unrelated BB.
4542 (value_replacement): Use reset_flow_sensitive_info.
4543 (minmax_replacement): Reset flow sensitive info on the def
4544 we move. Do not reset flow sensitive info in the whole BB
4545 we move the stmt to.
4546 (abs_replacement): Likewise.
4548 2017-12-07 Segher Boessenkool <segher@kernel.crashing.org>
4551 * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4552 rs6000_cpu to the given -mcpu=, or to the default processor.
4554 2017-12-07 Segher Boessenkool <segher@kernel.crashing.org>
4556 * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4557 * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4558 instead of rs6000_cpu_attr.
4559 (rs6000_adjust_cost): Ditto.
4560 (is_microcoded_insn): Ditto.
4561 (rs6000_adjust_priority): Ditto.
4562 (rs6000_issue_rate): Ditto.
4563 (rs6000_use_sched_lookahead): Ditto.
4564 (rs6000_use_sched_lookahead_guard): Ditto.
4565 (rs6000_sched_reorder): Ditto.
4566 (force_new_group): Ditto.
4567 * config/rs6000/rs6000.md (cpu attribute): Ditto.
4568 (group_ending_nop): Ditto.
4570 2017-12-07 Segher Boessenkool <segher@kernel.crashing.org>
4572 * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4573 * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4574 rs6000_tune. Use rs6000_tune instead of rs6000_cpu where appropriate.
4575 (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4577 (rs6000_reassociation_width): Ditto.
4578 (rs6000_emit_epilogue): Ditto.
4579 (rs6000_adjust_cost): Ditto.
4580 (is_microcoded_insn): Ditto.
4581 (is_cracked_insn): Ditto.
4582 (rs6000_adjust_priority): Ditto.
4583 (rs6000_sched_reorder): Ditto.
4584 (rs6000_sched_reorder2): Ditto.
4585 (insn_must_be_first_in_group): Ditto.
4586 (insn_must_be_last_in_group): Ditto.
4587 (rs6000_register_move_cost): Ditto.
4588 * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4591 2017-12-07 Julia Koval <julia.koval@intel.com>
4593 * config.gcc: Add vaesintrin.h.
4594 * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4596 * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4597 __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4599 * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4600 * config/i386/immintrin.h: Include vaesintrin.h.
4601 * config/i386/sse.md (vaesdec_<mode>): New pattern.
4602 * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4603 _mm_aesdec_epi128): New intrinsics.
4605 2017-12-06 David Malcolm <dmalcolm@redhat.com>
4607 * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4608 * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4609 c-family/c-spellcheck.cc.
4610 (best_macro_match::best_macro_match): Likewise.
4612 (struct edit_distance_traits<cpp_hashnode *>): Move to
4613 c-family/c-spellcheck.h.
4614 (class best_macro_match): Likewise.
4616 2017-12-06 Jakub Jelinek <jakub@redhat.com>
4618 PR tree-optimization/83293
4619 * gimple-ssa-strength-reduction.c (insert_initializers): Use
4620 GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4621 might insert into empty bb.
4624 * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4625 simplify for plus with :c added, and pointer_plus without that.
4626 ((T)P - (T)(P + A) -> -(T) A): Likewise. If type is integral
4627 with undefined overflow and the conversion is not widening,
4628 perform negation in utype and only convert to type afterwards.
4629 ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4630 simplify for plus with :c added, and pointer_plus without that.
4631 If type is integral with undefined overflow and the conversion is
4632 not widening, perform minus in utype and only convert to type
4633 afterwards. Move the last pointer_diff_expr simplify into the
4636 2017-12-06 Martin Sebor <msebor@redhat.com>
4638 PR tree-optimization/82646
4639 * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4640 strncpy, not maxlen.
4642 2017-12-06 Martin Sebor <msebor@redhat.com>
4644 * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4647 PR tree-optimization/83075
4648 * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4649 strncat/strncpy don't change length of source string.
4651 2017-12-06 Eric Botcazou <ebotcazou@adacore.com>
4654 2017-11-29 Martin Aberg <maberg@gaisler.com>
4656 * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4657 to prevent b2bst errata sequence.
4658 (sqrtdf2_fix): Likewise.
4660 2017-12-06 Jakub Jelinek <jakub@redhat.com>
4662 PR tree-optimization/81945
4663 * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4664 * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4665 to dest_cfun has orig_loop_num set, either remap it to the new
4666 loop number if the loop got moved too, or clear it.
4668 2017-12-05 Steve Ellcey <sellcey@cavium.com>
4670 * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4672 (thunderx2t99_nothing): New insn reservation.
4673 (thunderx2t99_mrs): New insn reservation.
4674 (thunderx2t99_multiple): New insn reservation.
4675 (thunderx2t99_alu_basi): Add bfx to reservation.
4676 (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4678 2017-12-05 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4681 * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4683 2017-12-05 Bin Cheng <bin.cheng@arm.com>
4685 * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4686 tree-ssa-pre.c::remove_dead_inserted_code.
4687 * tree-ssa-dce.h: New file.
4688 * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4689 (remove_dead_inserted_code): Move and rename to function
4690 tree-ssa-dce.c::simple_dce_from_worklist.
4691 (pass_pre::execute): Update use.
4693 2017-12-05 Richard Biener <rguenther@suse.de>
4695 PR tree-optimization/83277
4696 * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4697 to code-gen liveout vars.
4699 2017-12-05 Richard Sandiford <richard.sandiford@linaro.org>
4701 * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4702 (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4703 (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4706 2017-12-05 Max Filippov <jcmvbkbc@gmail.com>
4708 * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4710 (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4711 * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4714 2017-12-05 Richard Biener <rguenther@suse.de>
4716 * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4717 * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4718 (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4719 (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4721 2017-12-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4723 * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4725 2017-12-05 Martin Liska <mliska@suse.cz>
4726 Jakub Jelinek <jakub@redhat.com>
4728 * doc/invoke.texi: Document the options.
4729 * flag-types.h (enum sanitize_code): Add
4730 SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4731 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4732 of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4733 * opts.c: Define new sanitizer options.
4734 * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4735 (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4737 2017-12-05 Julia Koval <julia.koval@intel.com>
4739 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4740 OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4741 (ix86_handle_option): Handle -mavx512vnni.
4742 * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4743 * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4744 * config/i386/i386-c (__AVX512VNNI__): New.
4745 * config/i386/i386.c (ix86_target_string): Handle new option.
4746 (ix86_valid_target_attribute_inner_p): Handle new option.
4747 * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4748 * config/i386/i386.opt (mavx512vnni): New option.
4750 2017-12-01 Jan Hubicka <hubicka@ucw.cz>
4753 * athlon.md: Disable for generic.
4754 * haswell.md: Enable for generic.
4755 * i386.c (ix86_sched_init_global): Add core hooks for generic.
4756 * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4758 (ix86_adjust_cost): Move generic to haswell path.
4760 2017-12-04 Eric Botcazou <ebotcazou@adacore.com>
4762 * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4763 instead of MEM_P in a couple more places. Fix formatting issues.
4765 2017-12-04 Jim Wilson <jimw@sifive.com>
4767 * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4768 instead of GP_REG_LAST-1.
4769 (riscv_adjust_libcall_cfi_prologue): Likewise.
4770 (riscv_adjust_libcall_cri_epilogue): Likewise.
4771 * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4774 2017-12-04 Luis Machado <luis.machado@linaro.org>
4776 * ipa-pure-const.c (check_decl): Add missing newline.
4777 (state_from_flags): Likewise.
4779 2017-12-04 Jeff Law <law@redhat.com>
4781 PR tree-optimizatin/78496
4782 * gimple-ssa-evrp-analyze.h
4783 (evrp_range_analyzer::get_vr_values): Simplify.
4784 * gimple-ssa-evrp-analyze.c: Corresponding changes.
4785 * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4786 and gimple-ssa-evrp-analyze.h.
4787 (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4788 (simplify_stmt_for_jump_threading): Copy a blob of code from
4789 tree-vrp.c to use ranges to simplify statements.
4790 (dom_opt_dom_walker::before_dom_children): Call
4791 evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4792 (dom_opt_dom_walker::after_dom_children): Similarly for
4793 evrp_range_analyzer::leave.
4794 (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4797 * gimple-ssa-evrp-analyze.c
4798 (evrp_range_analyzer::extract_range_from_stmt): Always use
4799 vr_values::update_value_range so preexisting range info is
4800 medged with new range info, even if the new range is VR_VARYING.
4802 2017-12-04 Segher Boessenkool <segher@kernel.crashing.org>
4804 * combine.c: Adjust comment.
4805 (use_crosses_set_p): Delete.
4806 (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4807 (try_combine): Ditto.
4809 2017-12-04 Richard Biener <rguenther@suse.de>
4811 PR tree-optimization/83255
4812 * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4813 Re-add zero-iteration check.
4815 2017-12-04 Segher Boessenkool <segher@kernel.crashing.org>
4817 PR rtl-optimization/83245
4818 * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4819 hard registers as earlyclobber, also if not in an asm.
4821 2017-12-04 Segher Boessenkool <segher@kernel.crashing.org>
4825 2017-12-01 Segher Boessenkool <segher@kernel.crashing.org>
4828 * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4829 rs6000_cpu based on cpu_index, not tune_index.
4831 2017-12-04 Richard Biener <rguenther@suse.de>
4833 PR tree-optimization/83238
4834 * graphite-scop-detection.c (scop_detection::merge_sese): Make
4835 code match comment, rejecting invalid SESE regions.
4837 2017-12-03 John David Anglin <danglin@gcc.gnu.org>
4839 * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4840 require base operand is a REG_POINTER prior to reload on targets
4841 with non-equivalent space registers.
4843 2017-12-01 Jan Hubicka <hubicka@ucw.cz>
4845 * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
4846 (update_specialized_profile): Fix updating of counts.
4847 (perhaps_add_new_callers): Likewise.
4849 2017-12-01 Jan Hubicka <hubicka@ucw.cz>
4852 * x86-tune.def: Remove obsolette FIXMEs.
4853 (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
4854 (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
4855 X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
4857 (X86_TUNE_PAD_RETURNS): Disable for generic.
4859 2017-12-02 Jakub Jelinek <jakub@redhat.com>
4861 PR tree-optimization/83170
4862 PR tree-optimization/83241
4863 * gimple-ssa-store-merging.c
4864 (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
4865 gimple_vuse (ins_stmt) in case it has changed.
4866 (imm_store_chain_info::output_merged_store): Likewise.
4868 * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
4872 * tree-cfg.c (pass_warn_function_return::execute): Handle
4873 __builtin_ubsan_handle_missing_return like __builtin_unreachable
4874 with BUILTINS_LOCATION.
4878 * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
4879 is BLKmode for vector field with vector raw mode, use TYPE_MODE
4880 instead of DECL_MODE.
4882 * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
4883 last argument to rtx pointer.
4884 * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
4885 with OPERANDS. For AVX+ 128-bit VEX encoded instructions over 256-bit
4886 or 512-bit. If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
4887 depending on the chosen ISAs.
4888 * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
4889 *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
4890 *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
4892 * config/i386/sse.md (mov<mode>_internal): Likewise.
4893 * config/i386/mmx.md (*mov<mode>_internal): Likewise.
4895 2017-12-01 Segher Boessenkool <segher@kernel.crashing.org>
4897 * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
4899 2017-12-01 Segher Boessenkool <segher@kernel.crashing.org>
4901 * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
4902 parameters from prototype.
4903 * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
4904 parameters. Don't print a comment.
4905 (emit_fusion_gpr_load): Adjust.
4906 (emit_fusion_load_store): Adjust.
4907 * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
4908 * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
4909 comment on the second line.
4911 2017-12-01 Segher Boessenkool <segher@kernel.crashing.org>
4914 * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4915 rs6000_cpu based on cpu_index, not tune_index.
4917 2017-12-01 Segher Boessenkool <segher@kernel.crashing.org>
4919 * final.c (output_asm_name): Print insn_cost. Shorten output. Print
4920 which_alternative instead of which_alternative + 1.
4921 (output_asm_insn): Print an extra tab if the template is short.
4923 2017-12-01 Jim Wilson <jimw@sifive.com>
4925 * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
4927 * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
4928 * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
4930 * doc/tm.texi: Regenerate.
4932 2017-12-01 Michael Meissner <meissner@linux.vnet.ibm.com>
4935 * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
4936 whether we can allocate pseudos before trying to fix an address.
4937 * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
4938 memory address is indexed or indirect.
4939 (floatuns_<mode>si2_hw2): Likewise.
4941 2017-12-01 Jason Merrill <jason@redhat.com>
4943 * Makefile.in (TAGS): Add c-family/*.cc.
4945 2017-12-01 Wilco Dijkstra <wdijkstr@arm.com>
4947 * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
4948 (call_value_insn): Likewise.
4949 (sibcall_insn): Likewise.
4950 (sibcall_value_insn): Likewise.
4951 (movsi_aarch64): Likewise.
4952 (movdi_aarch64): Likewise.
4953 (add_losym_): Likewise.
4954 (ldr_got_small_): Likewise.
4955 (ldr_got_small_sidi): Likewise.
4956 (ldr_got_small_28k_): Likewise.
4957 (ldr_got_small_28k_sidi): Likewise.
4958 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4959 Move output_addr_const to symbolic case. Add error check.
4961 2017-12-01 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4963 * config/s390/predicates.md (plus16_Q_operand): New predicate.
4964 * config/s390/s390.md: Disable MVC merging peephole if it would
4965 disable operand forwarding.
4966 (new peephole2): Split MVCs if it would turn them into up to 2
4969 2017-12-01 Richard Biener <rguenther@suse.de>
4971 PR tree-optimization/83232
4972 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
4973 detection of same access. Instead of breaking the group here
4974 do not consider the duplicate. Add comment explaining real fix.
4976 2017-12-01 Jonathan Wakely <jwakely@redhat.com>
4978 * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
4980 2017-12-01 Sudakshina Das <sudi.das@arm.com>
4982 * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
4984 2017-12-01 Jakub Jelinek <jakub@redhat.com>
4986 * function.h (struct function): Remove cilk_frame_decl,
4987 is_cilk_function and calls_cilk_spawn fields.
4988 * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
4990 * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
4991 cilk_elemental field.
4992 * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
4993 * target.def: Adjust comment.
4994 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4995 Don't test cilk_elemental.
4997 PR tree-optimization/83233
4998 * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
4999 bswap_stat name for the struct.
5002 * tree.h (SWITCH_BREAK_LABEL_P): Define.
5003 * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5004 starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5005 SWITCH_BREAK_LABEL_P set on the label.
5006 (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5007 added for default case if it was missing and not all cases covered.
5008 Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5009 switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5011 * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5012 to avoid -Wimplicit-fallthrough warning.
5013 * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5014 FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5016 PR tree-optimization/83221
5017 * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5019 (init_reassoc): Formatting fix.
5022 * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5023 C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5027 * tree-cfg.c: Include asan.h.
5028 (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5030 2017-12-01 Sergey Shalnov <Sergey.Shalnov@intel.com>
5032 * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5034 2017-12-01 Sergey Shalnov <Sergey.Shalnov@intel.com>
5036 * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5037 vpcmpeqd instruction.
5039 2017-12-01 Sergey Shalnov <Sergey.Shalnov@intel.com>
5041 * config/i386/i386.c (standard_sse_constant_opcode): Fix
5042 registers type for 128bit mode.
5044 2017-11-30 Jakub Jelinek <jakub@redhat.com>
5046 * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5048 * gengtype-state.c (read_state_pair): Likewise.
5049 * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5050 * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5051 * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5052 * ifcvt.c (noce_try_store_flag_constants): Likewise.
5053 * tree-ssa-ccp.c (ccp_finalize): Likewise.
5054 * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5055 * builtins.c (fold_builtin_3): Likewise.
5056 * graphite-scop-detection.c
5057 (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5058 * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5061 2017-12-01 Maxim Ostapenko <m.ostapenko@samsung.com>
5064 * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5065 parameter. Return true if ignore_decl_rtl_set_p is true and other
5066 conditions are satisfied.
5067 * asan.h (asan_protect_global): Add new parameter.
5068 * varasm.c (categorize_decl_for_section): Pass true as second parameter
5069 to asan_protect_global calls.
5071 2017-11-30 Jim Wilson <jimw@sifive.com>
5073 * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5074 -mno-memcpy options. For -mplt, -mfdiv, -mdiv, -msave-restore, and
5075 -mstrict-align, add info on default value. Delete redundant lines for
5076 -mabi. Add missing -mexplicit-relocs docs.
5078 2017-11-30 Claudiu Zissulescu <claziss@synopsys.com>
5080 * config/arc/arc.md (trap): New pattern.
5082 2017-11-30 Claudiu Zissulescu <claziss@synopsys.com>
5084 * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5085 instruction to end into a delay slot.
5086 * config/arc/arc.md (cond_delay_insn): Check if the instruction
5087 can be placed into a delay slot against reg_note.
5089 2017-11-30 Claudiu Zissulescu <claziss@synopsys.com>
5091 * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5092 labels number of usages.
5094 2017-11-30 Claudiu Zissulescu <claziss@synopsys.com>
5096 * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5098 (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5100 2017-11-30 Jakub Jelinek <jakub@redhat.com>
5103 * internal-fn.c (expand_mul_overflow): Optimize unsigned
5104 multiplication by power of 2 constant into two shifts + comparison.
5106 2017-11-30 Jan Hubicka <hubicka@ucw.cz>
5109 * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5111 2017-11-30 Richard Biener <rguenther@suse.de>
5113 PR tree-optimization/83202
5114 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5115 allow_peel argument and guard peeling.
5116 (canonicalize_loop_induction_variables): Likewise.
5117 (canonicalize_induction_variables): Pass false.
5118 (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5119 peeling from cunrolli.
5121 2017-11-29 Segher Boessenkool <segher@kernel.crashing.org>
5123 * combine.c (try_combine): Print a message to dump file whenever
5124 I0, I1, or I2 cannot be combined into I3.
5126 2017-11-29 Segher Boessenkool <segher@kernel.crashing.org>
5128 PR rtl-optimization/83156
5129 PR rtl-optimization/82621
5130 * combine.c (try_combine): Don't split an I2 if one of the dests is
5131 set again before I3. Allow unused dests.
5133 2017-11-29 Segher Boessenkool <segher@kernel.crashing.org>
5135 * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5137 2017-11-29 Vladimir Makarov <vmakarov@redhat.com>
5139 PR rtl-optimization/80818
5140 * lra.c (collect_non_operand_hard_regs): New arg insn. Pass it
5141 recursively. Use insn code for clobber.
5142 (lra_set_insn_recog_data): Pass the new arg to
5143 collect_non_operand_hard_regs.
5144 (add_regs_to_insn_regno_info): Pass insn instead of uid. Use insn
5146 (lra_update_insn_regno_info): Pass insn to
5147 add_regs_to_insn_regno_info.
5149 2017-11-29 Jim Wilson <jimw@sifive.com>
5150 Andrew Waterman <andrew@sifive.com>
5152 * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5153 (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5154 SINGLE_SHIFT_COST. Case LT and ZERO_EXTEND, likewise. Case ASHIFT,
5155 use SINGLE_SHIFT_COST.
5156 * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5157 (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5159 2017-11-29 Julia Koval <julia.koval@intel.com>
5161 * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5162 _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5163 _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5164 _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5165 * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5166 _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5167 _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5168 _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5169 _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5170 _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5171 _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5172 * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5173 __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5174 __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5175 __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5176 __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5177 __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5178 __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5179 __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5180 __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5181 __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5182 __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5183 __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5184 __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5185 __builtin_ia32_vpshldv_v2di_mask,
5186 __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5187 * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5188 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5190 2017-11-29 Julia Koval <julia.koval@intel.com>
5192 * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5193 _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5194 _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5195 _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5196 * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5197 _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5198 _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5199 _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5200 _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5201 _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5202 _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5203 * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5204 V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5205 V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5206 V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5207 V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5208 V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5209 V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5210 * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5211 * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5212 vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5214 2017-11-29 Daniel Cederman <cederman@gaisler.com>
5216 * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5217 movsi_pic_gotdata_op instruction as a load for the UT699 errata
5220 2017-11-29 Martin Aberg <maberg@gaisler.com>
5222 * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5223 to prevent b2bst errata sequence.
5224 (sqrtdf2_fix): Likewise.
5226 2017-11-29 Daniel Cederman <cederman@gaisler.com>
5228 * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5229 (div_sqrt_insn_p): New function.
5230 (sparc_do_work_around_errata): Insert NOP instructions to
5231 prevent sequences that could trigger the TN-0013 errata for
5232 certain LEON3 processors.
5233 (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5234 (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5235 * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5236 (in_branch_delay): Prevent div and sqrt in delay slot if
5238 * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5240 2017-11-29 Daniel Cederman <cederman@gaisler.com>
5242 * config/sparc/sparc.c (atomic_insn_p): New function.
5243 (sparc_do_work_around_errata): Insert NOP instructions to
5244 prevent sequences that could trigger the TN-0010 errata for
5246 * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5247 instruction referable in atomic_insns_p.
5249 2017-11-29 Daniel Cederman <cederman@gaisler.com>
5251 * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5252 (atomic_compare_and_swap_leon3_1): Likewise.
5255 2017-11-29 Daniel Cederman <cederman@gaisler.com>
5257 * config/sparc/sparc.c (fpop_insn_p): New function.
5258 (sparc_do_work_around_errata): Insert NOP instructions to
5259 prevent sequences that could trigger the TN-0012 errata for
5261 (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5262 * config/sparc/sparc.md (fix_gr712rc): New attribute.
5263 (in_branch_annul_delay): Prevent floating-point instructions
5264 in delay slot of annulled integer branch.
5266 2017-11-29 Richard Biener <rguenther@suse.de>
5268 PR tree-optimization/83202
5269 * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5271 (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5272 nodes vectorized to the same stmts multiple times.
5273 (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5274 (vect_analyze_slp_instance): Adjust.
5275 (scalar_stmts_to_slp_tree_map_t): New typedef.
5276 (vect_schedule_slp_instance): Add a map recording the SLP node
5277 representing the vectorized stmts for a set of scalar stmts.
5278 Avoid code-generating redundancies.
5279 (vect_schedule_slp): Allocate map and pass it down.
5281 2017-11-29 Nathan Sidwell <nathan@acm.org>
5284 * tree.c (build_complex_type): Fix canonicalization. Only fill in
5287 2017-11-29 Wilco Dijkstra <wdijkstr@arm.com>
5289 * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5290 cases for printing LDP/STP memory addresses.
5291 (aarch64_print_address_internal): Renamed from
5292 aarch64_print_operand_address, added parameter, add Pmode check.
5293 (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5294 (aarch64_print_operand_address): Indirect to
5295 aarch64_print_address_internal.
5296 * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5297 'y' operand output specifier.
5299 2017-11-29 Jakub Jelinek <jakub@redhat.com>
5302 * tree.c (build_simple_mem_ref_loc): Handle
5303 get_addr_base_and_unit_offset returning a MEM_REF.
5306 * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5309 * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5312 2017-11-28 David Malcolm <dmalcolm@redhat.com>
5314 * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5316 (layout::move_to_column): Likewise.
5318 2017-11-28 Jakub Jelinek <jakub@redhat.com>
5321 * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5322 is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5324 2017-11-28 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5325 Martin Jambor <mjambor@suse.cz>
5328 * tree.h (expr_type_first_operand_type_p): Declare
5329 * tree.c (expr_type_first_operand_type_p): New function.
5330 * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5331 (ipa_value_from_jfunc): Adjust declaration.
5332 * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5333 Use it as result type for arithmetics, unless it is NULL in which case
5334 be more conservative.
5335 (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5336 ipa_get_jf_pass_through_result.
5337 (propagate_vals_across_pass_through): Likewise.
5338 (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5339 is to propagate_vals_across_pass_through.
5340 (propagate_constants_across_call): Pass PARM_TYPE to
5341 propagate_scalar_across_jump_function.
5342 (find_more_scalar_values_for_callers_subset): Pass parameter type to
5343 ipa_value_from_jfunc.
5344 (cgraph_edge_brings_all_scalars_for_node): Likewise.
5345 * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5346 to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5347 * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5348 target_type, pass it to ipa_value_from_jfunc.
5349 (update_indirect_edges_after_inlining): Pass parameter type to
5350 try_make_edge_direct_simple_call.
5352 2017-11-28 Jeff Law <law@redhat.com>
5354 * gimple-ssa-evrp-analyze.c
5355 (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5356 refine ranges if scev_initialized_p returns true.
5357 * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5359 2017-11-28 Julia Koval <julia.koval@intel.com>
5361 * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5362 _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5363 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5364 _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5365 * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5366 _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5367 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5368 _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5369 _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5370 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5371 _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5372 * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5373 __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5374 __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5375 __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5376 __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5377 __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5378 __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5379 __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5380 __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5381 __builtin_ia32_vpshrd_v2di_mask): New builtins.
5382 * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5384 2017-11-28 Julia Koval <julia.koval@intel.com>
5386 * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5387 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5388 _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5389 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5390 * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5391 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5392 _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5393 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5394 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5395 _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5396 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5397 * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5398 V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5399 V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5400 V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5401 V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5402 V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5403 V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5404 V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5405 V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5406 * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5407 __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5408 __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5409 __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5410 __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5411 __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5412 __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5413 __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5414 __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5415 __builtin_ia32_vpshld_v2di_mask): New builtins.
5416 * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5417 * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5419 2017-11-28 Richard Biener <rguenther@suse.de>
5421 PR tree-optimization/80776
5422 * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5424 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5426 (evrp_range_analyzer::record_ranges_from_incoming_edges):
5427 If the incoming edge is an effective fallthru because the other
5428 edge only reaches a __builtin_unreachable () then record ranges
5429 derived from the controlling condition in SSA info.
5430 (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5431 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5433 2017-11-28 Olivier Hainque <hainque@adacore.com>
5435 * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5436 on mingw build hosts.
5438 2017-11-28 Jakub Jelinek <jakub@redhat.com>
5441 * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5443 2017-11-28 Richard Biener <rguenther@suse.de>
5446 * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5447 copies generated from memcpy use a character array as reference
5450 2017-11-28 Julia Koval <julia.koval@intel.com>
5451 Sebastian Peryt <sebastian.peryt@intel.com>
5453 * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5454 c-family/c-cilkplus.o, c-family/array-notation-common.o,
5455 cilk-common.o, cilk.h, cilk-common.c): Remove.
5457 (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5458 * builtins.c (is_builtin_name): Remove cilkplus condition.
5459 (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5460 * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5461 cilk-builtins.def, cilkplus.def): Remove.
5462 * cif-code.def (CILK_SPAWN): Remove.
5463 * cilk-builtins.def: Delete.
5464 * cilk-common.c: Ditto.
5466 * cilkplus.def: Ditto.
5467 * config/darwin.h (fcilkplus): Delete.
5468 * cppbuiltin.c: Ditto.
5469 * doc/extend.texi: Remove cilkplus doc.
5470 * doc/generic.texi: Ditto.
5471 * doc/invoke.texi: Ditto.
5472 * doc/passes.texi: Ditto.
5473 * gcc.c (fcilkplus): Remove.
5474 * gengtype.c (cilk.h): Remove.
5475 * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5477 * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5479 * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5481 is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5482 gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5483 cilkplus conditions.
5484 * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5485 inline_read_section): Ditto.
5486 * ipa-inline-analysis.c (cilk.h): Remove.
5487 * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5488 * lto-wrapper.c (merge_and_complain, append_compiler_options,
5489 append_linker_options): Remove condition for fcilkplus.
5490 * lto/lto-lang.c (cilk.h): Remove.
5491 (lto_init): Remove condition for fcilkplus.
5492 * omp-expand.c (expand_cilk_for_call): Delete.
5493 (expand_omp_taskreg, expand_omp_for_static_chunk,
5494 expand_omp_for): Remove cilkplus
5496 (expand_cilk_for): Delete.
5497 * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5498 * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5499 execute_lower_omp, diagnose_sb_0): Ditto.
5500 * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5501 * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5502 * tree-nested.c: Ditto.
5503 * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5504 (dump_generic_node): Ditto.
5505 * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5506 * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5508 * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5510 2017-11-28 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5512 * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5513 * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5514 for vector mode and !TARGET_SIMD.
5516 2017-11-28 Jakub Jelinek <jakub@redhat.com>
5518 * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5520 * tree.h (SWITCH_LABELS): Remove.
5521 * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5522 assert SWITCH_BODY is non-NULL.
5523 * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5525 * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5527 PR tree-optimization/80788
5528 * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5529 has TREE_OVERFLOW set, call drop_tree_overflow.
5531 2017-11-28 Richard Biener <rguenther@suse.de>
5533 PR tree-optimization/83158
5534 * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5536 2017-11-28 Segher Boessenkool <segher@kernel.crashing.org>
5539 * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5540 TARGET_ISEL && !TARGET_MFCRF differently. Simplify code.
5542 2017-11-27 Segher Boessenkool <segher@kernel.crashing.org>
5544 * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5545 LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5547 2017-11-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5550 * varasm.c (compare_constant): Take the mode of the constants into
5551 account when comparing floating point constants.
5553 2017-11-27 Gerald Pfeifer <gerald@pfeifer.com>
5555 * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5556 from explicit instantiation of debug_helper.
5557 * vec.h (DEFINE_DEBUG_VEC): Ditto.
5559 2017-11-27 Richard Biener <rguenther@suse.de>
5561 * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5564 2017-11-27 Richard Biener <rguenther@suse.de>
5566 * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5567 (type_hash_canon): Also clear int_cst_hash_table entry for
5569 (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5571 2017-11-27 Tamar Christina <tamar.christina@arm.com>
5573 * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5574 * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5575 * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5577 2017-11-27 Markus Trippelsdorf <markus@trippelsdorf.de>
5579 * hash-map.h (gt_cleare_cache): Avoid UB.
5581 2017-11-27 Eric Botcazou <ebotcazou@adacore.com>
5583 * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5584 * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5585 (decide_unroll_runtime_iterations): Likewise.
5586 (decide_unroll_stupid): Likewise.
5588 2017-11-27 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5591 * config/i386/i386.md: Add a loop with incssp.
5593 2017-11-27 Martin Jambor <mjambor@suse.cz>
5595 PR tree-optimization/81248
5596 * tree-sra.c (splice_param_accesses): Remove size check.
5597 (decide_one_param_reduction): Fix size check.
5598 * gimple-pretty-print.c (dump_profile): Silence warning.
5599 * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5601 2017-11-27 Jakub Jelinek <jakub@redhat.com>
5604 * dbxout.c (lastlineno): New variable.
5605 (dbx_debug_hooks): Use dbxout_switch_text_section as
5606 switch_text_section debug hook.
5607 (dbxout_function_end): Switch to current_function_section
5608 rather than function_section. If crtl->has_bb_partition,
5609 output just one N_FUN, depending on in_cold_section_p.
5610 (dbxout_source_line): Remember last lineno in lastlineno.
5611 (dbxout_switch_text_section): New function.
5612 (dbxout_function_decl): Adjust dbxout_block caller.
5613 (dbx_block_with_cold_children): New function.
5614 (dbxout_block): Return true if any LBRAC/RBRAC have been
5615 emitted. Use dbx_block_with_cold_children at depth == 0
5616 in second partition. Add PARENT_BLOCKNUM argument, pass
5617 it optionally adjusted to children. Output LBRAC/RBRAC
5618 around recursive call only if the block is in the current
5619 partition, if not and anything was output, emit empty
5621 * final.c (final_scan_insn): Compute cold_function_name
5622 before calling switch_text_section debug hook. Call
5623 that hook even if dwarf2out_do_frame if not emitting
5627 * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5628 TREE_READONLY decls.
5630 2017-11-27 Markus Trippelsdorf <markus@trippelsdorf.de>
5632 PR rtl-optimization/82488
5633 * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5635 2017-11-26 Julia Koval <julia.koval@intel.com>
5637 * config/i386/i386.c (processor_target_table): Add skylake_cost for
5639 * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5642 2017-11-26 Julia Koval <julia.koval@intel.com>
5644 * config/i386/driver-i386.c (host_detect_local_cpu):
5645 Detect skylake-avx512.
5647 2017-11-26 Julia Koval <julia.koval@intel.com>
5649 * config.gcc: Add -march=cannonlake.
5650 * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5651 * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5652 * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5653 (PTA_CANNONLAKE): New.
5654 (processor_target_table): Add cannonlake.
5655 (ix86_option_override_internal): Ditto.
5656 (fold_builtin_cpu): Ditto.
5657 (get_builtin_code_for_version): Handle cannonlake.
5658 (M_INTEL_COREI7_CANNONLAKE): New.
5659 * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5660 * doc/invoke.texi: Add -march=cannonlake.
5662 2017-11-14 Boris Kolpackov <boris@codesynthesis.com>
5664 * plugin.c (add_new_plugin): Use platform-specific library extensions.
5665 (try_init_one_plugin): Alternative implementation for MinGW.
5666 * Makefile.in (plugin_implib): New.
5667 (gengtype-lex.c): Fix broken AIX workaround.
5668 * configure: Regenerate.
5669 * doc/plugins.texi: Document support for MinGW.
5671 2017-11-25 Jakub Jelinek <jakub@redhat.com>
5673 PR rtl-optimization/81553
5674 * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5675 to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5676 is a shift where C1 has different mode than the whole shift, use C1's
5677 mode for MULT rather than the shift's mode.
5680 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5681 builtins not enabled in the currently selected ISA.
5683 2017-11-24 Jackson Woodruff <jackson.woodruff@arm.com>
5685 PR tree-optimization/71026
5686 * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5687 (insert_reciprocals): Change to insert reciprocals before a division
5688 by a square and to insert the square of a reciprocal.
5689 (execute_cse_reciprocals_1): Change to consider division by a square.
5690 (register_division_in): Add importance parameter.
5692 2017-11-24 Richard Biener <rguenther@suse.de>
5694 PR tree-optimization/82402
5695 * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5696 set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5698 2017-11-24 Marc Glisse <marc.glisse@inria.fr>
5700 * match.pd (0-ptr): New transformation.
5702 2017-11-24 Jan Hubicka <hubicka@ucw.cz>
5705 * ipa-inline.c (inline_small_functions): Set current badnes correctly
5706 when skipping checking.
5708 2017-11-24 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/83128
5711 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5712 (vn_reference_lookup_3): Likewise.
5714 2017-11-24 Jakub Jelinek <jakub@redhat.com>
5717 * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5718 instead of pp_printf with HOST_WIDE_INT_PRINT_DEC. Avoid calling
5721 * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5722 ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5723 check for the latter with gimple_call_builtin_p. Do not handle
5724 BUILT_IN_STPNCPY_CHK which is not a pass through call.
5726 2017-11-24 Christophe Lyon <christophe.lyon@linaro.org>
5728 * config/arm/arm_neon.h: Fix pragma GCC push_options before
5731 2017-11-23 Julia Koval <julia.koval@intel.com>
5733 * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5734 _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5735 _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5736 _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5737 _mm512_maskz_expandloadu_epi16): New intrinsics.
5738 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5739 _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5740 _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5741 _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5742 _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5743 _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5744 _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5745 _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5746 _mm256_maskz_expandloadu_epi8): New intrinsics.
5747 * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5748 V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5749 V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5750 V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5751 * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5752 * config/i386/sse.md (VI248_VLBW): New iterator.
5753 (expand<mode>_mask, expand<mode>_maskz): New patterns.
5755 2017-11-23 Julia Koval <julia.koval@intel.com>
5757 * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5758 * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5759 _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5760 _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5761 _mm512_mask_compressstoreu_epi16): New.
5762 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5763 _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5764 _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5765 _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5766 _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5767 _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5768 _mm256_mask_compressstoreu_epi8): New.
5769 * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5770 VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5771 VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5772 VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5773 * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5774 __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5775 __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5776 __builtin_ia32_compresshi128_mask,
5777 __builtin_ia32_compressstoreuqi512_mask,
5778 __builtin_ia32_compressstoreuhi512_mask,
5779 __builtin_ia32_compressstoreuqi256_mask,
5780 __builtin_ia32_compressstoreuqi128_mask,
5781 __builtin_ia32_compressstoreuhi256_mask,
5782 __builtin_ia32_compressstoreuhi128_mask): New builtins.
5783 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5785 (ix86_expand_special_args_builtin): Handle new types.
5786 (s4fma_expand): Handle new builtin array.
5787 * config/i386/immintrin.h: Include new headers.
5788 * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5789 (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5791 2017-11-23 Jakub Jelinek <jakub@redhat.com>
5793 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5794 fixes. Declare temp and g variables at the top in order to avoid
5795 {} in most of the cases.
5797 2017-11-23 Marc Glisse <marc.glisse@inria.fr>
5799 * match.pd (ptr-0): New transformation.
5801 2017-11-23 Charles Baylis <charles.baylis@linaro.org>
5803 * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5804 (struct addr_mode_cost_table): New.
5805 (struct tune_params): Add field addr_mode_costs.
5806 * config/arm/arm.c (generic_addr_mode_costs): New.
5807 (arm_slowmul_tune): Initialise addr_mode_costs field.
5808 (arm_fastmul_tune): Likewise.
5809 (arm_strongarm_tune): Likewise.
5810 (arm_xscale_tune): Likewise.
5811 (arm_9e_tune): Likewise.
5812 (arm_marvell_pj4_tune): Likewise.
5813 (arm_v6t2_tune): Likewise.
5814 (arm_cortex_tune): Likewise.
5815 (arm_cortex_a8_tune): Likewise.
5816 (arm_cortex_a7_tune): Likewise.
5817 (arm_cortex_a15_tune): Likewise.
5818 (arm_cortex_a35_tune): Likewise.
5819 (arm_cortex_a53_tune): Likewise.
5820 (arm_cortex_a57_tune): Likewise.
5821 (arm_exynosm1_tune): Likewise.
5822 (arm_xgene1_tune): Likewise.
5823 (arm_cortex_a5_tune): Likewise.
5824 (arm_cortex_a9_tune): Likewise.
5825 (arm_cortex_a12_tune): Likewise.
5826 (arm_cortex_a73_tune): Likewise.
5827 (arm_v7m_tune): Likewise.
5828 (arm_cortex_m7_tune): Likewise.
5829 (arm_v6m_tune): Likewise.
5830 (arm_fa726te_tune): Likewise.
5831 (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5833 2017-11-23 Charles Baylis <charles.baylis@linaro.org>
5835 * config/arm/arm.c (arm_mem_costs): New function.
5836 (arm_rtx_costs_internal): Use arm_mem_costs.
5838 2017-11-23 Mark Wielaard <mark@klomp.org>
5840 * dwarf2out.c (init_sections_and_labels): Use generation to create
5841 unique ranges_section_label and ranges_base_label. Return generation.
5842 (output_rnglists): Add generation argument. Use generation to create
5843 unique ranges labels.
5844 (dwarf2out_finish): Get generation from init_sections_and_labels
5845 and pass generation to output_rnglists.
5847 2017-11-23 Mike Stump <mikestump@comcast.net>
5848 Eric Botcazou <ebotcazou@adacore.com>
5850 * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
5851 * cfgloop.h (struct loop): Add unroll field.
5852 * function.h (struct function): Add has_unroll bitfield.
5853 * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
5854 (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
5855 * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
5856 (pass_rtl_unroll_loops::gate): Likewise.
5857 * loop-unroll.c (decide_unrolling): Tweak note message. Skip loops
5858 for which loop->unroll==1.
5859 (decide_unroll_constant_iterations): Use note for consistency and
5860 take loop->unroll into account. Return early if loop->unroll is set.
5861 Fix thinko in existing test.
5862 (decide_unroll_runtime_iterations): Use note for consistency and
5863 take loop->unroll into account.
5864 (decide_unroll_stupid): Likewise.
5865 * lto-streamer-in.c (input_cfg): Read loop->unroll.
5866 * lto-streamer-out.c (output_cfg): Write loop->unroll.
5867 * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
5869 (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
5870 (print_loop): Print loop->unroll if set.
5871 * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
5872 * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
5873 * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
5875 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
5876 loop->unroll is set and smaller than the trip count. Otherwise bypass
5877 entirely the heuristics if loop->unroll is set. Remove dead note.
5878 Fix off-by-one bug in other note.
5879 (try_peel_loop): Bail out if loop->unroll is set. Fix formatting.
5880 (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
5882 (tree_unroll_loops_completely): Make static.
5883 (pass_complete_unroll::execute): Use correct type for variable.
5884 (pass_complete_unrolli::execute): Fix formatting.
5885 * tree.def (ANNOTATE_EXPR): Add 3rd operand.
5887 2017-11-23 Sergey Shalnov <Sergey.Shalnov@intel.com>
5889 * config/i386/i386.h (TARGET_PREFER_AVX256): Also
5890 enable when TARGET_PREFER_AVX128 is set.
5892 2017-11-23 Jan Hubicka <hubicka@ucw.cz>
5894 * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
5896 * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
5897 (cgraph_edge::maybe_hot_p): Use sreal frequencies.
5899 2017-11-23 Jan Hubicka <hubicka@ucw.cz>
5901 * ipa-fnsummary.c (record_modified_bb_info): Use sreal
5902 frequencies. Fix estimation of aggregate parameters.
5904 2017-11-23 Jan Hubicka <hubicka@ucw.cz>
5906 * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
5909 2017-11-23 Tom de Vries <tom@codesourcery.com>
5911 * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
5913 2017-11-23 Tom de Vries <tom@codesourcery.com>
5915 * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
5918 2017-11-23 Oleg Endo <olegendo@gcc.gnu.org>
5921 * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
5922 sibcall_value_pcrel_fdpic): Use local variable instead of
5924 (calli_tbr_rel): Add missing operand 2.
5925 (call_valuei_tbr_rel): Add missing operand 3.
5927 2017-11-23 Jakub Jelinek <jakub@redhat.com>
5930 * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
5931 bitpos/bitsize covering the whole destination, use store_expr only if
5932 the complex mode is the same. Otherwise, use expand_normal and if
5933 it returns CONCAT, subreg each part separately instead of trying to
5934 subreg the whole result.
5936 2017-11-23 Richard Biener <rguenther@suse.de>
5938 PR tree-optimization/23094
5939 * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
5940 come here from walking over backedges in the first iteration.
5941 (vn_reference_lookup_3): Skip clobbers that store the same value.
5943 2017-11-23 Richard Biener <rguenther@suse.de>
5945 PR tree-optimization/81403
5946 * tree-ssa-pre.c (get_representative_for): Add parameter specifying
5947 a block we need a leader relative to.
5948 (phi_translate_1): For nary processing require a leader from
5949 get_representative_for given we run expression simplification
5950 using match-and-simplify. Remove previous fix.
5952 2017-11-22 Jeff Law <law@redhat.com>
5954 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
5955 Use new method allocate_value_range rather than accessing the
5956 vrp_value_range_pool data member directly.
5957 * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
5958 to use extract_range_from_stmt method to avoid need for
5959 extract_range_from_assignment method.
5960 (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
5961 method rather than setting values_propgated data member directly.
5962 * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
5963 and values propagated data members and extract_range_from_assignment
5964 method. Reorder private data members to conform to standards.
5965 Add new methods set_lattice_propagation_complete and
5966 allocate_value_range.
5968 2017-11-22 Eric Botcazou <ebotcazou@adacore.com>
5970 PR rtl-optimization/83030
5971 * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
5972 for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
5973 (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
5974 * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
5975 CROSSING_JUMP_P flag is already set before setting it.
5976 * cfgrtl.c (fixup_partition_crossing): Likewise.
5977 * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
5980 2017-11-22 Jakub Jelinek <jakub@redhat.com>
5982 * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
5983 Handle the case where both arguments are using gen_const_vec_series.
5985 2017-11-22 David Malcolm <dmalcolm@redhat.com>
5988 * pretty-print.c (pp_format): Move quoting implementation to
5989 pp_begin_quote and pp_end_quote. Update pp_format_decoder call
5990 to pass address of "quote" local.
5991 (pp_begin_quote): New function.
5992 (pp_end_quote): New function.
5993 * pretty-print.h (printer_fn): Convert penultimate param from bool
5995 (pp_begin_quote): New decl.
5996 (pp_end_quote): New decl.
5997 * tree-diagnostic.c (default_tree_printer): Convert penultimate
5998 param from bool to bool *.
5999 * tree-diagnostic.h (default_tree_printer): Likewise.
6001 2017-11-22 Jeff Law <law@redhat.com>
6003 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6004 Initialize vr_values.
6005 (evrp_range_analyzer::try_find_new_range): Call methods attached to
6006 vr_values via vr_values class instance rather than delegators.
6007 (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6008 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6009 (evrp_range_analyzer::push_value_range): Likewise.
6010 (evrp_range_analyzer::pop_value_range): Likewise.
6011 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6012 most delegators. Those remaining are exposed as public interfaces.
6013 Make vr_values a pointer and private.
6014 (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6016 (evrp_range_analyzer::get_vr_value): New method.
6017 * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6018 (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6019 (class evrp_dom_walker): Attach evrp_folder class, initialize
6020 it in the ctor. Remove temporary delegators.
6021 (evrp_dom_walker::before_dom_children): Call methods in attached
6022 evrp_range_analyzer class via class instance pointer. Use
6023 free value_range_constant_singleton to remove need for
6024 op_with_constant_singleton_value delegator method. Do not
6025 create a vrp_prop class instance for every call! Narrow
6026 scope of a couple variables.
6027 (evrp_dom_walker::cleanup): Call methods in attached
6028 evrp_range_analyzer class via class instance pointer.
6029 * vr-values.h (class vr_values): Privatize many methods and
6032 2017-11-22 Richard Sandiford <richard.sandiford@linaro.org>
6034 * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6035 * cfgexpand.c (expand_debug_expr): Remove handling for them.
6036 * expr.c (expand_expr_real_2): Likewise.
6037 * fold-const.c (const_unop): Likewise.
6038 * optabs-tree.c (optab_for_tree_code): Likewise.
6039 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6040 * tree-inline.c (estimate_operator_cost): Likewise.
6041 * tree-pretty-print.c (dump_generic_node): Likewise.
6042 (op_code_prio): Likewise.
6043 (op_symbol_code): Likewise.
6044 * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6045 (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6046 * internal-fn.c (direct_internal_fn_optab): New function.
6047 (direct_internal_fn_array, direct_internal_fn_supported_p
6048 (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6049 * fold-const-call.c (fold_const_reduction): New function.
6050 (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6052 * tree-vect-loop.c: Include internal-fn.h.
6053 (reduction_code_for_scalar_code): Rename to...
6054 (reduction_fn_for_scalar_code): ...this and return an internal
6056 (vect_model_reduction_cost): Take an internal_fn rather than
6058 (vect_create_epilog_for_reduction): Likewise. Build calls rather
6060 (vectorizable_reduction): Use internal functions rather than tree
6061 codes for the reduction operation. Update calls to the functions
6063 * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6064 Use calls to internal functions rather than REDUC tree codes.
6065 * config/aarch64/aarch64-simd.md: Update comment accordingly.
6067 2017-11-22 Olivier Hainque <hainque@adacore.com>
6069 * config/vxworks.c (vxworks_override_options): Pick default
6070 dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6071 * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6072 DWARF_GNAT_ENCODINGS_DEFAULT.
6073 * config/vxworksae.h: Likewise.
6075 2017-11-22 Marc Glisse <marc.glisse@inria.fr>
6077 PR tree-optimization/83104
6078 * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6081 2017-11-22 Marek Polacek <polacek@redhat.com>
6082 H.J. Lu <hongjiu.lu@intel.com>
6083 Jason Merrill <jason@redhat.com>
6088 * calls.c (initialize_argument_information): Call
6089 warn_parameter_passing_abi target hook.
6090 (store_one_arg): Use 0 for empty record size. Don't push 0 size
6091 argument onto stack.
6092 (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6093 * common.opt: Update -fabi-version description.
6094 * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6095 (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6097 (ix86_is_empty_record): New function.
6098 (ix86_warn_parameter_passing_abi): New function.
6099 (TARGET_EMPTY_RECORD_P): Redefine.
6100 (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6101 * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6102 * doc/tm.texi: Regenerated.
6103 * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6104 TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6105 * dwarf2out.c (get_ultimate_context): Move to tree.c.
6106 * explow.c (hard_function_value): Call arg_int_size_in_bytes
6107 instead of int_size_in_bytes.
6108 * expr.c (copy_blkmode_to_reg): Likewise.
6109 * function.c (aggregate_value_p): Return 0 for empty types.
6110 (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6112 (locate_and_pad_parm): Call arg size_in_bytes instead
6114 * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6115 * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6116 * target.def (empty_record_p, warn_parameter_passing_abi): New target
6118 * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6119 (std_gimplify_va_arg_expr): Skip empty records. Call
6120 arg_size_in_bytes instead size_in_bytes.
6121 * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6122 * tree-core.h (tree_type_common): Add empty_flag.
6123 (tree_decl_common): Update comments.
6124 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6126 (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6127 * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6129 (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6130 * tree.c (default_is_empty_type): New function.
6131 (default_is_empty_record): New function.
6132 (arg_int_size_in_bytes): New function.
6133 (arg_size_in_bytes): New function.
6134 (get_ultimate_context): New function.
6135 * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6136 TRANSLATION_UNIT_WARN_EMPTY_P.
6137 (default_is_empty_record, arg_int_size_in_bytes,
6138 arg_size_in_bytes, get_ultimate_context): Declare.
6140 2017-11-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6142 * config/arm/arm.c (cmse_clear_registers): New function.
6143 (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6144 code by call to cmse_clear_registers.
6145 (cmse_nonsecure_entry_clear_before_return): Likewise.
6147 2017-11-22 Tamar Christina <tamar.christina@arm.com>
6149 * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6150 (vdot_s32, vdotq_s32): New.
6151 (vdot_lane_u32, vdotq_lane_u32): New.
6152 (vdot_lane_s32, vdotq_lane_s32): New.
6155 2017-11-22 Richard Sandiford <richard.sandiford@linaro.org>
6158 * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6159 for unsigned values with fewer HWIs than the precision.
6160 (test_overflow): New function.
6161 (wide_int_cc_tests): Call it.
6163 2017-11-22 Richard Sandiford <richard.sandiford@linaro.org>
6165 * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6166 has the same mode class as Pmode.
6168 2017-11-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6170 * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6171 padding_bits_to_clear_ptr.
6172 (cmse_nonsecure_entry_clear_before_return): Likewise.
6174 2017-11-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6176 * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6177 auto_sbitap instead of integer bitfield to control register needing
6180 2017-11-22 Jakub Jelinek <jakub@redhat.com>
6182 PR tree-optimization/83044
6183 * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6184 INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6185 that need the upper bound. Subtract offset from
6186 get_addr_base_and_unit_offset only if positive and subtract it
6187 before division by eltsize rather than after it.
6190 * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6191 debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6192 (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6196 * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6197 expand_binop with *mul_widen_optab, make sure at least one of the
6198 operands doesn't have VOIDmode.
6201 * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6203 PR rtl-optimization/82044
6204 PR tree-optimization/82042
6205 * dse.c (record_store): Check for overflow.
6206 (check_mem_read_rtx): Properly check for overflow if width == -1, call
6207 add_wild_read instead of clear_rhs_from_active_local_stores on
6208 overflow and log it into dump_file.
6210 2017-11-22 Richard Biener <rguenther@suse.de>
6212 * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6213 fake edges to exit when looking for a place to insert.
6214 * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6216 (insert_into_preds_of_block): Commit edge insertion immediately,
6217 assert that doesn't require new BBs.
6218 (fini_pre): Release expressions.
6219 (pass_pre::execute): Shuffle things around a bit, if the fn
6220 is too large do not compute AVAIL either as this is really the
6223 2017-11-22 Richard Biener <rguenther@suse.de>
6225 PR tree-optimization/83089
6226 * tree-if-conv.c (pass_if_conversion::execute): If anything
6227 changed reset SCEV and free the number of iteration estimates.
6229 2017-11-21 Martin Sebor <msebor@redhat.com>
6231 PR tree-optimization/82945
6232 * calls.h (warn_nonstring_bound): Remove unused function.
6234 2017-11-21 Martin Sebor <msebor@redhat.com>
6236 PR tree-optimization/82945
6237 * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6238 * calls.h (maybe_warn_nonstring_arg): Declare new function.
6239 * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6241 (initialize_argument_information): Call maybe_warn_nonstring_arg.
6242 * calls.h (get_attr_nonstring_decl): Declare new function.
6243 * doc/extend.texi (attribute nonstring): Update.
6244 * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6245 get_attr_nonstring_decl and handle it.
6246 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same. Improve
6247 detection of nul-termination.
6248 (strlen_to_stridx): Change to a pointer.
6249 (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6250 (pass_strlen::execute): Same.
6252 2017-11-21 Sergey Shalnov <Sergey.Shalnov@intel.com>
6254 * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6255 for the new option -mprefer-vector-width=[none|128|256|512].
6256 * config/i386/i386.c (ix86_target_string): remove old style options
6257 -mprefer-avx256 and make -mprefer-avx128 as alias.
6258 (ix86_option_override_internal): Apply defaults for the
6259 -mprefer-vector-width=[128|256] option.
6260 * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6261 Implement macros to work with -mprefer-vector-width=.
6262 * config/i386/i386.opt: Implemented option
6263 -mprefer-vector-width=[none|128|256|512].
6264 * doc/invoke.texi: Documentation for
6265 -mprefer-vector-width=[none|128|256|512].
6267 2017-11-21 Pat Haugen <pthaugen@us.ibm.com>
6269 * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6271 2017-11-21 Jonathan Wakely <jwakely@redhat.com>
6273 * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6275 2017-11-21 Marc Glisse <marc.glisse@inria.fr>
6277 * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6279 * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6280 * expr.c (expand_expr_real_2): Likewise.
6281 * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6282 fold_binary_loc): Likewise.
6283 * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6284 P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6285 (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6286 MINUS_EXPR transformations.
6287 * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6288 * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6289 * tree-inline.c (estimate_operator_cost): Likewise.
6290 * tree-pretty-print.c (dump_generic_node, op_code_prio,
6291 op_symbol_code): Likewise.
6292 * tree-vect-stmts.c (vectorizable_operation): Likewise.
6293 * vr-values.c (extract_range_from_binary_expr): Likewise.
6294 * varasm.c (initializer_constant_valid_p_1): Likewise.
6295 * tree.def: New tree code POINTER_DIFF_EXPR.
6297 2017-11-21 Uros Bizjak <ubizjak@gmail.com>
6299 * config/i386/i386.md (*bswap<mode>2_movbe): Add
6300 integer suffix to movbe mnemonic.
6301 (*bswaphi2_movbe): Ditto.
6302 (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6304 2017-11-21 Jakub Jelinek <jakub@redhat.com>
6307 * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6308 Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6309 Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6311 2017-11-21 Martin Liska <mliska@suse.cz>
6313 * tree-inline.c (expand_call_inline): Remove not needed
6316 2017-11-21 James Cowgill <James.Cowgill@imgtec.com>
6317 Jakub Jelinek <jakub@redhat.com>
6320 * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6321 Remove static keyword from f variable.
6323 2017-11-21 Jakub Jelinek <jakub@redhat.com>
6325 PR tree-optimization/83086
6326 * gimple-ssa-store-merging.c
6327 (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6328 rather than n.base_addr.
6330 2017-11-21 Martin Liska <mliska@suse.cz>
6332 PR rtl-optimization/82044
6333 PR tree-optimization/82042
6334 * dse.c (check_mem_read_rtx): Check for overflow.
6336 2017-11-21 Aldy Hernandez <aldyh@redhat.com>
6338 * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6341 2017-11-21 Claudiu Zissulescu <claziss@synopsys.com>
6343 * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6344 and its corresponding call arg location note.
6346 2017-11-21 Claudiu Zissulescu <claziss@synopsys.com>
6347 Andrew Burgess <andrew.burgess@embecosm.com>
6349 * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6351 (arc_return_slot_offset): Likewise.
6352 (arc_eh_return_address_location): New declaration.
6353 * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6354 (MUST_SAVE_REGISTER): Add exception handler case.
6355 (MUST_SAVE_RETURN_ADDR): Likewise.
6356 (arc_frame_pointer_required): Likewise.
6357 (arc_frame_pointer_needed): New function.
6358 (arc_compute_frame_size): Changed.
6359 (arc_expand_prologue): Likewise.
6360 (arc_expand_epilogue): Likewise.
6361 (arc_initial_elimination_offset): Likewise.
6362 (arc_return_slot_offset): Delete.
6363 (arc_eh_return_address_location): New function.
6364 (arc_builtin_setjmp_frame_value): Likewise.
6365 * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6366 (EH_RETURN_STACKADJ_RTX): Define.
6367 (EH_RETURN_HANDLER_RTX): Likewise.
6368 * config/arc/arc.md (eh_return): Delete.
6370 2017-11-21 Aldy Hernandez <aldyh@redhat.com>
6372 * print-tree.h (debug_vec_tree): Remove prototype.
6373 * gdbinit.in (pvt): Remove macro.
6375 2017-11-21 Jakub Jelinek <jakub@redhat.com>
6377 PR tree-optimization/83047
6378 * gimple-ssa-store-merging.c
6379 (imm_store_chain_info::output_merged_store): If the loads with the
6380 same vuse are in different basic blocks, for load_gsi pick a load
6381 location that is dominated by the other loads.
6384 * config/i386/i386.c (ix86_memmodel_check): Start
6385 -Winvalid-memory-model diagnostics with lowercase letter.
6388 * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6389 set in_cold_section_p to the partition containing loc_list->first.
6390 When seeing loc_list->last_before_switch node, update secname and
6391 perform range_across_switch second partition handling only after that.
6394 * run-rtl-passes.c: Include debug.h.
6395 (run_rtl_passes): Call debug_hooks->assembly_start.
6396 * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6400 * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6401 OPTAB_DIRECT in calls to expand_simple_binop.
6403 2017-11-20 David Malcolm <dmalcolm@redhat.com>
6406 * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6408 2017-11-20 David Malcolm <dmalcolm@redhat.com>
6411 * spellcheck.h (best_match::blithely_get_best_candidate): New
6414 2017-11-20 Jakub Jelinek <jakub@redhat.com>
6416 * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6417 with lower case letter.
6419 2017-11-20 Uros Bizjak <ubizjak@gmail.com>
6421 * config/i386/i386.md (bswaphi2): New expander.
6422 (*bswaphi2_movbe): New insn pattern.
6423 (bswaphi -> rorhi pepehole2): New peephole pattern.
6425 2017-11-20 Jeff Law <law@redhat.com>
6427 * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6428 * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6429 * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6430 * gimple-ssa-evrp.c: Remove bits moved into new files. Include
6431 gimple-ssa-evrp-analyze.h.
6433 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6434 set BB_VISITED here.
6435 (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6437 2017-11-20 Bin Cheng <bin.cheng@arm.com>
6439 * tree-predcom.c: Add general comment on Store-Store chains.
6440 (split_data_refs_to_components): Postpone clearing eliminate_store_p
6442 (get_chain_last_ref_at): Rename into...
6443 (get_chain_last_write_at): ...this.
6444 (get_chain_last_write_before_load): New function.
6445 (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6446 CT_STORE_STORE when write reference is added.
6447 (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6448 (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6449 (initialize_root_vars_store_elim_1): Ditto.
6450 (initialize_root_vars_store_elim_2): Ditto. Replace rhs once default
6451 definition is created.
6452 (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6453 chain by replacing it with dominant stored value.
6455 2017-11-20 Bin Cheng <bin.cheng@arm.com>
6457 * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6459 2017-11-20 Marc Glisse <marc.glisse@inria.fr>
6461 * vr-values.c (extract_range_from_binary_expr): Use a full range
6464 2017-11-20 Thomas Preud'homme <thomas.preudhomme@arm.com>
6466 * config/arm/arm.md (R4_REGNUM): Define constant.
6467 (nonsecure_call_internal): Remove r4 clobber.
6468 (nonsecure_call_value_internal): Likewise.
6469 * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6470 clobber and resequence match_operands.
6471 (nonsecure_call_value_reg_thumb1_v5): Likewise.
6472 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6473 (nonsecure_call_value_reg_thumb2): Likewise.
6475 2017-11-20 Jakub Jelinek <jakub@redhat.com>
6477 PR tree-optimization/78821
6478 * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6479 if base is TARGET_MEM_REF. If base is not MEM_REF, set base_addr
6480 to the address of the base rather than the base itself.
6481 (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6482 (find_bswap_or_nop_finalize): New function.
6483 (find_bswap_or_nop): Use it.
6484 (bswap_replace): Return a tree rather than bool, change first
6485 argument from gimple * to gimple_stmt_iterator, allow inserting
6486 into an empty sequence, allow ins_stmt to be NULL - then emit
6487 all stmts into gsi. Fix up MEM_REF address gimplification.
6488 (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6489 (struct store_immediate_info): Add N and INS_STMT non-static
6491 (store_immediate_info::store_immediate_info): Initialize them
6492 from newly added ctor args.
6493 (merged_store_group::apply_stores): Formatting fixes. Sort by
6495 (stmts_may_clobber_ref_p): For stores call also
6496 refs_anti_dependent_p.
6497 (gather_bswap_load_refs): New function.
6498 (imm_store_chain_info::try_coalesce_bswap): New method.
6499 (imm_store_chain_info::coalesce_immediate_stores): Use it.
6500 (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6501 (imm_store_chain_info::output_merged_store): Fail if number of
6502 new estimated stmts is bigger or equal than old. Handle LROTATE_EXPR
6503 and NOP_EXPR rhs_code.
6504 (pass_store_merging::process_store): Compute n and ins_stmt, if
6505 ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6506 LROTATE_EXPR rhs_code. Pass n and ins_stmt to store_immediate_info
6508 (pass_store_merging::execute): Calculate dominators.
6510 * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6511 BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6512 CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6513 init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6514 find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6515 class pass_optimize_bswap, bswap_replace,
6516 pass_optimize_bswap::execute): Moved to ...
6517 * gimple-ssa-store-merging.c: ... this file.
6518 Include optabs-tree.h.
6519 (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6520 init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6521 find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6522 anonymous namespace, remove static keywords.
6523 (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6524 (pass_optimize_bswap::execute): ... rather than here. Formatting fix.
6526 2017-11-20 Jan Hubicka <hubicka@ucw.cz>
6529 * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6531 2017-11-20 Aldy Hernandez <aldyh@redhat.com>
6533 * vec.h (debug_helper): New function.
6534 (DEFINE_DEBUG_VEC): New macro.
6535 * hash-set.h (debug_helper): New function.
6536 (DEFINE_DEBUG_HASH_SET): New macro.
6537 * cfg.c (debug_slim (edge)): New function.
6538 Call DEFINE_DEBUG_VEC for edges.
6539 Call DEFINE_DEBUG_HASH_SET for edges.
6540 * cfghooks.c (debug_slim (basic_block)): New function.
6541 Call DEFINE_DEBUG_VEC for basic blocks.
6542 Call DEFINE_DEBUG_HASH_SET for basic blocks.
6543 * print-tree.c (debug_slim): New function to handle trees.
6544 Call DEFINE_DEBUG_VEC for trees.
6545 Call DEFINE_DEBUG_HASH_SET for trees.
6546 (debug (vec<tree, va_gc>) &): Remove.
6547 (debug (<vec<tree, va_gc>) *): Remove.
6548 * print-rtl.c (debug_slim): New function to handle const_rtx.
6549 Call DEFINE_DEBUG_VEC for rtx_def.
6550 Call DEFINE_DEBUG_VEC for rtx_insn.
6551 Call DEFINE_DEBUG_HASH_SET for rtx_def.
6552 Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6553 * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6554 (debug (vec<rtx_insn *> *ptr): Remove.
6555 (debug_insn_vector): Remove.
6556 * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6558 2017-11-20 Tom de Vries <tom@codesourcery.com>
6560 PR rtl-optimization/82020
6561 * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6562 IF_THEN_ELSE condition.
6564 2017-11-19 Jeff Law <law@redhat.com>
6566 * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6567 of degenerates resulting from ignoring an edge.
6569 2017-11-19 Jan Hubicka <hubicka@ucw.cz>
6572 * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6574 2017-11-19 Jan Hubicka <hubicka@ucw.cz>
6577 * profile-count.c (profile_count::to_sreal_scale): Fix return value
6578 for uninitialied counts.
6580 2017-11-19 Jan Hubicka <hubicka@ucw.cz>
6583 * tree-inline.c (estimate_num_insns): Set to 1 at least.
6585 2017-11-19 Jan Hubicka <hubicka@ucw.cz>
6588 * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6590 2017-11-19 Tom de Vries <tom@codesourcery.com>
6592 * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6595 2017-11-19 Tom de Vries <tom@codesourcery.com>
6597 * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6600 2017-11-19 Tom de Vries <tom@codesourcery.com>
6602 * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6605 2017-11-19 Tom de Vries <tom@codesourcery.com>
6607 * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6609 * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6611 2017-11-19 Tom de Vries <tom@codesourcery.com>
6613 * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6615 * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6616 semicolon after MCORE_EXPORT_NAME call.
6618 2017-11-19 Tom de Vries <tom@codesourcery.com>
6621 * vmsdbgout.c (vmsdbgout_early_finish): New function.
6622 (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6624 2017-11-18 Jan Hubicka <hubicka@ucw.cz>
6626 * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6627 (cgraph_edge::clone): Cleanup updating of profile.
6628 * ipa-cp.c (update_profiling_info): Likewise.
6629 * ipa-inline-transform.c (inline_transform): Likewise.
6630 * ipa-inline.c (inline_small_functions): Add missing space to dump.
6631 * ipa-split.c (execute_split_functions): Do not split when function
6633 * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6634 * profile-count.c (profile_count::dump): Add global0.
6635 (profile_count::to_cgraph_frequency): Do not ICE when entry is
6637 (profile_count::to_sreal_scale): Likewise.
6638 (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6639 (profile_count::combine_with_ipa_count): New function.
6640 * profile-count.h (profile_guessed_global0adjusted): New.
6641 (profile_count::adjusted_zero): New.
6642 (profile_count::global0adjusted): New.
6643 (profile_count::combine_with_ipa_count): New.
6644 * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6645 correct profile of return block of split functions.
6646 (copy_cfg_body): Remove unused profile_count.
6647 (copy_body): Likewise.
6648 (expand_call_inline): Update.
6649 (tree_function_versioning): Update.
6651 2017-11-18 Aldy Hernandez <aldyh@redhat.com>
6653 * hash-set.h (hash_set::empty): New.
6654 * tree-ssa-threadbackward.h: Delete.
6655 * tree-ssa-threadbackward.c (class thread_jumps): New.
6656 Move max_threaded_paths into class.
6657 (fsm_find_thread_path): Remove arguments that are now in class.
6658 (profitable_jump_thread_path): Rename to...
6659 (thread_jumps::profitable_jump_thread_path): ...this.
6660 (convert_and_register_jump_thread_path): Rename to...
6661 (thread_jumps::convert_and_register_current_path): ...this.
6662 (check_subpath_and_update_thread_path): Rename to...
6663 (thread_jumps::check_subpath_and_update_thread_path): ...this.
6664 (register_jump_thread_path_if_profitable): Rename to...
6665 (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6666 (handle_phi): Rename to...
6667 (thread_jumps::handle_phi): ...this.
6668 (handle_assignment): Rename to...
6669 (thread_jumps::handle_assignment): ...this.
6670 (fsm_find_control_statement_thread_paths): Rename to...
6671 (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6672 (find_jump_threads_backwards): Rename to...
6673 (thread_jumps::find_jump_threads_backwards): ...this.
6674 Initialize path local data.
6675 (pass_thread_jumps::execute): Call find_jump_threads_backwards
6676 from within thread_jumps class.
6677 (pass_early_thread_jumps::execute): Same.
6679 2017-11-17 Jan Hubicka <hubicka@ucw.cz>
6681 * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6683 2017-11-17 Steve Ellcey <sellcey@cavium.com>
6685 * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6686 to canonical location.
6688 2017-11-17 Steve Ellcey <sellcey@cavium.com>
6691 * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6693 (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6695 2017-11-17 H.J. Lu <hongjiu.lu@intel.com>
6697 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6698 rcrt1.o%s/grcrt1.o%s for -static-pie.
6700 2017-11-17 Jan Hubicka <hubicka@ucw.cz>
6702 * i386.c (ix86_multiplication_cost, ix86_division_cost,
6703 ix86_shift_rotate_cost): Break out from ...
6704 (ix86_rtx_costs): ... here.
6705 (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6708 2017-11-17 Jan Hubicka <hubicka@ucw.cz>
6710 * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6711 when entry block was promoted unlikely.
6712 (estimate_bb_frequencies): Increase frequency scale.
6713 * profile-count.h (profile_count): Export precision info.
6715 2017-11-17 Jan Hubicka <hubicka@ucw.cz>
6717 * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6718 disturbin profile of entry block.
6720 2017-11-17 Jan Hubicka <hubicka@ucw.cz>
6722 * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6725 2017-11-17 Jan Hubicka <hubicka@ucw.cz>
6727 * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6728 * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6730 2017-11-17 Jeff Law <law@redhat.com>
6732 * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6733 from evrp_dom_walker class. Various methods moved into new class.
6734 (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6735 (evrp_range_analyzer::enter): New method.
6736 (evrp_range_analyzer::leave): New method.
6737 (evrp_dom_walker): Remove delegators no longer needed by this class.
6738 Replace vr_values data member with evrp_range_analyzer
6740 * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6741 method extracted from evrp_dom_walker::before_dom_children.
6742 (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6743 (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6745 * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6746 Add private copy constructor and move assignment operators.
6747 Privatize methods and class data where trivially possible.
6748 (evrp_dom_walker::cleanup): New function, extracted from
6749 execute_early_vrp. Simplify access to class data.
6751 * vr-values.h (get_output_for_vrp): Prototype.
6752 * vr-values.c (get_output_for_vrp): New function extracted from
6753 vrp_visit_assignment_or_call and extract_range_from_stmt.
6754 (vrp_visit_assignment_or_call): Use get_output_for_vrp. Simplify.
6756 2017-11-17 Luis Machado <luis.machado@linaro.org>
6758 * config/aarch64/aarch64.c
6759 (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6760 (qdf24xx_tunings) <autoprefetcher_model>: Set to
6761 tune_params::AUTOPREFETCHER_WEAK.
6763 2017-11-17 Tamar Christina <tamar.christina@arm.com>
6766 * config/arm/arm.c (arm_valid_target_attribute_rec):
6767 Parse "arch=" and "+<ext>".
6768 (arm_valid_target_attribute_tree): Re-init global options.
6769 (arm_option_override): Make non-static.
6770 (arm_options_perform_arch_sanity_checks): Make errors fatal.
6771 * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6772 (__ARM_FEATURE_CRC32): Support undef.
6773 * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6774 * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6776 2017-11-17 David Malcolm <dmalcolm@redhat.com>
6778 * gdbinit.in (break-on-diagnostic): New command.
6780 2017-11-17 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6782 * config/i386/i386.c (ix86_expand_epilogue): Change simple
6783 return to indirect jump for EH return if control-flow
6784 protection is enabled. Change explicit 'false' argument in
6785 pro_epilogue_adjust_stack with a value of
6787 * config/i386/i386.md (simple_return_indirect_internal):
6788 Remove SImode restriction to support 64-bit.
6790 2017-11-17 Segher Boessenkool <segher@kernel.crashing.org>
6792 * combine.c (added_notes_insn): New.
6793 (try_combine): Handle added_notes_insn like added_links_insn.
6794 Rewrite return value code.
6795 (distribute_notes): Set added_notes_insn to the earliest insn we added
6798 2017-11-17 Segher Boessenkool <segher@kernel.crashing.org>
6800 PR rtl-optimization/82621
6801 * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6802 dest of one of those SETs is unused.
6804 2017-11-17 Richard Biener <rguenther@suse.de>
6807 * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6808 * tree-pretty-print.c (dump_generic_node): Handle
6809 annot_expr_parallel_kind.
6810 * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6811 * gimplify.c (gimple_boolify): Likewise.
6813 2017-11-17 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6815 * config.gcc (extra_headers): Add cet.h for x86 targets.
6816 * config/i386/cet.h: New file.
6817 * doc/install.texi: Add --enable-cet/--disable-cet.
6819 2017-11-17 Richard Biener <rguenther@suse.de>
6821 PR tree-optimization/83017
6822 * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6823 (gen_parallel_loop): Properly count iterations.
6824 (parallelize_loops): Handle loop->can_be_parallel independent
6825 of flag_loop_parallelize_all. Make static profitability test match
6827 * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6828 * invoke.texi (parloops-min-per-thread): Document.
6830 2017-11-17 Vineet Gupta <vgupta@synopsys.com>
6832 * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6833 upstreaming review comments.
6835 2017-11-17 Tamar Christina <tamar.christina@arm.com>
6837 * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6838 with fast unaligned access.
6839 * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6841 2017-11-17 Thomas Preud'homme <thomas.preudhomme@arm.com>
6843 * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
6844 to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
6846 2017-11-17 Richard Biener <rguenther@suse.de>
6848 * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
6849 folding of references.
6851 2017-11-17 Qing Zhao <qing.zhao@oracle.com>
6854 * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
6855 of replacing call to strncmp with corresponding call to strcmp when
6858 2017-11-17 Sergey Shalnov <Sergey.Shalnov@intel.com>
6860 * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
6861 option prefer-avx256 for skylake-avx512 configuration.
6862 * config/i386/i386.c (ix86_option_override_internal): Ditto.
6863 (get_builtin_code_for_version): Ditto.
6865 2017-11-17 Chung-Ju Wu <jasonwucj@gmail.com>
6866 Monk Chiang <sh.chiang04@gmail.com>
6868 * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
6869 (FIXED_REGISTERS): Reserve more register numbers.
6870 (CALL_USED_REGISTERS): Likewise.
6871 (REG_ALLOC_ORDER): Likewise.
6872 (REG_CLASS_CONTENTS): Likewise.
6873 (REGISTER_NAMES): Likewise.
6875 2017-11-17 Chung-Ju Wu <jasonwucj@gmail.com>
6876 Kito Cheng <kito.cheng@gmail.com>
6878 * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
6880 * config/nds32/iterators.md: Add vector mode iterators and attributes.
6882 2017-11-16 Steven Munroe <munroesj@gcc.gnu.org>
6884 * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
6885 parameter list for vec_splats.
6887 2017-11-16 Joseph Myers <joseph@codesourcery.com>
6889 * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
6891 (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
6893 2017-11-16 Thomas Koenig <tkoenig@gcc.gnu.org>
6896 * doc/install.texi: Document incompatibility of Perl >=5.6.26
6897 with the required version of automake 1.11.6.
6899 2017-11-16 Pat Haugen <pthaugen@us.ibm.com>
6901 * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
6903 (DU_C2_3_power9): Correct reservation combinations.
6904 (FP_DIV_power9, VEC_DIV_power9): New.
6905 (power9-alu): Split out rotate/shift...
6906 (power9-rot): ...to here, correct dispatch resource.
6907 (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
6909 (power9-fp): Correct latency.
6910 (power9-sdiv): Add div/sqrt resource.
6911 (power9-ddiv): Correct latency, add div/sqrt resource.
6912 (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
6913 (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
6915 (power9-qpdiv, power9-qpmul): Adjust resource usage.
6917 2017-11-15 Michael Meissner <meissner@linux.vnet.ibm.com>
6919 * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
6920 switch statement mapping KF built-ins to TF built-ins if we don't
6921 have the proper ISA 3.0 assembler support.
6923 2017-11-16 Jan Hubicka <hubicka@ucw.cz>
6925 * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
6926 (lower_emutls_function_body): Do not compute it.
6928 2017-11-16 Jan Hubicka <hubicka@ucw.cz>
6930 * ipa-split.c (split_bb_info): Turn time to sreal.
6931 (split_point): Likewise.
6932 (dump_split_point): Likewise.
6933 (fine_split_points): Likewise.
6934 (execute_split_functions): Only zero split_bbs; turn time to sreals.
6936 2017-11-16 Jan Hubicka <hubicka@ucw.cz>
6938 * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
6941 2017-11-16 Jan Hubicka <hubicka@ucw.cz>
6943 * predict.c (combine_predictions_for_bb): Preserve zero predicted
6945 (expensive_function_p): Remove useless assert.
6946 (determine_unlikely_bbs): Propagate also forward; determine cold blocks
6948 2017-11-16 Martin Sebor <msebor@redhat.com>
6950 PR tree-optimization/82588
6951 PR tree-optimization/82583
6952 * tree-vrp.c (check_array_ref): Handle flexible array members,
6953 string literals, and inner indices.
6954 (search_for_addr_array): Add detail to diagnostics.
6956 2017-11-16 Nathan Sidwell <nathan@acm.org>
6960 * tree.h (COPY_DECL_RTL): Rename parms for clarity.
6961 (SET_DECL_ASSEMBLER_NAME): Forward to
6962 overwrite_decl_assembler_name.
6963 (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
6964 (overwrite_decl_assembler_name): Declare.
6965 * tree.c (overwrite_decl_assembler_name): New.
6966 * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
6967 (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
6968 (LANG_HOOKS_INITIALIZER): Add it.
6969 * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
6970 * langhooks.c (lhd_set_decl_assembler_name): Use
6971 SET_DECL_ASSEMBLER_NAME.
6972 (lhd_overwrite_decl_assembler_name): Default implementation.
6974 2017-11-16 Wilco Dijkstra <wdijkstr@arm.com>
6975 Jackson Woodruff <jackson.woodruff@arm.com>
6977 PR tree-optimization/71026
6978 * match.pd: Canonicalize constant multiplies in division.
6980 2017-11-16 Wilco Dijkstra <wdijkstr@arm.com>
6982 * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
6983 * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
6984 Remove OPT_fomit_frame_pointer entry.
6985 * common/config/arc/arc-common.c: Likewise.
6986 * common/config/arm/arm-common.c: Likewise.
6987 * common/config/avr/avr-common.c: Likewise.
6988 * common/config/c6x/c6x-common.c: Likewise.
6989 * common/config/cr16/cr16-common.c: Likewise.
6990 * common/config/cris/cris-common.c: Likewise.
6991 * common/config/epiphany/epiphany-common.c: Likewise.
6992 * common/config/fr30/fr30-common.c: Likewise.
6993 * common/config/frv/frv-common.c: Likewise.
6994 * common/config/ia64/ia64-common.c: Likewise.
6995 * common/config/iq2000/iq2000-common.c: Likewise.
6996 * common/config/lm32/lm32-common.c: Likewise.
6997 * common/config/m32r/m32r-common.c: Likewise.
6998 * common/config/mcore/mcore-common.c: Likewise.
6999 * common/config/microblaze/microblaze-common.c: Likewise.
7000 * common/config/mips/mips-common.c: Likewise.
7001 * common/config/mmix/mmix-common.c: Likewise.
7002 * common/config/mn10300/mn10300-common.c: Likewise.
7003 * common/config/nios2/nios2-common.c: Likewise.
7004 * common/config/pa/pa-common.c: Likewise.
7005 * common/config/pdp11/pdp11-common.c: Likewise.
7006 * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7007 * common/config/riscv/riscv-common.c: Likewise.
7008 * common/config/rs6000/rs6000-common.c: Likewise.
7009 * common/config/rx/rx-common.c: Likewise.
7010 * common/config/s390/s390-common.c: Likewise.
7011 * common/config/sh/sh-common.c: Likewise.
7012 * common/config/sparc/sparc-common.c: Likewise.
7013 * common/config/tilegx/tilegx-common.c: Likewise.
7014 * common/config/tilepro/tilepro-common.c: Likewise.
7015 * common/config/v850/v850-common.c: Likewise.
7016 * common/config/visium/visium-common.c: Likewise.
7017 * common/config/xstormy16/xstormy16-common.c: Likewise.
7018 * common/config/xtensa/xtensa-common.c: Likewise.
7019 * invoke.texi (-fomit-frame-pointer): Update documentation.
7021 2017-11-16 Jan Hubicka <hubicka@ucw.cz>
7023 * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7025 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7027 * cfg.c (scale_bbs_frequencies_int,
7028 cale_bbs_frequencies_gcov_type): Remove.
7029 * cfg.h (scale_bbs_frequencies_int,
7030 cale_bbs_frequencies_gcov_type): Remove.
7032 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7034 * tree-ssa-loop-manip.c
7035 (scale_dominated_blocks_in_loop): Update to profile counts.
7036 (tree_transform_and_unroll_loop): Likewise.
7038 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7040 * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7041 scale_bbs_frequencies_int.
7043 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7045 * final.c (compute_alignments): Use counts rather than frequencies.
7047 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7049 * cfgloopanal.c: Include sreal.h
7050 (average_num_loop_insns): Use counts and sreal for accounting.
7052 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7054 * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7057 2017-11-16 Chung-Ju Wu <jasonwucj@gmail.com>
7058 Kito Cheng <kito.cheng@gmail.com>
7060 * config/nds32/constraints.md: Provide more constraints.
7061 * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7062 REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7063 support constraints usage.
7065 2017-11-16 Chung-Ju Wu <jasonwucj@gmail.com>
7067 * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7069 2017-11-16 Chung-Ju Wu <jasonwucj@gmail.com>
7070 Kito Cheng <kito.cheng@gmail.com>
7072 * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7073 * config/nds32/nds32.opt: Refine the layout.
7074 * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7075 TARGET_EXT_STRING): Support new options.
7076 * config/nds32/nds32.h: Likewise.
7077 * config/nds32/nds32.md: Likewise.
7078 * config/nds32/nds32-predicates.c: Likewise.
7079 * config/nds32/constraints.md: Likewise.
7080 * common/config/nds32/nds32-common.c: Likewise.
7082 2017-11-16 Julia Koval <julia.koval@intel.com>
7085 * config/i386/gfniintrin.h: Add sse check.
7086 * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7088 2017-11-16 Julia Koval <julia.koval@intel.com>
7090 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7091 OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7092 (ix86_handle_option): Handle -mavx512vbmi2.
7093 * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7094 * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7095 * config/i386/i386-c.c (__AVX512VBMI2__): New.
7096 * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7097 (ix86_valid_target_attribute_inner_p): Ditto.
7098 * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7099 * config/i386/i386.opt (mavx512vbmi2): New option.
7100 * doc/invoke.texi: Add new option.
7102 2017-11-16 Julia Koval <julia.koval@intel.com>
7104 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7105 _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7106 _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7107 _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7108 _mm512_gf2p8mul_epi8): New intrinsics.
7109 * config/i386/i386-builtin-types.def
7110 (V64QI_FTYPE_V64QI_V64QI): New type.
7111 * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7112 __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7113 __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7114 __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7115 * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7116 * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7118 2017-11-15 Uros Bizjak <ubizjak@gmail.com>
7120 * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7121 explicitly as a stream of bytes.
7123 2017-11-15 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7125 * config/rs6000/altivec.h (vec_xst_be): New #define.
7126 * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7127 and externalize from *altivec_vperm_<mode>_internal.
7128 * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7130 (XL_BE_V8HI): Likewise.
7131 (XL_BE_V4SI): Likewise.
7132 (XL_BE_V4SI): Likewise.
7133 (XL_BE_V2DI): Likewise.
7134 (XL_BE_V4SF): Likewise.
7135 (XL_BE_V2DF): Likewise.
7136 (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7137 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7138 all array entries with these keys: VSX_BUILTIN_VEC_XL,
7139 VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST. Add entries for key
7140 VSX_BUILTIN_VEC_XST_BE.
7141 * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7142 (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7144 (altivec_init_builtins): Replace conditional calls to def_builtin
7145 for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7146 __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7147 with unconditional calls. Remove calls to def_builtin for
7148 __builtin_vsx_le_be_<mode>. Add a call to def_builtin for
7149 __builtin_vec_xst_be.
7150 * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7151 to define_expand, and add alternate RTL generation for P8.
7152 (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7153 vsx_ld_elemrev_v8hi.
7154 (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7155 add alternate RTL generation for P8.
7156 (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7157 vsx_ld_elemrev_v16qi.
7158 (vsx_st_elemrev_v8hi): Convert define_insn
7159 to define_expand, and add alternate RTL generation for P8.
7160 (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7161 vsx_st_elemrev_v8hi.
7162 (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7163 add alternate RTL generation for P8.
7164 (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7165 vsx_st_elemrev_v16qi.
7167 2017-11-15 H.J. Lu <hongjiu.lu@intel.com>
7170 * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7171 TARGET_AVX512ER check.
7172 (ix86_option_override_internal): Set MASK_VZEROUPPER if
7173 neither -mzeroupper nor -mno-zeroupper is used and
7174 TARGET_EMIT_VZEROUPPER is set.
7175 * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7176 * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7178 2017-11-15 Will Schmidt <will_schmidt@vnet.ibm.com>
7180 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7181 folding of vector compares.
7182 (fold_build_vec_cmp): New helper function.
7183 (fold_compare_helper): New helper function.
7184 (builtin_function_type): Add compare builtins to the list of functions
7185 having unsigned arguments. Cosmetic updates to comment indentation.
7186 * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7187 the not+eq combination.
7189 2017-11-15 Bin Cheng <bin.cheng@arm.com>
7191 PR tree-optimization/82726
7192 PR tree-optimization/70754
7193 * tree-predcom.c (order_drefs_by_pos): New function.
7194 (combine_chains): Move code setting has_max_use_after to...
7195 (try_combine_chains): ...here. New parameter. Sort combined chains
7196 according to position information.
7197 (tree_predictive_commoning_loop): Update call to above function.
7198 (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7200 2017-11-15 Bin Cheng <bin.cheng@arm.com>
7202 PR tree-optimization/82726
7204 2017-01-23 Bin Cheng <bin.cheng@arm.com>
7206 PR tree-optimization/70754
7207 * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7208 (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE. Insert
7209 combined stmt before it if not NULL.
7210 (combine_chains): Process refs reversely and compute dominance point
7214 2017-02-23 Bin Cheng <bin.cheng@arm.com>
7216 PR tree-optimization/79663
7217 * tree-predcom.c (combine_chains): Process refs in reverse order
7218 only for ZERO length chains, and add explaining comment.
7220 2017-11-15 Tamar Christina <tamar.christina@arm.com>
7222 * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7223 * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7224 * doc/invoke.texi (ARM Options): Add armv8.3-a.
7226 2017-11-15 Tamar Christina <tamar.christina@arm.com>
7228 * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7230 2017-11-15 Martin Liska <mliska@suse.cz>
7232 * tree-cfg.c (pass_warn_function_return::execute):
7233 Compare warn_return_type for greater than zero.
7235 2017-11-15 Sebastian Peryt <sebastian.peryt@intel.com>
7239 * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7240 to return true on Xeon and not on Xeon Phi.
7241 (ix86_check_avx256_register): Changed to ...
7242 (ix86_check_avx_upper_register): ... this. Add extra check for
7243 VALID_AVX512F_REG_OR_XI_MODE.
7244 (ix86_avx_u128_mode_needed): Changed
7245 ix86_check_avx256_register to ix86_check_avx_upper_register.
7246 (ix86_check_avx256_stores): Changed to ...
7247 (ix86_check_avx_upper_stores): ... this. Changed
7248 ix86_check_avx256_register to ix86_check_avx_upper_register.
7249 (ix86_avx_u128_mode_after): Changed
7250 avx_reg256_found to avx_upper_reg_found. Changed
7251 ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7252 (ix86_avx_u128_mode_entry): Changed
7253 ix86_check_avx256_register to ix86_check_avx_upper_register.
7254 (ix86_avx_u128_mode_exit): Ditto.
7255 * config/i386/i386.h: (host_detect_local_cpu): New define.
7257 2017-11-15 Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
7259 * config/arm/xgene1.md (xgene1): Split into automatons
7260 xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7261 (xgene1_f_load): Adjust reservations and/or types.
7262 (xgene1_f_store): Likewise.
7263 (xgene1_load_pair): Likewise.
7264 (xgene1_store_pair): Likewise.
7265 (xgene1_fp_load1): Likewise.
7266 (xgene1_load1): Likewise.
7267 (xgene1_store1): Likewise.
7268 (xgene1_move): Likewise.
7269 (xgene1_alu): Likewise.
7270 (xgene1_simd): Likewise.
7271 (xgene1_bfm): Likewise.
7272 (xgene1_neon_load1): Likewise.
7273 (xgene1_neon_store1): Likewise.
7274 (xgene1_neon_logic): Likewise.
7275 (xgene1_neon_st1): Likewise.
7276 (xgene1_neon_ld1r): Likewise.
7277 (xgene1_alu_cond): Added.
7278 (xgene1_shift_reg): Likwise.
7279 (xgene1_bfx): Likewise.
7280 (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7282 2017-11-15 Jakub Jelinek <jakub@redhat.com>
7285 * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7287 (can_widen_mult_without_libcall): New function.
7288 (expand_mul_overflow): If only checking unsigned mul overflow,
7289 not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7290 Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7291 no other way works. Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7292 (expand_DIVMOD): Formatting fix.
7293 * expmed.h (expand_mult): Add NO_LIBCALL argument.
7294 * expmed.c (expand_mult): Likewise. Use OPTAB_WIDEN rather
7295 than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7297 PR tree-optimization/82977
7298 * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7299 constructed temporary to strlen_to_stridx.put.
7301 2017-11-15 Martin Liska <mliska@suse.cz>
7303 * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7304 * configure: Regenerate.
7306 2017-11-15 Martin Liska <mliska@suse.cz>
7309 * config/sh/sh-mem.cc: Use proper probability for
7312 2017-11-14 Jeff Law <law@redhat.com>
7314 * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7315 the red zone for stack_clash_protection_final_dynamic_probe targets
7316 when the total dynamic stack size is zero bytes.
7318 * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7319 blocks is post order.
7321 2017-11-15 Alexandre Oliva <aoliva@redhat.com>
7323 * dumpfile.h (TDF_COMPARE_DEBUG): New.
7324 * final.c (rest_of_clean_state): Set it for the
7325 -fcompare-debug dump.
7326 * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7327 class when TDF_COMPARE_DEBUG is set.
7329 * dwarf2out.c (gen_producer_string): Discard
7332 2017-11-15 Joseph Myers <joseph@codesourcery.com>
7335 * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7336 * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7337 (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7338 (__TGMATH_REAL_2_3): Remove macros.
7339 (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7340 (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7342 (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7344 (remquo): Define using __TGMATH_REAL_3.
7346 2017-11-14 Jeff Law <law@redhat.com>
7348 * vr-values.c: New file with contents extracted from tree-vrp.c.
7349 * Makefile.in (OBJS): Add vr-values.o
7350 * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7351 (set_value_range, set_and_canonicalize_value_range): Likewise.
7352 (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7353 (value_range_constant_singleton, symbolic_range_p): Likewise.
7354 (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7355 (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7356 (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7357 (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7358 (find_case_label_range, find_case_label_index): Likewise.
7359 (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7360 (range_int_cst_singleton_p, value_inside_range): Likewise.
7361 (get_single_symbol): Likewise.
7362 (switch_update): Move structure definition here.
7363 (to_remove_edges, to_update_switch_stmts): Provide externs.
7364 * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7365 (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7366 (vrp_val_is_min, set_value_range): Likewise.
7367 (set_and_canonicalize_value_range, copy_value_range): Likewise.
7368 (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7369 (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7370 (range_is_nonnull, range_int_cst_p): Likewwise.
7371 (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7372 (get_single_symbol, operand_less_p): Likewise
7373 (compare_values_warnv, compare_values): Likewise.
7374 (value_inside_range, value_range_constant_singleton): Likewise.
7375 (zero_nonzero_bitgs_from_vr): Likewise.
7376 (extract_range_from_binary_expr_1): Likewise.
7377 (overflow_comparison_p): Likewise.
7378 (to_remove_edges, to_update_switch_stmts): Likewise.
7379 (find_case_label-index, find_case_label_range): Likewise.
7380 (switch_update, set_value_range_to_nonnegative): Remove.
7381 (set_value_range_to_truthvalue): Likewise.
7382 (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7383 (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7384 (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7385 (find_case_label_ranges, test_for_singularity): Likewise.
7386 (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7387 (x_vr_values): Move to its remaining use site.
7389 2017-11-10 Jeff Law <law@redhat.com>
7391 * vr-values.h (VR_INITIALIZER): Move #define here.
7392 * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7393 * Makefile.in (OBJS): Add tree-evrp.o
7394 * tree-vrp.h (assert_info): Move structure definition here.
7395 (set_value_range_to_varying): Prototype.
7396 (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7397 (infer_value_range, register_edge_assert_for): Likewise.
7398 (stmt_interesting_for_vrp): Likewise.
7399 * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7400 (set_value_range_to_varying): No longer static.
7401 (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7402 (infer_value_range, register_edge_assert_for): Likewise.
7404 2017-11-14 Michael Meissner <meissner@linux.vnet.ibm.com>
7406 * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7407 generate the XXBRD instruction.
7409 * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7410 (rs6000_builtin_type_compatible): Treat _Float128 and long double
7411 as being compatible if -mabi=ieeelongdouble.
7412 * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7413 to setup float128 built-ins with hardware support.
7414 (BU_FLOAT128_HW_2): Likewise.
7415 (BU_FLOAT128_HW_3): Likewise.
7416 (BU_FLOAT128_HW_VSX_1): Likewise.
7417 (BU_FLOAT128_HW_VSX_2): Likewise.
7418 (scalar_extract_expq): Change float128 built-in functions to
7419 accommodate having both KFmode and TFmode functions. Use the
7420 KFmode variant as the default.
7421 (scalar_extract_sigq): Likewise.
7422 (scalar_test_neg_qp): Likewise.
7423 (scalar_insert_exp_q): Likewise.
7424 (scalar_insert_exp_qp): Likewise.
7425 (scalar_test_data_class_qp): Likewise.
7426 (sqrtf128_round_to_odd): Delete processing the round to odd
7427 built-in functions as special built-in functions, and define them
7428 as float128 built-ins. Use the KFmode variant as the default.
7429 (truncf128_round_to_odd): Likewise.
7430 (addf128_round_to_odd): Likewise.
7431 (subf128_round_to_odd): Likewise.
7432 (mulf128_round_to_odd): Likewise.
7433 (divf128_round_to_odd): Likewise.
7434 (fmaf128_round_to_odd): Likewise.
7435 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7436 support for KFmode and TFmode xststdcqp calls.
7437 (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7438 point, switch the built-in handlers for the get/set float128
7439 exponent, get float128 mantissa, float128 test built-ins, and the
7440 float128 round to odd built-in functions. Eliminate creating the
7441 float128 round to odd built-in functions as special built-ins.
7442 (rs6000_init_builtins): Eliminate special creation of the float128
7443 round to odd built-in functions.
7444 * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7445 function insns to support both TFmode and KFmode variants.
7446 (xsxsigqp_<mode>): Likewise.
7447 (xsiexpqpf_<mode>): Likewise.
7448 (xsiexpqp_<mode>): Likewise.
7449 (xststdcqp_<mode>): Likewise.
7450 (xststdcnegqp_<mode>): Likewise.
7451 (xststdcqp_<mode>): Likewise.
7453 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7455 * tree-ssa-threadupdate.c (compute_path_counts): Remove
7456 unused path_in_freq_ptr parameter.
7457 (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7459 2017-11-14 Jan Hubicka <hubicka@ucw.cz>
7461 * ipa-inline.c (edge_badness): Dump sreal frequency.
7462 (compute_inlined_call_time): Match natural implementaiton ...
7463 * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7464 forgotten division by CGRAPH_FREQ_BASE.
7466 2017-11-14 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7468 * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7469 Solaris 11. Update comment.
7470 * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7472 * config/sol2.h (STARTFILE_SPEC): Likewise.
7473 * configure: Regenerate.
7475 2017-11-14 Carl Love <cel@us.ibm.com>
7477 * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7478 le_ and be_ prefixes to swap* variables. Remove
7479 if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7482 2017-11-14 Jason Merrill <jason@redhat.com>
7484 Support GTY((cache)) on hash_map.
7485 * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7486 (ggc_cache_remove): Override it instead of ggc_mx.
7487 * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7488 (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7489 * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7490 (simple_cache_map_traits): Override maybe_mx.
7491 * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7492 (hash_map): Friend gt_cleare_cache.
7493 (gt_cleare_cache): New.
7494 * tree.h (tree_cache_traits): New hash_map traits class.
7495 (tree_cache_map): New typedef.
7497 2017-11-14 Richard Biener <rguenther@suse.de>
7499 * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7500 paramter and handling.
7501 (cleanup_control_flow_bb): Likewise.
7502 (cleanup_control_flow_pre): New helper performing a DFS walk
7503 to call cleanup_control_flow_bb in PRE order.
7504 (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7505 via cleanup_control_flow_pre.
7507 2017-11-14 James Greenhalgh <james.greenhalgh@arm.com>
7509 * config/aarch64/aarch64-simd.md
7510 (aarch64_simd_bsl<mode>_internal): Remove DImode.
7511 (*aarch64_simd_bsl<mode>_alt): Likewise.
7512 (aarch64_simd_bsldi_internal): New.
7513 (aarch64_simd_bsldi_alt): Likewise.
7515 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7517 * tracer.c (better_p): Do not compare frequencies.
7518 * reg-stack.c (better_edge): Likewise.
7519 * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7522 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7524 * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7525 * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7526 * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7527 * lto-streamer-in.c (input_function): Use update_max_bb_count.
7528 * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7529 * predict.c (maybe_hot_frequency_p): Inline to ...
7530 (maybe_hot_count_p): ... here; rewrite to counts.
7531 (counts_to_freqs): Rename to ...
7532 (update_max_bb_count): ... this one.
7533 (expensive_function_p): Use counts.
7534 (estimate_bb_frequencies): Update.
7535 (rebuild_frequencies): Update.
7536 * predict.h (counts_to_freqs): Rename to ...
7537 (update_max_bb_count): ... this one.
7538 * profile.c (compute_branch_probabilities): Add debug info
7539 * tree-inline.c (expand_call_inline): Update debug info.
7540 (optimize_inline_calls): Use update_max_bb_count..
7541 (tree_function_versioning): Use update_max_bb_count..
7542 * value-prof.c (gimple_value_profile_transformations):
7543 Do not use update_max_bb_count.
7545 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7547 * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7548 always use frequencies.
7550 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7552 * bb-reorder.c: Remove frequencies from comments.
7553 (better_edge_p): Use profile counts.
7554 (find_traces): Dump profile counts.
7555 (rotate_loop): Use profile counts.
7556 (find_traces_1_round): Likewise.
7557 (connect_better_edge_p): Use counts instead of probabilities for
7559 (copy_bb_p): Drop early check for non-0 frequency.
7560 (sanitize_hot_paths): Update comments.
7562 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7564 * ipa-split.c (struct split_point): Add count.
7565 (consider_split): Do not compute incoming frequency; compute incoming
7566 count and store it to split_point.
7567 (split_function): Set count of the call to split part correctly.
7569 2017-11-13 Carl Love <cel@us.ibm.com>
7571 * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7573 2017-11-13 Tom Tromey <tom@tromey.com>
7575 * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7577 2017-11-13 Carl Love <cel@us.ibm.com>
7579 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7580 Add support for builtins:
7581 unsigned int vec_first_{,miss}_match_{,or_eos}index,
7582 vector {un,}signed {char,int,short},
7583 vector {un,}signed {char,int,short}) arguments.
7584 * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7585 VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7586 Add BU_P9V_AV_2 expansions for the builtins.
7587 * config/rs6000/altivec.h (vec_first_match_index,
7588 vec_first_mismatch_index, vec_first_match_or_eos_index,
7589 vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7590 * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7591 new extern declaration.
7592 * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7593 * config/rs6000/vsx.md (first_match_index_<mode>,
7594 first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7595 first_mismatch_or_eos_index_<mode>): Add define expand.
7596 (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7597 * doc/extend.texi: Update the built-in documenation file for the new
7600 2017-11-13 Michael Meissner <meissner@linux.vnet.ibm.com>
7602 * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7603 into the min/max operations for _Float<N> and _Float<N>X types.
7605 2017-11-13 Eric Botcazou <ebotcazou@adacore.com>
7608 * dwarf2out.c (do_eh_frame): New static variable.
7609 (dwarf2out_begin_prologue): Set it.
7610 (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7612 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7614 * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7616 * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7618 2017-11-13 Eric Botcazou <ebotcazou@adacore.com>
7621 * debug.h (dwarf2out_do_eh_frame): Declare.
7622 * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7623 (dwarf2out_do_frame): Use it.
7624 (dwarf2out_do_cfi_asm): Likewise.
7625 * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7626 (dwarf2out_assembly_start): Likewise.
7627 (dwarf2out_begin_prologue): Fix comment.
7628 * toplev.c (compile_file): Always call dwarf2out_frame_finish
7629 if the target needs either debug or unwind DWARF2 info.
7630 * lto-opts.c (lto_write_options): Do not save -fexceptions,
7631 -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7632 -ftrapping-math, -ftrapv and -fwrapv.
7634 2017-11-13 Jan Hubicka <hubicka@ucw.cz>
7636 * cgraph.c (cgraph_edge::sreal_frequency): New function.
7637 * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7638 * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7639 (estimate_edge_size_and_time): Likewise.
7640 (ipa_merge_fn_summary_after_inlining): Likewise.
7641 * ipa-inline.c (cgraph_freq_base_rec): Remove.
7642 (compute_uninlined_call_time): Use sreal_frequency.
7643 (compute_inlined_call_time): Likewise.
7644 (ipa_inline): Do not initialize cgraph_freq_base_rec.
7645 * profile-count.c: Include sreal.h.
7646 (profile_count::to_sreal_scale): New.
7647 * profile-count.h: Forward declare sreal.
7648 (profile_count::to_sreal_scale): Declare.
7650 2017-11-13 Nathan Sidwell <nathan@acm.org>
7652 * diagnostic.c (maybe_line_and_column): New.
7653 (diagnostic_get_location_text): Use it.
7654 (diagnostic_report_current_module): Likewise.
7655 (test_diagnostic_get_location_text): Add tests.
7657 2017-11-13 Luis Machado <luis.machado@linaro.org>
7659 * doc/md.texi (Specifying processor pipeline description): Fix
7660 incorrect latency for the div instruction example.
7662 2017-11-13 Jakub Jelinek <jakub@redhat.com>
7664 PR tree-optimization/78821
7665 * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7666 that bit_not_p is the same.
7667 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7668 (split_group): Count precisely bit_not_p bits in each statement.
7669 (invert_op): New function.
7670 (imm_store_chain_info::output_merged_store): Use invert_op to
7671 emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7672 but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7674 2017-11-13 Martin Liska <mliska@suse.cz>
7676 * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7677 (struct source_info): Likewise.
7678 (add_branch_counts): Likewise.
7679 (add_line_counts): Likewise.
7680 (function_summary): Likewise.
7681 (output_intermediate_line): Likewise.
7682 (generate_results): Likewise.
7684 2017-11-13 Martin Liska <mliska@suse.cz>
7686 * gcov.c (struct block_info): Remove typedef for block_t.
7687 (struct line_info): Likewise.
7688 (line_info::has_block): Likewise.
7689 (EXIT_BLOCK): Likewise.
7690 (unblock): Likewise.
7691 (circuit): Likewise.
7692 (get_cycles_count): Likewise.
7693 (process_file): Likewise.
7694 (read_graph_file): Likewise.
7695 (solve_flow_graph): Likewise.
7696 (find_exception_blocks): Likewise.
7697 (add_line_counts): Likewise.
7698 (accumulate_line_info): Likewise.
7699 (output_line_details): Likewise.
7701 2017-11-13 Martin Liska <mliska@suse.cz>
7703 * gcov.c (struct arc_info): Remove typedef for arc_t.
7704 (struct line_info): Likewise.
7705 (add_branch_counts): Likewise.
7706 (output_branch_count): Likewise.
7707 (function_info::~function_info): Likewise.
7708 (circuit): Likewise.
7709 (output_intermediate_line): Likewise.
7710 (read_graph_file): Likewise.
7711 (solve_flow_graph): Likewise.
7712 (find_exception_blocks): Likewise.
7713 (add_line_counts): Likewise.
7714 (accumulate_line_info): Likewise.
7715 (output_line_details): Likewise.
7716 (output_function_details): Likewise.
7718 2017-11-13 Martin Liska <mliska@suse.cz>
7720 * gcov.c (struct function_info): Remove typedef for function_t.
7721 (struct source_info): Likewise.
7722 (source_info::get_functions_at_location): Likewise.
7723 (solve_flow_graph): Likewise.
7724 (find_exception_blocks): Likewise.
7725 (add_line_counts): Likewise.
7726 (output_intermediate_file): Likewise.
7727 (process_file): Likewise.
7728 (generate_results): Likewise.
7729 (release_structures): Likewise.
7730 (read_graph_file): Likewise.
7731 (read_count_file): Likewise.
7732 (accumulate_line_counts): Likewise.
7733 (output_lines): Likewise.
7735 2017-11-13 Martin Liska <mliska@suse.cz>
7737 * gcov.c (function_info::function_info): Remove num_counts
7738 and add vector<gcov_type>.
7739 (function_info::~function_info): Use the vector.
7740 (process_file): Likewise.
7741 (read_graph_file): Likewise.
7742 (read_count_file): Likewise.
7743 (solve_flow_graph): Likewise.
7745 2017-11-13 Martin Liska <mliska@suse.cz>
7747 * gcov.c (function_info::is_artificial): New function.
7748 (process_file): Erase all artificial early.
7749 (generate_results): Skip as all artificial are already
7752 2017-11-13 Martin Liska <mliska@suse.cz>
7754 * gcov.c (read_graph_file): Store to global vector of functions.
7755 (read_count_file): Iterate the vector.
7756 (process_file): Likewise.
7757 (generate_results): Likewise.
7758 (release_structures): Likewise.
7760 2017-11-13 Jakub Jelinek <jakub@redhat.com>
7762 PR tree-optimization/82954
7763 * gimple-ssa-store-merging.c
7764 (imm_store_chain_info::coalesce_immediate_stores): If
7765 !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7767 2017-11-13 Richard Sandiford <richard.sandiford@linaro.org>
7769 * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7770 Upddate call to ENDIAN_LANE_N.
7771 (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7772 (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7773 (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7774 and use aarch64_endian_lane_rtx.
7775 (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7777 2017-11-12 Tom de Vries <tom@codesourcery.com>
7779 * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7781 2017-11-12 Tom de Vries <tom@codesourcery.com>
7783 * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7785 2017-11-12 Tom de Vries <tom@codesourcery.com>
7787 * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7789 * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7790 * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7791 * defaults.h (ASM_OUTPUT_LABELREF): Same.
7793 2017-11-11 Martin Sebor <msebor@redhat.com>
7796 * doc/extend.texi (attribute nonstring): Remove spurious argument.
7799 * prefic.c (translate_name): Replace strncpy with memcpy to
7800 avoid -Wstringop-truncation.
7802 2017-11-10 Jan Hubicka <hubicka@ucw.cz>
7804 * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7806 2017-11-10 Jan Hubicka <hubicka@ucw.cz>
7808 * predict.c (maybe_hot_frequency_p): Do not use cfun.
7810 2017-11-10 Jan Hubicka <hubicka@ucw.cz>
7812 * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7815 2017-11-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7817 * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7818 enable generating XXBRH if the value is in a vector register.
7819 (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7820 value is in a vector register.
7821 (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7822 register to register bswap64's instead of doing the GPR sequence
7823 used on previous machines.
7824 (bswapdi2_xxbrd): New insn.
7825 (bswapdi2_reg): Disallow on ISA 3.0.
7826 (register to register bswap64 splitter): Do not split the insn on
7827 ISA 3.0 systems that use XXBRD.
7829 2017-11-10 Martin Sebor <msebor@redhat.com>
7832 * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7834 (find_subframework_file): Same.
7836 2017-11-10 Jan Hubicka <hubicka@ucw.cz>
7838 * auto-profile.c (afdo_indirect_call): Drop frequency.
7839 * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7840 (cgraph_node::create_edge): Drop frequency argument.
7841 (cgraph_node::create_indirect_edge): Drop frequency argument.
7842 (cgraph_edge::make_speculative): Drop frequency arguments.
7843 (cgraph_edge::resolve_speculation): Do not update frequencies
7844 (cgraph_edge::dump_edge_flags): Do not dump frequency.
7845 (cgraph_node::dump): Check consistency in IPA mode.
7846 (cgraph_edge::maybe_hot_p): Use IPA counter.
7847 (cgraph_edge::verify_count_and_frequency): Rename to ...
7848 (cgraph_edge::verify_count): ... this one; drop frequency checking.
7849 (cgraph_node::verify_node): Update.
7850 * cgraph.h (struct cgraph_edge): Drop frequency.
7851 (cgraph_edge::frequency): New function.
7852 * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
7854 (cgraph_edge::rebuild_edges): Likewise.
7855 * cgraphclones.c (cgraph_edge::clone): Scale only counts.
7856 (duplicate_thunk_for_node): Do not pass frequency.
7857 (cgraph_node::create_clone): Scale only counts.
7858 (cgraph_node::create_virtual_clone): Do not pass frequency.
7859 (cgraph_node::create_edge_including_clones): Do not pass frequency.
7860 (cgraph_node::create_version_clone): Do not pass frequency.
7861 * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
7862 (cgraph_node::expand_thunk): Do not pass frequency.
7863 (cgraph_node::create_wrapper): Do not pass frequency.
7864 * gimple-iterator.c (update_call_edge_frequencies): Do not pass
7866 * gimple-streamer-in.c (input_bb): Scale only IPA counts.
7867 * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
7868 * ipa-cp.c (ipcp_lattice::print): Use frequency function.
7869 (gather_caller_stats): Use frequency function.
7870 (ipcp_cloning_candidate_p): Use frequency function.
7871 (ipcp_propagate_stage): Use frequency function.
7872 (get_info_about_necessary_edges): Use frequency function.
7873 (update_profiling_info): Update only IPA profile.
7874 (update_specialized_profile): Use frequency functoin.
7875 (perhaps_add_new_callers): Update only IPA profile.
7876 * ipa-devirt.c (ipa_devirt): Use IPA profile.
7877 * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
7878 (dump_ipa_call_summary): Use frequency function.
7879 (estimate_edge_size_and_time): Use frequency function.
7880 (ipa_merge_fn_summary_after_inlining): Use frequency function.
7881 * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
7882 * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
7883 (update_noncloned_counts): ... ths one; scale counts only.
7884 (clone_inlined_nodes): Do not scale frequency.
7885 (inline_call): Do not pass frequency.
7886 * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
7887 (compute_inlined_call_time): Use IPA profile.
7888 (want_inline_small_function_p): Use IPA profile.
7889 (want_inline_self_recursive_call_p): Use IPA profile.
7890 (edge_badness): Use IPA profile.
7891 (lookup_recursive_calls): Use IPA profile.
7892 (recursive_inlining): Do not pass frequency.
7893 (resolve_noninline_speculation): Do not update frequency.
7894 (inline_small_functions): Collect max of IPA profile.
7895 (dump_overall_stats): Dump IPA porfile.
7896 (dump_inline_stats): Dump IPA porfile.
7897 (ipa_inline): Collect IPA stats.
7898 * ipa-inline.h (clone_inlined_nodes): Update prototype.
7899 * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
7900 (ipa_propagate_frequency): Use frequency function.
7901 (ipa_profile): Cleanup.
7902 * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
7903 * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
7904 * lto-cgraph.c (lto_output_edge): Do not stream frequency.
7905 (input_node): Do not stream frequency.
7906 (input_edge): Do not stream frequency.
7907 (merge_profile_summaries): Scale only IPA profiles.
7908 * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
7909 * predict.c (drop_profile): Do not recompute frequency.
7910 * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
7911 (ipa_tm_insert_gettmclone_call): Do not pass frequency.
7912 * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
7913 * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
7914 * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
7915 * tree-inline.c (copy_bb): Do not scale frequency.
7916 (expand_call_inline): Do not scale frequency.
7917 (tree_function_versioning): Do not scale frequency.
7918 * ubsan.c (ubsan_create_edge): Do not pass frequency.
7920 2017-11-10 Julia Koval <julia.koval@intel.com>
7922 * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
7923 (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
7924 (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
7925 (_mm256_maskz_gf2p8affine_epi64_epi8)
7926 (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
7927 (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
7928 * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
7929 (__builtin_ia32_vgf2p8affineqb_v32qi)
7930 (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
7931 * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
7933 2017-11-10 Tamar Christina <tamar.christina@arm.com>
7937 (arm_option_override): Refactor.
7938 (arm_option_reconfigure_globals): New.
7939 (arm_options_perform_arch_sanity_checks): New.
7940 * config/arm/arm-protos.h (arm_option_reconfigure_globals):
7942 (arm_options_perform_arch_sanity_checks): Likewise
7944 2017-11-10 Pat Haugen <pthaugen@us.ibm.com>
7946 * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
7947 (power9-qpmul): New.
7948 * rs6000/rs6000.md ("type" attr): Add qmul.
7949 (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
7950 *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
7951 *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
7953 2017-11-10 Martin Sebor <msebor@redhat.com>
7956 * builtins.c (compute_objsize): Handle arrays that
7957 compute_builtin_object_size likes to fail for. Make extern.
7958 * builtins.h (compute_objsize): Declare.
7959 (check_strncpy_sizes): New function.
7960 (expand_builtin_strncpy): Call check_strncpy_sizes.
7961 * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
7962 -Wstringop-truncation.
7963 (gimple_fold_builtin_strncat): Same.
7964 * gimple.c (gimple_build_call_from_tree): Set call location.
7965 * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
7966 (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
7967 (handle_builtin_stxncpy, handle_builtin_strncat): Same.
7968 (handle_builtin_strlen): Use strlen_to_stridx.
7969 (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
7971 Use strlen_to_stridx.
7972 (pass_strlen::execute): Release strlen_to_stridx.
7973 * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
7974 (-Wstringop-truncation): Document new option.
7976 2017-11-10 Martin Liska <mliska@suse.cz>
7978 PR gcov-profile/82702
7979 * gcov.c (main): Handle intermediate files in a different
7981 (get_gcov_intermediate_filename): New function.
7982 (output_gcov_file): Remove support of intermediate files.
7983 (generate_results): Allocate intermediate file.
7984 (release_structures): Clean-up properly fn_end.
7985 (output_intermediate_file): Start iterating with line 1.
7987 2017-11-10 Jakub Jelinek <jakub@redhat.com>
7989 PR tree-optimization/82929
7990 * gimple-ssa-store-merging.c (struct store_immediate_info): Add
7991 ops_swapped_p non-static data member.
7992 (store_immediate_info::store_immediate_info): Clear it.
7993 (imm_store_chain_info::coalesce_immediate_stores): If swapping
7994 ops set ops_swapped_p.
7995 (count_multiple_uses): Handle ops_swapped_p.
7997 2017-11-10 Martin Liska <mliska@suse.cz>
7999 * coverage.c (coverage_init): Stream information about
8000 support of has_unexecuted_blocks.
8001 * doc/gcov.texi: Document that.
8002 * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8003 * gcov.c (read_graph_file): Likewise.
8004 (output_line_beginning): Fix a small issue with
8007 2017-11-10 Bin Cheng <bin.cheng@arm.com>
8009 * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8010 reference of trivial component.
8012 2017-11-10 Jakub Jelinek <jakub@redhat.com>
8015 * gimple-ssa-store-merging.c
8016 (pass_store_merging::terminate_all_aliasing_chains): For
8017 gimple_store_p stmts also call refs_output_dependent_p.
8019 PR rtl-optimization/82913
8020 * compare-elim.c (try_merge_compare): Punt if def_insn is not
8023 2017-11-09 Jeff Law <law@redhat.com>
8025 * vr-values.h: New file with vr_values class.
8026 * tree-vrp.c: Include vr-values.h
8027 (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8028 data objects into the vr_values class.
8029 (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8030 (get_value_range): Make it a member function within vr_values class.
8031 (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8032 (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8033 (op_with_constant_singleton_value_range): Likewise.
8034 (extract_range_for_var_from_comparison_expr): Likewise.
8035 (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8036 (extract_range_from_binary_expr): Likewise.
8037 (extract_range_from_unary_expr): Likewise.
8038 (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8039 (check_for_binary_op_overflow, extract_range_basic): Likewise.
8040 (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8041 (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8042 (compare_name_with_value, compare_names): Likewise.
8043 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8044 (vrp_evaluate_conditional_warnv_with_ops): Likewise. Remove prototype.
8045 (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8046 (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8047 (extract_range_from_phi_node): Likewise.
8048 (simplify_truth_ops_using_ranges): Likewise.
8049 (simplify_div_or_mod_using_ranges): Likewise.
8050 (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8051 (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8052 (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8053 (simplify_float_conversion_using_ranges): Likewise.
8054 (simplify_internal_call_using_ranges): Likewise.
8055 (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8056 (vrp_visit_assignment_or_call): Likewise. Smuggle class instance
8057 poitner via x_vr_values for calls into gimple folder.
8058 (vrp_initialize_lattice): Make this the vr_values ctor.
8059 (vrp_free_lattice): Make this the vr_values dtor.
8060 (set_vr_value): New function.
8061 (class vrp_prop): Add vr_values data member. Add various member
8062 functions as well as member functions that delegate to vr_values.
8063 (check_array_ref): Make a member function within vrp_prop class.
8064 (search_for_addr_array, vrp_initialize): Likewise.
8065 (vrp_finalize): Likewise. Revamp to avoid direct access to
8066 vr_value, values_propagated, etc.
8067 (check_array_bounds): Extract vrp_prop class instance pointer from
8068 walk info structure. Use it to call member functions.
8069 (check_all_array_refs): Make a member function within vrp_prop class.
8070 Smuggle class instance pointer via walk info structure.
8071 (x_vr_values): New local static.
8072 (vrp_valueize): Use x_vr_values to get class instance.
8073 (vr_valueize_1): Likewise.
8074 (class vrp_folder): Add vr_values data member. Add various member
8075 functions as well as member functions that delegate to vr_values.
8076 (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8077 (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8078 class instance from vr_values. Use it to call member functions.
8079 (vrp_dom_walker): Add vr_values data member.
8080 (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8081 instance via x_vr_values.
8082 (identify_jump_threads): Accept vr_values as argument. Store
8083 it into the walker structure.
8084 (evrp_dom_walker): Add vr_values class data member. Add various
8086 (evrp_dom_walker::try_find_new_range): Use vr_values data
8087 member to access the memory allocator.
8088 (evrp_dom_walker::before_dom_children): Store vr_values class
8089 instance into the vrp_folder class.
8090 (evrp_dom_walker::push_value_range): Rework to avoid direct
8091 access to num_vr_values and vr_value.
8092 (evrp_dom_walker::pop_value_range): Likewise.
8093 (execute_early_vrp): Remove call to vrp_initialize_lattice.
8094 Use vr_values to get to dump_all_value_ranges member function.
8095 Remove call to vrp_free_lattice. Call vrp_initialize, vrp_finalize,
8096 and simplify_cond_using_ranges_2 via vrp_prop class instance.
8097 Pass vr_values class instance down to identify_jump_threads.
8098 Remove call to vrp_free_lattice.
8099 (debug_all_value_ranges): Remove.
8101 * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8102 (vrp_folder): Likewise.
8104 * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8105 Get it from the existing bitmap instead.
8106 (vrp_intersect_ranges_1): Likewise.
8108 2017-11-09 Jakub Jelinek <jakub@redhat.com>
8110 * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8112 (store_immediate_info::store_immediate_info): Add bitnotp argument,
8113 set bit_not_p to it.
8114 (imm_store_chain_info::coalesce_immediate_stores): Break group
8115 if bit_not_p is different.
8116 (count_multiple_uses, split_group,
8117 imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8118 (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8119 (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8120 result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8121 to store_immediate_info ctor.
8123 2017-11-09 Jim Wilson <jimw@sifive.com>
8125 * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8126 (scan_prog_file): Likewise.
8128 2017-11-09 Jan Hubicka <hubicka@ucw.cz>
8130 * bb-reorder.c (max_entry_frequency): Remove.
8131 (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8132 connect_traces, push_to_next_round_p): Remove prototypes.
8133 (find_traces_1_round): Use counts only.
8134 (push_to_next_round_p): Likewise.
8135 (find_traces): Likewise.
8136 (rotate_loop): Likewise.
8137 (find_traces_1_round): Likewise.
8138 (connect_traces): Likewise.
8139 (edge_order): Likewise.
8141 2017-11-09 Thomas Preud'homme <thomas.preudhomme@arm.com>
8143 * config/arm/arm.c (output_return_instruction): Add comments to
8144 indicate requirement for cmse_nonsecure_entry return to account
8145 for the size of clearing instruction output here.
8146 (thumb_exit): Likewise.
8147 * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8148 return in hardfloat mode.
8150 2017-11-09 Segher Boessenkool <segher@kernel.crashing.org>
8152 * config/rs6000/rs6000.c (machine_function): Add a bool,
8153 "toc_is_wrapped_separately".
8154 (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8155 if it wasn't explicitly set or unset, we are optimizing for speed, and
8156 doing separate shrink-wrapping.
8157 (rs6000_get_separate_components): Enable the TOC component if
8158 saving the TOC register in the prologue.
8159 (rs6000_components_for_bb): Handle the TOC component.
8160 (rs6000_emit_prologue_components): Store the TOC register where needed.
8161 (rs6000_set_handled_components): Mark TOC as handled, if handled.
8162 (rs6000_emit_prologue): Don't save the TOC if that is already done.
8164 2017-11-09 Martin Jambor <mjambor@suse.cz>
8166 * ipa-param-manipulation.c: New file.
8167 * ipa-param-manipulation.h: Likewise.
8168 * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8169 (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8170 * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8171 (ipa_parm_adjustment): Likewise.
8172 (ipa_parm_adjustment_vec): Likewise.
8173 (ipa_get_vector_of_formal_parms): Moved declaration to
8174 ipa-param-manipulation.h.
8175 (ipa_get_vector_of_formal_parm_types): Likewise.
8176 (ipa_modify_formal_parameters): Likewise.
8177 (ipa_modify_call_arguments): Likewise.
8178 (ipa_combine_adjustments): Likewise.
8179 (ipa_dump_param_adjustments): Likewise.
8180 (ipa_modify_expr): Likewise.
8181 (ipa_get_adjustment_candidate): Likewise.
8182 * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8183 ipa-param-manipulation.c.
8184 (ipa_get_vector_of_formal_parm_types): Likewise.
8185 (ipa_modify_formal_parameters): Likewise.
8186 (ipa_modify_call_arguments): Likewise.
8187 (ipa_modify_expr): Likewise.
8188 (get_ssa_base_param): Likewise.
8189 (ipa_get_adjustment_candidate): Likewise.
8190 (index_in_adjustments_multiple_times_p): Likewise.
8191 (ipa_combine_adjustments): Likewise.
8192 (ipa_dump_param_adjustments): Likewise.
8193 * tree-sra.c: Also include ipa-param-manipulation.h
8194 * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8197 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8198 Alan Hayward <alan.hayward@arm.com>
8199 David Sherwood <david.sherwood@arm.com>
8201 * doc/sourcebuild.texi (vect_masked_store): Document.
8203 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8204 Alan Hayward <alan.hayward@arm.com>
8205 David Sherwood <david.sherwood@arm.com>
8207 * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8209 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8210 Alan Hayward <alan.hayward@arm.com>
8211 David Sherwood <david.sherwood@arm.com>
8213 * doc/sourcebuild.texi (vect_variable_length): Document.
8215 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8216 Alan Hayward <alan.hayward@arm.com>
8217 David Sherwood <david.sherwood@arm.com>
8219 * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8221 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8222 Alan Hayward <alan.hayward@arm.com>
8223 David Sherwood <david.sherwood@arm.com>
8225 * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8227 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8228 Alan Hayward <alan.hayward@arm.com>
8229 David Sherwood <david.sherwood@arm.com>
8231 * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8232 previously undocumented selectors.
8233 (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8235 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8237 * doc/rtl.texi (const_vector): Say that elements can be
8238 const_wide_ints too.
8239 * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8240 * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8241 (gen_vec_duplicate): Use it instead of CONSTANT_P.
8242 * optabs.c (expand_vector_broadcast): Likewise.
8244 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8245 Alan Hayward <alan.hayward@arm.com>
8246 David Sherwood <david.sherwood@arm.com>
8248 * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8249 scaled index even if the unscaled address was invalid.
8250 Don't increase the complexity of using a scale in that case.
8252 2017-11-09 Richard Sandiford <richard.sandiford@linaro.org>
8253 Alan Hayward <alan.hayward@arm.com>
8254 David Sherwood <david.sherwood@arm.com>
8256 * doc/rtl.texi: Rewrite the subreg rules so that they partition
8257 the inner register into REGMODE_NATURAL_SIZE bytes rather than
8258 UNITS_PER_WORD bytes.
8259 * emit-rtl.c (validate_subreg): Divide subregs into blocks
8260 based on REGMODE_NATURAL_SIZE of the inner mode.
8261 (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8262 !SCALAR_FLOAT_MODE_P cases. Use REGMODE_NATURAL_SIZE for the latter.
8263 * expmed.c (lowpart_bit_field_p): Divide the value up into
8264 chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8265 * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8266 whether something is likely to occupy more than one register.
8268 2017-11-09 Jan Hubicka <hubicka@ucw.cz>
8271 * ipa-inline-transform.c (update_noncloned_frequencies): Use
8272 profile_count::adjust_for_ipa_scaling.
8273 * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8274 * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8276 * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8278 2017-11-09 Jakub Jelinek <jakub@redhat.com>
8280 * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8281 (split_group): Add total_orig and total_new arguments, estimate the
8282 number of statements related to the store group without store merging
8283 and with store merging.
8284 (imm_store_chain_info::output_merged_store): Adjust split_group
8285 callers, punt if estimated number of statements with store merging
8286 is not smaller than estimated number of statements without it.
8288 (handled_load): Remove has_single_use checks.
8289 (pass_store_merging::process_store): Likewise.
8291 2017-11-09 Richard Biener <rguenther@suse.de>
8293 PR tree-optimization/82902
8294 * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8296 2017-11-09 Martin Liska <mliska@suse.cz>
8299 * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8302 2017-11-09 Martin Liska <mliska@suse.cz>
8304 PR tree-optimization/82669
8305 * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8307 2017-11-09 Martin Liska <mliska@suse.cz>
8309 PR gcov-profile/48463
8310 * coverage.c (coverage_begin_function): Output also end locus
8311 of a function and information whether the function is
8313 * gcov-dump.c (tag_function): Parse and print the information.
8314 * gcov.c (INCLUDE_MAP): Add include.
8315 (INCLUDE_SET): Likewise.
8316 (struct line_info): Move earlier in the source file because
8317 of vector<line_info> in function_info structure.
8318 (line_info::line_info): Likewise.
8319 (line_info::has_block): Likewise.
8320 (struct source_info): Add new member index.
8321 (source_info::get_functions_at_location): New function.
8322 (function_info::group_line_p): New function.
8323 (output_intermediate_line): New function.
8324 (output_intermediate_file): Use the mentioned function.
8325 (struct function_start): New.
8326 (struct function_start_pair_hash): Likewise.
8327 (process_file): Add code that identifies group functions.
8328 Assign lines either to global or function scope.
8329 (generate_results): Skip artificial functions.
8330 (find_source): Assign index for each source file.
8331 (read_graph_file): Read new flag artificial and end_line.
8332 (add_line_counts): Assign it either to global of function scope.
8333 (accumulate_line_counts): Isolate core of the function to
8334 accumulate_line_info and call it for both function and global
8336 (accumulate_line_info): New function.
8337 (output_line_beginning): Fix GNU coding style.
8338 (print_source_line): New function.
8339 (output_line_details): Likewise.
8340 (output_function_details): Likewise.
8341 (output_lines): Iterate both source (global) scope and function
8343 (struct function_line_start_cmp): New class.
8344 * doc/gcov.texi: Reflect changes in documentation.
8346 2017-11-09 Jakub Jelinek <jakub@redhat.com>
8349 * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8350 (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8351 and similarly for not instead of neg.
8353 2017-11-08 Andi Kleen <ak@linux.intel.com>
8355 * config/i386/i386.opt: Add -mforce-indirect-call.
8356 * config/i386/predicates.md: Check for flag_force_indirect_call.
8357 * doc/invoke.texi: Document -mforce-indirect-call
8359 2017-11-08 Kito Cheng <kito.cheng@gmail.com>
8361 * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8363 (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8364 config/riscv/riscv.c (predict.h): New include.
8365 (riscv_slow_unaligned_access_p): No longer static.
8366 (riscv_block_move_straight): Add require.
8367 * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8369 2017-11-08 Jakub Jelinek <jakub@redhat.com>
8372 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8373 <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8374 nonimmediate_operand predicate for operand 1 instead of
8377 2017-11-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8379 * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8381 * config/aarch64/constraints.md (Uml): New constraint.
8382 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8385 2017-11-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8387 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8388 of two vec_duplicates into a vec_concat.
8390 2017-11-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8392 * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8393 Simplify vec_merge of vec_duplicate and vec_concat.
8394 * config/aarch64/constraints.md (Utq): New constraint.
8395 * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8398 2017-11-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8400 * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8401 Simplify vec_merge of vec_duplicate and const_vector.
8402 * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8404 * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8405 mode iterator. Update predicate on operand 1 to
8406 handle non-const_vec constants. Delete constraints.
8407 (*aarch64_combinez_be<mode>): Likewise for operand 2.
8409 2017-11-08 Jakub Jelinek <jakub@redhat.com>
8411 PR tree-optimization/78821
8412 * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8414 (store_operand_info::store_operand_info): Initialize it to false.
8415 (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8416 ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8417 store in the group, and if chain_info is non-NULL, to ignore altogether
8419 (compatible_load_p): Fail if bit_not_p does not match.
8420 (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8421 (handled_load): Fill in bit_not_p. Handle BIT_NOT_EXPR.
8422 (pass_store_merging::process_store): Adjust
8423 terminate_all_aliasing_chains calls to pass NULL in all current spots,
8424 call terminate_all_aliasing_chains newly when adding a store into
8425 a chain with non-NULL chain_info.
8427 2017-11-08 Wilco Dijkstra <wdijkstr@arm.com>
8429 * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8431 2017-11-08 Wilco Dijkstra <wdijkstr@arm.com>
8433 * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8435 (aarch64_layout_frame): Initialise emit_frame_chain.
8436 (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8437 (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8439 2017-11-08 Martin Liska <mliska@suse.cz>
8441 * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8442 of gimple_call_internal_p.
8444 2017-11-07 Tom de Vries <tom@codesourcery.com>
8446 * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8448 2017-11-07 Tom de Vries <tom@codesourcery.com>
8450 * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8453 2017-11-08 Martin Liska <mliska@suse.cz>
8456 * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8458 2017-11-07 Eric Botcazou <ebotcazou@adacore.com>
8460 * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8462 2017-11-07 Jakub Jelinek <jakub@redhat.com>
8465 * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8466 (*cmp<mode>_ccz_1): New insn with $k alternative.
8469 * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8470 RTX_COMM_COMPARE as commutative as well.
8471 (ix86_binary_operator_ok): Formatting fix.
8472 * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8473 *<code><mode>3<mask_name><round_saeonly_name>,
8474 *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8475 *<s>mul<mode>3_highpart<mask_name>,
8476 *vec_widen_umult_even_v16si<mask_name>,
8477 *vec_widen_umult_even_v8si<mask_name>,
8478 *vec_widen_umult_even_v4si<mask_name>,
8479 *vec_widen_smult_even_v16si<mask_name>,
8480 *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8481 *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8482 *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8483 *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8484 *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8485 <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8486 *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8487 *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8488 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8489 !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8490 ix86_binary_operator_ok. Formatting fixes.
8491 (*<plusminus_insn><mode>3<mask_name><round_name>,
8492 *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8495 2017-11-07 Segher Boessenkool <segher@kernel.crashing.org>
8497 * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8498 ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8499 eq and ne if TARGET_ISEL.
8500 (cmp): New code_iterator.
8501 (UNS, UNSU_, UNSIK): New code_attrs.
8502 (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8503 ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8505 ("eq<mode>3"): ... this.
8506 ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8508 ("ne<mode>3"): ... this.
8510 2017-11-07 Julia Koval <julia.koval@intel.com>
8513 * common/config/i386/i386-common.c
8514 (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8515 (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8516 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8517 * config/i386/i386.opt: Ditto.
8518 * config/i386/i386.c (ix86_target_string): Ditto.
8519 (ix86_option_override_internal): Ditto.
8520 (ix86_init_mpx_builtins): Move MPX to args2.
8521 (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8522 * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8523 __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8524 __builtin_ia32_vgf2p8affineinvqb_v32qi,
8525 __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8526 __builtin_ia32_vgf2p8affineinvqb_v16qi,
8527 __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8529 2017-11-07 Uros Bizjak <ubizjak@gmail.com>
8532 * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8533 and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8534 (zero-extendsidi peephole2): Remove peephole.
8536 2017-11-07 Eric Botcazou <ebotcazou@adacore.com>
8539 * stor-layout.c: Include attribs.h.
8540 (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8541 explicit lookup of "aligned" attribute.
8543 2017-11-07 Andrew Waterman <andrew@sifive.com>
8545 * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8546 (riscv_expand_block_move): Likewise.
8547 * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8549 (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8550 (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8551 * config/riscv/riscv.c (riscv_block_move_straight): New function.
8552 (riscv_adjust_block_mem): Likewise.
8553 (riscv_block_move_loop): Likewise.
8554 (riscv_expand_block_move): Likewise.
8555 * config/riscv/riscv.md (movmemsi): New pattern.
8557 2017-11-07 Michael Clark <michaeljclark@mac.com>
8559 * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8560 (MUSL_DYNAMIC_LINKER): Likewise.
8562 2017-11-07 Richard Sandiford <richard.sandiford@linaro.org>
8564 * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8565 (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8568 2017-11-07 Richard Biener <rguenther@suse.de>
8570 * match.pd: Fix build.
8572 2017-11-07 Wilco Dijkstra <wdijkstr@arm.com>
8573 Jackson Woodruff <jackson.woodruff@arm.com>
8575 PR tree-optimization/71026
8576 * match.pd: Canonicalize negate in division.
8578 2017-11-07 Sudakshina Das <sudi.das@arm.com>
8581 * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8583 2017-11-07 Marc Glisse <marc.glisse@inria.fr>
8585 * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8586 (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8588 2017-11-07 Marc Glisse <marc.glisse@inria.fr>
8590 * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8591 non-scalar integral types.
8592 * match.pd (negate_expr_p): Handle MINUS_EXPR.
8593 (-(A-B), -(~A)): New transformations.
8595 2017-11-07 Tom de Vries <tom@codesourcery.com>
8597 * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8598 semicolon after "do {} while (0)".
8599 * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8600 * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8601 * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8602 * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8603 * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8605 2017-11-07 Tom de Vries <tom@codesourcery.com>
8607 * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8608 after "do {} while (0)".
8609 * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8610 * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8611 * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8612 * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8613 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8615 2017-11-07 Tom de Vries <tom@codesourcery.com>
8618 * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8620 (arm_rtx_costs_internal): Add missing semicolon after
8621 HANDLE_NARROW_SHIFT_ARITH call.
8623 2017-11-06 Segher Boessenkool <segher@kernel.crashing.org>
8625 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8626 disable isel if it was not set explicitly.
8628 2017-11-06 James Bowman <james.bowman@ftdichip.com>
8630 * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8631 (add_type_attribute) likewise.
8633 2017-11-06 H.J. Lu <hongjiu.lu@intel.com>
8635 * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8636 of struct ix86_frame.
8637 (ix86_initial_elimination_offset): Likewise.
8638 (ix86_expand_split_stack_prologue): Likewise.
8640 2017-11-06 Marc Glisse <marc.glisse@inria.fr>
8642 * tree-vrp.h (enum value_range_type): Update stale comment.
8644 2017-11-06 Richard Sandiford <richard.sandiford@linaro.org>
8645 Alan Hayward <alan.hayward@arm.com>
8646 David Sherwood <david.sherwood@arm.com>
8648 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8649 (aarch64_expand_vec_perm_const): Take the number of units too.
8650 * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8651 (aarch64_expand_vec_perm_const): Likewise.
8652 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8653 (vec_perm<mode>): Update accordingly.
8655 2017-11-06 Richard Sandiford <richard.sandiford@linaro.org>
8656 Alan Hayward <alan.hayward@arm.com>
8657 David Sherwood <david.sherwood@arm.com>
8659 * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8660 Take the number of units too.
8661 * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8662 (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8663 but check for a vector mode before rather than after the call.
8664 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8665 (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8666 (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8667 (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8668 (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8669 (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8670 (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8671 (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8672 (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8673 (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8674 (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8675 (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8676 (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8677 (aarch64_sqdmull2_n<mode>): Update accordingly.
8679 2017-11-06 Richard Sandiford <richard.sandiford@linaro.org>
8680 Alan Hayward <alan.hayward@arm.com>
8681 David Sherwood <david.sherwood@arm.com>
8683 * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8684 the number of units too.
8685 * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8686 * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8687 (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8688 (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8689 (vec_store_lanesxi<mode>): Update accordingly.
8691 2017-11-06 Richard Sandiford <richard.sandiford@linaro.org>
8692 Alan Hayward <alan.hayward@arm.com>
8693 David Sherwood <david.sherwood@arm.com>
8695 * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8696 * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8697 * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8698 of units rather than the mode.
8699 * config/aarch64/iterators.md (nunits): New mode attribute.
8700 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8701 Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8702 * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8703 (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8704 (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8705 (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8706 (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8707 (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8708 (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8709 (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8710 (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8711 (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8712 (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8713 (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8714 (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8715 (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8716 (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8717 (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8718 (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8719 (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8720 (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8721 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8722 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8723 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8724 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8725 (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8726 (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8727 (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8728 (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8729 (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8730 (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8731 (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8732 (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8733 (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8734 (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8735 (aarch64_simd_vec_setv2di): Likewise.
8737 2017-11-06 Carl Love <cel@us.ibm.com>
8739 * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8741 (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8742 * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8744 * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8746 * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8747 * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8748 Add power 8 macro expansions.
8749 (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8750 * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8751 power 8 instructions. (VSX_XXBR): Add iterator.
8753 2017-11-06 Wilco Dijkstra <wdijkstr@arm.com>
8755 * config/arm/arm.md (predicable_short_it): Change default to "no",
8756 improve documentation, remove uses that are identical to the default.
8757 (enabled_for_depr_it): Rename to enabled_for_short_it.
8758 * gcc/config/arm/arm-fixed.md (predicable_short_it):
8759 Remove default uses.
8760 * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8761 * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8762 * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8763 * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8765 2017-11-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8768 * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8769 float128 helper macros, which are no longer used after deleting
8770 the old 'q' built-in functions, and moving the round to odd
8771 built-in functions to being special built-in functions.
8772 (BU_FLOAT128_2): Likewise.
8773 (BU_FLOAT128_1_HW): Likewise.
8774 (BU_FLOAT128_2_HW): Likewise.
8775 (BU_FLOAT128_3_HW): Likewise.
8776 (FABSQ): Delete old 'q' built-in functions.
8777 (COPYSIGNQ): Likewise.
8778 (SQRTF128_ODD): Move round to odd built-in functions to be
8779 special built-in functions, so that we can handle
8780 -mabi=ieeelongdouble.
8781 (TRUNCF128_ODD): Likewise.
8782 (ADDF128_ODD): Likewise.
8783 (SUBF128_ODD): Likewise.
8784 (MULF128_ODD): Likewise.
8785 (DIVF128_ODD): Likewise.
8786 (FMAF128_ODD): Likewise.
8787 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8788 built-in names to 'f128'.
8789 * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8790 old 'q' built-in functions, as the machine independent code for
8791 'f128' built-in functions handles this.
8792 (rs6000_expand_builtin): Add expansion for float128 round to odd
8793 functions, keying off on -mabi=ieeelongdouble of whether to use
8794 the KFmode or TFmode variant.
8795 (rs6000_init_builtins): Initialize the _Float128 round to odd
8797 * doc/extend.texi (PowerPC Built-in Functions): Document the old
8798 _Float128 'q' built-in functions are now mapped into the new
8799 'f128' built-in functions.
8801 2017-11-06 David Edelsohn <dje.gcc@gmail.com>
8803 * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8804 (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8806 2017-11-06 David Malcolm <dmalcolm@redhat.com>
8809 * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8810 * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8811 * toplev.c: Include "ipa-fnsummary.h".
8812 (toplev::finalize): Call ipa_fnsummary_c_finalize.
8814 2017-11-06 Jakub Jelinek <jakub@redhat.com>
8816 PR tree-optimization/82838
8817 * gimple-ssa-store-merging.c
8818 (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8819 on a separate gimple_seq which is then appended to seq.
8821 2017-11-06 Jeff Law <law@redhat.com>
8824 * config/i386/i386.c (PROBE_INTERVAL): Remove.
8825 (get_probe_interval): New functions.
8826 (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8827 (ix86_adjust_stack_and_probe): Likewise.
8828 (output_adjust_stack_and_probe): Likewise.
8829 (ix86_emit_probe_stack_range): Likewise.
8830 (ix86_expand_prologue): Likewise.
8832 2017-11-06 Richard Sandiford <richard.sandiford@linaro.org>
8834 PR tree-optimization/82816
8835 * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8836 if the modes of the two types are the same.
8837 (convert_plusminus_to_widen): Likewise.
8839 2017-11-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8841 * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8843 (usadv16qi): New define_expand.
8844 (usadv8hi): New define_expand.
8846 2017-11-06 Jan Hubicka <hubicka@ucw.cz>
8849 * ipa-inline-transform.c (update_noncloned_frequencies): Always
8851 (inline_transform): Likewise.
8852 * predict.c (counts_to_freqs): Remove useless conditional.
8853 * profile-count.h (profile_count::apply_scale): Move sanity check.
8854 * tree-inline.c (copy_bb): Always scale.
8855 (copy_cfg_body): Likewise.
8857 2017-11-06 Christophe Lyon <christophe.lyon@linaro.org>
8860 * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
8862 (*cmp_ite0): Add enabled_for_depr_it attribute.
8863 (*cmp_ite1): Likewise.
8865 2017-11-06 Segher Boessenkool <segher@kernel.crashing.org>
8867 * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
8870 2017-11-06 Richard Sandiford <richard.sandiford@linaro.org>
8872 * tree-vrp.c (vrp_int_const_binop): Return true on success and
8873 return the value by pointer.
8874 (extract_range_from_multiplicative_op_1): Update accordingly.
8875 Return as soon as an operation fails.
8877 2017-11-05 Tom de Vries <tom@codesourcery.com>
8880 * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
8881 (DEF_SANITIZER_BUILTIN): ... here.
8882 (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
8883 DEF_SANITIZER_BUILTIN in if stmt. Add missing semicolon.
8885 2017-11-05 Tom de Vries <tom@codesourcery.com>
8888 * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
8890 (ASM_OUTPUT_CASE_LABEL): Add semicolon after
8891 ASM_OUTPUT_BEFORE_CASE_LABEL call.
8892 * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
8894 * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8895 * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8896 * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8898 2017-11-05 Tom de Vries <tom@codesourcery.com>
8901 * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
8904 2017-11-04 Michael Clark <michaeljclark@mac.com>
8906 * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
8907 config/riscv/riscv.md (addsi3): Use 'i' for immediates.
8909 (*addsi3_extended): Likewise.
8910 (*addsi3_extended2): Likewise.
8911 (<optab>si3): Likewise.
8912 (<optab>di3): Likewise.
8913 (<optab><mode>3): Likewise.
8914 (<*optabe>si3_internal): Likewise.
8915 (zero_extendqi<SUPERQI:mode>2): Likewise.
8916 (*add<mode>hi3): Likewise.
8917 (*xor<mode>hi3): Likewise.
8918 (<optab>di3): Likewise.
8919 (*<optab>si3_extend): Likewise.
8920 (*sge<u>_<X:mode><GPR:mode>): Likewise.
8921 (*slt<u>_<X:mode><GPR:mode>): Likewise.
8922 (*sle<u>_<X:mode><GPR:mode>): Likewise.
8924 2017-11-04 Andrew Waterman <andrew@sifive.com>
8926 * config/riscv/riscv.c (riscv_option_override): Conditionally set
8927 TARGET_STRICT_ALIGN based upon -mtune argument.
8929 2017-11-04 Andrew Waterman <andrew@sifive.com>
8931 * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
8933 2017-11-04 Daniel Santos <daniel.santos@pobox.com>
8935 * config/i386/i386.c (choose_basereg): Use optional scratch
8936 register and add assertion.
8937 (x86_emit_outlined_ms2sysv_save): Use scratch register when
8938 needed, and don't allocate stack.
8939 (ix86_expand_prologue): Rearrange where SSE saves/stub call is
8940 emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
8941 (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
8943 2017-11-03 Jeff Law <law@redhat.com>
8945 * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
8946 (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
8947 to probe at the start of a noreturn function.
8949 2017-11-03 Jakub Jelinek <jakub@redhat.com>
8951 PR tree-optimization/78821
8952 * gimple-ssa-store-merging.c: Update the file comment.
8953 (MAX_STORE_ALIAS_CHECKS): Define.
8954 (struct store_operand_info): New type.
8955 (store_operand_info::store_operand_info): New constructor.
8956 (struct store_immediate_info): Add rhs_code and ops data members.
8957 (store_immediate_info::store_immediate_info): Add rhscode, op0r
8958 and op1r arguments to the ctor, initialize corresponding data members.
8959 (struct merged_store_group): Add load_align_base and load_align
8961 (merged_store_group::merged_store_group): Initialize them.
8962 (merged_store_group::do_merge): Update them.
8963 (merged_store_group::apply_stores): Pick the constant for
8964 encode_tree_to_bitpos from one of the two operands, or skip
8965 encode_tree_to_bitpos if neither operand is a constant.
8966 (class pass_store_merging): Add process_store method decl. Remove
8967 bool argument from terminate_all_aliasing_chains method decl.
8968 (pass_store_merging::terminate_all_aliasing_chains): Remove
8969 var_offset_p argument and corresponding handling.
8970 (stmts_may_clobber_ref_p): New function.
8971 (compatible_load_p): New function.
8972 (imm_store_chain_info::coalesce_immediate_stores): Terminate group
8973 if there is overlap and rhs_code is not INTEGER_CST. For
8974 non-overlapping stores terminate group if rhs is not mergeable.
8975 (get_alias_type_for_stmts): Change first argument from
8976 auto_vec<gimple *> & to vec<gimple *> &. Add IS_LOAD, CLIQUEP and
8977 BASEP arguments. If IS_LOAD is true, look at rhs1 of the stmts
8978 instead of lhs. Compute *CLIQUEP and *BASEP in addition to the
8980 (get_location_for_stmts): Change first argument from
8981 auto_vec<gimple *> & to vec<gimple *> &.
8982 (struct split_store): Remove orig_stmts data member, add orig_stores.
8983 (split_store::split_store): Create orig_stores rather than orig_stmts.
8984 (find_constituent_stmts): Renamed to ...
8985 (find_constituent_stores): ... this. Change second argument from
8986 vec<gimple *> * to vec<store_immediate_info *> *, push pointers
8987 to info structures rather than the statements.
8988 (split_group): Rename ALLOW_UNALIGNED argument to
8989 ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
8990 it. Adjust find_constituent_stores caller.
8991 (imm_store_chain_info::output_merged_store): Handle rhs_code other
8992 than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
8993 get_location_for_stmts callers. Set MR_DEPENDENCE_CLIQUE and
8994 MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
8995 (mem_valid_for_store_merging): New function.
8996 (handled_load): New function.
8997 (pass_store_merging::process_store): New method.
8998 (pass_store_merging::execute): Use process_store method. Adjust
8999 terminate_all_aliasing_chains caller.
9001 2017-11-03 Wilco Dijkstra <wdijkstr@arm.com>
9003 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9004 Return true for more constants, symbols and label references.
9005 (aarch64_valid_floating_const): Remove unused function.
9007 2017-11-03 Wilco Dijkstra <wdijkstr@arm.com>
9010 * config/aarch64/aarch64.c (aarch64_layout_frame):
9011 Undo forcing of LR at bottom of frame.
9013 2017-11-03 Jeff Law <law@redhat.com>
9016 * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9017 for int_registers_saved.
9019 * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9020 extracted from tree-ssa-dom.c.
9021 * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9022 * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9023 (record_equivalences_from_incoming_edge): Add additional argument
9024 to single_pred_edge_ignoring_loop_edges call.
9025 * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9026 (uncprop_dom_walker::before_dom_children): Add additional argument
9027 to single_pred_edge_ignoring_loop_edges call.
9028 * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9029 single_pred_edge_ignoring_loop_edges rather than open coding.
9030 * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9032 2017-11-03 Marc Glisse <marc.glisse@inria.fr>
9034 * match.pd (-(-A)): Rewrite.
9036 2017-11-03 Segher Boessenkool <segher@kernel.crashing.org>
9038 * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9039 (rs6000_emit_int_cmove): New declaration.
9040 * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9041 (rs6000_emit_sISEL): Delete.
9042 (rs6000_emit_int_cmove): Make non-static.
9043 * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9044 instead of rs6000_emit_sISEL.
9046 2017-11-03 Jan Hubicka <hubicka@ucw.cz>
9048 * asan.c (create_cond_insert_point): Maintain profile.
9049 * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9051 * basic-block.h (struct basic_block_def): Remove frequency.
9052 (EDGE_FREQUENCY): Use to_frequency
9053 * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9055 (find_traces): Update to use to_frequency.
9056 (find_traces_1_round): Likewise; use only IPA counts.
9057 (bb_to_key): Likewise.
9058 (connect_traces): Use IPA counts only.
9059 (copy_bb_p): Update to use to_frequency.
9060 (fix_up_crossing_landing_pad): Likewise.
9061 (sanitize_hot_paths): Likewise.
9062 * bt-load.c (basic_block_freq): Likewise.
9063 * cfg.c (init_flow): Set count_max to uninitialized.
9064 (check_bb_profile): Remove frequencies; check counts.
9065 (dump_bb_info): Do not dump frequencies.
9066 (update_bb_profile_for_threading): Update counts only.
9067 (scale_bbs_frequencies_int): Likewise.
9068 (MAX_SAFE_MULTIPLIER): Remove.
9069 (scale_bbs_frequencies_gcov_type): Update counts only.
9070 (scale_bbs_frequencies_profile_count): Update counts only.
9071 (scale_bbs_frequencies): Update counts only.
9072 * cfg.h (struct control_flow_graph): Add count-max.
9073 (update_bb_profile_for_threading): Update prototype.
9074 * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9075 (find_many_sub_basic_blocks): Likewise.
9076 * cfgcleanup.c (try_forward_edges): Likewise.
9077 (try_crossjump_to_edge): Likewise.
9078 * cfgexpand.c (expand_gimple_cond): Likewise.
9079 (expand_gimple_tailcall): Likewise.
9080 (construct_init_block): Likewise.
9081 (construct_exit_block): Likewise.
9082 * cfghooks.c (verify_flow_info): Check consistency of counts.
9083 (dump_bb_for_graph): Do not dump frequencies.
9084 (split_block_1): Do not update frequencies.
9085 (split_edge): Do not update frequencies.
9086 (make_forwarder_block): Do not update frequencies.
9087 (duplicate_block): Do not update frequencies.
9088 (account_profile_record): Do not update frequencies.
9089 * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9090 for global heuristics.
9091 * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9092 (expected_loop_iterations_unbounded): Use counts only.
9093 * cfgloopmanip.c (scale_loop_profile): Simplify.
9094 (create_empty_loop_on_edge): Simplify
9096 (duplicate_loop_to_header_edge): Simplify
9097 * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9098 (update_br_prob_note): Take care of removing note when profile
9100 (relink_block_chain): Do not dump frequency.
9101 (rtl_account_profile_record): Use to_frequency.
9102 * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9103 (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9104 (cgraph_update_edges_for_call_stmt_node): Likewise.
9105 (cgraph_edge::verify_count_and_frequency): Update.
9106 (cgraph_node::verify_node): Temporarily disable frequency verification.
9107 * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9108 to_cgraph_frequency.
9109 (cgraph_edge::rebuild_edges): Convert to ipa counts.
9110 * cgraphunit.c (init_lowered_empty_function): Do not initialize
9112 (cgraph_node::expand_thunk): Update profile.
9113 * except.c (dw2_build_landing_pads): Do not update frequency.
9114 * final.c (compute_alignments): Use to_frequency.
9115 (dump_basic_block_info): Do not dump frequency.
9116 * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9117 (dump_gimple_bb_header): Do not dump frequency.
9118 * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9120 * gimple-streamer-in.c (input_bb): Do not stream frequency.
9121 * gimple-streamer-out.c (output_bb): Do not stream frequency.
9122 * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9123 (init_before_recovery): Do not update frequency.
9124 (sched_create_recovery_edges): Do not update frequency.
9125 * hsa-gen.c (convert_switch_statements): Do not update frequency.
9126 * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9127 (ipa_cp_c_finalize): Set max_count to uninitialized.
9128 * ipa-fnsummary.c (get_minimal_bb): Use counts.
9129 (param_change_prob): Use counts.
9130 * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9132 * ipa-split.c (consider_split): Use to_frequency.
9133 (split_function): Use to_frequency.
9134 * ira-build.c (loop_compare_func): Likewise.
9135 (mark_loops_for_removal): Likewise.
9136 (mark_all_loops_for_removal): Likewise.
9137 * loop-doloop.c (doloop_modify): Do not update frequency.
9138 * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9140 * lto-streamer-in.c (input_function): Update count_max.
9141 * omp-expand.c (expand_omp_taskreg): Update count_max.
9142 * omp-simd-clone.c (simd_clone_adjust): Update profile.
9143 * predict.c (maybe_hot_frequency_p): Use to_frequency.
9144 (maybe_hot_count_p): Use ipa counts only.
9145 (maybe_hot_bb_p): Simplify.
9146 (maybe_hot_edge_p): Simplify.
9147 (probably_never_executed): Do not take frequency argument.
9148 (probably_never_executed_bb_p): Do not pass frequency.
9149 (probably_never_executed_edge_p): Likewise.
9150 (combine_predictions_for_bb): Check that profile is nonzero.
9151 (propagate_freq): Do not set frequency.
9152 (drop_profile): Simplify.
9153 (counts_to_freqs): Simplify.
9154 (expensive_function_p): Use to_frequency.
9155 (propagate_unlikely_bbs_forward): Simplify.
9156 (determine_unlikely_bbs): Simplify.
9157 (estimate_bb_frequencies): Add hack to silence graphite issues.
9158 (compute_function_frequency): Use ipa counts.
9159 (pass_profile::execute): Update.
9160 (rebuild_frequencies): Use counts only.
9161 (force_edge_cold): Use counts only.
9162 * profile-count.c (profile_count::dump): Dump new count types.
9163 (profile_count::differs_from_p): Check compatiblity.
9164 (profile_count::to_frequency): New function.
9165 (profile_count::to_cgraph_frequency): New function.
9166 * profile-count.h (struct function): Declare.
9167 (enum profile_quality): Add profile_guessed_local and
9168 profile_guessed_global0.
9169 (class profile_proability): Decrease number of bits to 29;
9170 update from_reg_br_prob_note and to_reg_br_prob_note.
9171 (class profile_count: Update comment; decrease number of bits
9172 to 61. Check compatibility.
9173 (profile_count::compatible_p): New private member function.
9174 (profile_count::ipa_p): New member function.
9175 (profile_count::operator<): Handle global zero correctly.
9176 (profile_count::operator>): Handle global zero correctly.
9177 (profile_count::operator<=): Handle global zero correctly.
9178 (profile_count::operator>=): Handle global zero correctly.
9179 (profile_count::nonzero_p): New member function.
9180 (profile_count::force_nonzero): New member function.
9181 (profile_count::max): New member function.
9182 (profile_count::apply_scale): Handle IPA scalling.
9183 (profile_count::guessed_local): New member function.
9184 (profile_count::global0): New member function.
9185 (profile_count::ipa): New member function.
9186 (profile_count::to_frequency): Declare.
9187 (profile_count::to_cgraph_frequency): Declare.
9188 * profile.c (OVERLAP_BASE): Delete.
9189 (compute_frequency_overlap): Delete.
9190 (compute_branch_probabilities): Do not use compute_frequency_overlap.
9191 * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9192 * sched-ebb.c (rank): Use counts only.
9193 * shrink-wrap.c (handle_simple_exit): Use counts only.
9194 (try_shrink_wrapping): Use counts only.
9195 (place_prologue_for_one_component): Use counts only.
9196 * tracer.c (find_best_predecessor): Use to_frequency.
9197 (find_trace): Use to_frequency.
9198 (tail_duplicate): Use to_frequency.
9199 * trans-mem.c (expand_transaction): Do not update frequency.
9200 * tree-call-cdce.c: Do not update frequency.
9201 * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9202 (gimple_merge_blocks): Likewise.
9203 (gimple_split_edge): Likewise.
9204 (gimple_duplicate_sese_region): Likewise.
9205 (gimple_duplicate_sese_tail): Likewise.
9206 (move_sese_region_to_fn): Likewise.
9207 (gimple_account_profile_record): Likewise.
9208 (insert_cond_bb): Likewise.
9209 * tree-complex.c (expand_complex_div_wide): Likewise.
9210 * tree-eh.c (lower_resx): Update profile.
9211 * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9213 (initialize_cfun): Do not initialize frequencies
9214 (freqs_to_counts): Delete.
9215 (copy_cfg_body): Ignore count parameter.
9216 (copy_body): Update.
9217 (expand_call_inline): Update count_max.
9218 (optimize_inline_calls): Update count_max.
9219 (tree_function_versioning): Update count_max.
9220 * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9221 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9223 * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9224 * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9225 (try_peel_loop): Likewise.
9226 * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9228 * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9229 (tree_transform_and_unroll_loop): Do not use frequencies
9230 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9231 Use reliable prediction only.
9232 * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9233 * tree-ssa-sink.c (select_best_block): Use to_frequency.
9234 * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9235 probability scaling.
9236 * tree-ssa-threadupdate.c (create_block_for_threading): Do
9237 not update frequency
9238 (any_remaining_duplicated_blocks): Likewise.
9239 (update_profile): Likewise.
9240 (estimated_freqs_path): Delete.
9241 (freqs_to_counts_path): Delete.
9242 (clear_counts_path): Delete.
9243 (ssa_fix_duplicate_block_edges): Likewise.
9244 (duplicate_thread_path): Likewise.
9245 * tree-switch-conversion.c (gen_inbound_check): Use counts.
9246 * tree-tailcall.c (decrease_profile): Do not update frequency.
9247 (eliminate_tail_call): Likewise.
9248 * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9249 * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9250 (optimize_mask_stores): Likewise.
9251 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9252 * ubsan.c (ubsan_expand_null_ifn): Update profile.
9253 (ubsan_expand_ptr_ifn): Update profile.
9254 * value-prof.c (gimple_ic): Simplify.
9255 * value-prof.h (gimple_ic): Update prototype.
9256 * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9257 * ipa-inline.c (compute_uninlined_call_time): Be sure that
9259 (want_inline_self_recursive_call_p): Likewise.
9260 (resolve_noninline_speculation): Only cummulate defined counts.
9261 (inline_small_functions): Use nonzero_p.
9262 (ipa_inline): Do not access freed node.
9264 2017-11-03 Wilco Dijkstra <wdijkstr@arm.com>
9266 * config/aarch64/aarch64.c (aarch64_override_options_internal):
9267 Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9269 2017-11-03 Kito Cheng <kito.cheng@gmail.com>
9271 * config/riscv/riscv.c (riscv_legitimize_move): Handle
9272 non-legitimate address.
9274 2017-11-03 Segher Boessenkool <segher@kernel.crashing.org>
9276 * config/rs6000/rs6000.md (*lt0_disi): Delete.
9277 (*lt0_<mode>di, *lt0_<mode>si): New.
9279 2017-11-03 Segher Boessenkool <segher@kernel.crashing.org>
9281 * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9282 TARGET_PAIRED_FLOAT.
9284 2017-11-03 Siddhesh Poyarekar <siddhesh.poyarekar@linaro.org>
9285 Jim Wilson <jim.wilson@linaro.org>
9287 * config/aarch64/aarch64-cores.def (saphira): New CPU.
9288 * config/aarch64/aarch64-tune.md: Regenerated.
9289 * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9290 * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9292 2017-11-03 Cupertino Miranda <cmiranda@synopsys.com>
9294 * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9295 (arc_expand_prologue): Restore blink for millicode.
9296 * config/arc/linux.h (LINK_EH_SPEC): Defined.
9298 2017-11-03 Richard Sandiford <richard.sandiford@linaro.org>
9301 * config/i386/i386.c (ix86_vector_duplicate_value): Use
9302 gen_vec_duplicate after forcing the scalar into a register.
9304 2017-11-02 Segher Boessenkool <segher@kernel.crashing.org>
9306 * combine (try_combine): Print the insns input to try_combine to the
9309 2017-11-02 Steve Ellcey <sellcey@cavium.com>
9312 * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9313 Remove second argument from aarch64_process_target_attr call.
9314 * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9316 * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9318 (aarch64_handle_attr_arch): Remove second argument.
9319 (aarch64_handle_attr_cpu): Ditto.
9320 (aarch64_handle_attr_tune): Ditto.
9321 (aarch64_handle_attr_isa_flags): Ditto.
9322 (aarch64_process_one_target_attr): Ditto.
9323 (aarch64_process_target_attr): Ditto.
9324 (aarch64_option_valid_attribute_p): Remove second argument.
9325 on aarch64_process_target_attr call.
9327 2017-11-02 David Malcolm <dmalcolm@redhat.com>
9329 * diagnostic.c: Include "selftest-diagnostic.h".
9330 (selftest::assert_location_text): New function.
9331 (selftest::test_diagnostic_get_location_text): New function.
9332 (selftest::diagnostic_c_tests): Call it.
9334 2017-11-02 David Malcolm <dmalcolm@redhat.com>
9336 * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9337 * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9338 (class selftest::test_diagnostic_context): Move to...
9339 * selftest-diagnostic.c: New file.
9340 * selftest-diagnostic.h: New file.
9342 2017-11-02 James Bowman <james.bowman@ftdichip.com>
9344 * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9345 offset range for FT32B.
9346 * config/ft32/ft32.h: option "mcompress" enables relaxation.
9347 * config/ft32/ft32.md: Add TARGET_NOPM.
9348 * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9349 * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9351 2017-11-02 Wilco Dijkstra <wdijkstr@arm.com>
9353 * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9355 2017-11-02 Jeff Law <law@redhat.com>
9357 * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9358 virtual keyword on FINAL OVERRIDE members.
9360 * tree-ssa-propagate.h (ssa_propagation_engine): Group
9361 virtuals together. Add virtual destructor.
9362 (substitute_and_fold_engine): Similarly.
9364 2017-11-02 Jan Hubicka <hubicka@ucw.cz>
9366 * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9368 2017-11-02 Richard Biener <rguenther@suse.de>
9370 PR tree-optimization/82795
9371 * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9373 2017-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9375 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9377 (gcc_GAS_CHECK_FEATURE): Remove.
9378 * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9379 gcc_AC_INITFINI_ARRAY here. Update for Solaris 11.4 changes.
9380 * configure: Regenerate.
9382 2017-11-02 Claudiu Zissulescu <claziss@synopsys.com>
9384 * config/arc/arc.c (hwloop_optimize): Account for empty
9387 2017-11-02 Richard Biener <rguenther@suse.de>
9390 * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9391 Truncate ARRAY_REF index and element size.
9393 2017-11-01 Palmer Dabbelt <palmer@dabbelt.com>
9395 * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9397 2017-11-01 Jeff Law <law@redhat.com>
9399 * tree-ssa-ccp.c (ccp_folder): New class derived from
9400 substitute_and_fold_engine.
9401 (ccp_folder::get_value): New member function.
9402 (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9403 (ccp_fold_stmt): Remove prototype.
9404 (ccp_finalize): Call substitute_and_fold from the ccp_class.
9405 * tree-ssa-copy.c (copy_folder): New class derived from
9406 substitute_and_fold_engine.
9407 (copy_folder::get_value): Renamed from get_value.
9408 (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9409 * tree-vrp.c (vrp_folder): New class derived from
9410 substitute_and_fold_engine.
9411 (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9412 (vrp_folder::get_value): New member function.
9413 (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9414 (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9415 * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9416 provide a class interface to folder/substitute routines.
9417 (ssa_prop_fold_stmt_fn): Remove typedef.
9418 (ssa_prop_get_value_fn): Likewise.
9419 (subsitute_and_fold): Remove prototype.
9420 (replace_uses_in): Likewise.
9421 * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9422 Renamed from replace_uses_in. Call the virtual member function
9423 (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9424 (substitute_and_fold_dom_walker): Remove initialization of
9425 data member entries for calbacks. Add substitute_and_fold_engine
9426 member and initialize it.
9427 (substitute_and_fold_dom_walker::before_dom_children0: Use the
9428 member functions for get_value, replace_phi_args_in c
9429 replace_uses_in, and fold_stmt calls.
9430 (substitute_and_fold_engine::substitute_and_fold): Renamed from
9431 substitute_and_fold. Remove assert. Update ctor call.
9433 * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9434 (ssa_prop_visit_phi_fn): Likewise.
9435 (class ssa_propagation_engine): New class to provide an interface
9437 * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9439 (ssa_prop_visit_phi): Likewise.
9440 (ssa_propagation_engine::simulate_stmt): Moved into class.
9441 Call visit_phi/visit_stmt from the class rather than via
9442 file scoped static variables.
9443 (ssa_propagation_engine::simulate_block): Moved into class.
9444 (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9445 (ssa_propagation_engine::ssa_propagate): Similarly. No longer
9446 set file scoped statics for the visit_stmt/visit_phi callbacks.
9447 * tree-complex.c (complex_propagate): New class derived from
9448 ssa_propagation_engine.
9449 (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9450 (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9451 (tree_lower_complex): Call ssa_propagate via the complex_propagate
9453 * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9454 ssa_propagation_engine.
9455 (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9456 (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9457 (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9458 * tree-ssa-copy.c (copy_prop): New class derived from
9459 ssa_propagation_engine.
9460 (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9461 (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9462 (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9463 * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9464 (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9465 (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9466 (execute_vrp): Call ssa_propagate from the vrp_prop class.
9468 2017-11-01 Jakub Jelinek <jakub@redhat.com>
9470 PR rtl-optimization/82778
9471 PR rtl-optimization/82597
9472 * compare-elim.c (struct comparison): Add in_a_setter field.
9473 (find_comparison_dom_walker::before_dom_children): Remove killed
9474 bitmap and df_simulate_find_defs call, instead walk the defs.
9475 Compute last_setter and initialize in_a_setter. Merge definitions
9476 with first initialization for a few variables.
9477 (try_validate_parallel): Use insn_invalid_p instead of
9478 recog_memoized. Return insn rather than just the pattern.
9479 (try_merge_compare): Fix up comment. Don't uselessly test if
9480 in_a is a REG_P. Use cmp->in_a_setter instead of walking UD
9482 (execute_compare_elim_after_reload): Remove df_chain_add_problem
9485 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9486 Alan Hayward <alan.hayward@arm.com>
9487 David Sherwood <david.sherwood@arm.com>
9489 * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9490 aarch64_hard_regno_nregs to get the number of registers
9493 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9494 Alan Hayward <alan.hayward@arm.com>
9495 David Sherwood <david.sherwood@arm.com>
9497 * config/aarch64/constraints.md (Upl): Rename to...
9499 * config/aarch64/aarch64.md
9500 (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9503 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9504 Alan Hayward <alan.hayward@arm.com>
9505 David Sherwood <david.sherwood@arm.com>
9507 * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9508 (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9511 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9512 Alan Hayward <alan.hayward@arm.com>
9513 David Sherwood <david.sherwood@arm.com>
9515 * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9516 (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9517 (aarch64_evpc_dup): Generate rtl direcly, rather than using
9519 (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9520 of a single element.
9521 * config/aarch64/iterators.md: Add a comment above the permute
9522 unspecs to say that they are generated directly by
9523 aarch64_expand_vec_perm_const.
9524 * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9526 2017-11-01 Nathan Sidwell <nathan@acm.org>
9528 * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9530 2017-11-01 Palmer Dabbelt <palmer@dabbelt.com>
9532 * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9533 and medany code models, and describe what they do.
9535 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9537 Revert accidental duplicate:
9539 * combine.c (can_change_dest_mode): Reject changes in
9540 REGMODE_NATURAL_SIZE.
9542 2017-11-01 Segher Boessenkool <segher@kernel.crashing.org>
9544 PR rtl-optimization/64682
9545 PR rtl-optimization/69567
9546 PR rtl-optimization/69737
9547 PR rtl-optimization/82683
9548 * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9549 register mentioned in the note, drop the note, unless it came from I3,
9550 in which case it should go to I3 again.
9552 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9554 * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9555 and return false if not.
9556 (clear_bytes_written_by, live_bytes_read): Update accordingly.
9558 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9560 * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9561 range is known to be empty.
9563 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9564 Alan Hayward <alan.hayward@arm.com>
9565 David Sherwood <david.sherwood@arm.com>
9567 * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9568 and CONST_VECTOR_NUNITS instead of computing the number of units from
9569 the byte sizes of the vector and element.
9570 (simplify_binary_operation_1): Likewise.
9571 (simplify_const_binary_operation): Likewise.
9572 (simplify_ternary_operation): Likewise.
9574 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9575 Alan Hayward <alan.hayward@arm.com>
9576 David Sherwood <david.sherwood@arm.com>
9578 * var-tracking.c (INT_MEM_OFFSET): Replace with...
9579 (int_mem_offset): ...this new function.
9580 (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9581 (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9582 (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9585 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9586 Alan Hayward <alan.hayward@arm.com>
9587 David Sherwood <david.sherwood@arm.com>
9589 * lower-subreg.c (interesting_mode_p): New function.
9590 (compute_costs, find_decomposable_subregs, decompose_register)
9591 (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9592 (resolve_clobber, dump_choices): Use it.
9594 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9595 Alan Hayward <alan.hayward@arm.com>
9596 David Sherwood <david.sherwood@arm.com>
9598 * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9600 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9601 Alan Hayward <alan.hayward@arm.com>
9602 David Sherwood <david.sherwood@arm.com>
9604 * alias.c (find_base_value, find_base_term): Only process integer
9605 truncations. Check the precision rather than the size.
9607 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9608 Alan Hayward <alan.hayward@arm.com>
9609 David Sherwood <david.sherwood@arm.com>
9611 * machmode.h (is_narrower_int_mode): New function
9612 * optabs.c (expand_float, expand_fix): Use it.
9613 * dwarf2out.c (rotate_loc_descriptor): Likewise.
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 * rtl.h (narrower_subreg_mode): New function.
9620 * ira-color.c (update_costs_from_allocno): Use it.
9622 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9623 Alan Hayward <alan.hayward@arm.com>
9624 David Sherwood <david.sherwood@arm.com>
9626 * optabs-query.h (convert_optab_p): New function, split out from...
9627 (convert_optab_handler): ...here.
9628 (widening_optab_handler): Delete.
9629 (find_widening_optab_handler): Remove permit_non_widening parameter.
9630 (find_widening_optab_handler_and_mode): Likewise. Provide an
9631 override that operates on mode class wrappers.
9632 * optabs-query.c (widening_optab_handler): Delete.
9633 (find_widening_optab_handler_and_mode): Remove permit_non_widening
9634 parameter. Assert that the two modes are the same class and that
9635 the "from" mode is narrower than the "to" mode. Use
9636 convert_optab_handler instead of widening_optab_handler.
9637 * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9638 instead of widening_optab_handler.
9639 * expr.c (expand_expr_real_2): Update calls to
9640 find_widening_optab_handler.
9641 * optabs.c (expand_widen_pattern_expr): Likewise.
9642 (expand_binop_directly): Take the insn_code as a parameter.
9643 (expand_binop): Only call find_widening_optab_handler for
9644 conversion optabs; use optab_handler otherwise. Update calls
9645 to find_widening_optab_handler and expand_binop_directly.
9646 Use convert_optab_handler instead of widening_optab_handler.
9647 * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9648 find_widening_optab_handler and use scalar_mode rather than
9650 (convert_plusminus_to_widen): Likewise.
9652 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9653 Alan Hayward <alan.hayward@arm.com>
9654 David Sherwood <david.sherwood@arm.com>
9656 * machmode.h (fixed_size_mode): New class.
9657 * rtl.h (get_pool_mode): Return fixed_size_mode.
9658 * gengtype.c (main): Add fixed_size_mode.
9659 * target.def (get_raw_result_mode): Return a fixed_size_mode.
9660 (get_raw_arg_mode): Likewise.
9661 * doc/tm.texi: Regenerate.
9662 * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9663 * targhooks.c (default_get_reg_raw_mode): Likewise.
9664 * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9665 * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9666 * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9667 (msp430_get_raw_result_mode): Likewise.
9668 * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9669 * dbxout.c (dbxout_parms): Require fixed-size modes.
9670 * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9671 * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9672 * omp-low.c (lower_oacc_reductions): Likewise.
9673 * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9674 (simplify_subreg): Update accordingly.
9675 * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9676 (force_const_mem): Update accordingly. Return NULL_RTX for modes
9677 that aren't fixed-size.
9678 (get_pool_mode): Return a fixed_size_mode.
9679 (output_constant_pool_2): Take a fixed_size_mode.
9681 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9682 Alan Hayward <alan.hayward@arm.com>
9683 David Sherwood <david.sherwood@arm.com>
9685 * doc/rtl.texi (vec_series): Document.
9686 (const): Say that the operand can be a vec_series.
9687 * rtl.def (VEC_SERIES): New rtx code.
9688 * rtl.h (const_vec_series_p_1): Declare.
9689 (const_vec_series_p): New function.
9690 * emit-rtl.h (gen_const_vec_series): Declare.
9691 (gen_vec_series): Likewise.
9692 * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9693 (gen_vec_series): Likewise.
9694 * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9695 * simplify-rtx.c (simplify_unary_operation): Handle negations
9697 (simplify_binary_operation_series): New function.
9698 (simplify_binary_operation_1): Use it. Handle VEC_SERIES.
9699 (test_vector_ops_series): New function.
9700 (test_vector_ops): Call it.
9701 * config/powerpcspe/altivec.md (altivec_lvsl): Use
9702 gen_const_vec_series.
9703 (altivec_lvsr): Likewise.
9704 * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9706 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9707 Alan Hayward <alan.hayward@arm.com>
9708 David Sherwood <david.sherwood@arm.com>
9710 * doc/rtl.texi (const): Update description of address constants.
9711 Say that vector constants are allowed too.
9712 * common.md (E, F): Use CONSTANT_P instead of checking for
9714 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9715 checking for CONST_VECTOR.
9716 * expmed.c (make_tree): Use build_vector_from_val for a CONST
9718 * expr.c (expand_expr_real_2): Check for vector modes instead
9719 of checking for CONST_VECTOR.
9720 * rtl.h (const_vec_p): New function.
9721 (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9722 (unwrap_const_vec_duplicate): Handle them here too.
9724 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9725 David Malcolm <dmalcolm@redhat.com>
9726 Alan Hayward <alan.hayward@arm.com>
9727 David Sherwood <david.sherwood@arm.com>
9729 * rtl.h (vec_duplicate_p): New function.
9730 * selftest-rtl.c (assert_rtx_eq_at): New function.
9731 * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9732 (assert_rtx_eq_at): Declare.
9733 * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9734 * selftest-run-tests.c (selftest::run_tests): Call it.
9735 * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9736 (simplify_unary_operation_1): Recursively handle vector duplicates.
9737 (simplify_binary_operation_1): Likewise. Handle VEC_SELECTs of
9739 (simplify_subreg): Handle subregs of vector duplicates.
9740 (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9741 (selftest::simplify_rtx_c_tests): New functions.
9743 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9744 Alan Hayward <alan.hayward@arm.com>
9745 David Sherwood <david.sherwood@arm.com>
9747 * emit-rtl.h (gen_const_vec_duplicate): Declare.
9748 (gen_vec_duplicate): Likewise.
9749 * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9751 (gen_const_vector): ...here.
9752 (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9753 (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9754 whose elements are all equal.
9755 * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9756 * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9757 (simplify_relational_operation): Likewise.
9758 * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9760 (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9761 (aarch64_expand_vector_init): Likewise.
9762 * config/arm/arm.c (neon_vdup_constant): Likewise.
9763 (neon_expand_vector_init): Likewise.
9764 (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9765 (arm_block_set_unaligned_vect): Likewise.
9766 (arm_block_set_aligned_vect): Likewise.
9767 * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9768 * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9769 (expand_vec_perm_even_odd_pack): Likewise.
9770 (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9771 * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9772 * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9773 gen_const_vec_duplicate.
9774 * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9775 * config/mips/mips.c (mips_gen_const_int_vector): Use
9776 gen_const_vec_duplicate.
9777 (mips_expand_vector_init): Use CONST0_RTX.
9778 * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9779 (define_split): Use gen_const_vec_duplicate.
9780 * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9781 (define_split): Use gen_const_vec_duplicate.
9782 * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9783 (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9784 * config/spu/spu.c (spu_const): Likewise.
9786 2017-11-01 Richard Sandiford <richard.sandiford@linaro.org>
9787 Alan Hayward <alan.hayward@arm.com>
9788 David Sherwood <david.sherwood@arm.com>
9790 * combine.c (can_change_dest_mode): Reject changes in
9791 REGMODE_NATURAL_SIZE.
9793 2017-10-31 Sandra Loosemore <sandra@codesourcery.com>
9795 * configure.ac (--enable-libssp): New.
9796 (gcc_cv_libc_provides_ssp): Check for explicit setting before
9797 trying to determine target-specific default. Adjust indentation.
9798 * configure: Regenerated.
9799 * doc/install.texi (Configuration): Expand --disable-libssp
9802 2017-10-31 Daniel Santos <daniel.santos@pobox.com>
9804 config/i386/i386.c (ix86_expand_epilogue): Correct stack
9807 2017-10-31 Martin Jambor <mjambor@suse.cz>
9810 * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9812 2017-10-31 David Malcolm <dmalcolm@redhat.com>
9814 * auto-profile.c (autofdo_source_profile::read): Use
9815 UNKNOWN_LOCATION rather than 0.
9816 * diagnostic-core.h (warning_at_rich_loc): Rename to...
9817 (warning_at): ...this overload.
9818 (warning_at_rich_loc_n): Rename to...
9819 (warning_n): ...this overload.
9820 (error_at_rich_loc): Rename to...
9821 (error_at): ...this overload.
9822 (pedwarn_at_rich_loc): Rename to...
9823 (pedwarn): ...this overload.
9824 (permerror_at_rich_loc): Rename to...
9825 (permerror): ...this overload.
9826 (inform_at_rich_loc): Rename to...
9827 (inform): ...this overload.
9828 * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9829 (diagnostic_n_impl_richloc): Rename to...
9830 (diagnostic_n_impl): ...this rich_location *-based decl.
9831 (inform_at_rich_loc): Rename to...
9832 (inform): ...this, and add an assertion.
9833 (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9834 (warning_at_rich_loc): Rename to...
9835 (warning_at): ...this, and add an assertion.
9836 (warning_at_rich_loc_n): Rename to...
9837 (warning_n): ...this, and add an assertion.
9838 (warning_n): Update location_t-based implementation for removal of
9839 location_t-based diagnostic_n_impl.
9840 (pedwarn_at_rich_loc): Rename to...
9841 (pedwarn): ...this, and add an assertion.
9842 (permerror_at_rich_loc): Rename to...
9843 (permerror): ...this, and add an assertion.
9844 (error_n): Update for removal of location_t-based diagnostic_n_impl.
9845 (error_at_rich_loc): Rename to...
9846 (error_at): ...this, and add an assertion.
9847 * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
9848 (driver::do_spec_on_infiles): Likewise.
9849 * substring-locations.c (format_warning_va): Update for renaming
9850 of inform_at_rich_loc.
9852 2017-10-31 Michael Meissner <meissner@linux.vnet.ibm.com>
9854 * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
9855 _Float<N>X built-in functions so that the variant without the
9856 "__builtin_" prefix is only enabled for the GNU C and Objective C
9857 languages when they are in non-strict ANSI/ISO mode.
9858 (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
9859 * target.def (floatn_builtin_p): Add a target hook to control
9860 whether _Float<N> and _Float<N>X built-in functions without the
9861 "__builtin_" prefix are enabled, and return true for C and
9862 Objective C in the default hook. Include langhooks.h in
9864 * targhooks.h (default_floatn_builtin_p): Likewise.
9865 * targhooks.c (default_floatn_builtin_p): Likewise.
9866 * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
9867 floatn_builtin_p target hook.
9868 * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
9870 2017-10-31 Matthew Fortune <matthew.fortune@imgtec.com>
9871 Eric Botcazou <ebotcazou@adacore.com>
9873 PR rtl-optimization/81803
9874 * lra-constraints.c (curr_insn_transform): Also reload the whole
9875 register for a strict subreg no wider than a word if this is for
9876 a WORD_REGISTER_OPERATIONS target.
9878 2017-10-31 Jason Merrill <jason@redhat.com>
9880 * gdbinit.in: Skip over inlines from timevar.h.
9882 2017-10-31 Martin Liska <mliska@suse.cz>
9884 * doc/gcov.texi: Document new option.
9885 * gcov.c (print_usage): Likewise print it.
9886 (process_args): Support the argument.
9887 (format_count): New function.
9888 (format_gcov): Use the function.
9890 2017-10-31 Martin Liska <mliska@suse.cz>
9892 * gcov.c (struct name_map): do not use typedef.
9893 Define operator== and operator<.
9894 (name_search): Remove.
9895 (name_sort): Remove.
9896 (main): Do not allocate names.
9897 (process_file): Add vertical space.
9898 (generate_results): Use std::find.
9899 (release_structures): Do not release memory.
9900 (find_source): Use std::find.
9902 2017-10-31 Martin Liska <mliska@suse.cz>
9904 * gcov.c (struct line_info): Remove it's typedef.
9905 (line_info::line_info): Add proper ctor.
9906 (line_info::has_block): Do not use a typedef.
9907 (struct source_info): Do not use typedef.
9908 (circuit): Likewise.
9909 (get_cycles_count): Likewise.
9910 (output_intermediate_file): Iterate via vector iterator.
9911 (add_line_counts): Use std::vector methods.
9912 (accumulate_line_counts): Likewise.
9913 (output_lines): Likewise.
9915 2017-10-31 Martin Liska <mliska@suse.cz>
9917 * gcov.c (struct source_info): Remove typedef.
9918 (source_info::source_info): Add proper ctor.
9919 (accumulate_line_counts): Use struct, not it's typedef.
9920 (output_gcov_file): Likewise.
9921 (output_lines): Likewise.
9922 (main): Do not allocate an array.
9923 (output_intermediate_file): Use size of vector container.
9924 (process_file): Resize the vector.
9925 (generate_results): Do not preallocate, use newly added vector
9927 (release_structures): Do not release sources.
9928 (find_source): Use vector methods.
9929 (add_line_counts): Do not use typedef.
9931 2017-10-31 Martin Liska <mliska@suse.cz>
9933 * doc/gcov.texi: Document that.
9934 * gcov.c (add_line_counts): Mark lines with a non-executed
9936 (output_line_beginning): Handle such lines.
9937 (output_lines): Pass new argument.
9938 (output_intermediate_file): Print it in intermediate format.
9940 2017-10-31 Martin Liska <mliska@suse.cz>
9942 * color-macros.h: New file.
9943 * diagnostic-color.c: Factor out color related to macros to
9945 * doc/gcov.texi: Document -k option.
9946 * gcov.c (INCLUDE_STRING): Include string.h.
9947 (print_usage): Add -k option.
9948 (process_args): Parse it.
9949 (pad_count_string): New function.
9950 (output_line_beginning): Likewise.
9951 (DEFAULT_LINE_START): New macro.
9952 (output_lines): Support color output.
9954 2017-10-31 Martin Liska <mliska@suse.cz>
9956 PR gcov-profile/82633
9957 * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
9958 their interaction with GCOV infrastructure.
9959 * configure.ac: Add -fkeep-{inline,static}-functions to
9961 * configure: Regenerate.
9963 2017-10-31 Uros Bizjak <ubizjak@gmail.com>
9966 * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
9968 2017-10-31 Segher Boessenkool <segher@kernel.crashing.org>
9971 * config/rs6000/rs6000.md (allocate_stack): Force update interval
9972 into a register if it does not fit into an immediate offset field.
9974 2017-10-31 Olivier Hainque <hainque@adacore.com>
9976 * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
9978 2017-10-31 Julia Koval <julia.koval@intel.com>
9980 * config.gcc: Add gfniintrin.h.
9981 * config/i386/gfniintrin.h: New.
9982 * config/i386/i386-builtin-types.def
9983 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9984 __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
9985 __builtin_ia32_vgf2p8affineinvqb_v32qi,
9986 __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
9987 __builtin_ia32_vgf2p8affineinvqb_v16qi,
9988 __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
9989 * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
9990 V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
9991 V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
9992 V64QI_FTYPE_V64QI_V64QI_INT): New types.
9993 * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
9994 * config/i386/immintrin.h: Include gfniintrin.h.
9995 * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
9997 2017-10-30 Eric Botcazou <ebotcazou@adacore.com>
9999 * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10001 2017-10-30 Wilco Dijkstra <wdijkstr@arm.com>
10003 * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10004 (arm_ashldi3_1bit): Remove pattern.
10005 (ashrdi3): Remove shift by 1 expansion.
10006 (arm_ashrdi3_1bit): Remove pattern.
10007 (lshrdi3): Remove shift by 1 expansion.
10008 (arm_lshrdi3_1bit): Remove pattern.
10009 * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10010 cost of ashldi3 by 1.
10011 * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10012 (<shift>di3_neon): Likewise.
10014 2017-10-30 Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
10016 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10017 both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10018 and (*aarch64_simd_mov<VQ:mode>).
10019 (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10020 pattern alternative.
10021 (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10022 attributes to match pattern alternative.
10024 2017-10-30 Steven Munroe <munroesj@gcc.gnu.org>
10026 * config.gcc (powerpc*-*-*): Add emmintrin.h.
10027 * config/rs6000/emmintrin.h: New file.
10028 * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10030 2017-10-30 Wilco Dijkstra <wdijkstr@arm.com>
10032 * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10033 * (movdi_vfp_cortexa8): Remove pattern.
10035 2017-10-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10037 * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10039 (Specific, alpha*-dec-osf5.1): Remove.
10040 (Specific, mips-sgi-irix5): Remove.
10041 (Specific, mips-sgi-irix6): Remove.
10043 2017-10-30 Jakub Jelinek <jakub@redhat.com>
10045 PR middle-end/22141
10046 * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10047 arguments to clear_bit_region_be.
10049 2017-10-30 Jim Wilson <wilson@tuliptree.org>
10051 * gimplify.c: Include memmodel.h.
10053 2017-10-30 Martin Jambor <mjambor@suse.cz>
10055 * omp-grid.c (grid_attempt_target_gridification): Also insert a
10056 condition whether loop should be executed at all.
10058 2017-10-30 Will Schmidt <will_schmidt@vnet.ibm.com>
10060 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10061 gimple folding of vec_madd() intrinsics.
10062 * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10063 fmav8hi4. (altivec_vmladduhm): Rename to fmav8hi4.
10064 * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10066 2017-10-30 Richard Biener <rguenther@suse.de>
10068 PR tree-optimization/82762
10070 2017-10-23 Richard Biener <rguenther@suse.de>
10072 PR tree-optimization/82129
10074 2017-08-01 Richard Biener <rguenther@suse.de>
10076 PR tree-optimization/81181
10077 * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10078 (compute_antic): ... end of iteration here.
10080 2017-10-30 Joseph Myers <joseph@codesourcery.com>
10082 * doc/invoke.texi (C Dialect Options): Document -std=c17,
10083 -std=iso9899:2017 and -std=gnu17.
10084 * doc/standards.texi (C Language): Document C17 support.
10085 * doc/cpp.texi (Overview): Mention -std=c17.
10086 (Standard Predefined Macros): Document C11 and C17 values of
10087 __STDC_VERSION__. Do not refer to C99 support as incomplete.
10088 * doc/extend.texi (Inline): Do not list individual options for
10089 standards newer than C99.
10090 * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10092 * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10095 2017-10-30 Maxim Ostapenko <m.ostapenko@samsung.com>
10097 * asan.c (asan_finish_file): Align asan globals array by shadow
10100 2017-10-30 Jakub Jelinek <jakub@redhat.com>
10102 PR middle-end/22141
10103 * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10104 (struct store_immediate_info): Add bitregion_start and bitregion_end
10106 (store_immediate_info::store_immediate_info): Add brs and bre
10107 arguments and initialize bitregion_{start,end} from those.
10108 (struct merged_store_group): Add bitregion_start, bitregion_end,
10109 align_base and mask fields. Drop unnecessary struct keyword from
10110 struct store_immediate_info. Add do_merge method.
10111 (clear_bit_region_be): Use memset instead of loop storing zeros.
10112 (merged_store_group::do_merge): New method.
10113 (merged_store_group::merge_into): Use do_merge. Allow gaps in between
10114 stores as long as the surrounding bitregions have no gaps.
10115 (merged_store_group::merge_overlapping): Use do_merge.
10116 (merged_store_group::apply_stores): Test that bitregion_{start,end}
10117 is byte aligned, rather than requiring that start and width are
10118 byte aligned. Drop unnecessary struct keyword from
10119 struct store_immediate_info. Allocate and populate also mask array.
10120 Make start of the arrays relative to bitregion_start rather than
10121 start and size them according to bitregion_{end,start} difference.
10122 (struct imm_store_chain_info): Drop unnecessary struct keyword from
10123 struct store_immediate_info.
10124 (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10125 (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10126 struct keyword from struct store_immediate_info.
10127 (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10128 between stores as long as the surrounding bitregions have no gaps.
10130 (struct split_store): Add orig non-static data member.
10131 (split_store::split_store): Initialize orig to false.
10132 (find_constituent_stmts): Return store_immediate_info *, non-NULL
10133 if there is exactly a single original stmt. Change stmts argument
10134 to pointer from reference, if NULL, don't push anything to it. Add
10135 first argument, use it to optimize skipping over orig stmts that
10136 are known to be before bitpos already. Simplify.
10137 (split_group): Return unsigned int count how many stores are or
10138 would be needed rather than a bool. Add allow_unaligned argument.
10139 Change split_stores argument from reference to pointer, if NULL,
10140 only do a dry run computing how many stores would be produced.
10141 Rewritten algorithm to use both alignment and misalign if
10142 !allow_unaligned and handle bitfield stores with gaps.
10143 (imm_store_chain_info::output_merged_store): Set start_byte_pos
10144 from bitregion_start instead of start. Compute allow_unaligned
10145 here, if true, do 2 split_group dry runs to compute which one
10146 produces fewer stores and prefer aligned if equal. Punt if
10147 new count is bigger or equal than original before emitting any
10148 statements, rather than during that. Remove no longer needed
10149 new_ssa_names tracking. Replace num_stmts with
10150 split_stores.length (). Use 32-bit stack allocated entries
10151 in split_stores auto_vec. Try to reuse original store lhs/rhs1
10152 if possible. Handle bitfields with gaps.
10153 (pass_store_merging::execute): Ignore bitsize == 0 stores.
10154 Compute bitregion_{start,end} for the stores and construct
10155 store_immediate_info with that. Formatting fixes.
10157 2017-10-30 Uros Bizjak <ubizjak@gmail.com>
10160 * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10161 UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10163 2017-10-29 Jim Wilson <wilson@tuliptree.org>
10165 * gimplify.c: Include tm_p.h.
10167 * common.opt (gcoff): Re-add as ignored option.
10168 (gcoff1, gcoff2, gcoff3): Likewise.
10170 * Makefile.in (OBJS): Delete sdbout.o.
10171 (GTFILES): Delete $(srcdir)/sdbout.c.
10172 * debug.h: Delete sdb_debug_hooks.
10173 * final.c: Delete sdbout.h include.
10174 (final_scan_insn): Delete SDB_DEBUG check.
10175 (rest_of_clean_state): Likewise.
10176 * output.h: Delete sdb_begin_function_line.
10177 * sdbout.c: Delete.
10178 * sdbout.h: Delete.
10179 * toplev.c: Delete sdbout.h include.
10180 (process_options): Delete SDB_DEBUG check.
10181 * tree-core.h (tree_type_common): Delete pointer field of
10183 * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10184 TYPE_SYMTAB_POINTER.
10185 * tree.h (TYPE_SYMTAB_POINTER): Delete.
10186 (TYPE_SYMTAB_IS_POINTER): Delete.
10187 (TYPE_SYMTAB_IS_DIE): Renumber.
10188 * xcoffout.c: Refer to former sdbout.c file.
10189 (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10191 * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10192 * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10193 (Debugging Options): Delete -gcoff.
10194 (-gstabs): Delete SDB reference.
10196 (-gcoff@var{level}): Delete.
10197 * doc/passes.texi (Debugging information output): Delete SDB and
10198 sdbout.c references.
10199 * doc/tm.texi: Regenerate.
10200 * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10201 (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10203 (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10204 (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10205 (SDB_DEBUGGING_INFO): Delete.
10206 (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10207 SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10208 * target.def (output_source_filename): Delete COFF reference.
10210 * common.opt (gcoff): Delete.
10211 (gxcoff+): Update Negative chain.
10212 * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10214 * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10215 * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10216 * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10217 and SDB references.
10218 (expand_function_start): Change sdb reference to past tense.
10219 (expand_function_end): Change sdb reference to past tense.
10220 * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10221 * opts.c (debug_type_names): Delete coff entry.
10222 (common_handle_option): Delete OPT_gcoff case.
10223 * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10225 * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10226 * config/cris/cris.h: Delete SDB reference in comment.
10227 * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10228 (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10229 * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10230 * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10232 (ix86_expand_prologue): Likewise.
10233 * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10234 * config/ia64/ia64.h: Likewise.
10235 * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10236 * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10238 * config/mmix/mmix.h: Likewise.
10239 * config/nds32/nds32.c: Likewise.
10240 * config/stormy/storym16.h: Likewise.
10241 * config/visium/visium.h: Likewise.
10242 * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10244 2017-10-28 Sandra Loosemore <sandra@codesourcery.com>
10246 * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10247 * config/nios2/nios2.c (nios2_initial_elimination_offset): Make
10248 FRAME_POINTER_REGNUM point at high end of local var area.
10250 2017-10-27 Eric Botcazou <ebotcazou@adacore.com>
10252 * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10253 Move comment around. Do not reset best_edge for a copiable
10254 destination if the copy would cause a partition change.
10255 (better_edge_p): Remove redundant check.
10257 2017-10-27 Uros Bizjak <ubizjak@gmail.com>
10259 * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10261 2017-10-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10263 * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10264 for math functions that have _Float<N> and _Float<N>X variants.
10265 (mathfn_built_in_2): Add support for math functions that have
10266 _Float<N> and _Float<N>X variants.
10267 (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10268 (expand_builtin_mathfn_ternary): Add support for fma with
10269 _Float<N> and _Float<N>X variants.
10270 (expand_builtin): Likewise.
10271 (fold_builtin_3): Likewise.
10272 * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10273 create math function _Float<N> and _Float<N>X variants as external
10275 (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10276 DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10277 the __builtin_ prefix and if not strict ansi, without the prefix.
10278 (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10279 (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10280 (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10281 (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10282 (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10283 (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10284 * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10285 function signatures for fma _Float<N> and _Float<N>X variants.
10286 (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10287 (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10288 (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10289 (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10290 (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10291 (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10292 * gencfn-macros.c (print_case_cfn): Add support for math functions
10293 that have _Float<N> and _Float<N>X variants.
10294 (print_define_operator_list): Likewise.
10295 (fltfn_suffixes): Likewise.
10297 * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10298 for math functions that have _Float<N> and _Float<N>X variants.
10299 (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10300 and _Float<N>X variants.
10301 (COPYSIGN): Likewise.
10304 * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10305 copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10307 (integer_valued_read_call_p): Likewise.
10308 * fold-const-call.c (fold_const_call_ss): Likewise.
10309 (fold_const_call_sss): Add support for copysign, fmin, and fmax
10310 _Float<N> and _Float<N>X variants.
10311 (fold_const_call_ssss): Add support for fma _Float<N> and
10312 _Float<N>X variants.
10313 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10314 support for copysign and fma _Float<N> and _Float<N>X variants.
10315 (backprop::process_builtin_call_use): Likewise.
10316 * tree-call-cdce.c (can_test_argument_range); Add support for
10317 sqrt _Float<N> and _Float<N>X variants.
10318 (edom_only_function): Likewise.
10319 (get_no_error_domain): Likewise.
10320 * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10321 * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10322 copysign _Float<N> and _Float<N>X variants.
10323 * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10324 handled by machine independent code.
10325 (FMAF128): Likewise.
10326 * doc/cpp.texi (Common Predefined Macros): Document defining
10327 __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10328 fma _Float<N> and _Float<N>X variants.
10330 2017-10-27 Uros Bizjak <ubizjak@gmail.com>
10333 * config/i386/i386-modes.def (CCFPU): Remove definition.
10334 * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10335 (ix86_cc_modes_compatible): Ditto.
10336 (ix86_expand_carry_flag_compare): Ditto.
10337 (ix86_expand_int_movcc): Ditto.
10338 (ix86_expand_int_addcc): Ditto.
10339 (ix86_reverse_condition): Ditto.
10340 (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10341 Return true/false for unordered/ordered fp comparisons.
10342 (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10343 (ix86_prepare_fp_compare_args): Update for rename.
10344 (ix86_expand_fp_compare): Update for rename. Generate unordered
10345 compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10346 (ix86_expand_sse_compare_and_jump): Ditto.
10347 * config/i386/predicates.md (fcmov_comparison_operator):
10348 Remove CCFPU mode handling.
10349 (ix86_comparison_operator): Ditto.
10350 (ix86_carry_flag_operator): Ditto.
10351 * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10352 (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10353 (*cmpu<mode>_cc_i387): Ditto.
10354 (FPCMP): Remove mode iterator.
10355 (unord): Remove mode attribute.
10356 (unord_subst): New define_subst transformation
10357 (unord): New define_subst attribute.
10358 (unordered): Ditto.
10359 (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10360 (*cmpi<unord>xf_i387): Ditto.
10361 * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10362 from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10363 using unord_subst transformation.
10364 * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10365 (round_saeonly): Also handle CCFP mode.
10366 * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10367 Remove UNSPEC_SAHF unspec handling.
10369 2017-10-27 Jan Hubicka <hubicka@ucw.cz>
10371 * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10373 2017-10-27 Jeff Law <law@redhat.com>
10375 * gimple-ssa-sprintf.c: Include domwalk.h.
10376 (class sprintf_dom_walker): New class, derived from dom_walker.
10377 (sprintf_dom_walker::before_dom_children): New function.
10378 (struct call_info): Moved into sprintf_dom_walker class
10379 (compute_formath_length, handle_gimple_call): Likewise.
10380 (sprintf_length::execute): Call the dominator walker rather
10381 than walking the statements.
10383 * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10384 gimple statement locations.
10385 (check_array_bounds): Corresponding changes. Get the statement's
10386 location directly from wi->stmt.
10388 2017-10-27 Palmer Dabbelt <palmer@dabbelt.com>
10391 * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10393 2017-10-27 Jan Hubicka <hubicka@ucw.cz>
10395 * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10396 X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10398 2017-10-27 Jakub Jelinek <jakub@redhat.com>
10401 * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10402 * config/i386/i386.c (maybe_get_pool_constant): Removed.
10403 (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10404 maybe_get_pool_constant.
10405 * config/i386/predicates.md (zero_extended_scalar_load_operand):
10408 2017-10-27 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10410 * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10411 2.26 caveat. Update gas and gld versions.
10412 (Specific, *-*-solaris2*): Update binutils version. Remove caveat
10415 2017-10-27 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10417 * cgraph.h (set_malloc_flag): Declare.
10418 * cgraph.c (set_malloc_flag_1): New function.
10419 (set_malloc_flag): Likewise.
10420 * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10421 * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10423 (read_ipa_call_summary): Add support for reading is_return_callee.
10424 (write_ipa_call_summary): Stream is_return_callee.
10425 * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10426 * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10427 ipa-prop.h, ipa-fnsummary.h.
10428 (pure_const_names): Change to static.
10429 (malloc_state_e): Define.
10430 (malloc_state_names): Define.
10431 (funct_state_d): Add field malloc_state.
10432 (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10433 (check_retval_uses): New function.
10434 (malloc_candidate_p): Likewise.
10435 (analyze_function): Add support for malloc attribute.
10436 (pure_const_write_summary): Stream malloc_state.
10437 (pure_const_read_summary): Add support for reading malloc_state.
10438 (dump_malloc_lattice): New function.
10439 (propagate_malloc): New function.
10440 (warn_function_malloc): New function.
10441 (ipa_pure_const::execute): Call propagate_malloc and
10442 ipa_free_fn_summary.
10443 (pass_local_pure_const::execute): Add support for malloc attribute.
10444 * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10445 * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10447 2017-10-27 Martin Liska <mliska@suse.cz>
10449 PR gcov-profile/82457
10450 * doc/invoke.texi: Document that one needs a non-strict ISO mode
10451 for fork-like functions to be properly instrumented.
10453 2017-10-27 Richard Biener <rguenther@suse.de>
10455 PR middle-end/81659
10456 * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10457 info when we redirected EH.
10459 2017-10-26 Michael Collison <michael.collison@arm.com>
10461 * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10463 (<optab>_trunchf<GPI:mode>2: New pattern.
10464 (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10465 * config/aarch64/iterators.md (wv): New mode attribute.
10466 (vf, VF): New mode attributes.
10467 (vgp, VGP): New mode attributes.
10468 (s): Update attribute with SImode and DImode prefixes.
10470 2017-10-26 Sandra Loosemore <sandra@codesourcery.com>
10472 * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10473 * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10474 * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10475 (nios2_option_overide): Initialize it. Don't allow R0-relative
10476 addressing with PIC.
10477 (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10478 (nios2_symbolic_constant_p): Likewise.
10479 (nios2_legitimate_address_p): Likewise.
10480 (nios2_r0rel_section_name_p): New.
10481 (nios2_symbol_ref_in_r0rel_data_p): New.
10482 (nios2_emit_move_sequence): Handle r0rel_constant_p.
10483 (r0rel_constant_p): New.
10484 (nios2_print_operand_address): Handle r0rel_constant_p.
10485 (nios2_cdx_narrow_form_p): Likewise.
10486 * config/nios2/nios2.opt (mr0rel-sec=): New option.
10487 * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10488 (Nios II Options): Document -mr0rel-sec.
10490 2017-10-26 Sandra Loosemore <sandra@codesourcery.com>
10492 * config/nios2/nios2.c: Include xregex.h.
10493 (nios2_gprel_sec_regex): New.
10494 (nios2_option_overide): Initialize it. Don't allow GP-relative
10495 addressing with PIC.
10496 (nios2_small_section_name_p): Check for regex match.
10497 * config/nios2/nios2.opt (mgprel-sec=): New option.
10498 * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10499 (Nios II Options): Document -mgprel-sec.
10501 2017-10-26 Jim Wilson <wilson@tuliptree.org>
10503 * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10505 2017-10-26 Tom de Vries <tom@codesourcery.com>
10507 PR tree-optimization/82707
10508 * gimple.c (gimple_copy): Fix unsharing of
10509 GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10511 2017-10-26 Olga Makhotina <olga.makhotina@intel.com>
10513 * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10514 _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10515 _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10516 _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10517 _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10518 _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10519 _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10520 _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10521 _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10522 _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10523 _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10524 _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10525 _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10526 _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10527 _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10528 _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10529 _mm512_mask_cmpunord_ps_mask): New intrinsics.
10531 2017-10-26 Michael Meissner <meissner@linux.vnet.ibm.com>
10533 * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10535 * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10536 * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10537 warning to rs6000.c. Remove the Undocumented flag, since it has
10539 (-mabi=ibmlongdouble): Likewise.
10540 * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10541 already set, set the default format for long double.
10542 (rs6000_debug_reg_global): Print whether long double is IBM or
10544 (rs6000_option_override_internal): Rework setting long double
10545 format. Only warn if the user is changing the long double default
10546 and they did not use -Wno-psabi.
10547 * doc/invoke.texi (PowerPC options): Update the documentation for
10548 -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10550 2017-10-26 Richard Sandiford <richard.sandiford@linaro.org>
10551 Alan Hayward <alan.hayward@arm.com>
10552 David Sherwood <david.sherwood@arm.com>
10554 * rtl.h (wider_subreg_mode): New function.
10555 * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10556 rather than an unsigned int *.
10557 * ira-color.c (regno_max_ref_width): Replace with...
10558 (regno_max_ref_mode): ...this new variable.
10559 (coalesced_pseudo_reg_slot_compare): Update accordingly.
10560 Use wider_subreg_mode.
10561 (ira_sort_regnos_for_alter_reg): Likewise. Take a machine_mode *
10562 rather than an unsigned int *.
10563 * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10564 (process_alt_operands): Likewise.
10565 (invariant_p): Likewise.
10566 * lra-spills.c (assign_mem_slot): Likewise.
10567 (add_pseudo_to_slot): Likewise.
10568 * lra.c (collect_non_operand_hard_regs): Likewise.
10569 (add_regs_to_insn_regno_info): Likewise.
10570 * reload1.c (regno_max_ref_width): Replace with...
10571 (regno_max_ref_mode): ...this new variable.
10572 (reload): Update accordingly. Update call to
10573 ira_sort_regnos_for_alter_reg.
10574 (alter_reg): Update to use regno_max_ref_mode. Call wider_subreg_mode.
10575 (init_eliminable_invariants): Update to use regno_max_ref_mode.
10576 (scan_paradoxical_subregs): Likewise.
10578 2017-10-26 Wilco Dijkstra <wdijkstr@arm.com>
10580 * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10581 (aarch64_frame): Add emit_frame_chain boolean.
10582 * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10583 Move eh_return case to aarch64_layout_frame.
10584 (aarch64_layout_frame): Initialize emit_frame_chain.
10585 (aarch64_expand_prologue): Use emit_frame_chain.
10587 2017-10-26 Wilco Dijkstra <wdijkstr@arm.com>
10589 * config/aarch64/aarch64.c (aarch64_layout_frame):
10590 Ensure LR is always stored at the bottom of the callee-saves.
10591 Remove rarely used frame layout which saves callee-saves at top of
10592 frame, so the store of LR can be used as a valid probe in all cases.
10594 2017-10-26 Wilco Dijkstra <wdijkstr@arm.com>
10596 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10597 Improve unaligned TImode/TFmode base/offset split.
10599 2017-10-26 Richard Sandiford <richard.sandiford@linaro.org>
10600 Alan Hayward <alan.hayward@arm.com>
10601 David Sherwood <david.sherwood@arm.com>
10603 * caller-save.c (mark_referenced_regs): Use read_modify_subreg_p.
10604 * combine.c (find_single_use_1): Likewise.
10605 (expand_field_assignment): Likewise.
10606 (move_deaths): Likewise.
10607 * lra-constraints.c (simplify_operand_subreg): Likewise.
10608 (curr_insn_transform): Likewise.
10609 * lra.c (collect_non_operand_hard_regs): Likewise.
10610 (add_regs_to_insn_regno_info): Likewise.
10611 * rtlanal.c (reg_referenced_p): Likewise.
10612 (covers_regno_no_parallel_p): Likewise.
10614 2017-10-26 Richard Sandiford <richard.sandiford@linaro.org>
10616 * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10617 Don't print any bits outside the precision of the value.
10618 * wide-int.cc (test_printing): Add some new tests.
10620 2017-10-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10622 * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10623 supports -xbrace_comment option.
10624 * configure: Regenerate.
10625 * config.in: Regenerate.
10626 * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10627 (ASM_CPU_SPEC): Use it.
10629 2017-10-26 Richard Sandiford <richard.sandiford@linaro.org>
10631 * target.def (static_rtx_alignment): New hook.
10632 * targhooks.h (default_static_rtx_alignment): Declare.
10633 * targhooks.c (default_static_rtx_alignment): New function.
10634 * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10635 * doc/tm.texi: Regenerate.
10636 * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10637 instead of targetm.constant_alignment. Remove call to
10638 set_mem_attributes.
10639 * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10640 (cris_preferred_mininum_alignment): New function, split out from...
10641 (cris_constant_alignment): ...here.
10642 (cris_static_rtx_alignment): New function.
10643 * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10645 (ix86_constant_alignment): ...here.
10646 (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10647 * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10648 (mmix_static_rtx_alignment): New function.
10649 * config/spu/spu.c (spu_static_rtx_alignment): New function.
10650 (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10652 2017-10-26 Tamar Christina <tamar.christina@arm.com>
10655 * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10656 Add flag_trapping_math and flag_fp_int_builtin_inexact.
10658 2017-10-25 Palmer Dabbelt <palmer@dabbelt.com>
10660 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10661 * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10662 mark as a sign-extending load.
10663 (local_pic_load_u): Define.
10665 2017-10-25 Eric Botcazou <ebotcazou@adacore.com>
10667 PR middle-end/82062
10668 * fold-const.c (operand_equal_for_comparison_p): Also return true
10669 if ARG0 is a simple variant of ARG1 with narrower precision.
10670 (fold_ternary_loc): Always pass unstripped operands to the predicate.
10672 2017-10-25 Jan Hubicka <hubicka@ucw.cz>
10674 * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10676 * i386.h (processor_costs): Add gather_static, gather_per_elt,
10677 scatter_static, scatter_per_elt.
10678 * x86-tune-costs.h: Add new cost entries.
10680 2017-10-25 Richard Biener <rguenther@suse.de>
10682 * tree-ssa-sccvn.h (vn_eliminate): Declare.
10683 * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10684 class pass_fre): Move to ...
10685 * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10686 class pass_fre): ... here and adjust for statistics.
10688 2017-10-25 Jakub Jelinek <jakub@redhat.com>
10691 * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10692 for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10693 (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10694 * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10697 2017-10-25 Richard Biener <rguenther@suse.de>
10699 * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10700 el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10701 eliminate_push_avail, eliminate_insert): Move inside...
10702 (class eliminate_dom_walker): ... this class in preparation
10704 (fini_eliminate): Remove by merging with ...
10705 (eliminate): ... this function. Adjust for class changes.
10706 (pass_pre::execute): Remove fini_eliminate call.
10707 (pass_fre::execute): Likewise.
10709 2017-10-24 Jakub Jelinek <jakub@redhat.com>
10712 * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10713 (VPERMI2, VPERMI2I): New mode iterators.
10714 (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10715 (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10717 (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10718 mode iterator. Remove 3 old define_insn patterns.
10719 (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10720 (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10721 VPERMI2 mode iterator, remove the other two expanders.
10722 (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10723 to use VPERMI2 mode iterator, add another alternative for vpermi2*
10724 instructions, remove the other two patterns.
10725 (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10726 mode iterator, remove the other two patterns.
10727 * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10728 (ix86_expand_vec_perm_vpermt2): ... this. Swap mask and op0
10729 arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10730 and adjust argument order accordingly.
10731 (ix86_expand_vec_perm): Adjust caller.
10732 (expand_vec_perm_1): Likewise.
10733 (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10734 (expand_vec_perm_vpermt2_vpshub2): ... this.
10735 (ix86_expand_vec_perm_const_1): Adjust caller.
10736 (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10739 * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10740 (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10741 (vec_shl_<mode>): Remove unused expander.
10742 (avx512bw_<shift_insn><mode>3): New define_insn.
10743 (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10744 (<sse2_avx2>_<shift_insn><mode>3): ... this. New define_insn.
10746 2017-10-24 Paolo Carlini <paolo.carlini@oracle.com>
10749 * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10752 2017-10-24 Wilco Dijkstra <wdijkstr@arm.com>
10754 PR rtl-optimization/82396
10755 * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10756 (autopref_multipass_init): Simplify initialization.
10757 (autopref_rank_data): Simplify sort order.
10758 * gcc/sched-int.h (autopref_multipass_data_): Remove
10759 multi_mem_insn_p, min_offset and max_offset.
10761 2017-10-24 Wilco Dijkstra <wdijkstr@arm.com>
10763 PR middle-end/60580
10764 * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10765 Check special value of flag_omit_frame_pointer.
10766 (aarch64_can_eliminate): Likewise.
10767 (aarch64_override_options_after_change_1): Simplify handling of
10768 -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10770 2017-10-24 Richard Biener <rguenther@suse.de>
10772 PR tree-optimization/82697
10773 * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10774 zero for conditional load and unconditional store.
10776 2017-10-24 H.J. Lu <hongjiu.lu@intel.com>
10778 * doc/install.texi: Document bootstrap-cet.
10780 2017-10-24 H.J. Lu <hongjiu.lu@intel.com>
10783 * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10784 ENDBR instruction at function entrance if function is only
10787 2017-10-24 Jakub Jelinek <jakub@redhat.com>
10790 * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10791 patterns to better describe from which operation the CF is computed.
10792 (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10793 * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10794 one LTU with [DT]Imode and another one with [SD]Imode. If arg0
10795 is 0, use _0 suffixed expanders instead of emitting a comparison
10798 2017-10-06 Sergey Shalnov <Sergey.Shalnov@intel.com>
10800 * config/i386/i386.md(*movsf_internal, *movdf_internal):
10801 Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10803 2017-10-24 Eric Botcazou <ebotcazou@adacore.com>
10805 PR middle-end/82569
10806 * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10807 * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10808 * loop-iv.c (iv_get_reaching_def): Likewise.
10809 * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10810 variable is promoted and the partition contains undefined values.
10812 2017-10-23 Sandra Loosemore <sandra@codesourcery.com>
10814 * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10816 (nios2_address_cost): Define.
10817 (nios2_legitimize_address): Recognize (exp + constant) directly.
10818 (TARGET_ADDRESS_COST): Define.
10820 2017-10-23 Sandra Loosemore <sandra@codesourcery.com>
10822 * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10823 (nios2_symbolic_memory_operand_p): Declare.
10824 (nios2_split_large_constant): Declare.
10825 (nios2_split_symbolic_memory_operand): Declare.
10826 * config/nios2/nios2.c: Adjust includes.
10827 (nios2_symbolic_constant_allowed): New.
10828 (nios2_symbolic_constant_p): New.
10829 (nios2_plus_symbolic_constant_p): New.
10830 (nios2_valid_addr_expr_p): Recognize addresses involving
10831 symbolic constants.
10832 (nios2_legitimate_address_p): Likewise, also LO_SUM.
10833 (nios2_symbolic_memory_operand_p): New.
10834 (nios2_large_constant_p): New.
10835 (nios2_split_large_constant): New.
10836 (nios2_split_plus_large_constant): New.
10837 (nios2_split_symbolic_memory_operand): New.
10838 (nios2_legitimize_address): Code refactoring. Handle addresses
10839 involving symbolic constants.
10840 (nios2_emit_move_sequence): Likewise.
10841 (nios2_print_operand): Improve error output.
10842 (nios2_print_operand_address): Handle LO_SUM.
10843 (nios2_cdx_narrow_form_p): Likewise.
10844 * config/nios2/nios2.md (movqi_internal): Add splitter for memory
10845 operands involving symbolic constants.
10846 (movhi_internal, movsi_internal): Likewise.
10847 (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
10848 (extendhisi2, extendqi<mode>2): Likewise.
10850 2017-10-23 Sandra Loosemore <sandra@codesourcery.com>
10852 * tree-pass.h (PROP_rtl_split_insns): Define.
10853 * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
10855 2017-10-23 Sandra Loosemore <sandra@codesourcery.com>
10857 * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
10859 2017-10-23 Jakub Jelinek <jakub@redhat.com>
10862 * target.def (const_not_ok_for_debug_p): Default to
10863 default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
10864 * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
10865 * targhooks.c (default_const_not_ok_for_debug_p): New function.
10866 * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
10867 which targetm.const_not_ok_for_debug_p returned true.
10868 * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
10870 * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
10872 * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
10873 * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
10874 UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
10876 (ix86_const_not_ok_for_debug_p): New function.
10877 (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
10878 (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
10880 2017-10-23 David Malcolm <dmalcolm@redhat.com>
10883 * system.h: Conditionally include "unique-ptr.h" if
10884 INCLUDE_UNIQUE_PTR is defined.
10885 * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
10886 of defining INCLUDE_UNIQUE_PTR before including "system.h".
10888 2017-10-23 Sebastian Perta <sebastian.perta@renesas.com>
10890 * config/rl78/rl78.md: New define_expand "subdi3".
10892 2017-10-23 H.J. Lu <hongjiu.lu@intel.com>
10895 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
10896 DF_REF_INSN if DF_REF_INSN_INFO is false.
10898 2017-10-23 Jan Hubicka <hubicka@ucw.cz>
10900 * i386.c (dimode_scalar_chain::compute_convert_gain): Use
10901 xmm_move instead of sse_move.
10902 (sse_store_index): New function.
10903 (ix86_register_move_cost): Be more sensible about mismatch stall;
10904 model AVX moves correctly; make difference between sse->integer and
10906 (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
10907 moves; make difference between SSE and AVX.
10908 * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
10909 and zmm_move. Increase size of sse load and store tables;
10910 add unaligned load and store tables; add ssemmx_to_integer.
10911 * x86-tune-costs.h: Update all entries according to real
10912 move latencies from Agner Fog's manual and chip documentation.
10914 2017-10-23 Jakub Jelinek <jakub@redhat.com>
10917 * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
10918 * config/i386/constraints.md (Wf): New constraint.
10919 * config/i386/i386.md (UNSPEC_SBB): New unspec.
10920 (cmp<dwi>_doubleword): Removed.
10921 (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
10922 (sub<mode>3_carry_ccgz): Use unspec instead of compare.
10923 * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
10924 expand with cmp<dwi>_doubleword. For LTU and GEU use
10925 sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
10927 * common.opt (gcolumn-info): Enable by default.
10928 * doc/invoke.texi (gcolumn-info): Document new default.
10930 2017-10-23 Richard Biener <rguenther@suse.de>
10932 PR tree-optimization/82672
10933 * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
10934 Fold the stmt if we propagated into it.
10936 2017-10-23 Richard Biener <rguenther@suse.de>
10938 * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
10939 (bitmap_remove_expr_from_set): ... this. All callers call this
10940 for non-constant values.
10941 (bitmap_set_subtract): Rename to...
10942 (bitmap_set_subtract_expressions): ... this. Adjust and
10944 (bitmap_set_contains_value): Remove superfluous check.
10945 (bitmap_set_replace_value): Inline into single caller ...
10946 (bitmap_value_replace_in_set): ... here and simplify.
10947 (dependent_clean): Merge into ...
10948 (clean): ... this using an overload. Adjust.
10949 (prune_clobbered_mems): Adjust.
10950 (compute_antic_aux): Likewise.
10951 (compute_partial_antic_aux): Likewise.
10953 2017-10-23 Richard Biener <rguenther@suse.de>
10955 PR tree-optimization/82129
10957 2017-08-01 Richard Biener <rguenther@suse.de>
10959 PR tree-optimization/81181
10960 * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10961 (compute_antic): ... end of iteration here.
10963 2017-10-23 Richard Sandiford <richard.sandiford@linaro.org>
10965 * target.def (starting_frame_offset): New hook.
10966 * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
10967 (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
10968 * doc/tm.texi.in: Regenerate.
10969 * hooks.h (hook_hwi_void_0): Declare.
10970 * hooks.c (hook_hwi_void_0): New function.
10971 * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
10972 STARTING_FRAME_OFFSET.
10973 * builtins.c (expand_builtin_setjmp_receiver): Likewise.
10974 * reload1.c (reload): Likewise.
10975 * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
10976 instead of STARTING_FRAME_OFFSET.
10977 * function.c (try_fit_stack_local): Likewise.
10978 (assign_stack_local_1): Likewise
10979 (instantiate_virtual_regs): Likewise.
10980 * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
10981 * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
10982 * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
10983 * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
10984 * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
10985 * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
10986 * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
10987 * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
10988 * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
10989 * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
10990 * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
10991 * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
10992 * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
10993 * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
10994 * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
10995 * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
10996 * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
10997 * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
10998 * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
10999 * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11000 * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11001 * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11002 * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11003 * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11004 * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11005 * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11006 * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11007 * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11008 * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11009 * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11010 * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11011 * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11012 * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11013 * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11014 * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11015 * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11016 * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11017 * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11018 * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11019 * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11020 * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11021 * config/avr/avr.c (avr_starting_frame_offset): Make static and
11022 return a HOST_WIDE_INT.
11023 (avr_builtin_setjmp_frame_value): Use it instead of
11024 STARTING_FRAME_OFFSET.
11025 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11026 * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11027 * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11029 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11030 * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11031 * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11032 (TARGET_CONSTANT_ALIGNMENT): Redefine.
11033 * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11034 * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11035 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11036 * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11037 * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11038 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11039 * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11040 * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11042 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11043 * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11044 * config/mips/mips.c (mips_compute_frame_info): Refer to
11045 TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11046 (mips_starting_frame_offset): New function.
11047 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11048 * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11049 * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11050 * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11051 and return a HOST_WIDE_INT.
11052 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11053 (mmix_initial_elimination_offset): Refer to
11054 TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11055 * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11056 * config/pa/pa.c (pa_starting_frame_offset): New function.
11057 (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11058 (pa_expand_prologue): Likewise.
11059 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11060 * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11061 !FRAME_GROWS_DOWNWARD handling to...
11062 (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11063 * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11064 !FRAME_GROWS_DOWNWARD handling to...
11065 (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11066 * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11067 !FRAME_GROWS_DOWNWARD handling to...
11068 (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11069 * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11071 (rs6000_starting_frame_offset): New function.
11072 * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11073 !FRAME_GROWS_DOWNWARD handling to...
11074 (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11075 * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11076 !FRAME_GROWS_DOWNWARD handling to...
11077 (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11078 * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11079 !FRAME_GROWS_DOWNWARD handling to...
11080 (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11081 * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11082 (rs6000_starting_frame_offset): New function.
11083 * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11084 * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11085 * config/vax/vax.c (vax_starting_frame_offset): New function.
11086 (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11087 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11088 * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11089 * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11090 (TARGET_STARTING_FRAME_OFFSET): Redefine.
11091 * system.h (STARTING_FRAME_OFFSET): Poison.
11093 2017-10-23 Richard Sandiford <richard.sandiford@linaro.org>
11095 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11096 SCALAR_TYPE_MODE instead of TYPE_MODE.
11098 2017-10-23 Richard Sandiford <richard.sandiford@linaro.org>
11099 Alan Hayward <alan.hayward@arm.com>
11100 David Sherwood <david.sherwood@arm.com>
11102 * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11104 2017-10-23 Richard Sandiford <richard.sandiford@linaro.org>
11105 Alan Hayward <alan.hayward@arm.com>
11106 David Sherwood <david.sherwood@arm.com>
11108 * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11110 2017-10-23 Richard Biener <rguenther@suse.de>
11112 PR tree-optimization/82129
11113 * tree-ssa-pre.c (bitmap_set_and): Remove.
11114 (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11115 canonicalizing expressions in the set to those with lowest
11116 ID rather than taking that from the first edge.
11118 2017-10-23 Richard Sandiford <richard.sandiford@linaro.org>
11120 * combine.c (rtx_equal_for_field_assignment_p): Use
11121 byte_lowpart_offset.
11123 2017-10-22 Richard Sandiford <richard.sandiford@linaro.org>
11124 Alan Hayward <alan.hayward@arm.com>
11125 David Sherwood <david.sherwood@arm.com>
11127 * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11128 to a SUBREG_PROMOTED_VAR.
11130 2017-10-22 Richard Sandiford <richard.sandiford@linaro.org>
11131 Alan Hayward <alan.hayward@arm.com>
11132 David Sherwood <david.sherwood@arm.com>
11134 * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11135 (expand_debug_source_expr): Likewise.
11136 * combine.c (combine_simplify_rtx): Likewise.
11137 * cse.c (fold_rtx): Likewise.
11138 * optabs.c (expand_float): Likewise.
11139 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11140 (simplify_binary_operation_1): Likewise.
11142 2017-10-22 Richard Sandiford <richard.sandiford@linaro.org>
11143 Alan Hayward <alan.hayward@arm.com>
11144 David Sherwood <david.sherwood@arm.com>
11146 * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11147 (record_promoted_value): Likewise.
11148 * expr.c (expand_expr_real_2): Likewise.
11149 * ree.c (update_reg_equal_equiv_notes): Likewise.
11150 (combine_set_extension): Likewise.
11151 * rtlanal.c (low_bitmask_len): Likewise.
11152 * simplify-rtx.c (neg_const_int): Likewise.
11153 (simplify_binary_operation_1): Likewise.
11155 2017-10-22 Richard Sandiford <richard.sandiford@linaro.org>
11156 Alan Hayward <alan.hayward@arm.com>
11157 David Sherwood <david.sherwood@arm.com>
11159 * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11160 * regcprop.c (maybe_mode_change): Likewise.
11161 * reload1.c (alter_reg): Likewise.
11163 2017-10-22 Richard Sandiford <richard.sandiford@linaro.org>
11165 * inchash.h (inchash::hash::add_wide_int): New function.
11166 * lto-streamer-out.c (hash_tree): Use it.
11168 2017-10-22 Richard Sandiford <richard.sandiford@linaro.org>
11170 * inchash.h (inchash::hash::add_wide_int): Rename to...
11171 (inchash::hash::add_hwi): ...this.
11172 * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11173 (polymorphic_call_target_hasher::hash): Likewise.
11174 * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11175 (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11176 (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11177 * lto-streamer-out.c (hash_tree): Likewise.
11178 * optc-save-gen.awk: Likewise.
11179 * tree.c (add_expr): Likewise.
11181 2017-10-22 Uros Bizjak <ubizjak@gmail.com>
11184 * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11185 for ordered inequality comparisons even with TARGET_IEEE_FP.
11187 2017-10-22 Uros Bizjak <ubizjak@gmail.com>
11190 * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11191 * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11192 Expand with cmp<dwi>_doubleword.
11194 2017-10-21 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11196 * extend.texi: Add x86 specific to 'nocf_check' attribute.
11197 List CET intrinsics.
11198 * invoke.texi: Add -mcet, -mibt, -mshstk options. Add x86
11199 specific to -fcf-protection option.
11201 2017-10-21 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11203 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11204 (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11205 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11206 (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11207 (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11208 * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11209 (extra_objs): Add cet.o for Linux/x86 targets.
11210 (tmake_file): Add i386/t-cet for Linux/x86 targets.
11211 * config/i386/cet.c: New file.
11212 * config/i386/cetintrin.h: Likewise.
11213 * config/i386/t-cet: Likewise.
11214 * config/i386/cpuid.h (bit_SHSTK): New.
11215 (bit_IBT): Likewise.
11216 * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11217 pass IBT and SHSTK bits.
11218 * config/i386/i386-builtin-types.def
11219 (VOID_FTYPE_UNSIGNED_PVOID): New.
11220 (VOID_FTYPE_UINT64_PVOID): Likewise.
11221 * config/i386/i386-builtin.def: Add CET intrinsics.
11222 * config/i386/i386-c.c (ix86_target_macros_internal): Add
11223 OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11224 * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11225 * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11227 * config/i386/i386.c (rest_of_insert_endbranch): New.
11228 (pass_data_insert_endbranch): Likewise.
11229 (pass_insert_endbranch): Likewise.
11230 (make_pass_insert_endbranch): Likewise.
11231 (ix86_notrack_prefixed_insn_p): Likewise.
11232 (ix86_target_string): Add -mibt, -mshstk flags.
11233 (ix86_option_override_internal): Add flag_cf_protection
11235 (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11236 (ix86_print_operand): Add 'notrack' prefix output.
11237 (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11238 (ix86_expand_builtin): Expand CET intrinsics.
11239 (x86_output_mi_thunk): Add 'endbranch' instruction.
11240 * config/i386/i386.h (TARGET_IBT): New.
11241 (TARGET_IBT_P): Likewise.
11242 (TARGET_SHSTK): Likewise.
11243 (TARGET_SHSTK_P): Likewise.
11244 * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11245 UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11246 UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11247 (builtin_setjmp_setup): New pattern.
11248 (builtin_longjmp): Likewise.
11249 (rdssp<mode>): Likewise.
11250 (incssp<mode>): Likewise.
11251 (saveprevssp): Likewise.
11252 (rstorssp): Likewise.
11253 (wrss<mode>): Likewise.
11254 (wruss<mode>): Likewise.
11255 (setssbsy): Likewise.
11256 (clrssbsy): Likewise.
11257 (nop_endbr): Likewise.
11258 * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11260 * config/i386/immintrin.h: Include <cetintrin.h>.
11261 * config/i386/linux-common.h
11262 (file_end_indicate_exec_stack_and_cet): New prototype.
11263 (TARGET_ASM_FILE_END): New.
11265 2017-10-20 Jan Hubicka <hubicka@ucw.cz>
11267 * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11268 latencies instead of having separate table; make difference between
11269 integer and float costs.
11270 * i386.h (processor_costs): Remove scalar_stmt_cost,
11271 scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11272 scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11274 * x86-tune-costs.h: Remove entries which has been removed in
11275 procesor_costs from all tables; make cond_taken_branch_cost
11276 and cond_not_taken_branch_cost COST_N_INSNS based.
11278 2017-10-20 Jan Hubicka <hubicka@ucw.cz>
11280 * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11282 2017-10-20 Jakub Jelinek <jakub@redhat.com>
11284 * config/i386/i386.md (isa): Remove fma_avx512f.
11285 * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11286 <avx512>_fmadd_<mode>_mask3<round_name>,
11287 <avx512>_fmsub_<mode>_mask<round_name>,
11288 <avx512>_fmsub_<mode>_mask3<round_name>,
11289 <avx512>_fnmadd_<mode>_mask<round_name>,
11290 <avx512>_fnmadd_<mode>_mask3<round_name>,
11291 <avx512>_fnmsub_<mode>_mask<round_name>,
11292 <avx512>_fnmsub_<mode>_mask3<round_name>,
11293 <avx512>_fmaddsub_<mode>_mask<round_name>,
11294 <avx512>_fmaddsub_<mode>_mask3<round_name>,
11295 <avx512>_fmsubadd_<mode>_mask<round_name>,
11296 <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11297 (*vec_widen_umult_even_v16si<mask_name>,
11298 *vec_widen_smult_even_v16si<mask_name>): Likewise.
11299 (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11301 2017-10-20 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11303 * extend.texi: Add 'nocf_check' documentation.
11304 * gimple.texi: Add second parameter to
11305 gimple_build_call_from_tree.
11306 * invoke.texi: Add -fcf-protection documentation.
11307 * rtl.texi: Add REG_CALL_NOTRACK documenation.
11309 2017-10-20 Richard Biener <rguenther@suse.de>
11311 PR tree-optimization/82473
11312 * tree-vect-loop.c (vectorizable_reduction): Properly get at
11313 the largest input type.
11315 2017-10-20 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11317 * c-attribs.c (handle_nocf_check_attribute): New function.
11318 (c_common_attribute_table): Add 'nocf_check' handling.
11319 * gimple-parser.c: Add second argument NULL to
11320 gimple_build_call_from_tree.
11321 * attrib.c (comp_type_attributes): Check nocf_check attribute.
11322 * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11324 * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11325 * common.opt: Add fcf-protection flag.
11326 * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11327 * flag-types.h: Add enum cf_protection_level.
11328 * gimple.c (gimple_build_call_from_tree): Add second parameter.
11329 Add 'nocf_check' attribute propagation to gimple call.
11330 * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11331 (gimple_build_call_from_tree): Update prototype.
11332 (gimple_call_nocf_check_p): New function.
11333 (gimple_call_set_nocf_check): Likewise.
11334 * gimplify.c: Add second argument to gimple_build_call_from_tree.
11335 * ipa-icf.c: Add nocf_check attribute in statement hash.
11336 * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11337 * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11338 * toplev.c (process_options): Add flag_cf_protection handling.
11340 2017-10-19 Jan Hubicka <hubicka@ucw.cz>
11342 * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11344 2017-10-20 Richard Biener <rguenther@suse.de>
11346 PR tree-optimization/82603
11347 * tree-if-conv.c (predicate_mem_writes): Make sure to only
11348 remove false predicated stores.
11350 2017-10-20 Richard Biener <rguenther@suse.de>
11352 * graphite-isl-ast-to-gimple.c
11353 (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11354 Remove return value and simplify, dump copied stmt after lhs
11356 (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11357 Reduce dump verbosity.
11358 (gsi_insert_earliest): Likewise.
11359 (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11360 * graphite.c (print_global_statistics): Adjust dumping.
11361 (print_graphite_scop_statistics): Likewise.
11362 (print_graphite_statistics): Do not dump loops here.
11363 (graphite_transform_loops): But here.
11365 2017-10-20 Nicolas Roche <roche@adacore.com>
11367 * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11368 * configure: Regenerate.
11370 2017-10-20 Jakub Jelinek <jakub@redhat.com>
11373 * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11374 functions when optimizing replace GIMPLE_RETURN stmts with
11375 calls to __builtin_unreachable ().
11378 * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11379 for -fsanitize=thread link of executables.
11380 (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11381 link of executables.
11384 * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11385 New mode iterators.
11386 (<shift_insn><mode>3<mask_name>): Change the last of the 3
11387 define_insns for logical vector shifts to use VI248_AVX512BW
11388 iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11389 condition, useless isa and prefix attributes. Change the first
11390 2 of these define_insns to ...
11391 (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11392 define_insn for avx512vl.
11393 (<shift_insn><mode>3): ... and this, new define_insn without
11394 masking for non-avx512vl.
11397 * config/i386/sse.md (*andnot<mode>3,
11398 <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11399 (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11400 and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11401 not applied use empty suffix even for TARGET_AVX512VL.
11402 * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11403 is applied, supply evex,evex or evex,evex,evex instead of just
11406 2017-10-20 Julia Koval <julia.koval@intel.com>
11408 * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11409 (OPTION_MASK_ISA_GFNI_UNSET): New.
11410 (ix86_handle_option): Handle OPT_mgfni.
11411 * config/i386/cpuid.h (bit_GFNI): New.
11412 * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11413 * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11414 * config/i386/i386.c (ix86_target_string): Add -mgfni.
11415 (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11416 * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11417 * config/i386/i386.opt: Add mgfni.
11419 2017-10-20 Orlando Arias <oarias@knights.ucf.edu>
11421 * config/msp430/msp430.c (msp430_option_override): Disable
11422 -fdelete-null-pointer-checks.
11423 * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11425 2017-10-19 Jan Hubicka <hubicka@ucw.cz>
11427 * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11428 of x87 and SSE instructions.
11430 2017-10-19 Jan Hubicka <hubicka@ucw.cz>
11432 * asan.c (create_cond_insert_point): Do not update edge count.
11433 * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11434 (afdo_propagate_circuit): Likewise.
11435 (afdo_calculate_branch_prob): Likewise.
11436 (afdo_annotate_cfg): Likewise.
11437 * basic-block.h (struct edge_def): Remove count.
11438 (edge_def::count): New accessor.
11439 * bb-reorder.c (rotate_loop): Update.
11440 (find_traces_1_round): Update.
11441 (connect_traces): Update.
11442 (sanitize_hot_paths): Update.
11443 * cfg.c (unchecked_make_edge): Update.
11444 (make_single_succ_edge): Update.
11445 (check_bb_profile): Update.
11446 (dump_edge_info): Update.
11447 (update_bb_profile_for_threading): Update.
11448 (scale_bbs_frequencies_int): Update.
11449 (scale_bbs_frequencies_gcov_type): Update.
11450 (scale_bbs_frequencies_profile_count): Update.
11451 (scale_bbs_frequencies): Update.
11452 * cfganal.c (connect_infinite_loops_to_exit): Update.
11453 * cfgbuild.c (compute_outgoing_frequencies): Update.
11454 (find_many_sub_basic_blocks): Update.
11455 * cfgcleanup.c (try_forward_edges): Update.
11456 (try_crossjump_to_edge): Update
11457 * cfgexpand.c (expand_gimple_cond): Update
11458 (expand_gimple_tailcall): Update
11459 (construct_exit_block): Update
11460 * cfghooks.c (verify_flow_info): Update
11461 (redirect_edge_succ_nodup): Update
11462 (split_edge): Update
11463 (make_forwarder_block): Update
11464 (duplicate_block): Update
11465 (account_profile_record): Update
11466 * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11467 * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11468 * cfgloopmanip.c (scale_loop_profile): Update.
11470 (lv_adjust_loop_entry_edge): Update.
11471 * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11472 (force_nonfallthru_and_redirect): Update.
11473 (purge_dead_edges): Update.
11474 (rtl_flow_call_edges_add): Update.
11475 * cgraphunit.c (init_lowered_empty_function): Update.
11476 (cgraph_node::expand_thunk): Update.
11477 * gimple-pretty-print.c (dump_probability): Update.
11478 (dump_edge_probability): Update.
11479 * gimple-ssa-isolate-paths.c (isolate_path): Update.
11480 * haifa-sched.c (sched_create_recovery_edges): Update.
11481 * hsa-gen.c (convert_switch_statements): Update.
11482 * ifcvt.c (dead_or_predicable): Update.
11483 * ipa-inline-transform.c (inline_transform): Update.
11484 * ipa-split.c (split_function): Update.
11485 * ipa-utils.c (ipa_merge_profiles): Update.
11486 * loop-doloop.c (add_test): Update.
11487 * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11488 * lto-streamer-in.c (input_cfg): Update.
11489 (input_function): Update.
11490 * lto-streamer-out.c (output_cfg): Update.
11491 * modulo-sched.c (sms_schedule): Update.
11492 * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11493 * predict.c (maybe_hot_edge_p): Update.
11494 (unlikely_executed_edge_p): Update.
11495 (probably_never_executed_edge_p): Update.
11496 (dump_prediction): Update.
11497 (drop_profile): Update.
11498 (propagate_unlikely_bbs_forward): Update.
11499 (determine_unlikely_bbs): Update.
11500 (force_edge_cold): Update.
11501 * profile.c (compute_branch_probabilities): Update.
11502 * reg-stack.c (better_edge): Update.
11503 * shrink-wrap.c (handle_simple_exit): Update.
11504 * tracer.c (better_p): Update.
11505 * trans-mem.c (expand_transaction): Update.
11506 (split_bb_make_tm_edge): Update.
11507 * tree-call-cdce.c: Update.
11508 * tree-cfg.c (gimple_find_sub_bbs): Update.
11509 (gimple_split_edge): Update.
11510 (gimple_duplicate_sese_region): Update.
11511 (gimple_duplicate_sese_tail): Update.
11512 (gimple_flow_call_edges_add): Update.
11513 (insert_cond_bb): Update.
11514 (execute_fixup_cfg): Update.
11515 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11516 * tree-complex.c (expand_complex_div_wide): Update.
11517 * tree-eh.c (lower_resx): Update.
11518 (unsplit_eh): Update.
11519 (cleanup_empty_eh_move_lp): Update.
11520 * tree-inline.c (copy_edges_for_bb): Update.
11521 (freqs_to_counts): Update.
11522 (copy_cfg_body): Update.
11523 * tree-ssa-dce.c (remove_dead_stmt): Update.
11524 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11525 * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11526 * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11527 (unloop_loops): Update.
11528 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11529 * tree-ssa-loop-split.c (connect_loops): Update.
11530 (split_loop): Update.
11531 * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11532 * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11533 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11534 * tree-ssa-reassoc.c (branch_fixup): Update.
11535 * tree-ssa-tail-merge.c (replace_block_by): Update.
11536 * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11537 (compute_path_counts): Update.
11538 (update_profile): Update.
11539 (recompute_probabilities): Update.
11540 (update_joiner_offpath_counts): Update.
11541 (estimated_freqs_path): Update.
11542 (freqs_to_counts_path): Update.
11543 (clear_counts_path): Update.
11544 (ssa_fix_duplicate_block_edges): Update.
11545 (duplicate_thread_path): Update.
11546 * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11547 (case_bit_test_cmp): Update.
11548 (collect_switch_conv_info): Update.
11549 (gen_inbound_check): Update.
11550 (do_jump_if_equal): Update.
11551 (emit_cmp_and_jump_insns): Update.
11552 * tree-tailcall.c (decrease_profile): Update.
11553 (eliminate_tail_call): Update.
11554 * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11555 (vect_do_peeling): Update.
11556 * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11557 * ubsan.c (ubsan_expand_null_ifn): Update.
11558 (ubsan_expand_ptr_ifn): Update.
11559 * value-prof.c (gimple_divmod_fixed_value): Update.
11560 (gimple_mod_pow2): Update.
11561 (gimple_mod_subtract): Update.
11562 (gimple_ic): Update.
11563 (gimple_stringop_fixed_value): Update.
11565 2017-10-19 Uros Bizjak <ubizjak@gmail.com>
11568 * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11570 2017-10-19 Alexander Monakov <amonakov@ispras.ru>
11572 PR rtl-optimization/82395
11573 * ira-color.c (allocno_priority_compare_func): Fix comparison step
11574 based on non_spilled_static_chain_regno_p.
11576 2017-10-19 Uros Bizjak <ubizjak@gmail.com>
11578 * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11579 (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11580 (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11582 2017-10-19 Martin Sebor <msebor@redhat.com>
11584 PR tree-optimization/82596
11585 * tree.c (array_at_struct_end_p): Handle STRING_CST.
11587 2017-10-19 Eric Botcazou <ebotcazou@adacore.com>
11589 * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11590 (get_mem_refs_of_builtin_call): Likewise.
11591 * builtins.c (expand_builtin_apply): Adjust call to
11592 allocate_dynamic_stack_space.
11593 (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11594 the third argument to allocate_dynamic_stack_space, otherwise -1.
11595 (expand_builtin): Deal with all alloca variants.
11596 (is_inexpensive_builtin): Likewise.
11597 * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11598 * calls.c (special_function_p): Deal with all alloca variants.
11599 (initialize_argument_information): Adjust call to
11600 allocate_dynamic_stack_space.
11601 (expand_call): Likewise.
11602 * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11603 * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11604 * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11605 use it for the stack usage computation.
11606 * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11607 * function.c (gimplify_parameters): Call build_alloca_call_expr.
11608 * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11609 Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11610 (in_loop_p): Remove first argument and useless check.
11611 (pass_walloca::execute): Remove useless test and adjust call to above.
11612 * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11613 * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11614 (gimplify_call_expr): Deal with all alloca variants.
11615 * hsa-gen.c (gen_hsa_alloca): Likewise.
11616 (gen_hsa_insns_for_call): Likewise.
11617 * ipa-pure-const.c (special_builtin_state): Likewise.
11618 * tree-chkp.c (chkp_build_returned_bound): Likewise.
11619 * tree-object-size.c (alloc_object_size): Likewise.
11620 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11621 (call_may_clobber_ref_p_1): Likewise.
11622 * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11623 (ccp_fold_stmt): Likewise.
11624 (optimize_stack_restore): Likewise.
11625 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11626 (mark_all_reaching_defs_necessary_1): Likewise.
11627 (propagate_necessity): Likewise.
11628 (eliminate_unnecessary_stmts): Likewise.
11629 * tree.c (build_common_builtin_nodes): Build
11630 BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11631 (build_alloca_call_expr): New function.
11632 * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11633 (CASE_BUILT_IN_ALLOCA): Likewise.
11634 (build_alloca_call_expr): Declare.
11635 * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11637 2017-10-19 Eric Botcazou <ebotcazou@adacore.com>
11640 * dwarf2out.c (new_die_raw): New static inline function.
11641 (new_die): Use it to create the DIE.
11642 (add_AT_external_die_ref): Likewise.
11643 (clone_die): Likewise.
11644 (clone_as_declaration): Likewise.
11645 (dwarf2out_vms_debug_main_pointer): Likewise.
11646 (base_type_die): Likewise. Remove early return for corner cases.
11647 Do not call add_pubtype on the DIE here.
11648 (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11649 (modified_type_die): Adjust the lookup for reverse order DIEs. Skip
11650 typedefs for base types with DW_AT_endianity. Make sure a DIE with
11651 native order exists for base types, attach the DIE manually and call
11652 add_pubtype on it. Do not equate a reverse order DIE to the type.
11654 2017-10-19 Richard Earnshaw <rearnsha@arm.com>
11656 * config/arm/arm.c (align_ok_ldrd_strd): New function.
11657 (mem_ok_for_ldrd_strd): New parameter align. Extract the alignment of
11659 (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11661 2017-10-19 Jakub Jelinek <jakub@redhat.com>
11663 * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN. Or
11664 SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11665 * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11666 BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11667 * opts.c (sanitizer_opts): Add builtin.
11668 * ubsan.c (instrument_builtin): New function.
11669 (pass_ubsan::execute): Call it.
11670 (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11671 * doc/invoke.texi: Document -fsanitize=builtin.
11673 * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11674 builtins, store max (log2 (align), 0) into uchar field instead of
11675 align into uptr field.
11676 (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11677 store uchar 0 field instead of uptr 0 field.
11678 (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11679 instead of passing one address of struct with 2 locations pass
11680 two addresses of structs with 1 location each.
11681 * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11682 BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11683 BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11684 BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11685 (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11686 BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11687 BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11688 BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11690 2017-10-19 Martin Liska <mliska@suse.cz>
11693 * file-find.c (remove_prefix): Remove.
11694 * file-find.h (remove_prefix): Likewise.
11695 * gcc-ar.c: Remove smartness of lookup.
11697 2017-10-19 Segher Boessenkool <segher@kernel.crashing.org>
11699 * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11700 *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11701 *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11703 2017-10-19 Jakub Jelinek <jakub@redhat.com>
11706 * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11707 (setcc + and to xor + setcc): New peephole2.
11709 2017-10-19 Tom de Vries <tom@codesourcery.com>
11711 * doc/sourcebuild.texi (Test Directives, Variants of
11712 dg-require-support): Add dg-require-stack-size.
11714 2017-10-19 Martin Liska <mliska@suse.cz>
11717 * gimplify.c (gimplify_decl_expr): Do not instrument variables
11718 that have a large alignment.
11719 (gimplify_target_expr): Likewise.
11721 2017-10-18 Segher Boessenkool <segher@kernel.crashing.org>
11723 PR rtl-optimization/82602
11724 * ira.c (rtx_moveable_p): Return false for volatile asm.
11726 2017-10-18 Uros Bizjak <ubizjak@gmail.com>
11729 * config/i386/i386-modes.def (CCGZ): New CC mode.
11730 * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11731 * config/i386/predicates.md (ix86_comparison_operator):
11733 * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11734 Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11735 with double-word subtraction.
11736 (put_condition_code): Handle CCGZmode.
11738 2017-10-18 Aldy Hernandez <aldyh@redhat.com>
11740 * wide-int.cc (debug (const wide_int &)): New.
11741 (debug (const wide_int *)): New.
11742 (debug (const widest_int &)): New.
11743 (debug (const widest_int *)): New.
11745 2017-10-18 Vladimir Makarov <vmakarov@redhat.com>
11747 PR middle-end/82556
11748 * lra-constraints.c (curr_insn_transform): Use non-input operand
11749 instead of output one for matched reload.
11751 2017-10-18 Bin Cheng <bin.cheng@arm.com>
11753 * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11754 (tree-ssa-loop-ivopts.h): New header file.
11755 (struct builtin_info): New fields.
11756 (classify_builtin_1): Compute and record base and offset parts for
11757 memset builtin partition by calling strip_offset.
11758 (offset_cmp, fuse_memset_builtins): New functions.
11759 (finalize_partitions): Fuse adjacent memset partitions by calling
11761 * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11762 Expose the interface.
11763 * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11765 2017-10-18 Bin Cheng <bin.cheng@arm.com>
11767 PR tree-optimization/82574
11768 * tree-loop-distribution.c (find_single_drs): New parameter. Check
11769 that data reference must be executed exactly once per iteration
11770 against the outermost loop in nest.
11771 (classify_partition): Update call to above function.
11773 2017-10-18 Richard Biener <rguenther@suse.de>
11775 PR tree-optimization/82591
11776 * graphite.c (graphite_transform_loops): Move code gen message
11778 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11779 Here. Handle scop_to_isl_ast failing.
11780 (scop_to_isl_ast): Limit the number of ISL operations.
11782 2017-10-18 Richard Biener <rguenther@suse.de>
11784 * graphite-isl-ast-to-gimple.c
11785 (translate_isl_ast_to_gimple::set_rename): Simplify.
11786 (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11787 (graphite_copy_stmts_from_block): ... here.
11788 (copy_bb_and_scalar_dependences): Simplify.
11789 (add_parameters_to_ivs_params): Canonicalize.
11790 (generate_entry_out_of_ssa_copies): Simplify.
11791 * graphite-sese-to-poly.c (extract_affine_name): Simplify
11792 by passing in ISL dimension.
11793 (parameter_index_in_region_1): Rename to ...
11794 (parameter_index_in_region): ... this.
11795 (extract_affine): Adjust assert, pass down parameter index.
11796 (add_param_constraints): Use range-info when available.
11797 (build_scop_context): Adjust.
11798 * sese.c (new_sese_info): Adjust.
11799 (free_sese_info): Likewise.
11800 * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11801 Remove unused typedefs.
11802 (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11804 2017-10-18 Martin Liska <mliska@suse.cz>
11806 * combine.c (simplify_compare_const): Add gcc_fallthrough.
11808 2017-10-18 Robin Dapp <rdapp@linux.vnet.ibm.com>
11810 * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11811 (s390_sched_init): Do not reset s390_sched_state if we entered the
11812 current basic block via a fallthru edge and all others are unlikely.
11814 2017-10-18 Robin Dapp <rdapp@linux.vnet.ibm.com>
11816 * config/s390/s390.c (NUM_SIDES): New variable.
11817 (LONGRUNNING_THRESHOLD): New variable.
11818 (LATENCY_FACTOR): New variable.
11819 (s390_sched_score): Decrease score for long-running instructions on
11821 (s390_sched_variable_issue): Perform bookkeeping for long-running
11824 2017-10-18 Richard Biener <rguenther@suse.de>
11826 * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11827 Simplify with removal of the parameter rename map.
11828 (set_rename): Likewise.
11829 (should_copy_to_new_region): Likewise.
11830 (graphite_copy_stmts_from_block): Likewise.
11831 (copy_bb_and_scalar_dependences): Remove initialization of
11832 unused copied_bb_map.
11833 (copy_def): Remove.
11834 (copy_internal_parameters): Likewise.
11835 (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11836 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11837 Use INTEGRAL_TYPE_P.
11838 (parameter_index_in_region_1): Rename to ...
11839 (assign_parameter_index_in_region): ... this. Assert we have
11840 a parameter we handle.
11841 (scan_tree_for_params): Adjust.
11842 * sese.h (parameter_rename_map_t): Remove.
11843 (struct sese_info_t): Remove unused parameter_rename_map and
11844 copied_bb_map members.
11845 * sese.c (new_sese_info): Adjust.
11846 (free_sese_info): Likewise.
11848 2017-10-18 Martin Liska <mliska@suse.cz>
11851 * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
11852 on an abnormal edge.
11854 2017-10-18 Sebastian Huber <sebastian.huber@embedded-brains.de>
11856 * doc/invoke.texi (ffunction-sections and fdata-sections):
11859 2017-10-17 Eric Botcazou <ebotcazou@adacore.com>
11861 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
11862 the use statement can throw internally.
11864 2017-10-17 Eric Botcazou <ebotcazou@adacore.com>
11866 * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
11867 any RTX present on the RHS of a SET.
11868 * compare-elim.c (try_eliminate_compare): Restore comment.
11870 2017-10-17 Jakub Jelinek <jakub@redhat.com>
11872 * langhooks.h (struct lang_hooks): Document that tree_size langhook
11873 may be also called on tcc_type nodes.
11874 * langhooks.c (lhd_tree_size): Likewise.
11876 2017-10-17 David Malcolm <dmalcolm@redhat.com>
11878 * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
11879 format_warning_at_substring.
11880 (maybe_warn): Convert source_range * param to a location_t. Pass
11881 UNKNOWN_LOCATION rather than NULL to fmtwarn.
11882 (format_directive): Remove code to extract source_ranges and
11883 source_range * in favor of just a location_t.
11884 (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
11886 * substring-locations.c (format_warning_va): Convert
11887 source_range * param to a location_t.
11888 (format_warning_at_substring): Likewise.
11889 * substring-locations.h (format_warning_va): Likewise.
11890 (format_warning_at_substring): Likewise.
11892 2017-10-17 Jan Hubicka <hubicka@ucw.cz>
11894 * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
11896 * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
11897 and scatter/gather ops.
11899 * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
11900 vec_gather_load and vec_scatter_store.
11901 * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
11902 * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
11903 * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
11904 * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
11905 * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
11906 * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
11908 2017-10-17 Uros Bizjak <ubizjak@gmail.com>
11910 * reg-stack.c (compare_for_stack_reg): Add bool argument.
11911 Detect FTST instruction and handle its register pops. Only pop
11912 second operand if can_pop_second_op is true.
11913 (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
11914 set can_pop_second_op to false in the compare_for_stack_reg call.
11916 * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
11917 output_fp_compare for stack register operands.
11918 * config/i386/i386.c (output_fp_compare): Do not output SSE compare
11919 instructions here. Do not emit stack register pops here. Assert
11920 that FCOMPP pops next to top stack register. Rewrite function.
11922 2017-10-17 Nathan Sidwell <nathan@acm.org>
11924 PR middle-end/82577
11925 * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
11926 use DECL_ASSEMBLER_NAME_RAW.
11928 PR middle-end/82546
11929 * tree.c (tree_code_size): Reformat. Punt to lang hook for unknown
11932 2017-10-17 Qing Zhao <qing.zhao@oracle.com>
11933 Wilco Dijkstra <wilco.dijkstra@arm.com>
11935 * builtins.c (expand_builtin_update_setjmp_buf): Add a
11936 converstion to Pmode from the buf_addr.
11938 2017-10-17 Richard Biener <rguenther@suse.de>
11940 * graphite-dependences.c (scop_get_reads_and_writes): Change
11941 output parameters to references.
11943 2017-10-17 Jackson Woodruff <jackson.woodruff@arm.com>
11945 PR 71026/tree-optimization
11946 * fold-const.c (distribute_real_division): Removed.
11947 (fold_binary_loc): Remove calls to distribute_real_divison.
11949 2017-10-17 Richard Biener <rguenther@suse.de>
11951 * graphite-scop-detection.c
11952 (scop_detection::stmt_has_simple_data_refs_p): Always use
11953 the full nest as region.
11954 (try_generate_gimple_bb): Likewise.
11955 * sese.c (scalar_evolution_in_region): Simplify now that
11956 SCEV can handle instantiation in regions.
11957 * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
11958 in the non-loop part of a function if requested.
11960 2017-10-17 Richard Biener <rguenther@suse.de>
11962 PR tree-optimization/82563
11963 * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
11965 (graphite_regenerate_ast_isl): Call it.
11966 * graphite-scop-detection.c (build_scops): Remove entry edge split.
11968 2017-10-17 Jakub Jelinek <jakub@redhat.com>
11970 PR tree-optimization/82549
11971 * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
11972 Formatting fixes. Instead of calling make_bit_field_ref with negative
11975 2017-10-17 Olga Makhotina <olga.makhotina@intel.com>
11977 * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
11978 _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
11979 _mm_maskz_reduce_ss): New.
11980 * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
11981 __builtin_ia32_reducess_mask): Ditto..
11982 (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
11983 * config/i386/sse.md (reduces<mode>): Renamed to ...
11984 (reduces<mode><mask_scalar_name>): ... this.
11985 (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
11987 (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
11988 %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
11990 2017-10-16 David Malcolm <dmalcolm@redhat.com>
11992 * Makefile.in (OBJS): Add unique-ptr-tests.o.
11993 * selftest-run-tests.c (selftest::run_tests): Call
11994 selftest::unique_ptr_tests_cc_tests.
11995 * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
11996 * unique-ptr-tests.cc: New file.
11998 2017-10-16 Vladimir Makarov <vmakarov@redhat.com>
12001 * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12003 * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12004 (make_hard_regno_born, make_hard_regno_dead): Update
12005 bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12007 2017-10-16 Jeff Law <law@redhat.com>
12009 * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12011 2017-10-16 Jan Hubicka <hubicka@ucw.cz>
12013 * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12015 2017-10-16 Olivier Hainque <hainque@adacore.com>
12017 * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12018 with_cpu if we were configured for an e500v2 target cpu name.
12020 2017-10-16 Thomas Preud'homme <thomas.preudhomme@arm.com>
12022 * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12023 * doc/invoke.texi: Document +nodsp as a valid extension for
12026 2017-10-16 Martin Liska <mliska@suse.cz>
12028 * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12029 (test_set_range): Likewise.
12030 (test_range_functions): Rename to ...
12031 (test_bit_in_range): ... this.
12032 (sbitmap_c_tests): Add new test.
12034 2017-10-16 Tamar Christina <tamar.christina@arm.com>
12036 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12038 (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12039 (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12041 2017-10-16 Tamar Christina <tamar.christina@arm.com>
12043 * config/aarch64/aarch64-builtins.c
12044 (aarch64_types_quadopu_lane_qualifiers): New.
12045 (TYPES_QUADOPU_LANE): New.
12046 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12047 (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12048 (aarch64_<sur>dot_laneq<vsi2qi>): New.
12049 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12050 (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12051 * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12052 (Vdottype, DOTPROD): New.
12053 (sur): Add SDOT and UDOT.
12055 2017-10-16 Tamar Christina <tamar.christina@arm.com>
12057 * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12058 (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12059 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12060 Add TARGET_DOTPROD.
12061 * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12062 * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12063 Enable TARGET_DOTPROD.
12064 (cortex-a75.cortex-a55): Likewise.
12065 * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12067 2017-10-16 Tamar Christina <tamar.christina@arm.com>
12069 * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12070 (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12072 * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12074 * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12075 (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12076 * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12077 (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12078 * config/arm/types.md (neon_dot, neon_dot_q): New.
12079 * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12081 2017-10-16 Tamar Christina <tamar.christina@arm.com>
12083 * config/arm/arm.h (TARGET_DOTPROD): New.
12084 * config/arm/arm.c (arm_arch_dotprod): New.
12085 (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12086 * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12087 * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12088 (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12089 (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12090 * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12091 * doc/invoke.texi (armv8.2-a): Document dotprod
12093 2017-10-14 Jan Hubicka <hubicka@ucw.cz>
12095 * i386.c (ix86_vec_cost): New function.
12096 (ix86_rtx_costs): Handle vector operations better.
12097 * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12098 * x86-tune-costs.h: Add new costs to all tables.
12100 2017-10-14 Jan Hubicka <hubicka@ucw.cz>
12102 * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12104 * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12105 divsd, sqrtss and sqrtsd
12106 * x86-tune-costs.h: Add new entries to all costs.
12107 (znver1_cost): Fix to match real instruction latencies.
12109 2017-10-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12110 Michael Collison <michael.collison@arm.com>
12112 * compare-elim.c: Include emit-rtl.h.
12113 (can_merge_compare_into_arith): New function.
12114 (try_validate_parallel): Likewise.
12115 (try_merge_compare): Likewise.
12116 (try_eliminate_compare): Call the above when no previous clobber
12118 (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12121 2017-10-14 Jakub Jelinek <jakub@redhat.com>
12123 PR middle-end/62263
12124 PR middle-end/82498
12125 * tree-ssa-phiopt.c (value_replacement): Comment fix. Handle
12126 up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12128 PR middle-end/62263
12129 PR middle-end/82498
12130 * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12131 to be any operand_equal_p operands. For & (B - 1) require
12132 B to be power of 2. Recognize
12133 (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12135 2017-10-14 Uros Bizjak <ubizjak@gmail.com>
12138 * optabs.c (expand_memory_blockage): Fix call of
12139 targetm.have_memory_blockage.
12141 2017-10-14 Jakub Jelinek <jakub@redhat.com>
12144 * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12145 x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12146 objects to extra_objs instead of overwriting it.
12148 2017-10-14 Uros Bizjak <ubizjak@gmail.com>
12150 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12151 Use any_fp_register_operand as operand[3] predicate. Simplify
12152 equality test for operands[2] and operands[4] memory location.
12153 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12154 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12155 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12156 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12157 any_fp_register_operand as operand[1] predicate. Simplify
12158 equality test for operands[0] and operands[3] memory location.
12159 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12160 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12161 (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12163 2017-10-14 Uros Bizjak <ubizjak@gmail.com>
12165 * target-insns.def: Add memory_blockage.
12166 * optabs.c (expand_memory_blockage): New function.
12167 (expand_asm_memory_barrier): Rename ...
12168 (expand_asm_memory_blockage): ... to this.
12169 (expand_mem_thread_fence): Call expand_memory_blockage
12170 instead of expand_asm_memory_barrier.
12171 (expand_mem_singnal_fence): Ditto.
12172 (expand_atomic_load): Ditto.
12173 (expand_atomic_store): Ditto.
12174 * doc/md.texi (Standard Pattern Names For Generation):
12175 Document memory_blockage instruction pattern.
12177 2017-10-13 Sebastian Perta <sebastian.perta@renesas.com>
12179 * config/rl78/rl78.c (rl78_emit_libcall): New function.
12180 * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12181 * config/rl78/rl78.md: New define_expand "adddi3".
12183 2017-10-13 Jan Hubicka <hubicka@ucw.cz>
12185 * cfghooks.c (verify_flow_info): Disable check that all probabilities
12188 2017-10-13 Jeff Law <law@redhat.com>
12190 * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12192 2017-10-13 Jakub Jelinek <jakub@redhat.com>
12195 * internal-fn.c (expand_mul_overflow): If both operands have
12196 the same highpart of -1 or 0 and the topmost bit of lowpart
12197 is different, overflow is if res <= 0 rather than res < 0.
12199 2017-10-13 Pat Haugen <pthaugen@us.ibm.com>
12201 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12202 TARGET_P9_VECTOR code for unaligned_load case.
12204 2017-10-13 Jan Hubicka <hubicka@ucw.cz>
12206 * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12208 2017-10-13 Nathan Sidwell <nathan@acm.org>
12210 * tree-core.h (tree_contains_struct): Make bool.
12211 * tree.c (tree_contains_struct): Likewise.
12212 * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12213 (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12214 MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12215 MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12216 MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12218 2017-10-13 Richard Biener <rguenther@suse.de>
12220 * graphite-isl-ast-to-gimple.c
12221 (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12222 parameters and dominance check.
12223 (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12224 (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12225 (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12226 Do not update SSA form here or do intermediate IL verification.
12227 * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12228 (graphite_initialize): Remove check on the number of loops in
12229 the function and inline into graphite_transform_loops.
12230 (graphite_finalize): Inline into graphite_transform_loops.
12231 (graphite_transform_loops): Perform SSA update and IL verification
12233 * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12235 2017-10-13 Richard Biener <rguenther@suse.de>
12237 * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12238 graphite_expression_type_precision): Avoid global constructor
12240 (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12241 (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12242 (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12243 (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12244 (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12245 * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12247 2017-10-13 H.J. Lu <hongjiu.lu@intel.com>
12250 * config/i386/i386.h (ix86_red_zone_size): New.
12251 * config/i386/i386.md (push peephole2s): Replace
12252 "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12254 2017-10-13 Richard Sandiford <richard.sandiford@linaro.org>
12255 Alan Hayward <alan.hayward@arm.com>
12256 David Sherwood <david.sherwood@arm.com>
12258 * combine.c (can_change_dest_mode): Reject changes in
12259 REGMODE_NATURAL_SIZE.
12261 2017-10-13 Richard Sandiford <richard.sandiford@linaro.org>
12262 Alan Hayward <alan.hayward@arm.com>
12263 David Sherwood <david.sherwood@arm.com>
12265 * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12266 (expand_debug_source_expr): Likewise.
12267 * combine.c (combine_simplify_rtx): Likewise.
12268 * cse.c (fold_rtx): Likewise.
12269 * fwprop.c (canonicalize_address): Likewise.
12270 * targhooks.c (default_shift_truncation_mask): Likewise.
12272 2017-10-13 Richard Sandiford <richard.sandiford@linaro.org>
12273 Alan Hayward <alan.hayward@arm.com>
12274 David Sherwood <david.sherwood@arm.com>
12276 * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12277 (widened_mode): Likewise.
12278 (expand_unop): Likewise.
12279 * ree.c (transform_ifelse): Likewise.
12280 (merge_def_and_ext): Likewise.
12281 (combine_reaching_defs): Likewise.
12282 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12284 2017-10-13 Richard Sandiford <richard.sandiford@linaro.org>
12285 Alan Hayward <alan.hayward@arm.com>
12286 David Sherwood <david.sherwood@arm.com>
12288 * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12289 * combine.c (gen_lowpart_for_combine): Likewise.
12290 * dwarf2out.c (rtl_for_decl_location): Likewise.
12291 * final.c (alter_subreg): Likewise.
12292 * rtlhooks.c (gen_lowpart_general): Likewise.
12293 (gen_lowpart_if_possible): Likewise.
12295 2017-10-13 Richard Sandiford <richard.sandiford@linaro.org>
12296 Alan Hayward <alan.hayward@arm.com>
12297 David Sherwood <david.sherwood@arm.com>
12299 * calls.c (expand_call): Use subreg_lowpart_offset.
12300 * cse.c (cse_insn): Likewise.
12301 * regcprop.c (copy_value): Likewise.
12302 (copyprop_hardreg_forward_1): Likewise.
12304 2017-10-13 Jakub Jelinek <jakub@redhat.com>
12307 * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12308 *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12309 =Q constraints to +Q and into insn condition add check
12310 that operands[0] and operands[1] are equal.
12311 (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12312 =Q constraints to +Q and into insn condition add check
12313 that operands[0] is equal to either operands[1] or operands[2].
12316 * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12317 instead of handling MINUS_EXPR twice (once for each argument),
12318 canonicalize operand order and handle just once, use rtype where
12319 possible. Handle (A << B) | (A >> (-B & (Z - 1))).
12322 * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12323 any values of __C while still being pattern recognizable as a simple
12324 rotate instruction.
12326 2017-10-13 Richard Biener <rguenther@suse.de>
12328 PR tree-optimization/82451
12330 2017-10-02 Richard Biener <rguenther@suse.de>
12332 PR tree-optimization/82355
12333 * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12334 a mapping for the enclosing loop but avoid generating one for
12335 the loop tree root.
12336 (copy_bb_and_scalar_dependences): Remove premature codegen
12337 error on PHIs in blocks duplicated into multiple places.
12338 * graphite-scop-detection.c
12339 (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12340 in the region use it as loop and nest to analyze the DR in.
12341 (try_generate_gimple_bb): Likewise.
12342 * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12343 (add_loop_constraints): For blocks in a loop not in the region
12344 create a dimension with a single iteration.
12345 * sese.h (gbb_loop_at_index): Remove assert.
12347 * cfgloop.c (loop_preheader_edge): For the loop tree root
12348 return the single successor of the entry block.
12349 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12350 Reset the SCEV hashtable and niters.
12351 * graphite-scop-detection.c
12352 (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12353 assert that we only have POLYNOMIAL_CHREC that vary in loops
12354 contained in the region.
12355 (scop_detection::graphite_can_represent_expr): Adjust.
12356 (scop_detection::stmt_has_simple_data_refs_p): For loops
12357 not in the region set loop to NULL. The nest is now the
12358 entry edge to the region.
12359 (try_generate_gimple_bb): Likewise.
12360 * sese.c (scalar_evolution_in_region): Adjust for
12361 instantiate_scev change.
12362 * tree-data-ref.h (graphite_find_data_references_in_stmt):
12363 Make nest parameter the edge into the region.
12364 (create_data_ref): Likewise.
12365 * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12366 entry edge into a region and adjust instantiate_scev calls.
12367 (create_data_ref): Likewise.
12368 (graphite_find_data_references_in_stmt): Likewise.
12369 (find_data_references_in_stmt): Pass the loop preheader edge
12370 from the nest argument.
12371 * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12372 parameter the edge into the region.
12373 (instantiate_parameters): Use the loop preheader edge as entry.
12374 * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12376 (get_instantiated_value_entry): Make instantiate_below parameter
12377 the edge into the region.
12378 (instantiate_scev_name): Likewise. Adjust dominance checks,
12379 when we cannot use loop-based instantiation instantiate by
12380 walking use-def chains.
12381 (instantiate_scev_poly): Adjust.
12382 (instantiate_scev_binary): Likewise.
12383 (instantiate_scev_convert): Likewise.
12384 (instantiate_scev_not): Likewise.
12385 (instantiate_array_ref): Remove.
12386 (instantiate_scev_3): Likewise.
12387 (instantiate_scev_2): Likewise.
12388 (instantiate_scev_1): Likewise.
12389 (instantiate_scev_r): Do not blindly handle N-operand trees.
12390 Do not instantiate array-refs. Handle all constants and invariants.
12391 (instantiate_scev): Make instantiate_below parameter
12392 the edge into the region.
12393 (resolve_mixers): Use the loop preheader edge for the region
12394 parameter to instantiate_scev_r.
12395 * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12397 2017-10-13 Richard Biener <rguenther@suse.de>
12399 PR tree-optimization/82525
12400 * graphite-isl-ast-to-gimple.c
12401 (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12403 (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12404 Fail code generation when we cannot represent the isl integer.
12405 (binary_op_to_tree): Elide modulo operations that are no-ops
12406 in the type we code generate. Remove now superfluous code
12409 2017-10-13 Richard Biener <rguenther@suse.de>
12411 * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12412 (scop_detection::harmful_loop_in_region): Remove premature
12413 IV type restriction.
12414 (scop_detection::graphite_can_represent_scev): We can handle
12415 pointer IVs just fine.
12417 2017-10-13 Alan Modra <amodra@gmail.com>
12419 * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12420 "Clobbers and Scratch Registers". Add paragraph on
12421 alternative to clobbers for scratch registers and OpenBLAS
12424 2017-10-13 Alan Modra <amodra@gmail.com>
12426 * doc/extend.texi (Clobbers): Correct vax example. Delete old
12427 example of a memory input for a string of known length. Move
12428 commentary out of table. Add a number of new examples
12429 covering array memory inputs.
12431 2017-10-12 Martin Liska <mliska@suse.cz>
12433 PR tree-optimization/82493
12434 * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12435 (test_range_functions): New function.
12436 (sbitmap_c_tests): Likewise.
12437 * selftest-run-tests.c (selftest::run_tests): Run new tests.
12438 * selftest.h (sbitmap_c_tests): New function.
12440 * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12442 2017-10-12 Michael Meissner <meissner@linux.vnet.ibm.com>
12444 * config/rs6000/amo.h: Fix spacing issue.
12446 2017-10-12 Jakub Jelinek <jakub@redhat.com>
12449 * config/i386/i386.md (*ashl<mode>3_mask_1,
12450 *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12451 *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12454 2017-10-12 Jan Hubicka <hubicka@ucw.cz>
12456 * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12457 (profile_probability): Set max_probability
12458 to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12460 * profile-count.c (profile_probability::differs_from_p): Do not
12461 rely on max_probaiblity == 10000
12463 2017-10-12 Jeff Law <law@redhat.com>
12465 * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12468 2017-10-12 Martin Sebor <msebor@redhat.com>
12472 * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12473 (handle_alias_pairs): Call it.
12474 * common.opt (-Wattribute-alias): New option.
12475 * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12476 * doc/invoke.texi (-Wattribute-alias): Document.
12478 2017-10-12 Vladimir Makarov <vmakarov@redhat.com>
12481 2017-10-11 Vladimir Makarov <vmakarov@redhat.com>
12483 * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12485 * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12486 (make_hard_regno_born, make_hard_regno_dead): Update
12487 bb_killed_pseudos and bb_gen_pseudos.
12489 2017-10-12 Jan Hubicka <hubicka@ucw.cz>
12491 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12493 2017-10-12 Uros Bizjak <ubizjak@gmail.com>
12495 * config/alpha/alpha.c (alpha_split_conditional_move):
12496 Use std::swap instead of manually swapping.
12497 (alpha_stdarg_optimize_hook): Ditto.
12498 (alpha_canonicalize_comparison): Ditto.
12500 2017-10-12 Bin Cheng <bin.cheng@arm.com>
12502 * tree-loop-distribution.c (struct builtin_info): New struct.
12503 (struct partition): Refactor fields into struct builtin_info.
12504 (partition_free): Free struct builtin_info.
12505 (build_size_arg_loc, build_addr_arg_loc): Delete.
12506 (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12507 information from struct builtin_info.
12508 (find_single_drs): New function refactored from classify_partition.
12509 Also moved builtin validity checks to this function.
12510 (compute_access_range, alloc_builtin): New functions.
12511 (classify_builtin_st, classify_builtin_ldst): New functions.
12512 (classify_partition): Refactor code into functions find_single_drs,
12513 classify_builtin_st and classify_builtin_ldst.
12514 (distribute_loop): Don't do runtime alias check when distributing
12516 (find_seed_stmts_for_distribution): New function.
12517 (pass_loop_distribution::execute): Refactor code finding seed
12518 stmts into above function. Support distribution for the innermost
12519 two-level loop nest. Adjust dump information.
12521 2017-10-12 Bin Cheng <bin.cheng@arm.com>
12523 * tree-loop-distribution.c: Adjust the general comment.
12524 (NUM_PARTITION_THRESHOLD): New macro.
12525 (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12526 (classify_partition): Skip builtin pattern of loop nest's inner loop.
12527 (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12528 in call to build_partition_graph.
12529 (finalize_partitions): New parameter. Make loop distribution more
12530 conservative by fusing more partitions.
12531 (distribute_loop): Don't do runtime alias check in case of loop nest
12533 (find_seed_stmts_for_distribution): New function.
12534 (prepare_perfect_loop_nest): New function.
12535 (pass_loop_distribution::execute): Refactor code finding seed stmts
12536 and loop nest into above functions. Support loop nest distribution.
12537 Adjust dump information accordingly.
12539 2017-10-12 Bin Cheng <bin.cheng@arm.com>
12541 * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12542 and set PTYPE_SEQUENTIAL for merged partition.
12544 2017-10-12 Richard Biener <rguenther@suse.de>
12546 PR tree-optimization/69728
12548 2017-09-19 Richard Biener <rguenther@suse.de>
12550 PR tree-optimization/69728
12551 * graphite-sese-to-poly.c (schedule_error): New global.
12552 (add_loop_schedule): Handle empty domain by failing the
12554 (build_original_schedule): Handle schedule_error.
12556 * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12557 domain by returning an unchanged schedule.
12559 2017-10-12 Jakub Jelinek <jakub@redhat.com>
12561 * genrecog.c (validate_pattern): For VEC_SELECT verify that
12562 CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12564 2017-10-12 Aldy Hernandez <aldyh@redhat.com>
12566 * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12569 2017-10-12 Jakub Jelinek <jakub@redhat.com>
12572 * expr.c (store_field): Don't optimize away bitsize == 0 store
12573 from CALL_EXPR with addressable return type.
12575 2017-10-11 Segher Boessenkool <segher@kernel.crashing.org>
12577 * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12578 * config/rs6000/rs6000.md (sel): Delete mode attribute.
12579 (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12580 *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12581 TARGET_ISEL instead of TARGET_ISEL<sel>.
12583 2017-10-11 David Edelsohn <dje.gcc@gmail.com>
12585 * config/rs6000/rs6000.c
12586 (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12588 2017-10-11 Segher Boessenkool <segher@kernel.crashing.org>
12590 * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12592 (reg_or_cint_operand): Fix comment.
12593 (reg_or_zero_operand): New predicate.
12594 * config/rs6000/rs6000-protos.h (output_isel): Delete.
12595 * config/rs6000/rs6000.c (output_isel): Delete.
12596 * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12597 instead of reg_or_cint_operand. Output instruction directly (not via
12599 (isel_unsigned_<mode>): Ditto.
12600 (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12601 gpc_reg_operand. Add an instruction alternative for this. Output
12602 instruction directly.
12603 (*isel_reversed_unsigned_<mode>): Ditto.
12605 2017-10-11 Uros Bizjak <ubizjak@gmail.com>
12607 * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12608 (TARGET_CANONICALIZE_COMPARISON): Define.
12610 2017-10-11 Qing Zhao <qing.zhao@oracle.com>
12613 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12614 Check whether the dest is REG before adding REG_EQUIV note.
12616 2017-10-11 Vladimir Makarov <vmakarov@redhat.com>
12619 * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12621 * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12622 (make_hard_regno_born, make_hard_regno_dead): Update
12623 bb_killed_pseudos and bb_gen_pseudos.
12625 2017-10-11 Nathan Sidwell <nathan@acm.org>
12627 * incpath.h (enum incpath_kind): Name enum, prefix values.
12628 (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12629 * incpath.c (heads, tails): Use INC_MAX.
12630 (add_env_var_paths, add_standard_paths): Use incpath_kind.
12631 (merge_include_chains, split_quote_chain,
12632 register_include_chains): Update incpath_kind names.
12633 (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12634 * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12636 (add_framework_path, darwin_register_objc_includes): Likewise.
12637 * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12639 2017-10-11 Uros Bizjak <ubizjak@gmail.com>
12641 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12642 Do not use float_operator operator predicate.
12643 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12644 * config/i386/predicates.md (float_operator): Remove predicate.
12646 2017-10-11 Uros Bizjak <ubizjak@gmail.com>
12648 * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12649 (*jccxf_i387): Ditto.
12650 (*jcc<mode>_i387): Ditto.
12651 (*jccu<mode>_i387): Ditto.
12652 (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12653 (*jcc_*_i387 splitters): Remove.
12654 * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12655 * config/i386/i386.c (ix86_split_fp_branch): Remove.
12656 * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12659 2017-10-11 Jan Hubicka <hubicka@ucw.cz>
12661 * profile-count.h (slow_safe_scale_64bit): New function.
12662 (safe_scale_64bit): New inline.
12663 (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12664 * profile-count.c: Include wide-int.h
12665 (slow_safe_scale_64bit): New.
12667 2017-10-11 Nathan Sidwell <nathan@acm.org>
12669 * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12670 HAS_DECL_ASSEMBLER_NAME_P.
12671 * gimple-expr.c (gimple_decl_printable_name: Check
12672 HAS_DECL_ASSEMBLER_NAME_P too.
12673 * ipa-utils.h (type_in_anonymous_namespace_p): Check
12674 DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12675 (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12676 * passes.c (rest_of_decl_compilation): Check
12677 HAS_DECL_ASSEMBLER_NAME_P too.
12678 * recog.c (verify_changes): Likewise.
12679 * tree-pretty-print.c (dump_decl_name): Likewise.
12680 * tree-ssa-structalias.c (alias_get_name): Likewise. Reimplement.
12682 * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12683 (SET_DECL_ASSEMBLER_NAME): Use it.
12684 (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12685 (COPY_DECL_ASSEMBLER_NAME): Likewise.
12686 * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12688 2017-10-11 Jan Hubicka <hubicka@ucw.cz>
12690 * config.gcc (i386, x86_64): Add extra objects.
12691 * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12692 (ix86_min_insn_size): Declare.
12693 (ix86_issue_rate): Declare.
12694 (ix86_adjust_cost): Declare.
12695 (ia32_multipass_dfa_lookahead): Declare.
12696 (ix86_macro_fusion_p): Declare.
12697 (ix86_macro_fusion_pair_p): Declare.
12698 (ix86_bd_has_dispatch): Declare.
12699 (ix86_bd_do_dispatch): Declare.
12700 (ix86_core2i7_init_hooks): Declare.
12701 (ix86_atom_sched_reorder): Declare.
12702 * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12703 (COSTS_N_BYTES): Move to x86-tune-costs.h.
12704 (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12705 (rip_relative_addr_p): Rename to ...
12706 (ix86_rip_relative_addr_p): ... this one; export.
12707 (memory_address_length): Update.
12708 (ix86_issue_rate): Move to x86-tune-sched.c.
12709 (ix86_flags_dependent): Move to x86-tune-sched.c.
12710 (ix86_agi_dependent): Move to x86-tune-sched.c.
12711 (exact_dependency_1): Move to x86-tune-sched.c.
12712 (exact_store_load_dependency): Move to x86-tune-sched.c.
12713 (ix86_adjust_cost): Move to x86-tune-sched.c.
12714 (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12715 (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12716 (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12717 (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12718 (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12719 (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12720 (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12721 (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12722 (min_insn_size): Rename to ...
12723 (ix86_min_insn_size): ... this one; export.
12724 (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12725 (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12726 (core2i7_first_cycle_multipass_backtrack): Move to
12727 x86-tune-sched-core.c.
12728 (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12729 (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12730 (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12731 (ix86_avoid_jump_mispredicts): Update.
12732 (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12733 (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12734 (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12735 (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12736 (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12737 (MAX_INSN): Move to ix86-tune-sched-bd.c.
12738 (MAX_IMM): Move to ix86-tune-sched-bd.c.
12739 (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12740 (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12741 (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12742 (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12743 (MAX_STORE): Move to ix86-tune-sched-bd.c.
12744 (BIG): Move to ix86-tune-sched-bd.c.
12745 (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12746 (enum insn_path): Move to ix86-tune-sched-bd.c.
12747 (get_mem_group): Move to ix86-tune-sched-bd.c.
12748 (is_cmp): Move to ix86-tune-sched-bd.c.
12749 (dispatch_violation): Move to ix86-tune-sched-bd.c.
12750 (is_branch): Move to ix86-tune-sched-bd.c.
12751 (is_prefetch): Move to ix86-tune-sched-bd.c.
12752 (init_window): Move to ix86-tune-sched-bd.c.
12753 (allocate_window): Move to ix86-tune-sched-bd.c.
12754 (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12755 (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12756 (process_end_window): Move to ix86-tune-sched-bd.c.
12757 (allocate_next_window): Move to ix86-tune-sched-bd.c.
12758 (find_constant): Move to ix86-tune-sched-bd.c.
12759 (get_num_immediates): Move to ix86-tune-sched-bd.c.
12760 (has_immediate): Move to ix86-tune-sched-bd.c.
12761 (get_insn_path): Move to ix86-tune-sched-bd.c.
12762 (get_insn_group): Move to ix86-tune-sched-bd.c.
12763 (count_num_restricted): Move to ix86-tune-sched-bd.c.
12764 (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12765 (add_insn_window): Move to ix86-tune-sched-bd.c.
12766 (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12767 (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12768 (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12769 (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12770 (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12771 (do_dispatch): Move to ix86-tune-sched-bd.c.
12772 (has_dispatch): Move to ix86-tune-sched-bd.c.
12773 * config/i386/t-i386: Add new object files.
12774 * config/i386/x86-tune-costs.h: New file.
12775 * config/i386/x86-tune-sched-atom.c: New file.
12776 * config/i386/x86-tune-sched-bd.c: New file.
12777 * config/i386/x86-tune-sched-core.c: New file.
12778 * config/i386/x86-tune-sched.c: New file.
12780 2017-10-11 Liu Hao <lh_mouse@126.com>
12782 * pretty-print.c [_WIN32] (colorize_init): Remove. Use
12783 the generic version below instead.
12784 (should_colorize): Recognize Windows consoles as terminals
12786 * pretty-print.c [__MINGW32__] (write_all): New function.
12787 [__MINGW32__] (find_esc_head): Likewise.
12788 [__MINGW32__] (find_esc_terminator): Likewise.
12789 [__MINGW32__] (eat_esc_sequence): Likewise.
12790 [__MINGW32__] (mingw_ansi_fputs): New function that handles
12792 (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12795 2017-10-11 Richard Biener <rguenther@suse.de>
12797 * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12798 Properly call analyze_scalar_evolution with the loop of the stmt.
12800 2017-10-11 Richard Biener <rguenther@suse.de>
12802 * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12803 * tree-core.h (tree_base): Add chrec_var union member.
12804 * tree.h (CHREC_VAR): Remove.
12805 (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12806 * tree-chrec.h (build_polynomial_chrec): Adjust.
12807 * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12808 * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12810 2017-10-11 Marc Glisse <marc.glisse@inria.fr>
12812 * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12813 * match.pd: ... here.
12814 ((T) X == (T) Y): Relax condition.
12816 2017-10-11 Bin Cheng <bin.cheng@arm.com>
12818 PR tree-optimization/82472
12819 * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12821 (break_alias_scc_partitions): Update postorder number.
12823 2017-10-11 Martin Liska <mliska@suse.cz>
12826 * opts.c (parse_no_sanitize_attribute): Do not use error_value
12828 * opts.h (parse_no_sanitize_attribute): Remove last argument.
12830 2017-10-11 Martin Liska <mliska@suse.cz>
12832 * print-rtl.c (print_insn): Move declaration of idbuf
12833 to same scope as name.
12835 2017-10-11 Martin Liska <mliska@suse.cz>
12840 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12841 volatile arguments.
12843 2017-10-11 Martin Liska <mliska@suse.cz>
12846 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12847 volatile arguments.
12849 2017-10-11 Adhemerval Zanella <adhemerval.zanella@linaro.org>
12851 * config.gcc (default_gnu_indirect_function): Default to yes for
12852 arm*-*-linux* with glibc.
12854 2017-10-11 Richard Biener <rguenther@suse.de>
12856 * tree-scalar-evolution.c (get_scalar_evolution): Handle
12857 default-defs and types we do not want to analyze.
12858 (interpret_loop_phi): Replace unreachable code with an assert.
12859 (compute_scalar_evolution_in_loop): Remove and inline ...
12860 (analyze_scalar_evolution_1): ... here, replacing condition with
12861 what makes the intent clearer. Remove handling of cases
12862 get_scalar_evolution now handles.
12864 2017-10-10 Jim Wilson <wilson@tuliptree.org>
12866 PR rtl-optimization/81434
12867 * haifa-sched.c (prune_ready_list): Init min_cost_group to 0. Update
12868 comment for main loop. In sched_group_found if, also add checks for
12869 pass and min_cost_group.
12871 2017-10-10 Segher Boessenkool <segher@kernel.crashing.org>
12873 * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
12874 (rs6000_insn_cost): New function.
12875 * config/rs6000/rs6000.md (cost): New attribute.
12877 2017-10-10 Jakub Jelinek <jakub@redhat.com>
12878 H.J. Lu <hongjiu.lu@intel.com>
12882 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
12883 OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
12884 __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
12885 (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
12886 to OPTION_MASK_ISA_AVX512VL - builtins that have both
12887 OPTION_MASK_ISA_MMX and some other bit set require both
12888 mmx and the ISAs without the mmx bit.
12889 * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
12890 __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
12891 __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
12892 __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
12893 __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
12894 __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
12895 __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
12896 __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
12897 __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
12898 __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
12899 __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
12900 __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
12901 __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
12902 __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
12903 Add OPTION_MASK_ISA_MMX.
12905 2017-10-10 Andreas Tobler <andreast@gcc.gnu.org>
12907 * config.gcc (armv7*-*-freebsd*): New target.
12908 (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
12910 2017-10-10 Jan Hubicka <hubicka@ucw.cz>
12912 * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
12913 X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
12916 2017-10-10 Richard Sandiford <richard.sandiford@linaro.org>
12918 * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
12919 a template parameter.
12920 (WIDE_INT_REF_FOR): Update accordingly.
12921 * tree.h (wi::int_traits <const_tree>): Delete.
12922 (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
12923 (wi::to_widest, wi::to_offset): Use them. Expand commentary.
12924 (wi::tree_to_wide_ref): New typedef.
12925 (wi::to_wide): New function.
12926 * calls.c (get_size_range): Use wi::to_wide when operating on
12927 trees as wide_ints.
12928 * cgraph.c (cgraph_node::create_thunk): Likewise.
12929 * config/i386/i386.c (ix86_data_alignment): Likewise.
12930 (ix86_local_alignment): Likewise.
12931 * dbxout.c (stabstr_O): Likewise.
12932 * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
12933 * expr.c (const_vector_from_tree): Likewise.
12934 * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
12935 * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
12936 (fold_negate_expr_1, int_const_binop_1, const_binop)
12937 (fold_convert_const_int_from_real, optimize_bit_field_compare)
12938 (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
12939 (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
12940 (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
12941 (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
12942 (fold_not_const, round_up_loc): Likewise.
12943 * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
12944 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
12945 (alloca_call_type): Likewise.
12946 * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
12947 * godump.c (go_output_typedef): Likewise.
12948 * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
12949 * internal-fn.c (get_min_precision): Likewise.
12950 * ipa-cp.c (ipcp_store_vr_results): Likewise.
12951 * ipa-polymorphic-call.c
12952 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
12953 * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
12954 (ipa_modify_call_arguments): Likewise.
12955 * match.pd: Likewise.
12956 * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
12957 * print-tree.c (print_node_brief, print_node): Likewise.
12958 * stmt.c (expand_case): Likewise.
12959 * stor-layout.c (layout_type): Likewise.
12960 * tree-affine.c (tree_to_aff_combination): Likewise.
12961 * tree-cfg.c (group_case_labels_stmt): Likewise.
12962 * tree-data-ref.c (dr_analyze_indices): Likewise.
12963 (prune_runtime_alias_test_list): Likewise.
12964 * tree-dump.c (dequeue_and_dump): Likewise.
12965 * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
12966 * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
12967 * tree-pretty-print.c (dump_generic_node): Likewise.
12968 * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
12969 (simple_iv_with_niters): Likewise.
12970 * tree-ssa-address.c (addr_for_mem_ref): Likewise.
12971 * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
12972 * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
12973 * tree-ssa-loop-niter.c (split_to_var_and_offset)
12974 (refine_value_range_using_guard, number_of_iterations_ne_max)
12975 (number_of_iterations_lt_to_ne, number_of_iterations_lt)
12976 (get_cst_init_from_scev, record_nonwrapping_iv)
12977 (scev_var_range_cant_overflow): Likewise.
12978 * tree-ssa-phiopt.c (minmax_replacement): Likewise.
12979 * tree-ssa-pre.c (compute_avail): Likewise.
12980 * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
12981 (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
12982 * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
12983 * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
12984 * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
12985 * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
12986 (dump_case_nodes, try_switch_expansion): Likewise.
12987 * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
12988 (vect_do_peeling): Likewise.
12989 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12990 * tree-vect-stmts.c (vectorizable_load): Likewise.
12991 * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
12992 (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
12993 (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
12994 (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
12995 (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
12996 (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
12997 (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
12998 (evrp_dom_walker::before_dom_children): Likewise.
12999 * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13000 (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13001 (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13002 (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13003 (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13004 (get_range_pos_neg): Likewise.
13005 * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13006 * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13007 * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13008 * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13009 * config/avr/avr.c (avr_fold_builtin): Likewise.
13010 * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13011 * config/msp430/msp430.c (msp430_attr): Likewise.
13012 * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13013 * config/powerpcspe/powerpcspe-c.c
13014 (altivec_resolve_overloaded_builtin): Likewise.
13015 * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13016 (rs6000_expand_ternop_builtin): Likewise.
13017 * config/rs6000/rs6000-c.c
13018 (altivec_resolve_overloaded_builtin): Likewise.
13019 * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13020 (rs6000_expand_ternop_builtin): Likewise.
13021 * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13023 2017-10-10 Bin Cheng <bin.cheng@arm.com>
13025 * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13026 when copying loop nest with only one inner loop.
13028 2017-10-10 Richard Biener <rguenther@suse.de>
13030 * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13031 blocks if SCEV is active.
13032 * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13034 (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13035 (scev_initialize): Assert we are not yet initialized.
13037 2017-10-10 Bin Cheng <bin.cheng@arm.com>
13039 * tree-loop-distribution.c (generate_loops_for_partition): Remove
13040 inner loop's exit stmt by making it always exit the loop, otherwise
13041 we would generate an infinite empty loop.
13043 2017-10-10 Bin Cheng <bin.cheng@arm.com>
13045 * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13046 renaming variables in new preheader if it's deleted.
13048 2017-10-10 Bin Cheng <bin.cheng@arm.com>
13050 * tree-loop-distribution.c (struct partition): Remove unused field
13051 loops of the structure.
13052 (partition_alloc, partition_free): Ditto.
13053 (build_rdg_partition_for_vertex): Ditto.
13055 2017-10-09 Jeff Law <law@redhat.com>
13057 * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13058 return type to match prototype and documentation.
13060 2010-10-09 Segher Boessenkool <segher@kernel.crashing.org>
13062 * config/rs6000/rs6000.c (processor_costs): Move to ...
13063 * config/rs6000/rs6000.h: ... here.
13064 (rs6000_cost): Declare.
13066 2017-10-09 Eric Botcazou <ebotcazou@adacore.com>
13068 * except.c (setjmp_fn): New global variable.
13069 (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13070 (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13071 if DONT_USE_BUILTIN_SETJMP is defined.
13073 2017-10-09 Segher Boessenkool <segher@kernel.crashing.org>
13075 * target.def (insn_cost): New hook.
13076 * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13077 * doc/tm.texi: Regenerate.
13078 * rtlanal.c (insn_cost): Use the new hook.
13080 2017-10-09 Segher Boessenkool <segher@kernel.crashing.org>
13082 * combine.c (combine_validate_cost): Compute the new insn_cost,
13083 not just pattern_cost.
13084 (try_combine): Adjust comment.
13086 2017-10-09 Segher Boessenkool <segher@kernel.crashing.org>
13088 * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13090 * combine.c (uid_insn_cost): Adjust comment.
13091 (combine_validate_cost): Adjust comment. Use pattern_cost instead
13093 (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13094 * dse.c (find_shift_sequence): Ditto.
13095 * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13096 (bb_valid_for_noce_process_p): Use pattern_cost.
13097 * rtl.h (insn_rtx_cost): Delete.
13098 (pattern_cost): New prototype.
13099 (insn_cost): New prototype.
13100 * rtlanal.c (insn_rtx_cost): Rename to...
13101 (pattern_cost): ... this.
13104 2017-10-09 Uros Bizjak <ubizjak@gmail.com>
13106 * config/i386/i386.md (*jcc_2): Remove insn pattern.
13107 (*jcc<mode>_0_r_i387): Ditto.
13108 (*jccxf_r_i387): Ditto.
13109 (*jcc<mode>_r_i387): Ditto.
13110 (*jccu<mode>_r_i387): Ditto.
13111 (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13112 (*jcc): Rename from *jcc_1.
13114 2017-10-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13116 * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13117 deferred rescans after the lvx/stvx recombination pre-pass.
13119 2017-10-09 Michael Meissner <meissner@linux.vnet.ibm.com>
13121 * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13122 memory operation instruction support.
13123 * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13124 (rs6000-ibm-aix[789]*): Likewise.
13125 * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13126 Document new functions.
13128 2017-10-09 Richard Biener <rguenther@suse.de>
13130 PR tree-optimization/82397
13131 * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13132 equality only for semantically equal trees.
13134 2017-10-09 Richard Biener <rguenther@suse.de>
13136 PR tree-optimization/82449
13137 * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13138 * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13139 allow constant addresses.
13140 * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13143 2017-10-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13145 * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13148 2017-10-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13151 * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13154 2017-10-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13157 * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13159 2017-10-09 Jakub Jelinek <jakub@redhat.com>
13162 * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13163 s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13165 2017-10-09 Richard Sandiford <richard.sandiford@linaro.org>
13167 * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13168 (WI_BINARY_PREDICATE_RESULT): Likewise.
13169 (wi::binary_traits::operator_result): New type.
13170 (wi::binary_traits::predicate_result): Likewise.
13171 (generic_wide_int::operator~, unary generic_wide_int::operator-)
13172 (generic_wide_int::operator==, generic_wide_int::operator!=)
13173 (generic_wide_int::operator&, generic_wide_int::and_not)
13174 (generic_wide_int::operator|, generic_wide_int::or_not)
13175 (generic_wide_int::operator^, generic_wide_int::operator+
13176 (binary generic_wide_int::operator-, generic_wide_int::operator*):
13178 (operator~, unary operator-, operator==, operator!=, operator&)
13179 (operator|, operator^, operator+, binary operator-, operator*): New
13181 * expr.c (get_inner_reference): Use wi::bit_and_not.
13182 * fold-const.c (fold_binary_loc): Likewise.
13183 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13184 * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13185 (bit_value_binop): Likewise.
13186 * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13187 * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13188 (extract_range_from_binary_expr_1): Likewise.
13189 (masked_increment): Likewise.
13190 (simplify_bit_ops_using_ranges): Likewise.
13192 2017-10-09 Martin Jambor <mjambor@suse.cz>
13195 * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13196 * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13197 (hsa_type_for_scalar_tree_type): Use it. Always force min32int for
13199 (hsa_fixup_mov_insn_type): New function.
13200 (hsa_op_with_type::get_in_type): Use it.
13201 (hsa_build_append_simple_mov): Likewise. Allow sub-32bit
13202 immediates in an assert.
13203 (hsa_op_with_type::extend_int_to_32bit): New method.
13204 (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13205 types. Convert to dest type if necessary.
13206 (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13207 (reg_for_gimple_ssa): Pass false as min32int to
13208 hsa_type_for_scalar_tree_type.
13209 (gen_hsa_addr): Fixup type when creating addresable temporary.
13210 (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13211 (gen_hsa_unary_operation): Extend operands and convert to dest type if
13212 necessary. Call hsa_fixup_mov_insn_type.
13213 (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13214 extend operands and convert to dest type if necessary.
13215 (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13216 to dest type if necessary.
13217 (set_output_in_type): Call hsa_fixup_mov_insn_type. Just ude dest
13218 if conversion nt necessary and size matches.
13219 (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13220 to dest type if necessary.
13221 (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13222 (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13224 (gen_hsa_clrsb): Likewise.
13225 (gen_hsa_ffs): Likewise.
13226 (gen_hsa_divmod): Extend operands and convert to dest type if
13228 (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13230 2017-10-08 Segher Boessenkool <segher@kernel.crashing.org>
13232 * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13233 Remove empty default arguments. Use a brace block as output
13235 (conditional return): Ditto.
13237 (indirect_jump): Ditto. Use b%T0 instead of bctr/blr.
13238 (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13240 (group_ending_nop): Ditto.
13241 (doloop_end): Ditto.
13242 (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13243 (splitters for those): Ditto.
13245 2017-10-08 Segher Boessenkool <segher@kernel.crashing.org>
13247 * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13248 a conditional jump (and the compare for it) so that pc_rtx is the
13250 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13251 for the deleted and renamed ctr<mode>_internal[234] patterns.
13252 * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13253 Delete second conditional return pattern.
13254 (ctr<mode>_internal2): Delete this second bdnz pattern.
13255 (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13256 (ctr<mode>_internal4): Delete this second bdz pattern.
13258 2017-10-08 Eric Botcazou <ebotcazou@adacore.com>
13260 * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13261 (always_initialized_rtx_for_ssa_name_p): New predicate.
13262 * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13263 (finish_out_of_ssa): Free new field of SA.
13264 * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13265 * tree-ssa-coalesce.c: Include tree-ssa.h.
13266 (get_parm_default_def_partitions): Remove extern keyword.
13267 (get_undefined_value_partitions): New function.
13268 * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13269 not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13270 uninitialized bits.
13271 * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13273 2017-10-08 Eric Botcazou <ebotcazou@adacore.com>
13275 * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13277 2017-10-08 Jan Hubicka <hubicka@ucw.cz>
13279 * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13280 for targets that preffer 128bit.
13282 2017-10-08 Jan Hubicka <hubicka@ucw.cz>
13284 * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13286 2017-10-08 Olivier Hainque <hainque@adacore.com>
13288 * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13289 on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13291 (thumb_set_return_address): Likewise.
13293 2017-10-08 Olivier Hainque <hainque@adacore.com>
13295 * common/config/arm/arm-common.c (arm_except_unwind_info):
13296 Handle DWARF2_UNWIND_INFO.
13298 2017-10-07 Michael Collison <michael.collison@arm.com>
13300 * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13303 2017-10-07 Eric Botcazou <ebotcazou@adacore.com>
13305 * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13306 of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13307 * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13308 defined, force the creation of a new block for a dispatch label.
13310 2017-10-07 Jan Hubicka <hubicka@ucw.cz>
13312 * invoke.texi (Wsuggest-attribute=cold): Document.
13313 * common.opt (Wsuggest-attribute=cold): New
13314 * ipa-pure-const.c (warn_function_cold): New function.
13315 * predict.c (compute_function_frequency): Use it.
13316 * predict.h (warn_function_cold): Declare.
13318 2017-10-06 Jan Hubicka <hubicka@ucw.cz>
13320 * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13323 2017-10-06 Martin Liska <mliska@suse.cz>
13325 * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13326 keyword for member functions.
13327 (struct sanopt_tree_couple): New struct.
13328 (struct sanopt_tree_couple_hash): New function.
13329 (struct sanopt_ctx): Add new hash_map.
13330 (has_dominating_ubsan_ptr_check): New function.
13331 (record_ubsan_ptr_check_stmt): Likewise.
13332 (maybe_optimize_ubsan_ptr_ifn): Likewise.
13333 (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13334 (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13336 2017-10-06 Sudakshina Das <sudi.das@arm.com>
13339 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13340 aarch64_simd_valid_immediate on CONST_VECTORs.
13341 (aarch64_reg_or_bic_imm): Likewise.
13343 2017-10-06 Wilco Dijkstra <wdijkstr@arm.com>
13345 PR rtl-optimization/82396
13346 * haifa-sched.c (ready_sort_real): Disable qsort checking.
13348 2017-10-06 Sebastian Pop <sebpop@gmail.com>
13350 * graphite-dependences.c (scop_get_reads): Move code to...
13351 (scop_get_must_writes): Move code to...
13352 (scop_get_may_writes): Move code to...
13353 (scop_get_reads_and_writes): ... here.
13354 (scop_get_dependences): Call scop_get_reads_and_writes.
13356 2017-10-06 Jakub Jelinek <jakub@redhat.com>
13358 PR tree-optimization/82434
13359 * fold-const.h (can_native_encode_type_p,
13360 can_native_encode_string_p): Remove.
13361 * fold-const.c (native_encode_int): Formatting fixes. If ptr is NULL,
13362 don't encode anything, just return what would be otherwise returned.
13363 (native_encode_fixed, native_encode_complex, native_encode_vector):
13365 (native_encode_string): Likewise. Inline by hand
13366 can_native_encode_string_p.
13367 (can_native_encode_type_p): Remove.
13368 (can_native_encode_string_p): Remove.
13369 * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13370 STRING_CSTs using can_native_encode_string_p, test all
13371 CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13372 * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13373 argument from native_encode_expr.
13374 (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13375 (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13378 2017-10-06 Richard Biener <rguenther@suse.de>
13380 PR tree-optimization/82397
13381 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13382 operand_equal_p but rely on data_ref_compare_tree for detecting
13384 (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13385 to match up with dr_group_sort_cmp.
13387 2017-10-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13390 * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13392 * config/s390/s390-builtin-types.def: Regenerate.
13394 2017-10-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13397 * config/s390/s390-builtin-types.def: Regenerate.
13398 * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13399 Change flag from B_VXE to B_VX.
13400 (s390_vec_min_dbl): Remove B_VXE flag.
13402 2017-10-06 Richard Biener <rguenther@suse.de>
13404 * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13405 (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13406 translate_isl_ast_to_gimple::is_valid_rename,
13407 translate_isl_ast_to_gimple::get_rename,
13408 translate_isl_ast_to_gimple::get_def_bb_for_const,
13409 translate_isl_ast_to_gimple::get_new_name,
13410 translate_isl_ast_to_gimple::collect_all_ssa_names,
13411 translate_isl_ast_to_gimple::copy_loop_phi_args,
13412 translate_isl_ast_to_gimple::collect_all_ssa_names,
13413 translate_isl_ast_to_gimple::copy_loop_phi_args,
13414 translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13415 translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13416 translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13417 translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13418 translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13419 translate_isl_ast_to_gimple::copy_cond_phi_args,
13420 translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13421 translate_isl_ast_to_gimple::edge_for_new_close_phis,
13422 translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13423 translate_isl_ast_to_gimple::rename_uses,
13424 translate_isl_ast_to_gimple::rename_all_uses): Remove.
13425 (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13426 (set_rename_for_each_def): Likewise.
13427 (graphite_copy_stmts_from_block): Handle debug stmt resetting
13428 here. Handle rewriting SCEV analyzable uses here.
13429 (copy_bb_and_scalar_dependences): Generate code for PHI
13431 (graphite_regenerate_ast_isl): Adjust.
13432 * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13433 (add_write, add_read): New functions.
13434 (build_cross_bb_scalars_def): Use it and simplify.
13435 (build_cross_bb_scalars_use): Likewise.
13436 (graphite_find_cross_bb_scalar_vars): Inline into...
13437 (try_generate_gimple_bb): ...here. Add dependences for PHIs,
13438 simulating out-of-SSA. Compute liveout and add dependencies.
13439 (build_scops): Force an empty entry block.
13440 * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13442 (sese_build_liveouts): Declare.
13443 (sese_trivially_empty_bb_p): Likewise.
13444 * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13445 compute liveout and debug_liveout.
13446 (sese_bad_liveouts_use): Remove.
13447 (sese_reset_debug_liveouts_bb): Likewise.
13448 (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13449 (sese_build_liveouts): Build liveout and debug_liveout and store
13451 (new_sese_info): Adjust.
13452 (free_sese_info): Likewise.
13453 (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13454 do not build liveout here.
13455 (move_sese_in_condition): Adjust region entry.
13456 (scev_analyzable_p): Match up with chrec_apply requirements.
13457 (sese_trivially_empty_bb_p): New.
13458 * tree-into-ssa.c (get_reaching_def): Properly support generating
13459 default-defs for incremental rewrite of anonymous names.
13461 2017-10-06 Richard Biener <rguenther@suse.de>
13463 * graphite-sese-to-poly.c (extract_affine): For casts increasing
13464 precision do not perform modulo reduction.
13466 2017-10-06 Richard Biener <rguenther@suse.de>
13468 PR tree-optimization/82436
13469 * tree-vect-slp.c (vect_supported_load_permutation_p): More
13470 conservatively choose the vectorization factor when checking
13471 whether we can perform the required load permutation.
13472 (vect_transform_slp_perm_load): Assert when we may not fail.
13474 2017-10-05 Segher Boessenkool <segher@kernel.crashing.org>
13476 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13477 message for incompatible -msdata=* and -mcall-* options.
13479 2017-10-05 Jan Hubicka <hubicka@ucw.cz>
13481 * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13482 rate for post-reload scheduling.
13484 2017-10-05 Tamar Christina <tamar.christina@arm.com>
13486 * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13488 2017-10-05 Jan Hubicka <hubicka@ucw.cz>
13490 * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13491 to improve monte carlo in scimark.
13493 2017-10-05 Jan Hubicka <hubicka@ucw.cz>
13495 * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13496 pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13497 athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13498 pentium4_cost, nocona_cost): Set reassociation width to 1.
13499 (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13500 width to 2 for fp operations and 1 otherwise.
13501 (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13503 (atom_cost): Set reassociation width to 2.
13504 (slm_cost, generic_cost): Set fp reassociation width
13505 to 2 and 1 otherwise.
13506 (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13507 (core_cost): Set fp reassociation width to 4 and vector to 2.
13508 (ix86_reassociation_width): Rewrite using cost table; special case
13509 plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13510 and TARGET_AVX128_OPTIMAL.
13511 * config/i386/i386.h (processor_costs): Add
13512 reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13513 (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13514 TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13515 * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13516 (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13517 (X86_TUNE_VECTOR_PARALLEL_EXECUTION): Remove.
13519 2017-10-05 Nathan Sidwell <nathan@acm.org>
13521 * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13523 2017-10-05 Tamar Christina <tamar.christina@arm.com>
13525 * config/arm/arm.c (arm_test_fpu_data): New.
13526 (arm_run_selftests): Call arm_test_fpu_data.
13528 2017-10-04 Nathan Sidwell <nathan@acm.org>
13530 * toplev.c (toplev::main): Remove excess parens on pretty_printer
13532 * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13534 2017-10-04 Sudakshina Das <sudi.das@arm.com>
13536 * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13537 check type for aarch64_simd_valid_immediate.
13538 (aarch64_output_simd_mov_immediate): Update prototype.
13539 (aarch64_simd_valid_immediate): Update prototype.
13540 * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13541 support for ORR-immediate.
13542 (and<mode>3): modified pattern to add support for BIC-immediate.
13543 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13544 now checks for valid immediate for BIC and ORR based on new enum
13546 (aarch64_output_simd_mov_immediate): Function now used to output
13547 BIC/ORR imm as well based on new enum argument.
13548 * config/aarch64/constraints.md (Do): New vector immediate constraint.
13550 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13551 (aarch64_reg_or_bic_imm): Likewise.
13553 2017-10-04 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13555 * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13556 ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13558 2017-10-04 Wilco Dijkstra <wdijkstr@arm.com>
13562 PR rtl-optimization/82396
13563 * haifa-sched.c (autopref_multipass_init): Simplify
13565 (autopref_rank_data): Simplify sort order.
13566 * sched-int.h (autopref_multipass_data_): Remove
13567 multi_mem_insn_p, min_offset and max_offset.
13569 2017-10-04 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13571 * doc/sourcebuild.texi: Document vect_peeling_profitable.
13573 2017-10-04 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13575 * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13576 vect_doubleint_cvt.
13578 2017-10-04 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13580 * doc/sourcebuild.texi: Document vect_long_mult.
13582 2017-10-04 Richard Sandiford <richard.sandiford@linaro.org>
13584 PR tree-optimization/82413
13585 * fold-const.c (build_range_check): Use widest_int when comparing
13586 the maximum ETYPE value with HIGH.
13588 2017-10-04 Wilco Dijkstra <wdijkstr@arm.com>
13590 PR rtl-optimization/82396
13591 * haifa-sched.c (autopref_multipass_init): Simplify
13593 (autopref_rank_data): Simplify sort order.
13594 * sched-int.h (autopref_multipass_data_): Remove
13595 multi_mem_insn_p, min_offset and max_offset.
13597 2017-10-04 Jakub Jelinek <jakub@redhat.com>
13599 PR tree-optimization/82381
13600 * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13601 oeN->rank first. Return 1 or -1 if one op is SSA_NAME and the other
13604 PR tree-optimization/82374
13605 * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13606 DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13607 DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13608 current_function_decl to the new decl.
13610 2017-10-03 Michael Meissner <meissner@linux.vnet.ibm.com>
13612 * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13613 helper macro for IEEE float128 hardware built-in functions.
13614 (SQRTF128_ODD): Add built-in functions with the round-to-odd
13616 (TRUNCF128_ODD): Likewise.
13617 (ADDF128_ODD): Likewise.
13618 (SUBF128_ODD): Likewise.
13619 (MULF128_ODD): Likewise.
13620 (DIVF128_ODD): Likewise.
13621 (FMAF128_ODD): Likewise.
13622 * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13623 UNSPEC_TRUNC_ROUND_TO_ODD.
13624 (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13625 (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13626 floating point round to odd instructions.
13627 (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13628 (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13629 (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13630 (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13631 (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13632 (trunc<mode>sf2_hw): Change the truncate with round to odd
13633 expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13634 (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13635 to odd hardware instructions.
13636 (sub<mode>3_odd): Likewise.
13637 (mul<mode>3_odd): Likewise.
13638 (div<mode>3_odd): Likewise.
13639 (sqrt<mode>2_odd): Likewise.
13640 (fma<mode>4_odd): Likewise.
13641 (fms<mode>4_odd): Likewise.
13642 (nfma<mode>4_odd): Likewise.
13643 (nfms<mode>4_odd): Likewise.
13644 (trunc<mode>df2_odd): Change the truncate with round to odd
13645 expansion to use UNSPEC_TRUNC_ROUND_TO_ODD. Add a generator
13647 * doc/extend.texi (PowerPC built-in functions): Update documentation
13648 for existing IEEE float128-bit built-in functions. Add built-in
13649 functions that generate the IEEE 128-bit floating point round to
13652 2017-10-03 Segher Boessenkool <segher@kernel.crashing.org>
13654 PR rtl-optimization/77729
13655 * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13656 to (X&(C1&~C2))|C2 transformations.
13658 2017-10-03 Martin Jambor <mjambor@suse.cz>
13660 PR tree-optimization/82363
13661 * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13662 mismatch, mark lacc written regardless of racc.
13664 2017-10-03 Jakub Jelinek <jakub@redhat.com>
13666 PR tree-optimization/82381
13667 * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13668 stmt_to_insert nor wheather SSA_NAMEs are default defs.
13669 Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13670 fallthrough into reassoc_stmt_dominates_stmt_p.
13673 * combine.c (combine_instructions): Don't combine in unreachable
13676 2017-10-02 Peter Bergner <bergner@vnet.ibm.com>
13679 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13680 function to not use the have_cpu variable. Do not set cpu_index,
13681 rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13682 or the default cpu.
13683 (rs6000_valid_attribute_p): Remove duplicate initializations of
13684 old_optimize and func_optimize.
13685 (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13686 (rs6000_activate_target_options): Make global.
13687 * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13690 2017-10-02 Jakub Jelinek <jakub@redhat.com>
13692 * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13693 if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13694 Set *poffset to 0 and *psize and *pmax_size to -1 if
13695 *poffset + *psize overflows in HOST_WIDE_INT.
13697 PR tree-optimization/82387
13698 PR tree-optimization/82388
13699 PR tree-optimization/82389
13700 * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13701 instead of live_bytes non-NULL.
13703 2017-10-02 Georg-Johann Lay <avr@gjlay.de>
13706 * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13709 2017-10-02 Richard Biener <rguenther@suse.de>
13711 * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13712 -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13713 * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13715 2017-10-02 Richard Sandiford <richard.sandiford@linaro.org>
13717 * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13718 requested precision matches the type's.
13719 * calls.c (alloc_max_size): Calculate the new candidate size as
13720 a widest_int and use wi::to_widest when comparing it with the
13721 current candidate size.
13722 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13723 zero rather than integer_zero_node.
13724 * match.pd: Check for a no-op conversion before using wi::add
13725 rather than after. Use tree_to_uhwi when summing small shift
13726 counts into an unsigned int.
13728 2017-10-02 Richard Sandiford <richard.sandiford@linaro.org>
13729 Alan Hayward <alan.hayward@arm.com>
13730 David Sherwood <david.sherwood@arm.com>
13733 * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13734 (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13735 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13736 POINTER_AND_FP_REGS.
13738 2017-10-02 Richard Biener <rguenther@suse.de>
13740 PR tree-optimization/82355
13741 * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13742 a mapping for the enclosing loop but avoid generating one for
13743 the loop tree root.
13744 (copy_bb_and_scalar_dependences): Remove premature codegen
13745 error on PHIs in blocks duplicated into multiple places.
13746 * graphite-scop-detection.c
13747 (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13748 in the region use it as loop and nest to analyze the DR in.
13749 (try_generate_gimple_bb): Likewise.
13750 * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13751 (add_loop_constraints): For blocks in a loop not in the region
13752 create a dimension with a single iteration.
13753 * sese.h (gbb_loop_at_index): Remove assert.
13755 2017-10-01 Kevin Buettner <kevinb@redhat.com>
13757 * omp-expand.c (adjust_context_scope): New function.
13758 (expand_parallel_call): Call adjust_context_scope.
13760 2017-10-01 Jeff Law <law@redhat.com>
13762 * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13763 dom_opt_dom_walker class with direct access to private members.
13764 Add comments. Call test_for_singularity.
13765 (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13766 (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13767 m_dummy_cond anymore.
13768 (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13770 (pass_dominator:execute): Build the dummy_cond here and pass it
13771 to the dom_opt_dom_walker ctor.
13772 (test_for_singularity): New function.
13774 2017-09-30 Krister Walfridsson <krister.walfridsson@gmail.com>
13775 Maya Rashish <coypu@sdf.org>
13777 * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13778 netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13779 (alpha*-*-netbsd*) Use nbsd_tm_file.
13780 (arm*-*-netbsdelf*) Likewise.
13781 (i[34567]86-*-netbsdelf*) Likewise.
13782 (x86_64-*-netbsd*) Likewise.
13783 (mips*-*-netbsd*) Likewise.
13784 (powerpc-*-netbsd*) Likewise.
13785 (sh*-*-netbsd*) Likewise.
13786 (sparc-*-netbsdelf*) Likewise.
13787 (sparc64-*-netbsd*) Likewise.
13788 (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13790 (vax-*-netbsdelf*) Likewise.
13791 * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13792 (UINT_FAST8_TYPE) Likewise.
13793 (INT_FAST16_TYPE) Check CHAR_FAST16.
13794 (UINT_FAST16_TYPE) Likewise.
13796 2017-09-30 Jakub Jelinek <jakub@redhat.com>
13799 * config/i386/i386.md
13800 (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13801 (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13802 *divmodsi4_zext_2): New define_insn_and_split.
13803 (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13804 (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13805 (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13806 *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13807 New define_insn_and_split.
13808 (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13809 * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13810 operands[1] having DImode when mode is SImode.
13812 * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13813 always SImode for DIV and MOD in REG_EQUAL notes.
13815 2017-09-29 Yury Gribov <tetra2005@gmail.com>
13817 PR middle-end/82319
13818 * match.pd: Fix handling of NaNs in pattern.
13820 2017-09-29 Jeff Law <law@redhat.com>
13822 * sbitmap.c (bitmap_bit_in_range_p): New function.
13823 * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13824 * tree-ssa-dse.c (live_bytes_read): New function.
13825 (dse_classify_store): Ignore reads of dead bytes.
13827 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13828 typos and whitespace errors.
13829 * config/i386/predicates.md (address_no_seg_operand): Likewise.
13830 * config/s390/s390.c (s390_emit_prologue): Likewise.
13832 2017-09-29 Vladimir Makarov <vmakarov@redhat.com>
13835 * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13836 with a symbol for LRA.
13838 2017-09-29 Vladimir Makarov <vmakarov@redhat.com>
13840 PR rtl-optimization/82338
13841 * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13843 2017-09-29 Alexander Monakov <amonakov@ispras.ru>
13845 * genmodes.c (calc_wider_mode): Suppress qsort macro.
13846 * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
13847 (qsort_chk): Declare.
13848 * vec.c [CHECKING_P] (qsort_chk_error): New static function.
13849 (qsort_chk): New function.
13851 2017-09-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13853 PR tree-optimization/82337
13854 * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
13855 phi definition if the PHI result appears in an abnormal PHI.
13856 (find_basis_for_base_expr): Don't record a basis if the LHS of the
13857 basis appears in an abnormal PHI.
13859 2017-09-29 Richard Biener <rguenther@suse.de>
13861 * graphite-isl-ast-to-gimple.c
13862 (translate_isl_ast_to_gimple::set_codegen_error): New function.
13863 (binary_op_to_tree): Use it.
13864 (get_rename_from_scev): Likewise.
13865 (copy_loop_phi_nodes): Likewise.
13866 (copy_bb_and_scalar_dependences): Likewise.
13867 (translate_pending_phi_nodes): Likewise.
13869 2017-09-29 Jakub Jelinek <jakub@redhat.com>
13872 * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
13873 for movabsq $(i32 << shift), r64.
13875 2017-09-28 Uros Bizjak <ubizjak@gmail.com>
13877 * config/i386/i386.c (ix86_print_operand_address_as): Do not check
13878 index when encoding %esp as %rsp to avoid 0x67 prefix.
13880 2017-09-28 Sergey Shalnov <Sergey.Shalnov@intel.com>
13882 * config/i386/i386.md (*movsf_internal, *movdf_internal):
13883 Return 256-bit AVX modes for TARGET_PREFER_AVX256.
13885 2017-09-28 Thomas Preud'homme <thomas.preudhomme@arm.com>
13887 * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
13888 Extensions with more than 16 double VFP registers.
13889 (cmse_nonsecure_entry_clear_before_return): Remove second entry of
13890 to_clear_mask and all code related to it. Replace the remaining
13891 entry by a sbitmap and adapt code accordingly.
13893 2017-09-28 Henry Linjamäki <henry.linjamaki@parmance.com>
13895 * brig-builtins.def: Change pure attributes to const.
13897 2017-09-28 Joseph Myers <joseph@codesourcery.com>
13899 * config.gcc (default_gnu_indirect_function): Default to yes for
13900 sparc*-*-linux* with glibc.
13902 2017-09-28 Joseph Myers <joseph@codesourcery.com>
13904 * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
13905 (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
13906 when creating .init_array and .fini_array sections with priority
13909 2017-09-27 Christophe Lyon <christophe.lyon@linaro.org>
13912 * config/aarch64/aarch64.c
13913 (aarch64_builtin_support_vector_misalignment): Always return false
13914 when misalignment is unknown.
13916 2017-09-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
13918 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
13919 this function to return false if the definition used by the swap
13920 instruction is artificial, or if the memory address from which the
13921 constant value is loaded is not represented by a base address held
13922 in a register or if the base address register is a frame or stack
13923 pointer. Additionally, return false if the base address of the
13924 loaded constant is a SYMBOL_REF but is not considered to be a
13926 (replace_swapped_load_constant): New function.
13927 (rs6000_analyze_swaps): Add a new pass to replace a swap of a
13928 loaded constant vector with a load of a swapped constant vector.
13930 2017-09-27 Carl Love <cel@us.ibm.com>
13932 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
13933 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
13934 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
13937 2017-09-27 Alexander Monakov <amonakov@ispras.ru>
13939 * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
13940 first, always call autopref_rank_data otherwise.
13942 2017-09-27 Richard Biener <rguenther@suse.de>
13944 * graphite-scop-detection.c (find_scop_parameters): Move
13945 loop bound handling ...
13946 (gather_bbs::before_dom_children): ... here, avoiding the need
13947 to build scop_info->loop_nest.
13948 (record_loop_in_sese): Remove.
13949 * sese.h (sese_info_t::loop_nest): Remove.
13950 * sese.c (new_sese_info): Do not allocate loop_nest.
13951 (free_sese_info): Do not free loop_nest.
13953 2017-09-27 Jakub Jelinek <jakub@redhat.com>
13956 * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
13957 lhs from calls if the lhs has addressable type.
13959 2017-09-27 Richard Biener <rguenther@suse.de>
13961 * graphite.h (scop::max_alias_set): New member.
13962 * graphite-scop-detection.c: Remove references to non-existing
13963 --param in comments.
13964 (build_alias_sets): Record the maximum alias set used for drs.
13965 (build_scops): Support zero as unlimited for
13966 --param graphite-max-arrays-per-scop.
13967 * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
13968 and inline into ...
13969 (build_poly_sr_1): ... here. Compute alias set based on the
13970 maximum alias set used for drs rather than
13971 PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
13973 2017-09-27 Richard Biener <rguenther@suse.de>
13975 * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
13976 --param loop-block-tile-size=0 to disable tiling.
13978 2017-09-27 Richard Biener <rguenther@suse.de>
13980 * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
13981 (graphite-max-nb-scop-params): Document special value zero.
13982 * domwalk.h (dom_walker::STOP): New symbolical constant.
13983 (dom_walker::dom_walker): Add optional parameter for bb to
13985 (dom_walker::~dom_walker): Declare.
13986 (dom_walker::before_dom_children): Document STOP return value.
13987 (dom_walker::m_user_bb_to_rpo): New member.
13988 (dom_walker::m_bb_to_rpo): Likewise.
13989 * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
13990 mapping here if not provided by the user.
13991 (dom_walker::~dom_walker): Free bb to RPO mapping if not
13992 provided by the user.
13993 (dom_walker::STOP): Define.
13994 (dom_walker::walk): Do not compute bb to RPO mapping here.
13995 Support STOP return value from before_dom_children to stop
13997 * graphite-optimize-isl.c (optimize_isl): If the schedule
13998 is the same still generate code if -fgraphite-identity
13999 or -floop-parallelize-all are given.
14000 * graphite-scop-detection.c: Include cfganal.h.
14001 (gather_bbs::gather_bbs): Get and pass through bb to RPO
14003 (gather_bbs::before_dom_children): Return STOP for BBs
14005 (build_scops): Compute bb to RPO mapping and pass it to
14006 the domwalk. Treat --param graphite-max-nb-scop-params=0
14007 as not limiting the number of params.
14008 * graphite.c (graphite_initialize): Remove limit on the
14009 number of basic-blocks in a function.
14010 * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14011 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14012 default value of 10.
14014 2017-09-26 Michael Meissner <meissner@linux.vnet.ibm.com>
14016 * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14017 Adjust code to eliminate needing to do the shift right 32-bits
14018 operation after XSCVDPSPN.
14020 2017-09-26 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14022 * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14023 ((X / Y) != 0 -> X >= Y): Likewise.
14025 2017-09-26 Carl Love <cel@us.ibm.com>
14027 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14028 P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14029 vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14030 void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14031 * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14032 * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14033 definitions and overloading.
14034 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14035 statement for P9V_BUILTIN_XST_LEN_R.
14036 (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14037 * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14038 define_expand and define_insn for the instructions and builtins.
14039 * doc/extend.texi: Update the built-in documentation file for the new
14040 built-in functions.
14041 * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14042 define_insn for the instructions
14044 2017-09-26 Krister Walfridsson <krister.walfridsson@gmail.com>
14047 * gcc/config/netbsd-protos.h: New file.
14048 * gcc/config/netbsd.c: New file.
14049 * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14050 * gcc/config/t-netbsd: New file.
14051 * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14052 (tmake_file) Add t-netbsd.
14053 (extra_objs) Add netbsd.o.
14055 2017-09-26 Janus Weil <janus@gcc.gnu.org>
14059 * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14061 2017-09-26 Michael Meissner <meissner@linux.vnet.ibm.com>
14063 * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14064 sign extension from a vector register to a GPR by doing a 32-bit
14065 direct move and then an EXTSW.
14066 (extendsi<mode>2 splitter): Likewise.
14067 (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14068 right or vector extract after doing XSCVDPSPN. Use
14069 zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14071 (movdi_from_sf_zero_ext): Likewise.
14072 (reload_gpr_from_vsxsf): Likewise.
14073 (p8_mfvsrd_4_disf): Delete, no longer used.
14074 (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14075 then needing to move the SFmode to a GPR to use the XSCVDPSP
14076 instruction instead of FRSP and XSCVDPSPN.
14077 * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14078 it is adjacent to the other XSCVSPDP insns.
14079 (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14080 SFmode to be in traditional Altivec registers.
14081 (vsx_xscvdpspn): Eliminate useless alternative constraint.
14082 (vsx_xscvspdpn): Likewise.
14083 (vsx_xscvspdpn_scalar): Likewise.
14085 2017-09-26 Martin Jambor <mjambor@suse.cz>
14087 * tree-sra.c (compare_access_positions): Put integral types first,
14088 stabilize sorting of integral types, remove conditions putting
14089 non-full-precision integers last.
14090 (sort_and_splice_var_accesses): Disable scalarization if a
14091 non-integert would be represented by a non-full-precision integer.
14093 2017-09-26 Joseph Myers <joseph@codesourcery.com>
14095 * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14096 * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14097 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14098 * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14099 Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14100 conditionals inside the function instead of around it. Call
14101 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14102 (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14104 2017-09-26 Richard Biener <rguenther@suse.de>
14106 * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14108 (scop_detection::build_scop_breadth): ... this. Removed.
14109 (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14110 (scop_detection::harmful_stmt_in_bb): Likewise.
14111 (scop_detection::graphite_can_represent_stmt): Likewise.
14112 (scop_detection::loop_body_is_valid_scop): Likewise. Remove recursion.
14113 (scop_detection::can_represent_loop): Remove recursion, fold in ...
14114 (scop_detection::can_represent_loop_1): ... this. Removed.
14115 (scop_detection::harmful_loop_in_region): Simplify after inlining
14116 the above and remove more quadraticness.
14117 (build_scops): Adjust.
14118 * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14121 2017-09-26 Jakub Jelinek <jakub@redhat.com>
14124 * config/i386/i386.c (ix86_print_operand_address_as): Only test
14125 REGNO (base) == SP_REG if base is a REG.
14127 PR middle-end/35691
14128 * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14129 if it is different SSA_NAME.
14130 (optimize_range_tests_cmp_bitwise): New function.
14131 (optimize_range_tests): Call it.
14133 2017-09-26 Richard Biener <rguenther@suse.de>
14135 PR tree-optimization/82321
14136 * graphite.c (canonicalize_loop_closed_ssa): Properly check
14137 for the def being inside the loop.
14139 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14141 * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14143 * config/s390/s390-builtins.def: Fix constraint on op4.
14145 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14147 * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14148 independent expanders.
14149 * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14150 ("vec_ordered", "vec_unordered"): New expanders.
14152 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14154 * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14157 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14159 * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14160 vec_unpacks_lo_v16qi.
14161 ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14163 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14165 * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14166 ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14167 ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14169 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14171 * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14173 * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14175 ("*vec_slb<mode>"): New insn pattern.
14176 ("vec_shr_<mode>"): New expander.
14177 * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14178 and force the shift count operand to V16QImode.
14179 ("vec_srb<mode>"): Set shift count mode to V16QI.
14181 2017-09-26 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14183 * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14184 ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14185 ("vec_widen_smult_hi_<mode>"): New expander definitions.
14187 2017-09-26 Richard Earnshaw <rearnsha@arm.com>
14190 * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14192 2017-09-26 Richard Biener <rguenther@suse.de>
14194 PR tree-optimization/82320
14195 * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14198 2017-09-25 Jeff Law <law@redhat.com>
14200 * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14201 prototype for new argument.
14202 * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14203 mostly extracted from rs6000_emit_allocate_stack.
14204 (rs6000_emit_probe_stack_range_stack_clash): New function.
14205 (rs6000_emit_allocate_stack): Call
14206 rs6000_emit_probe_stack_range_stack_clash as needed.
14207 (rs6000_emit_probe_stack_range): Add additional argument
14208 to call to gen_probe_stack_range{si,di}.
14209 (output_probe_stack_range): New.
14210 (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14211 (output_probe_stack_range_stack_clash): New.
14212 (rs6000_emit_prologue): Emit notes into dump file as requested.
14213 * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14214 (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14215 Add additional operand and pass it to output_probe_stack_range.
14217 2017-09-25 Bin Cheng <bin.cheng@arm.com>
14219 PR tree-optimization/82163
14220 * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14221 (checking_verify_loop_closed_ssa): New parameter.
14222 * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14223 (check_loop_closed_ssa_stmt): Delete.
14224 (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14225 (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14226 (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14229 2017-09-25 Pekka Jaaskelainen <pekka@parmance.com>
14231 * brig-builtins.def: Treat HSAIL barrier builtins as
14232 setjmp/longjump style functions.
14234 2017-09-25 Richard Sandiford <richard.sandiford@linaro.org>
14236 * target.def (constant_alignment): New hook.
14237 * defaults.h (CONSTANT_ALIGNMENT): Delete.
14238 * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14239 (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14240 * doc/tm.texi: Regenerate.
14241 * targhooks.h (default_constant_alignment): Declare.
14242 (constant_alignment_word_strings): Likewise.
14243 * targhooks.c (default_constant_alignment): New function.
14244 (constant_alignment_word_strings): Likewise.
14245 * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14246 instead of CONSTANT_ALIGNMENT.
14247 * varasm.c (align_variable, get_variable_align, build_constant_desc)
14248 (force_const_mem): Likewise.
14249 * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14250 * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14251 (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14252 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14253 * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14255 * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14256 * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14257 constant_alignment_word_strings.
14258 * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14259 (CONSTANT_ALIGNMENT): Likewise.
14260 * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14261 (arm_constant_alignment): New function.
14262 * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14263 * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14264 constant_alignment_word_strings.
14265 * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14266 * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14267 constant_alignment_word_strings.
14268 * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14269 * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14270 (cris_constant_alignment): New function.
14271 * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14272 * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14273 (epiphany_constant_alignment): New function.
14274 * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14275 * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14276 constant_alignment_word_strings.
14277 * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14278 * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14279 constant_alignment_word_strings.
14280 * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14281 * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14282 constant_alignment_word_strings.
14283 * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14284 * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14285 * config/i386/i386.c (ix86_constant_alignment): Make static.
14286 Use the same interface as the target hook.
14287 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14288 * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14289 * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14290 constant_alignment_word_strings.
14291 * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14292 * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14293 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14294 * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14295 * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14296 constant_alignment_word_strings.
14297 * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14298 * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14299 constant_alignment_word_strings.
14300 * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14301 * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14302 constant_alignment_word_strings.
14303 * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14304 * config/microblaze/microblaze.c (microblaze_constant_alignment):
14306 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14307 * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14308 * config/mips/mips.c (mips_constant_alignment): New function.
14309 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14310 * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14311 * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14312 * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14313 (mmix_constant_alignment): Make static. Use the same interface
14314 as the target hook.
14315 * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14316 * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14317 constant_alignment_word_strings.
14318 * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14319 * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14320 constant_alignment_word_strings.
14321 * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14322 * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14323 constant_alignment_word_strings.
14324 * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14325 * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14326 (rs6000_constant_alignment): New function.
14327 * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14328 * config/riscv/riscv.c (riscv_constant_alignment): New function.
14329 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14330 * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14331 * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14332 (rs6000_constant_alignment): New function.
14333 * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14334 * config/s390/s390.c (s390_constant_alignment): New function.
14335 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14336 * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14337 * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14338 constant_alignment_word_strings.
14339 * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14340 * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14341 (sparc_constant_alignment): New function.
14342 * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14343 * config/spu/spu.c (spu_constant_alignment): New function.
14344 (TARGET_CONSTANT_ALIGNMENT): Redefine.
14345 * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14346 * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14347 constant_alignment_word_strings.
14348 * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14349 * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14350 constant_alignment_word_strings.
14351 * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14352 * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14353 constant_alignment_word_strings.
14354 * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14355 * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14356 (visium_constant_alignment): New function.
14357 * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14358 * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14359 (xtensa_constant_alignment): New function.
14360 * system.h (CONSTANT_ALIGNMENT): Poison.
14362 2017-09-25 Will Schmidt <will_schmidt@vnet.ibm.com>
14364 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14365 for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14366 (rs6000_builtin_valid_without_lhs): New helper function.
14367 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14368 Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14370 2017-09-25 Richard Sandiford <richard.sandiford@linaro.org>
14372 * target.h (vec_perm_indices): Use unsigned short rather than
14374 (auto_vec_perm_indices): Likewise.
14375 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14376 Use unsigned int rather than unsigned char.
14377 * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14379 2017-09-25 Richard Biener <rguenther@suse.de>
14381 * cfgloop.h (sort_sibling_loops): Declare.
14382 * cfgloop.c (sort_sibling_loops_cmp): New helper.
14383 (sort_sibling_loops): New function sorting the sibling loop list
14385 * graphite.c (graphite_transform_loops): Sort sibling loops.
14387 2017-09-25 Richard Sandiford <richard.sandifird@linaro.org>
14389 * target.def (vec_perm_const_ok): Change sel parameter to
14391 * optabs-query.c (can_vec_perm_p): Update accordingly.
14392 * doc/tm.texi: Regenerate.
14393 * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14394 auto_vec_perm_indices and remove separate nelt field.
14395 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14396 (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14397 (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14398 (aarch64_expand_vec_perm_const): Update accordingly.
14399 (aarch64_vectorize_vec_perm_const_ok): Likewise. Change sel
14400 to vec_perm_indices.
14401 * config/arm/arm.c (expand_vec_perm_d): Change perm to
14402 auto_vec_perm_indices and remove separate nelt field.
14403 (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14404 (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14405 (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14407 (arm_vectorize_vec_perm_const_ok): Likewise. Change sel
14408 to vec_perm_indices.
14409 * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14410 sel to vec_perm_indices.
14411 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14412 * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14413 * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14415 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14418 2017-09-25 Pierre-Marie de Rodat <derodat@adacore.com>
14421 * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14422 on the FUNCTION_DECL function context if it has a DIE that is a
14425 2017-09-25 Richard Biener <rguenther@suse.de>
14427 PR tree-optimization/82285
14428 * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14431 2017-09-25 Tom de Vries <tom@codesourcery.com>
14435 * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14438 2017-09-25 Richard Biener <rguenther@suse.de>
14440 * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14441 ISL errors other than isl_error_quota happen. Dump if the
14442 schedule is the same.
14443 * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14444 errors instead of aborting inside ISL.
14446 2017-09-25 Iain Sandoe <iain@codesourcery.com>
14449 * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14450 of libgcc_eh for m64.
14451 * config/i386/darwin64.h: Likewise.
14453 2017-09-25 Richard Biener <rguenther@suse.de>
14455 PR middle-end/82144
14456 * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14457 attribute for incomplete types nor twice for complete ones.
14459 2017-09-24 Uros Bizjak <ubizjak@gmail.com>
14462 * config/i386/i386.c (ix86_print_operand_address_as): Encode
14463 %esp as %rsp to avoid 0x67 prefix if there is no index or base
14466 2017-09-23 Uros Bizjak <ubizjak@gmail.com>
14469 * config/i386/i386.opt (mprefer-avx256): Use
14470 ix86_target_flags variable.
14471 * config/i386/i386.c (ix86_target_string): Move
14472 -mprefer-avx256 to flag2_opts.
14474 2017-09-22 Jakub Jelinek <jakub@redhat.com>
14476 PR middle-end/35691
14477 * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14478 and x != -1 | y != -1 into (x & y) != -1.
14480 2017-09-22 Steve Ellcey <sellcey@cavium.com>
14482 * config.gcc: Add new case statement to set
14483 default_gnu_indirect_function. Remove it from x86_64-*-linux*,
14484 i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14485 s390x-*-linux* case statements. Added aarch64 to the list of
14486 supported architectures.
14488 2017-09-22 Richard Sandiford <richard.sandiford@linaro.org>
14490 PR tree-optimization/82289
14491 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14492 STMT_VINFO_RELEVANT_P.
14494 2017-09-22 Richard Sandiford <richard.sandiford@linaro.org>
14495 Alan Hayward <alan.hayward@arm.com>
14496 David Sherwood <david.sherwood@arm.com>
14498 * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14499 for VR_RANGE only; don't allow VR_ANTI_RANGE.
14500 (extract_range_from_binary_expr_1): Don't call
14501 extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14503 2017-09-22 Richard Sandiford <richard.sandiford@linaro.org>
14504 Alan Hayward <alan.hayward@arm.com>
14505 David Sherwood <david.sherwood@arm.com>
14507 * target.def (preferred_vector_alignment): New hook.
14508 * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14510 * doc/tm.texi: Regenerate.
14511 * targhooks.h (default_preferred_vector_alignment): Declare.
14512 * targhooks.c (default_preferred_vector_alignment): New function.
14513 * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14515 (DR_TARGET_ALIGNMENT): New macro.
14516 (aligned_access_p): Update commentary.
14517 (vect_known_alignment_in_bytes): New function.
14518 * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14520 (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14521 Calculate the misalignment based on the target alignment rather than
14523 (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14524 rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14525 (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14526 the target alignment, rather than masking the element misalignment
14527 with the number of elements in a vector. Also use the target
14528 alignment when calculating the maximum number of peels.
14529 (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14530 instead of TYPE_ALIGN_UNIT.
14531 (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14532 Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14533 (vect_create_addr_base_for_vector_ref): Update call accordingly.
14534 (vect_create_data_ref_ptr): Likewise.
14535 (vect_setup_realignment): Realign by ANDing with
14536 -DR_TARGET_MISALIGNMENT.
14537 * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14538 the number of peels based on DR_TARGET_ALIGNMENT.
14539 * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14540 with the guaranteed alignment boundary when deciding whether
14542 (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14543 relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14544 (ensure_base_align): Remove stmt_info parameter. Get the
14545 target base alignment from DR_TARGET_ALIGNMENT.
14546 (vectorizable_store): Update call accordingly. Interpret
14547 DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14549 (vectorizable_load): Likewise.
14551 2017-09-22 Richard Sandiford <richard.sandiford@linaro.org>
14552 Alan Hayward <alan.hayward@arm.com>
14553 David Sherwood <david.sherwood@arm.com>
14555 * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14556 * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14557 (vect_enhance_data_refs_alignment): Likewise.
14559 2017-09-22 Richard Earnshaw <richard.earnshaw@arm.com>
14561 * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14562 error. Only quit immediately if parsing is complete.
14563 (BEGIN): Initialize fatal_err and parse_done.
14564 (begin fpu, end fpu): Check number of arguments.
14565 (begin arch, end arch): Likewise.
14566 (begin cpu, end cpu): Likewise.
14567 (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14568 (optalias): Likewise.
14570 2017-09-22 Richard Earnshaw <richard.earnshaw@arm.com>
14572 * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14573 * config/arm/arm-isa.h: Delete. Move definitions to ...
14574 * arm-cpus.in: ... here. Use new feature and fgroup values.
14575 * config/arm/arm.c (arm_option_override): Use lower case for feature
14577 * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14578 (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14579 * config/arm/parsecpu.awk (END): Add new command 'isa'.
14581 (print_isa_bits_for): New function.
14582 (gen_isa): New function.
14583 (gen_comm_data): Use print_isa_bits_for.
14584 (define feature): New keyword.
14585 (define fgroup): New keyword.
14586 * config/arm/t-arm (TM_H): Remove.
14587 (GTM_H): Add arm-isa.h.
14588 (arm-isa.h): Add rule to generate file.
14589 * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14590 case for feature bit names.
14592 2017-09-22 Richard Biener <rguenther@suse.de>
14594 * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14596 (graphite_regenerate_ast_isl): Do not reset SCEV. Move debug
14597 print of no dependency loops ...
14598 * graphite.c (graphite_transform_loops): ... here.
14599 (canonicalize_loop_closed_ssa_form): Work from inner to outer
14601 (same_close_phi_node, remove_duplicate_close_phi,
14602 make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14603 (canonicalize_loop_closed_ssa): ... here and simplify.
14604 * graphite-optimize-isl.c: Include tree-vectorizer.h.
14605 (optimize_isl): Use dump_printf_loc to tell when we stopped
14606 optimizing because of an ISL timeout.
14608 2017-09-22 Richard Biener <rguenther@suse.de>
14610 PR tree-optimization/82291
14611 * tree-if-conv.c (predicate_mem_writes): Make sure to
14612 remove writes in blocks predicated with false.
14614 2017-09-22 Richard Biener <rguenther@suse.de>
14616 * sese.c: Include cfganal.h.
14617 (if_region_set_false_region): Remove.
14618 (create_if_region_on_edge): Likewise.
14619 (move_sese_in_condition): Re-implement without destroying
14622 2017-09-22 Richard Biener <rguenther@suse.de>
14624 * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14625 Verify both BBs contain loop PHI nodes before dispatching to
14626 copy_loop_phi_args.
14627 (graphite_regenerate_ast_isl): Do not recompute dominators,
14628 do not verify three times. Restructure for clarity.
14629 * graphite-scop-detection.c (same_close_phi_node,
14630 remove_duplicate_close_phi, make_close_phi_nodes_unique,
14631 defined_in_loop_p, canonicalize_loop_closed_ssa,
14632 canonicalize_loop_closed_ssa_form): Simplify, remove excess
14633 checking and SSA rewrite, move to ...
14634 * graphite.c: ... here. Include ssa.h and tree-ssa-loop-manip.h.
14635 (graphite_initialize): Do not pass in ctx, do not reset the
14636 SCEV cache, compute only dominators.
14637 (graphite_transform_loops): Allocate ISL ctx after
14638 graphite_initialize. Call canonicalize_loop_closed_ssa_form.
14639 Maintain post-dominators only around build_scops.
14640 * sese.c (if_region_set_false_region): Make static. Free
14641 and recompute dominators.
14642 (move_sese_in_condition): Assert we don't get called with
14643 post-dominators computed.
14644 * sese.h (if_region_set_false_region): Remove.
14646 2017-09-22 Sergey Shalnov <sergey.shalnov@intel.com>
14648 * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14649 mode attribute for TARGET_AVX512VL.
14651 2017-09-21 Sergey Shalnov <sergey.shalnov@intel.com>
14653 * config/i386/i386.opt (mprefer-avx256): New option.
14654 * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14656 (ix86_preferred_simd_mode): Return 256-bit AVX modes
14657 for TARGET_PREFER_AVX256.
14658 * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14660 2017-09-21 Jeff Law <law@redhat.com>
14662 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14663 Fix dump output if the only stack space is for pushed registers.
14665 2017-09-21 Richard Sandiford <richard.sandiford@linaro.org>
14667 * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14670 2017-09-21 Martin Sebor <msebor@redhat.com>
14673 * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14674 code (in C++) or code that triggers warnings.
14676 2017-09-21 Eric Botcazou <ebotcazou@adacore.com>
14678 * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14680 2017-09-21 Segher Boessenkool <segher@kernel.crashing.org>
14682 * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14683 * sched-rgn.c: Ditto.
14684 * sel-sched-ir.c: Ditto.
14686 2017-09-21 Alexander Monakov <amonakov@ispras.ru>
14688 * toplev.h (set_random_seed): Adjust return type.
14689 * toplev.c (init_local_tick): Move eager initialization of random_seed
14690 to get_random_seed. Adjust comment.
14691 (init_random_seed): Inline to get_random_seed, delete.
14692 (get_random_seed): Initialize random_seed lazily.
14693 (set_random_seed): Do not return previous value.
14694 (print_switch_value): Do not call get_random_seed.
14696 2017-09-21 Evgeny Kudryashov <kudryashov@ispras.ru>
14698 * cgraph.c (delete_function_version): New, broken out from...
14699 (cgraph_node::delete_function_version): ...here. Rename to
14700 cgraph_node::delete_function_version_by_decl. Update all uses.
14701 (cgraph_node::remove): Call delete_function_version.
14703 2017-09-21 Jakub Jelinek <jakub@redhat.com>
14706 * tree-inline.c (expand_call_inline): Emit clobber stmts for
14707 VAR_DECLs to which addressable non-volatile parameters are mapped
14708 and for id->retvar after the return value assignment. Clear
14709 id->retval and id->retbnd after inlining.
14711 2017-09-21 Richard Biener <rguenther@suse.de>
14713 PR tree-optimization/82276
14714 PR tree-optimization/82244
14715 * tree-vrp.c (build_assert_expr_for): Set
14716 SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14718 (remove_range_assertions): Revert earlier change.
14720 2017-09-21 Wilco Dijkstra <wdijkstr@arm.com>
14723 * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14725 2017-09-21 Richard Biener <rguenther@suse.de>
14727 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14728 Restore valid IL after code generation errors.
14729 * graphite.c (graphite_transform_loops): Diagnose code
14730 generation issues as MSG_MISSED_OPTIMIZATION and continue
14731 with processing SCOPs.
14733 2017-09-21 Richard Sandiford <richard.sandiford@linaro.org>
14734 Alan Hayward <alan.hayward@arm.com>
14735 David Sherwood <david.sherwood@arm.com>
14737 * calls.c (compute_argument_addresses): Use simplify_gen_binary
14738 rather than choosing between plus_constant and gen_rtx_<CODE>.
14739 * expr.c (emit_push_insn): Likewise.
14740 (expand_expr_real_2): Likewise.
14742 2017-09-21 Richard Sandiford <richard.sandiford@linaro.org>
14743 Alan Hayward <alan.hayward@arm.com>
14744 David Sherwood <david.sherwood@arm.com>
14746 * loop-unroll.c (split_iv): Call copy_rtx on the step.
14748 2017-09-21 Richard Sandiford <richard.sandiford@linaro.org>
14749 Alan Hayward <alan.hayward@arm.com>
14750 David Sherwood <david.sherwood@arm.com>
14752 * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14753 calling tree_to_uhwi.
14755 2017-09-21 Richard Sandiford <richard.sandiford@linaro.org>
14756 Alan Hayward <alan.hayward@arm.com>
14757 David Sherwood <david.sherwood@arm.com>
14759 * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14760 INTEGER_CST rather than a negative test for ADDR_EXPR.
14762 2017-09-21 Richard Sandiford <richard.sandiford@linaro.org>
14763 Alan Hayward <alan.hayward@arm.com>
14764 David Sherwood <david.sherwood@arm.com>
14766 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14767 int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14769 2017-09-21 Richard Biener <rguenther@suse.de>
14771 PR tree-optimization/71351
14772 * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14773 graphite_create_new_loop_guard): Remove, fold remaining parts
14775 (translate_isl_ast_node_for): ... here and simplify.
14777 2017-09-21 Jakub Jelinek <jakub@redhat.com>
14780 * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14781 with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14782 latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14783 alternative always use QI mode, for -Os imov (=R,R) alternative
14784 always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14787 2017-09-20 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14788 Jeff Law <law@redhat.com>
14790 * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14791 (allocate_stack_space): New function, partially extracted from
14792 s390_emit_prologue.
14793 (s390_emit_prologue): Track offset to most recent stack probe.
14794 Code to allocate space moved into allocate_stack_space.
14795 Dump actions when no stack is allocated.
14796 (s390_prologue_plus_offset): New function.
14797 (s390_emit_stack_probe): Likewise.
14799 2017-09-20 Alexandre Oliva <aoliva@redhat.com>
14801 * common.opt (Wa, Wl, Wp, g, gz=): Add
14803 (gno-column-info): Remove.
14804 (gcolumn-info): Drop RejectNegative.
14805 (gno-): New prefix.
14806 (gno-record-gcc-switches): Remove.
14807 (grecord-gcc-switches): Drop RejectNegative.
14808 (gno-split-dwarf): Remove.
14809 (gsplit-dwarf): Drop RejectNegative.
14810 (gno-strict-dwarf): Remove.
14811 (gstrict-dwarf): Drop RejectNegative.
14812 * config/darwin.opt (gfull, gused): Add RejectNegative.
14813 * dwarf2out.c (gen_producer_string): Drop
14814 gno-record-gcc-switches handler.
14815 * optc-gen.awk: Add g to prefixes with negative forms.
14816 * opts-common.c (remapping_prefix_p): New.
14817 (find_opt): Check it.
14818 (generate_canonical_option): Test g prefix.
14819 (option_map): Add -gno- mapping.
14820 (add_misspelling_candidates): Check remapping_prefix_p.
14822 2017-09-20 Jeff Law <law@redhat.com>
14824 * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14825 thinko in stack clash protection support.
14827 * explow.c (compute_stack_clash_protection_loop_data): Use
14828 CONST_INT_P instead of explicit test. Verify object is a
14829 CONST_INT_P before looking at INTVAL.
14830 (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14831 instead of explicit test.
14833 2017-09-20 Segher Boessenkool <segher@kernel.crashing.org>
14836 * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14837 address instead of to r1 and r11.
14839 2017-09-20 Sebastian Peryt <sebastian.peryt@intel.com>
14841 * config.gcc: Support "knm".
14842 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14843 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14845 * config/i386/i386.c (m_KNM): Define.
14846 (processor_target_table): Add "knm".
14848 (ix86_option_override_internal): Add "knm".
14849 (ix86_issue_rate): Add PROCESSOR_KNM.
14850 (ix86_adjust_cost): Ditto.
14851 (ia32_multipass_dfa_lookahead): Ditto.
14852 (get_builtin_code_for_version): Handle PROCESSOR_KNM.
14853 (fold_builtin_cpu): Add M_INTEL_KNM.
14854 * config/i386/i386.h (processor_costs): Define TARGET_KNM.
14855 (processor_type): Add PROCESSOR_KNM.
14856 * config/i386/x86-tune.def: Add m_KNM.
14857 * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
14859 2017-09-20 Richard Biener <rguenther@suse.de>
14861 PR tree-optimization/80213
14862 * graphite-scop-detection.c (trivially_empty_bb_p): Labels
14863 are allowed in empty BBs as well.
14864 (canonicalize_loop_closed_ssa): Also look for other complex
14866 (scop_detection::get_sese): Include the loop-closed PHI block
14868 (scop_detection::merge_sese): Remove code adding extra blocks.
14869 (scop_detection::region_has_one_loop): Adjust for get_sese changes.
14870 (build_scops): Assert the final returned scop is invalid.
14872 2017-09-20 Richard Biener <rguenther@suse.de>
14874 PR tree-optimization/82264
14875 * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
14877 (vn_phi_lookup): Likewise.
14878 (vn_phi_insert): Likewise.
14880 2017-09-20 Jakub Jelinek <jakub@redhat.com>
14882 * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
14883 that fits into uhwi or shwi, add DW_AT_const_value regardless
14884 of early_dwarf without going through RTL, using add_AT_unsigned
14887 * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
14888 (DEBUG_LTO_ABBREV_SECTION): Likewise.
14889 (DEBUG_LTO_MACINFO_SECTION): Likewise.
14890 (DEBUG_MACRO_SECTION): Likewise.
14891 (DEBUG_LTO_MACRO_SECTION): Likewise.
14892 (DEBUG_STR_DWO_SECTION): Likewise.
14893 (DEBUG_LTO_STR_DWO_SECTION): Likewise.
14894 (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
14895 (DEBUG_LTO_DWO_LINE_SECTION): Define.
14896 (DEBUG_LTO_LINE_STR_SECTION): Define.
14897 (init_sections_and_labels): Initialize debug_line_str_section
14898 variable. Initialize debug_loc_section for -gdwarf-5 to
14899 DEBUG_LOCLISTS_SECTION. Formatting fixes.
14901 2017-09-20 Richard Biener <rguenther@suse.de>
14903 * graphite-sese-to-poly.c (extract_affine): Properly handle
14904 POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
14906 2017-09-20 Richard Biener <rguenther@suse.de>
14908 PR tree-optimization/81373
14909 * graphite-scop-detection.c (build_cross_bb_scalars_def):
14910 Force SESE live-out defs to be handled even if they are
14913 2017-09-19 Jeff Law <law@redhat.com>
14915 * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
14916 nothing for stack adjustments with REG_STACK_CHECK.
14917 * sched-deps.c (parse_add_or_inc): Reject insns with
14918 REG_STACK_CHECK from dependency breaking.
14919 * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
14920 (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
14921 * reg-notes.def (STACK_CHECK): New note.
14923 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
14924 (ix86_expand_prologue): Dump stack clash info as needed.
14925 Call ix86_adjust_stack_and_probe_stack_clash as needed.
14927 * function.c (dump_stack_clash_frame_info): New function.
14928 * function.h (dump_stack_clash_frame_info): Prototype.
14929 (enum stack_clash_probes): New enum.
14931 * config/alpha/alpha.c (alpha_expand_prologue): Also check
14932 flag_stack_clash_protection.
14933 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
14934 (arm_expand_prologue, thumb1_expand_prologue): Likewise.
14935 (arm_frame_pointer_required): Likewise.
14936 * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
14937 (ia64_expand_prologue): Likewise.
14938 * config/mips/mips.c (mips_expand_prologue): Likewise.
14939 * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
14940 * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
14941 (sparc_flat_expand_prologue): Likewise.
14942 * config/spu/spu.c (spu_expand_prologue): Likewise.
14944 * explow.c: Include "params.h".
14945 (anti_adjust_stack_and_probe_stack_clash): New function.
14946 (get_stack_check_protect): Likewise.
14947 (compute_stack_clash_protection_loop_data): Likewise.
14948 (emit_stack_clash_protection_loop_start): Likewise.
14949 (emit_stack_clash_protection_loop_end): Likewise.
14950 (allocate_dynamic_stack_space): Use get_stack_check_protect.
14951 Use anti_adjust_stack_and_probe_stack_clash.
14952 * explow.h (compute_stack_clash_protection_loop_data): Prototype.
14953 (emit_stack_clash_protection_loop_start): Likewise.
14954 (emit_stack_clash_protection_loop_end): Likewise.
14955 * rtl.h (get_stack_check_protect): Prototype.
14956 * target.def (stack_clash_protection_final_dynamic_probe): New hook.
14957 * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
14958 * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
14960 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
14962 * doc/tm.texi: Rebuilt.
14963 * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
14964 get_stack_check_protect.
14965 * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
14966 * config/arm/arm.c (arm_expand_prologue): Likewise.
14967 (arm_frame_pointer_required): Likewise.
14968 * config/i386/i386.c (ix86_expand_prologue): Likewise.
14969 * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
14970 * config/mips/mips.c (mips_expand_prologue): Likewise.
14971 * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
14972 * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
14973 * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
14974 (sparc_flat_expand_prologue): Likewise.
14976 * common.opt (-fstack-clash-protection): New option.
14977 * flag-types.h (enum stack_check_type): Note difference between
14978 -fstack-check= and -fstack-clash-protection.
14979 * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
14980 (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
14981 * toplev.c (process_options): Issue warnings/errors for cases
14982 not handled with -fstack-clash-protection.
14983 * doc/invoke.texi (-fstack-clash-protection): Document new option.
14984 (-fstack-check): Note additional problem with -fstack-check=generic.
14985 Note that -fstack-check is primarily for Ada and refer users
14986 to -fstack-clash-protection for stack-clash-protection.
14987 Document new params for stack clash protection.
14989 2017-09-19 Uros Bizjak <ubizjak@gmail.com>
14991 * config/i386/i386.c (ix86_split_long_move): Do not handle
14992 address used for LEA in a special way.
14994 2017-09-19 Segher Boessenkool <segher@kernel.crashing.org>
14996 * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
14998 2017-09-19 Martin Sebor <msebor@redhat.com>
15001 * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15002 of incompatible types.
15004 2017-09-19 Will Schmidt <will_schmidt@vnet.ibm.com>
15006 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15007 for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15008 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15009 Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15011 2017-09-19 Richard Biener <rguenther@suse.de>
15013 PR tree-optimization/82244
15014 * tree-vrp.c (remove_range_assertions): Do not propagate
15015 a constant to abnormals but replace the assert with a copy.
15017 2017-09-19 Alexander Monakov <amonakov@ispras.ru>
15019 PR rtl-optimization/57878
15020 PR rtl-optimization/68988
15021 * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15022 avoidance test involving non_reload_pseudos. Move frequency test
15023 below the general fragmentation avoidance test.
15025 2017-09-19 Richard Biener <rguenther@suse.de>
15027 PR tree-optimization/69728
15028 * graphite-sese-to-poly.c (schedule_error): New global.
15029 (add_loop_schedule): Handle empty domain by failing the
15031 (build_original_schedule): Handle schedule_error.
15033 2017-09-19 Richard Biener <rguenther@suse.de>
15035 * graphite-scop-detection.c (scop_detection::can_represent_loop):
15036 Do not iterate to sibling loops but only to siblings of inner
15039 2017-09-18 Andreas Schwab <schwab@linux-m68k.org>
15042 * config/m68k/m68k.md (moveq feeding equality comparison): Check
15043 that the registers are different.
15045 2017-09-18 Uros Bizjak <ubizjak@gmail.com>
15047 * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15048 to processor_model and "amdfam17h" to arch_names_table.
15049 * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15051 2017-09-18 Jakub Jelinek <jakub@redhat.com>
15054 * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15056 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15057 Alan Hayward <alan.hayward@arm.com>
15058 David Sherwood <david.sherwood@arm.com>
15060 * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15062 * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15064 * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15065 parameter. Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15066 vect_schedule_slp_instance.
15067 (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15068 Update call to vect_slp_analyze_node_operations. Simplify return
15070 (vect_slp_analyze_bb_1): Update call accordingly.
15071 (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15072 Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15073 (vect_schedule_slp): Update call accordingly.
15075 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15076 Alan Hayward <alan.hayward@arm.com>
15077 David Sherwood <david.sherwood@arm.com>
15079 * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15080 with types that aren't in fact scalar.
15082 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15084 * tree-vect-slp.c (vect_record_max_nunits): New function,
15086 (vect_build_slp_tree_1): ...here.
15087 (vect_build_slp_tree_2): Call it for phis too.
15089 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15091 * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15092 to vect_get_vec_def_for_operand when getting the mask operand.
15094 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15095 Alan Hayward <alan.hayward@arm.com>
15096 David Sherwood <david.sherwood@arm.com>
15098 * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15101 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15102 Alan Hayward <alan.hayward@arm.com>
15103 David Sherwood <david.sherwood@arm.com>
15105 * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15106 calculation for vector booleans.
15108 2017-09-18 Richard Sandiford <richard.sandiford@linaro.org>
15109 Alan Hayward <alan.hayward@arm.com>
15110 David Sherwood <david.sherwood@arm.com>
15112 * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15114 (vect_transform_stmt): ...here.
15115 (vect_analyze_stmt): Use it instead of calling
15116 vectorizable_live_operation directly.
15118 2017-09-18 Cesar Philippidis <cesar@codesourcery.com>
15120 * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15121 non-SIMT targets in acc vector loops.
15123 2017-09-18 Claudiu Zissulescu <claziss@synopsys.com>
15125 * configure.ac: Add arc and check if assembler supports gdwarf2.
15126 * configure: Regenerate.
15128 2017-09-18 Richard Biener <rguenther@suse.de>
15130 PR tree-optimization/82220
15131 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15132 epilogue niters from the min_profitable_iters compute.
15134 2017-09-18 Jakub Jelinek <jakub@redhat.com>
15137 * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15138 changes. Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15139 (ix86_init_pic_reg): Revert 2017-09-01 changes.
15141 2017-09-18 Eric Botcazou <ebotcazou@adacore.com>
15144 * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15145 switching to a new text section.
15147 2017-09-18 Richard Biener <rguenther@suse.de>
15149 * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15151 (build_alias_set): Reject aliases with no access function.
15153 2017-09-18 Richard Biener <rguenther@suse.de>
15155 PR tree-optimization/79622
15156 * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15158 (build_cross_bb_scalars_use): Likewise.
15160 2017-09-18 Pierre-Marie de Rodat <derodat@adacore.com>
15162 * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15164 2017-09-18 Alan Modra <amodra@gmail.com>
15167 * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15168 stack_pointer_rtx for count 0. Update comments. Break up
15169 large rtl expression.
15171 2017-09-17 Daniel Santos <daniel.santos@pobox.com>
15173 * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15174 Increase to 20 bytes.
15175 (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15176 (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15177 or avx version of the stub.
15179 2017-09-17 H.J. Lu <hongjiu.lu@intel.com>
15182 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15183 compute the minimum stack alignment. Also update preferred stack
15184 boundary for leaf functions.
15186 2017-09-16 Richard Sandiford <richard.sandiford@linaro.org>
15188 PR tree-optimization/82228
15189 * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15192 2017-09-16 Chung-Ju Wu <jasonwucj@gmail.com>
15194 * common/config/nds32/nds32-common.c
15195 (nds32_option_optimization_table): Refine formatting.
15196 (nds32_option_optimization_table): Use -fsched-pressure and
15197 -fomit-frame-pointer for specific optimization level.
15199 2017-09-16 Chung-Ju Wu <jasonwucj@gmail.com>
15201 * config/nds32/nds32.c: Refine formatting and comments.
15202 * config/nds32/nds32.h: Likewise.
15203 * config/nds32/nds32.md: Likewise.
15204 * config/nds32/nds32-cost.c: Likewise.
15205 * config/nds32/nds32-isr.c: Likewise.
15206 * config/nds32/nds32-md-auxiliary.c: Likewise.
15207 * config/nds32/nds32-multiple.md: Likewise.
15208 * config/nds32/nds32-predicates.c: Likewise.
15210 2017-09-15 Andrew Sutton <andrew.n.sutton@gmail.com>
15211 Jakub Jelinek <jakub@redhat.com>
15213 Add support for -std=c++2a.
15214 * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15216 * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15218 2017-09-15 Steve Ellcey <sellcey@cavium.com>
15221 * doc/extend.texi (Common Function Attributes): Add
15222 references to ARM, AArch64, and S/390 specific attributes.
15223 (Function Specific Option Pragmas): Add AArch64 and S/390
15224 to list of back ends that support the target pragma.
15226 2017-09-15 Nathan Sidwell <nathan@acm.org>
15228 * doc/standards.texi: Fix C++17 description. Update URLs for
15231 2017-09-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
15233 * common.opt (Wcast-align=strict): New warning option.
15234 * doc/invoke.texi: Document -Wcast-align=strict.
15236 2017-09-15 Pierre-Marie de Rodat <derodat@adacore.com>
15238 * cgraph.h (cgraph_thunk_info): Add comments.
15239 * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15240 assert for VIRTUAL_* arguments stricter.
15242 2017-09-15 Jackson Woodruff <jackson.woodruff@arm.com>
15244 PR tree-optimization/71026
15245 * match.pd: Move RDIV patterns from fold-const.c
15246 * fold-const.c (distribute_real_division): Removed.
15247 (fold_binary_loc): Remove calls to distribute_real_divison.
15249 2017-09-15 Jakub Jelinek <jakub@redhat.com>
15251 * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15252 c++1z and gnu++1z as deprecated. Change other references to
15253 -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15254 Change -Wc++1z-compat to -Wc++17-compat.
15255 * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15256 * dwarf2out.c (highest_c_language): Handle C++17.
15257 (gen_compile_unit_die): Likewise.
15259 2017-09-15 Jakub Jelinek <jakub@redhat.com>
15261 PR rtl-optimization/82192
15262 * combine.c (make_extraction): Don't look through non-paradoxical
15263 SUBREGs or TRUNCATE if pos + len is or might be bigger than
15266 2017-09-15 Richard Sandiford <richard.sandiford@linaro.org>
15267 Alan Hayward <alan.hayward@arm.com>
15268 David Sherwood <david.sherwood@arm.com>
15270 * target.def (function_arg_offset): New hook.
15271 * targhooks.h (default_function_arg_offset): Declare.
15272 * targhooks.c (default_function_arg_offset): New function.
15273 * function.c (locate_and_pad_parm): Use
15274 targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15275 * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15276 (TARGET_FUNCTION_ARG_OFFSET): ...this.
15277 * doc/tm.texi: Regenerate.
15278 * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15279 * config/spu/spu.c (spu_function_arg_offset): New function.
15280 (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15281 * system.h (FUNCTION_ARG_OFFSET): Poison.
15283 2017-09-15 Richard Sandiford <richard.sandiford@linaro.org>
15284 Alan Hayard <alan.hayward@arm.com>
15285 David Sherwood <david.sherwood@arm.com>
15287 * target.def (truly_noop_truncation): New hook.
15288 (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15289 than TRULY_NOOP_TRUNCATION.
15290 * hooks.h (hook_bool_uint_uint_true): Declare.
15291 * hooks.c (hook_bool_uint_uint_true): New function.
15292 * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15293 (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15294 * doc/tm.texi: Regenerate.
15295 * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15296 rather than TRULY_NOOP_TRUNCATION in comments.
15297 (simplify_comparison): Likewise.
15298 (record_truncated_value): Likewise.
15299 * expmed.c (extract_bit_field_1): Likewise.
15300 (extract_split_bit_field): Likewise.
15301 * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15302 instead of TRULY_NOOP_TRUNCATION.
15303 * function.c (assign_parm_setup_block): Likewise.
15304 * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15305 * rtlhooks.c: Include target.h.
15306 * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15307 * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15308 * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15309 * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15310 * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15311 * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15312 * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15313 * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15314 * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15315 * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15316 * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15317 * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15318 * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15319 * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15320 * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15321 * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15322 * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15323 * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15324 * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15325 * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15326 * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15327 * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15328 * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15329 * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15330 * config/mips/mips.c (mips_truly_noop_truncation): New function.
15331 (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15332 * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15333 rather than TRULY_NOOP_TRUNCATION in comments.
15334 * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15335 * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15336 * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15337 * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15338 * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15339 * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15340 * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15341 * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15342 * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15343 * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15344 * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15345 * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15346 rather than TRULY_NOOP_TRUNCATION in comments.
15347 * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15348 * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15349 * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15350 * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15351 * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15352 TRULY_NOOP_TRUNCATION condition.
15353 (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15354 (TRULY_NOOP_TRUNCATION): Delete.
15355 * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15356 * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15357 * config/spu/spu.c (spu_truly_noop_truncation): New function.
15358 (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15359 * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15360 * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15361 * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15362 (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15363 * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15364 rather than TRULY_NOOP_TRUNCATION in comments.
15365 * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15366 * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15367 * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15368 * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15369 * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15370 * system.h (TRULY_NOOP_TRUNCATION): Poison.
15372 2017-09-15 Christophe Lyon <christophe.lyon@linaro.org>
15375 * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15376 (*cmp_ior): Likewise.
15377 (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15378 (*ior_scc_scc_cmp): Likewise.
15379 (*and_scc_scc): Likewise.
15380 (*and_scc_scc_cmp): Likewise.
15382 2017-09-15 Richard Sandiford <richard.sandiford@linaro.org>
15383 Alan Hayard <alan.hayward@arm.com>
15384 David Sherwood <david.sherwood@arm.com>
15386 * target.def (can_change_mode_class): New hook.
15387 (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15388 (hard_regno_nregs): Likewise.
15389 * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15390 * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15391 * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15392 (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15393 (LOAD_EXTEND_OP): Update accordingly.
15394 * doc/tm.texi: Regenerate.
15395 * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15396 CANNOT_CHANGE_MODE_CLASS.
15397 * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15398 (REG_CAN_CHANGE_MODE_P): ...this new macro.
15399 * combine.c (simplify_set): Update accordingly.
15400 * emit-rtl.c (validate_subreg): Likewise.
15401 * recog.c (general_operand): Likewise.
15402 * regcprop.c (mode_change_ok): Likewise.
15403 * reload1.c (choose_reload_regs): Likewise.
15404 (inherit_piecemeal_p): Likewise.
15405 * rtlanal.c (simplify_subreg_regno): Likewise.
15406 * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15407 instead of CANNOT_CHANGE_MODE_CLASS.
15408 (reload_cse_simplify_operands): Likewise.
15409 * reload.c (push_reload): Use targetm.can_change_mode_class
15410 instead of CANNOT_CHANGE_MODE_CLASS.
15411 (push_reload): Likewise. Also use REG_CAN_CHANGE_MODE_P instead of
15412 REG_CANNOT_CHANGE_MODE_P.
15413 * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15414 * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15415 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15416 * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15417 * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15418 (arm_can_change_mode_class): New function.
15419 * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15420 than CANNOT_CHANGE_MODE_CLASS in comments.
15421 * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15422 * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15423 * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15424 (ix86_can_change_mode_class): ...this new function, inverting the
15425 sense of the return value.
15426 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15427 * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15428 * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15429 (ia64_can_change_mode_class): New function.
15430 * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15431 * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15432 * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15433 (m32c_can_change_mode_class): ...this new function, inverting the
15434 sense of the return value.
15435 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15436 * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15437 * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15438 * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15439 (mips_can_change_mode_class): ...this new function, inverting the
15440 sense of the return value.
15441 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15442 * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15443 * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15444 (msp430_can_change_mode_class): New function.
15445 * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15446 * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15447 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15448 * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15449 * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15450 * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15451 * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15452 (pa_cannot_change_mode_class): Replace with...
15453 (pa_can_change_mode_class): ...this new function, inverting the
15454 sense of the return value.
15455 (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15456 than CANNOT_CHANGE_MODE_CLASS in comments.
15457 * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15458 * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15459 * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15460 (pdp11_cannot_change_mode_class): Replace with...
15461 (pdp11_can_change_mode_class): ...this new function, inverting the
15462 sense of the return value.
15463 * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15464 * config/powerpcspe/powerpcspe-protos.h
15465 (rs6000_cannot_change_mode_class_ptr): Delete.
15466 * config/powerpcspe/powerpcspe.c
15467 (rs6000_cannot_change_mode_class_ptr): Delete.
15468 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15469 (rs6000_option_override_internal): Assign to
15470 targetm.can_change_mode_class instead of
15471 rs6000_cannot_change_mode_class_ptr.
15472 (rs6000_cannot_change_mode_class): Replace with...
15473 (rs6000_can_change_mode_class): ...this new function, inverting the
15474 sense of the return value.
15475 (rs6000_debug_cannot_change_mode_class): Replace with...
15476 (rs6000_debug_can_change_mode_class): ...this new function.
15477 * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15478 * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15479 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15480 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15481 * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15483 * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15484 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15485 (rs6000_option_override_internal): Assign to
15486 targetm.can_change_mode_class instead of
15487 rs6000_cannot_change_mode_class_ptr.
15488 (rs6000_cannot_change_mode_class): Replace with...
15489 (rs6000_can_change_mode_class): ...this new function, inverting the
15490 sense of the return value.
15491 (rs6000_debug_cannot_change_mode_class): Replace with...
15492 (rs6000_debug_can_change_mode_class): ...this new function.
15493 * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15494 * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15495 * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15496 (s390_can_change_mode_class): ...this new function, inverting the
15497 sense of the return value.
15498 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15499 * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15500 * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15501 * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15502 (sh_cannot_change_mode_class): Replace with...
15503 (sh_can_change_mode_class): ...this new function, inverting the
15504 sense of the return value.
15505 * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15506 * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15507 (sparc_can_change_mode_class): New function.
15508 * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15509 * config/spu/spu.c (spu_can_change_mode_class): New function.
15510 (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15511 * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15512 * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15513 (visium_can_change_mode_class): New function.
15514 * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15516 2017-09-15 Richard Biener <rguenther@suse.de>
15518 PR tree-optimization/82217
15519 * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15520 but not undefined case.
15522 2017-09-15 Jakub Jelinek <jakub@redhat.com>
15525 * postreload.c (reload_cse_simplify_operands): Skip
15526 NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15528 2017-09-15 Richard Biener <rguenther@suse.de>
15530 PR tree-optimization/68823
15531 * graphite-scop-detection.c (build_alias_set): If we have a
15532 possible dependence check whether we can handle them by just
15533 looking at the DRs DR_ACCESS_FNs.
15534 (build_scops): If build_alias_set fails, fail the SCOP.
15536 2017-09-14 Michael Meissner <meissner@linux.vnet.ibm.com>
15538 * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15539 to support float128 built-in functions that require the ISA 3.0
15541 (BU_FLOAT128_3_HW): Likewise.
15542 (SQRTF128): Add support for the IEEE 128-bit square root and fma
15543 built-in functions.
15544 (FMAF128): Likewise.
15546 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15547 support for built-in functions that need the ISA 3.0 IEEE 128-bit
15548 floating point instructions.
15549 (rs6000_invalid_builtin): Likewise.
15550 (rs6000_builtin_mask_names): Likewise.
15551 * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15552 (RS6000_BTM_FLOAT128_HW): Likewise.
15553 (RS6000_BTM_COMMON): Likewise.
15554 * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15556 * doc/extend.texi (RS/6000 built-in functions): Document the
15557 IEEE 128-bit floating point square root and fused multiply-add
15558 built-in functions.
15560 2017-09-14 Pat Haugen <pthaugen@us.ibm.com>
15562 * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15563 reg (r2) isn't in the set of registers defined in the prologue.
15565 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15566 Alan Hayward <alan.hayward@arm.com>
15567 David Sherwood <david.sherwood@arm.com>
15569 * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15570 (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15571 (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15572 (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15573 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15575 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15576 max_vectorization_factor.
15577 (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15579 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15580 Alan Hayward <alan.hayward@arm.com>
15581 David Sherwood <david.sherwood@arm.com>
15583 * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15584 (vect_worthwhile_without_simd_p): Declare.
15585 * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15586 (vectorizable_reduction): Use it.
15587 * tree-vect-stmts.c (vectorizable_shift): Likewise.
15588 (vectorizable_operation): Likewise.
15590 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15591 Alan Hayward <alan.hayward@arm.com>
15592 David Sherwood <david.sherwood@arm.com>
15594 * tree-vectorizer.h (vect_get_num_copies): New function.
15595 * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15596 * tree-vect-loop.c (vectorizable_reduction): Likewise.
15597 (vectorizable_induction): Likewise.
15598 (vectorizable_live_operation): Likewise.
15599 * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15600 (vectorizable_bswap): Likewise.
15601 (vectorizable_call): Likewise.
15602 (vectorizable_conversion): Likewise.
15603 (vectorizable_assignment): Likewise.
15604 (vectorizable_shift): Likewise.
15605 (vectorizable_operation): Likewise.
15606 (vectorizable_store): Likewise.
15607 (vectorizable_load): Likewise.
15608 (vectorizable_condition): Likewise.
15609 (vectorizable_comparison): Likewise.
15610 (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15612 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15613 Alan Hayward <alan.hayward@arm.com>
15614 David Sherwood <david.sherwood@arm.com>
15616 * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15617 of vect_init_vector.
15619 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15620 Alan Hayward <alan.hayward@arm.com>
15621 David Sherwood <david.sherwood@arm.com>
15623 * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15624 an inline wrapper that provides a location.
15625 (gimple_build_vector): Likewise.
15626 * gimple-fold.c (gimple_build_vector_from_val): New function.
15627 (gimple_build_vector): Likewise.
15628 * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15629 functions to build the initial value. Always return a gimple value.
15630 (get_initial_defs_for_reduction): Likewise. Only compute
15632 (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15633 vect_init_vector on the results from get_initial_def(s)_for_reduction.
15634 (vectorizable_induction): Use gimple_build_vector rather than
15637 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15638 Alan Hayward <alan.hayward@arm.com>
15639 David Sherwood <david.sherwood@arm.com>
15641 * target.h (vec_perm_indices): New typedef.
15642 (auto_vec_perm_indices): Likewise.
15643 * optabs-query.h: Include target.h
15644 (can_vec_perm_p): Take a vec_perm_indices *.
15645 * optabs-query.c (can_vec_perm_p): Likewise.
15646 (can_mult_highpart_p): Update accordingly. Use auto_vec_perm_indices.
15647 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15648 * tree-vect-generic.c (lower_vec_perm): Likewise.
15649 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15650 (vect_grouped_load_supported): Likewise.
15651 (vect_shift_permute_load_chain): Likewise.
15652 (vect_permute_store_chain): Use auto_vec_perm_indices.
15653 (vect_permute_load_chain): Likewise.
15654 * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15655 (fold_ternary_loc): Update accordingly. Use auto_vec_perm_indices.
15656 Update uses of can_vec_perm_p.
15657 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15658 mode with a number of elements. Take a vec_perm_indices *.
15659 (vect_create_epilog_for_reduction): Update accordingly.
15660 Use auto_vec_perm_indices.
15661 (have_whole_vector_shift): Likewise. Update call to can_vec_perm_p.
15662 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15663 (vect_transform_slp_perm_load): Likewise.
15664 (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15665 * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15666 (vect_gen_perm_mask_checked): Likewise.
15667 * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15668 (vect_gen_perm_mask_checked): Likewise.
15669 (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15670 (vectorizable_store): Likewise.
15671 (vectorizable_load): Likewise.
15672 (perm_mask_for_reverse): Likewise. Update call to can_vec_perm_p.
15673 (vectorizable_bswap): Likewise.
15675 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15676 Alan Hayward <alan.hayward@arm.com>
15677 David Sherwood <david.sherwood@arm.com>
15679 * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15680 * tree.c (build_vector): Likewise.
15681 (build_vector_from_ctor): Update accordingly.
15682 (build_vector_from_val): Likewise.
15683 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15684 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15685 * tree-vect-generic.c (add_rshift): Likewise.
15686 (expand_vector_divmod): Likewise.
15687 (optimize_vector_constructor): Likewise.
15688 * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15689 (vect_transform_slp_perm_load): Likewise.
15690 (vect_schedule_slp_instance): Likewise.
15691 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15692 (vectorizable_call): Likewise.
15693 (vect_gen_perm_mask_any): Likewise. Add elements in order.
15694 * expmed.c (make_tree): Likewise.
15695 * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15696 a vector passed to build_vector.
15697 (fold_convert_const): Likewise.
15698 (exact_inverse): Likewise.
15699 (fold_ternary_loc): Likewise.
15700 (fold_relational_const): Likewise.
15701 (const_binop): Likewise. Use VECTOR_CST_ELT directly when operating
15702 on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15703 (const_unop): Likewise. Store the reduction accumulator in a
15704 variable rather than an array.
15705 (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15706 (fold_vec_perm): Update calls accordingly. Use auto_vec<tree> for
15707 the new vector, rather than constructing it after the input arrays.
15708 (native_interpret_vector): Use auto_vec<tree> when building
15709 a vector passed to build_vector. Add elements in order.
15710 * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15711 auto_vec<tree> when building a vector passed to build_vector.
15712 (vect_create_epilog_for_reduction): Likewise.
15713 (vectorizable_induction): Likewise.
15714 (get_initial_def_for_reduction): Likewise. Fix indentation of
15716 * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15718 (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15719 passed to build_vector.
15721 2017-09-14 Richard Sandiford <richard.sandiford@linaro.org>
15722 Alan Hayward <alan.hayward@arm.com>
15723 David Sherwood <david.sherwood@arm.com>
15725 * tree-core.h (tree_base::u): Add an "nelts" field.
15726 (tree_vector): Use VECTOR_CST_NELTS as the length.
15727 * tree.c (tree_size): Likewise.
15728 (make_vector): Initialize VECTOR_CST_NELTS.
15729 * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15730 * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15731 TYPE_VECTOR_SUBPARTS.
15732 * expr.c (const_vector_mask_from_tree): Consistently use "units"
15733 as the number of units, setting it from VECTOR_CST_NELTS.
15734 (const_vector_from_tree): Likewise.
15735 * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15736 TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15737 (fold_negate_expr_1): Likewise.
15738 (fold_convert_const): Likewise.
15739 (const_binop): Likewise. Differentiate the number of output and
15741 (const_unop): Likewise.
15742 (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15743 in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15744 in cases that did the opposite.
15746 2017-09-14 Richard Biener <rguenther@suse.de>
15748 * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15751 2017-09-14 Eric Botcazou <ebotcazou@adacore.com>
15753 * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15755 2017-09-14 Jakub Jelinek <jakub@redhat.com>
15758 * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15759 if and where to split a bb, except for splitting before debug insn
15760 sequences followed by non-label real insn. Delete debug insns
15761 in between basic blocks.
15763 * combine.c (make_compound_operation_int): Formatting fixes.
15765 * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15766 * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15767 * config/netbsd.h (LINK_EH_SPEC): Likewise.
15768 * config/sol2.h (LINK_EH_SPEC): Likewise.
15769 * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15770 * config/s390/linux.h (LINK_SPEC): Likewise.
15771 * config/freebsd.h (LINK_EH_SPEC): Likewise.
15772 * config/openbsd.h (LINK_EH_SPEC): Likewise.
15773 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15774 * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15775 * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15776 * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15777 * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15778 * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15780 2017-09-13 Jakub Jelinek <jakub@redhat.com>
15782 * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15784 (ENDFILE_LINUX_SPEC): Likewise.
15785 (LINK_EH_SPEC): Likewise.
15786 * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15787 (LINK_OS_LINUX_SPEC32): Likewise.
15788 (LINK_OS_LINUX_SPEC64): Likewise.
15789 * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15790 (LINK_OS_LINUX_SPEC): Likewise.
15792 2017-09-13 Martin Liska <mliska@suse.cz>
15794 PR middle-end/82154
15795 * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15796 CASE_HIGH is NULL_TREE.
15798 2017-09-13 Richard Sandiford <richard.sandiford@linaro.org>
15799 Alan Hayward <alan.hayward@arm.com>
15800 David Sherwood <david.sherwood@arm.com>
15802 * target.def (secondary_memory_needed): New hook.
15803 (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15804 instead of SECONDARY_MEMORY_NEEDED.
15805 (secondary_memory_needed_mode): Likewise.
15806 * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15807 * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15808 * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15809 (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15810 (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15811 * doc/tm.texi: Regenerate.
15812 * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15813 * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15814 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15815 * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15816 * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15817 * config/i386/i386.c (inline_secondary_memory_needed): Put the
15818 mode argument first and change the reg_class arguments to reg_class_t.
15819 (ix86_secondary_memory_needed): Likewise. Remove the strict parameter.
15820 Make static. Update the call to inline_secondary_memory_needed.
15821 (ix86_register_move_cost): Update the call to
15822 inline_secondary_memory_needed.
15823 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15824 * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15826 * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15827 TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15829 * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15830 * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15831 * config/mips/mips.c (mips_secondary_memory_needed): Make static
15832 and match hook interface. Add comment from mips.h.
15833 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15834 * config/mmix/mmix.md (truncdfsf2): Refer to
15835 TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15837 * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15838 (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15839 * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15840 (pa_secondary_memory_needed): New function.
15841 * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15842 * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15843 * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15844 (pdp11_secondary_memory_needed): Make static and match hook interface.
15845 * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
15846 * config/powerpcspe/powerpcspe-protos.h
15847 (rs6000_secondary_memory_needed_ptr): Delete.
15848 * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
15850 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15851 (rs6000_option_override_internal): Assign to
15852 targetm.secondary_memory_needed rather than
15853 rs6000_secondary_memory_needed_ptr.
15854 (rs6000_secondary_memory_needed): Match hook interface.
15855 (rs6000_debug_secondary_memory_needed): Likewise.
15856 * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
15857 * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
15858 (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
15859 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15860 * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
15861 * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
15863 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
15864 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15865 (rs6000_option_override_internal): Assign to
15866 targetm.secondary_memory_needed rather than
15867 rs6000_secondary_memory_needed_ptr.
15868 (rs6000_secondary_memory_needed): Match hook interface.
15869 (rs6000_debug_secondary_memory_needed): Likewise.
15870 * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
15871 * config/s390/s390.c (s390_secondary_memory_needed): New function.
15872 (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15873 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
15874 * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15875 (sparc_secondary_memory_needed): New function.
15876 * lra-constraints.c (check_and_process_move): Refer to
15877 TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15879 (curr_insn_transform): Likewise.
15880 (process_alt_operands): Use targetm.secondary_memory_needed
15881 instead of TARGET_SECONDARY_MEMORY_NEEDED.
15882 (check_secondary_memory_needed_p): Likewise.
15883 (choose_split_class): Likewise.
15884 * reload.c: Unconditionally include code that was previously
15885 conditional on SECONDARY_MEMORY_NEEDED.
15886 (push_secondary_reload): Use targetm.secondary_memory_needed
15887 instead of TARGET_SECONDARY_MEMORY_NEEDED.
15888 (push_reload): Likewise.
15889 * reload1.c: Unconditionally include code that was previously
15890 conditional on SECONDARY_MEMORY_NEEDED.
15891 (choose_reload_regs): Use targetm.secondary_memory_needed
15892 instead of TARGET_SECONDARY_MEMORY_NEEDED.
15893 (gen_reload): Likewise.
15894 * system.h (SECONDARY_MEMORY_NEEDED): Poison.
15896 2017-09-13 Richard Sandiford <richard.sandiford@linaro.org>
15897 Alan Hayward <alan.hayward@arm.com>
15898 David Sherwood <david.sherwood@arm.com>
15900 * target.def (secondary_memory_needed_mode): New hook:
15901 * targhooks.c (default_secondary_memory_needed_mode): Declare.
15902 * targhooks.h (default_secondary_memory_needed_mode): New function.
15903 * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
15904 (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
15905 * doc/tm.texi: Regenerate.
15906 * lra-constraints.c (check_and_process_move): Use
15907 targetm.secondary_memory_needed_mode instead of
15908 TARGET_SECONDARY_MEMORY_NEEDED_MODE.
15909 (curr_insn_transform): Likewise.
15910 * reload.c (get_secondary_mem): Likewise.
15911 * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15912 * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
15914 (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15915 * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15916 * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
15917 (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15918 * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
15920 * config/powerpcspe/powerpcspe-protos.h
15921 (rs6000_secondary_memory_needed_mode): Delete.
15922 * config/powerpcspe/powerpcspe.c
15923 (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15924 (rs6000_secondary_memory_needed_mode): Make static.
15925 * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15926 * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
15928 * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
15930 (rs6000_secondary_memory_needed_mode): Make static.
15931 * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15932 * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
15933 (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15934 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15935 * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
15937 (sparc_secondary_memory_needed_mode): New function.
15938 * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
15940 2017-09-13 Jackson Woodruff <jackson.woodruff@arm.com>
15942 * config/aarch64/constraints.md (Umq): New constraint.
15943 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
15945 (mov<mode>): Update condition.
15947 2017-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15949 * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
15950 when bitposition is the same.
15952 2017-09-13 Richard Biener <rguenther@suse.de>
15954 * dwarf2out.c (output_die_symbol): Remove.
15955 (output_die): Do not output a DIEs symbol.
15957 2017-09-13 Richard Biener <rguenther@suse.de>
15959 PR middle-end/82128
15960 * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
15961 default-def to avoid breaking iterator update with the weird
15962 interaction with cgraph_update_edges_for_call_stmt_node.
15964 2017-09-13 Richard Biener <rguenther@suse.de>
15966 * tree-cfg.c (verify_gimple_assign_binary): Add verification
15967 for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
15968 VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
15969 (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
15971 2017-09-13 Kugan Vivekanandarajah <kuganv@linaro.org>
15973 * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
15974 Disable pc relative literal load irrespective of
15975 TARGET_FIX_ERR_A53_84341 for default.
15977 2017-09-12 Eric Botcazou <ebotcazou@adacore.com>
15979 * config/sparc/sparc.c (output_return): Output the source location of
15980 the insn in the delay slot, if any.
15981 (output_sibcall): Likewise.
15983 2017-09-12 H.J. Lu <hongjiu.lu@intel.com>
15986 * common.opt (-static-pie): New alias.
15987 (shared): Negate static-pie.
15988 (-no-pie): Update help text.
15990 (static-pie): New option.
15991 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
15992 -static-pie support.
15993 (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
15994 (LINK_EH_SPEC): Likewise.
15995 (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15996 * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
15997 * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
15998 * gcc.c (LINK_COMMAND_SPEC): Likewise.
15999 (init_gcc_specs): Likewise.
16000 (init_spec): Likewise.
16001 (display_help): Update help message for -pie.
16002 * doc/invoke.texi: Update -pie, -no-pie and -static. Document
16005 2017-09-12 Wilco Dijkstra <wdijkstr@arm.com>
16007 * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16008 (movdi_aarch64): Likewise.
16009 (movti_aarch64): Likewise.
16011 2017-09-12 Simon Wright <simon@pushface.org>
16014 * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16015 calculation of the minor version, always output as 0.
16017 2017-09-12 Jakub Jelinek <jakub@redhat.com>
16020 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16021 ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16022 on it early, rather than manual conversion late. For
16023 ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16024 instead of performing manual conversion.
16026 2017-09-12 Carl Love <cel@us.ibm.com>
16028 * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16029 vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16031 * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16032 VMULOSW): Add definitions.
16033 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16034 ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16035 ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16036 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16037 builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16039 2017-09-12 James Greenhalgh <james.greenhalgh@arm.com>
16041 * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16043 (movti_aarch64): Likewise.
16044 (movdf_aarch64): Likewise.
16045 (movtf_aarch64): Likewise.
16046 (load_pairdi): Likewise.
16047 (store_pairdi): Likewise.
16048 (load_pairdf): Likewise.
16049 (store_pairdf): Likewise.
16050 (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16051 (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16052 (ldr_got_small_<mode>): Likewise.
16053 (ldr_got_small_28k_<mode>): Likewise.
16054 (ldr_got_tiny): Likewise.
16055 * config/aarch64/iterators.md (ldst_sz): New.
16056 (ldpstp_sz): Likewise.
16057 * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16059 (thunderx_load): Split load_8 to load_16.
16060 * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16062 (thunderx2t99_storepair_basic): Split store_8 to store_16.
16063 * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16064 (xgene1_store_pair): Split store_8 to store_16.
16065 * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16066 (falkor_st_0_st_sd): Split store_8 to store_16.
16068 2017-09-12 James Greenhalgh <james.greenhalgh@arm.com>
16070 * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16071 and store1/2/3/4 to store_4/8/12/16.
16072 * config/aarch64/aarch64.md: Update for rename.
16073 * config/arm/arm.md: Likewise.: Likewise.
16074 * config/arm/arm.c: Likewise.
16075 * config/arm/thumb1.md: Likewise.
16076 * config/arm/thumb2.md: Likewise.
16077 * config/arm/vfp.md: Likewise.
16078 * config/arm/arm-generic.md: Likewise.
16079 * config/arm/arm1020e.md: Likewise.
16080 * config/arm/arm1026ejs.md: Likewise.
16081 * config/arm/arm1136jfs.md: Likewise.
16082 * config/arm/arm926ejs.md: Likewise.
16083 * config/arm/cortex-a15.md: Likewise.
16084 * config/arm/cortex-a17.md: Likewise.
16085 * config/arm/cortex-a5.md: Likewise.
16086 * config/arm/cortex-a53.md: Likewise.
16087 * config/arm/cortex-a57.md: Likewise.
16088 * config/arm/cortex-a7.md: Likewise.
16089 * config/arm/cortex-a8.md: Likewise.
16090 * config/arm/cortex-a9.md: Likewise.
16091 * config/arm/cortex-m4.md: Likewise.
16092 * config/arm/cortex-m7.md: Likewise.
16093 * config/arm/cortex-r4.md: Likewise.
16094 * config/arm/exynos-m1.md: Likewise.
16095 * config/arm/fa526.md: Likewise.
16096 * config/arm/fa606te.md: Likewise.
16097 * config/arm/fa626te.md: Likewise.
16098 * config/arm/fa726te.md: Likewise.
16099 * config/arm/fmp626.md: Likewise.
16100 * config/arm/iwmmxt.md: Likewise.
16101 * config/arm/ldmstm.md: Likewise.
16102 * config/arm/marvell-pj4.md: Likewise.
16103 * config/arm/xgene1.md: Likewise.
16104 * config/aarch64/thunderx.md: Likewise.
16105 * config/aarch64/thunderx2t99.md: Likewise.
16106 * config/aarch64/falkor.md: Likewise.
16108 2017-09-12 Martin Liska <mliska@suse.cz>
16110 * attribs.c (private_lookup_attribute): New function.
16111 * attribs.h (private_lookup_attribute): Declared here.
16112 (lookup_attribute): Called from this place.
16114 2017-09-12 Richard Biener <rguenther@suse.de>
16116 PR tree-optimization/82157
16117 * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16118 stmts with side-effects.
16120 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16121 Alan Hayward <alan.hayward@arm.com>
16122 David Sherwood <david.sherwood@arm.com>
16124 * target.def (hard_regno_nregs): New hook.
16125 (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16126 * targhooks.h (default_hard_regno_nregs): Declare.
16127 * targhooks.c (default_hard_regno_nregs): New function.
16128 * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16129 (TARGET_HARD_REGNO_NREGS): ...this hook.
16130 (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16131 (CLASS_MAX_NREGS): Likewise.
16132 * doc/tm.texi: Regenerate.
16133 * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16134 instead of HARD_REGNO_NREGS.
16135 * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16136 HARD_REGNO_NREGS in the comment.
16137 * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16138 * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16139 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16140 Return an unsigned int.
16141 (TARGET_HARD_REGNO_NREGS): Redefine.
16142 * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16143 * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16144 * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16145 (arc_hard_regno_nregs): New function.
16146 * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16147 * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16148 (arm_hard_regno_nregs): New function.
16149 * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16150 * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16151 * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16152 (TARGET_HARD_REGNO_NREGS): Redefine.
16153 * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16154 * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16155 (HARD_REGNO_NREGS): Delete.
16156 * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16157 (cr16_hard_regno_nregs): New function.
16158 (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16159 * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16160 * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16161 (cris_hard_regno_nregs): New function.
16162 * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16163 * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16164 (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16165 * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16166 (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16167 * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16168 * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16169 (frv_hard_regno_nregs): Make static. Take and return an
16171 (frv_class_max_nregs): Remove outdated copy of documentation.
16172 * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16173 * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16174 * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16175 * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16176 * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16177 * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16178 (TARGET_HARD_REGNO_NREGS): Redefine.
16179 * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16180 (CLASS_MAX_NREGS): Update comment.
16181 * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16182 (ia64_hard_regno_nregs): New function.
16183 * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16184 * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16185 * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16186 * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16187 * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16189 (m32c_hard_regno_nregs): Likewise. Make static.
16190 (TARGET_HARD_REGNO_NREGS): Redefine.
16191 * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16192 * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16193 * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16194 (m68k_hard_regno_nregs): New function.
16195 * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16196 * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16197 * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16198 * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16199 * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16200 Take and return an unsigned int.
16201 (TARGET_HARD_REGNO_NREGS): Redefine.
16202 * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16203 (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16204 * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16205 * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16206 * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16207 * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16208 * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16209 (msp430_hard_regno_nregs): Make static. Take and return an
16211 * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16212 * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16213 * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16214 (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16215 * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16216 * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16217 * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16218 (TARGET_HARD_REGNO_NREGS): Redefine.
16219 * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16220 (PA_HARD_REGNO_NREGS): ...this.
16221 * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16222 (PA_HARD_REGNO_NREGS): ...this.
16223 * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16224 (pa_hard_regno_nregs): New function.
16225 * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16226 * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16227 (pdp11_hard_regno_nregs): New function.
16228 * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16229 * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16230 (rs6000_hard_regno_nregs_hook): New function.
16231 * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16232 * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16233 * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16234 Take and return an unsigned int. Move earlier in file.
16235 (TARGET_HARD_REGNO_NREGS): Redefine.
16236 * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16237 * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16238 * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16239 (rl78_hard_regno_nregs): Make static. Take and return an
16241 * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16242 * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16243 (rs6000_hard_regno_nregs_hook): New function.
16244 * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16245 * config/rx/rx.c (rx_hard_regno_nregs): New function.
16246 (TARGET_HARD_REGNO_NREGS): Redefine.
16247 * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16248 * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16249 instead of HARD_REGNO_NREGS.
16250 (s390_hard_regno_nregs): New function.
16251 (s390_hard_regno_mode_ok): Add comment from s390.h.
16252 (TARGET_HARD_REGNO_NREGS): Redefine.
16253 * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16254 * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16255 (sh_hard_regno_nregs): New function.
16256 (sh_pass_in_reg_p): Use it.
16257 * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16258 * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16259 (sparc_hard_regno_nregs): New function.
16260 * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16261 * config/spu/spu.c (spu_hard_regno_nregs): New function.
16262 (spu_function_arg_advance): Use it, supplying a valid register number.
16263 (TARGET_HARD_REGNO_NREGS): Redefine.
16264 * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16265 * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16266 * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16267 * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16268 * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16269 * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16270 (CLASS_MAX_NREGS): Remove copy of old documentation.
16271 * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16272 (visium_hard_regno_nregs): New function.
16273 (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16274 * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16275 * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16276 xtensa_hard_regno_nregs): New function.
16277 * system.h (HARD_REGNO_NREGS): Poison.
16279 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16281 * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16282 hard_regno_nregs instead of HARD_REGNO_NREGS.
16283 (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16284 * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16285 (c6x_expand_epilogue): Likewise.
16286 * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16287 (frv_read_iacc_argument): Likewise.
16288 * config/sh/sh.c: Include regs.h.
16289 (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16290 (regs_used): Likewise.
16291 (output_stack_adjust): Likewise.
16292 * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16293 * expmed.c: Include regs.h.
16294 (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16295 * ree.c: Include regs.h.
16296 (combine_reaching_defs): Use hard_regno_nregs instead of
16298 (add_removable_extension): Likewise.
16300 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16302 * regs.h (hard_regno_nregs): Turn into a function.
16303 (end_hard_regno): Update accordingly.
16304 * caller-save.c (setup_save_areas): Likewise.
16305 (save_call_clobbered_regs): Likewise.
16306 (replace_reg_with_saved_mem): Likewise.
16307 (insert_restore): Likewise.
16308 (insert_save): Likewise.
16309 * combine.c (can_change_dest_mode): Likewise.
16310 (move_deaths): Likewise.
16311 (distribute_notes): Likewise.
16312 * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16313 * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16314 (rs6000_split_multireg_move): Likewise.
16315 (rs6000_register_move_cost): Likewise.
16316 (rs6000_memory_move_cost): Likewise.
16317 * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16318 (rs6000_split_multireg_move): Likewise.
16319 (rs6000_register_move_cost): Likewise.
16320 (rs6000_memory_move_cost): Likewise.
16321 * cselib.c (cselib_reset_table): Likewise.
16322 (cselib_lookup_1): Likewise.
16323 * emit-rtl.c (set_mode_and_regno): Likewise.
16324 * function.c (aggregate_value_p): Likewise.
16325 * ira-color.c (setup_profitable_hard_regs): Likewise.
16326 (check_hard_reg_p): Likewise.
16327 (calculate_saved_nregs): Likewise.
16328 (assign_hard_reg): Likewise.
16329 (improve_allocation): Likewise.
16330 (calculate_spill_cost): Likewise.
16331 * ira-emit.c (modify_move_list): Likewise.
16332 * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16333 (ira_hard_reg_in_set_p): Likewise.
16334 * ira.c (setup_reg_mode_hard_regset): Likewise.
16335 (clarify_prohibited_class_mode_regs): Likewise.
16336 (check_allocation): Likewise.
16337 * lra-assigns.c (find_hard_regno_for_1): Likewise.
16338 (lra_setup_reg_renumber): Likewise.
16339 (setup_try_hard_regno_pseudos): Likewise.
16340 (spill_for): Likewise.
16341 (assign_hard_regno): Likewise.
16342 (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16343 * lra-constraints.c (in_class_p): Likewise.
16344 (lra_constraint_offset): Likewise.
16345 (simplify_operand_subreg): Likewise.
16346 (lra_constraints): Likewise.
16347 (split_reg): Likewise.
16348 (split_if_necessary): Likewise.
16349 (invariant_p): Likewise.
16350 (inherit_in_ebb): Likewise.
16351 * lra-lives.c (process_bb_lives): Likewise.
16352 * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16353 (get_hard_regs): Likewise.
16354 (do_remat): Likewise.
16355 * lra-spills.c (assign_spill_hard_regs): Likewise.
16356 * mode-switching.c (create_pre_exit): Likewise.
16357 * postreload.c (reload_combine_recognize_pattern): Likewise.
16358 * recog.c (peep2_find_free_register): Likewise.
16359 * regcprop.c (kill_value_regno): Likewise.
16360 (set_value_regno): Likewise.
16361 (copy_value): Likewise.
16362 (maybe_mode_change): Likewise.
16363 (find_oldest_value_reg): Likewise.
16364 (copyprop_hardreg_forward_1): Likewise.
16365 * regrename.c (check_new_reg_p): Likewise.
16366 (regrename_do_replace): Likewise.
16367 * reload.c (push_reload): Likewise.
16368 (combine_reloads): Likewise.
16369 (find_dummy_reload): Likewise.
16370 (operands_match_p): Likewise.
16371 (find_reloads): Likewise.
16372 (find_equiv_reg): Likewise.
16373 (reload_adjust_reg_for_mode): Likewise.
16374 * reload1.c (count_pseudo): Likewise.
16375 (count_spilled_pseudo): Likewise.
16376 (find_reg): Likewise.
16377 (clear_reload_reg_in_use): Likewise.
16378 (free_for_value_p): Likewise.
16379 (allocate_reload_reg): Likewise.
16380 (choose_reload_regs): Likewise.
16381 (reload_adjust_reg_for_temp): Likewise.
16382 (emit_reload_insns): Likewise.
16383 (delete_output_reload): Likewise.
16384 * rtlanal.c (subreg_get_info): Likewise.
16385 * sched-deps.c (sched_analyze_reg): Likewise.
16386 * sel-sched.c (init_regs_for_mode): Likewise.
16387 (mark_unavailable_hard_regs): Likewise.
16388 (choose_best_reg_1): Likewise.
16389 (verify_target_availability): Likewise.
16390 * valtrack.c (dead_debug_insert_temp): Likewise.
16391 * var-tracking.c (track_loc_p): Likewise.
16392 (emit_note_insn_var_location): Likewise.
16393 * varasm.c (make_decl_rtl): Likewise.
16394 * reginfo.c (choose_hard_reg_mode): Likewise.
16395 (init_reg_modes_target): Refer directly to
16396 this_target_regs->x_hard_regno_nregs.
16398 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16400 * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16401 instead of hard_regno_nregs.
16403 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16405 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16406 end_hard_regno instead of hard_regno_nregs.
16407 * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16408 * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16409 * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16410 * ira-color.c (improve_allocation): Likewise.
16411 * lra-assigns.c (find_hard_regno_for_1): Likewise.
16412 * lra-lives.c (mark_regno_live): Likewise.
16413 (mark_regno_dead): Likewise.
16414 * lra-remat.c (operand_to_remat): Likewise.
16415 * lra.c (collect_non_operand_hard_regs): Likewise.
16416 * postreload.c (reload_combine_note_store): Likewise.
16417 (move2add_valid_value_p): Likewise.
16418 * reload.c (regno_clobbered_p): Likewise.
16420 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16422 * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16424 * config/v850/v850.c (v850_reorg): Likewise.
16425 * reload.c (refers_to_regno_for_reload_p): Likewise.
16426 (find_equiv_reg): Likewise.
16427 * reload1.c (reload_reg_reaches_end_p): Likewise.
16429 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16431 * caller-save.c (add_used_regs): Use REG_NREGS instead of
16433 * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16434 * config/arm/arm.c (output_move_neon): Likewise.
16435 (arm_attr_length_move_neon): Likewise.
16436 (neon_split_vcombine): Likewise.
16437 * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16438 (c6x_mark_reg_written): Likewise.
16439 (c6x_dwarf_register_span): Likewise.
16440 * config/i386/i386.c (ix86_save_reg): Likewise.
16441 * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16442 (rws_access_reg): Likewise.
16443 * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16444 * mode-switching.c (create_pre_exit): Likewise.
16445 * ree.c (combine_reaching_defs): Likewise.
16446 (add_removable_extension): Likewise.
16447 * regcprop.c (find_oldest_value_reg): Likewise.
16448 (copyprop_hardreg_forward_1): Likewise.
16449 * reload.c (reload_inner_reg_of_subreg): Likewise.
16450 (push_reload): Likewise.
16451 (combine_reloads): Likewise.
16452 (find_dummy_reload): Likewise.
16453 (reload_adjust_reg_for_mode): Likewise.
16454 * reload1.c (find_reload_regs): Likewise.
16455 (forget_old_reloads_1): Likewise.
16456 (reload_reg_free_for_value_p): Likewise.
16457 (reload_adjust_reg_for_temp): Likewise.
16458 (emit_reload_insns): Likewise.
16459 (delete_output_reload): Likewise.
16460 * sel-sched.c (choose_best_reg_1): Likewise.
16461 (choose_best_pseudo_reg): Likewise.
16463 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16464 Alan Hayward <alan.hayward@arm.com>
16465 David Sherwood <david.sherwood@arm.com>
16467 * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16468 * target.def (slow_unaligned_access): New hook.
16469 * targhooks.h (default_slow_unaligned_access): Declare.
16470 * targhooks.c (default_slow_unaligned_access): New function.
16471 * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16472 (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16473 * doc/tm.texi: Regenerate.
16474 * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16475 * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16476 * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16478 * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16479 * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16481 (rs6000_slow_unaligned_access): New function.
16482 (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16483 (expand_block_compare): Likewise.
16484 (expand_strn_compare): Likewise.
16485 (rs6000_rtx_costs): Likewise.
16486 * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16487 (riscv_slow_unaligned_access): Likewise.
16488 * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16489 (riscv_slow_unaligned_access_p): ...this and make static.
16490 (riscv_option_override): Update accordingly.
16491 (riscv_slow_unaligned_access): New function.
16492 (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16493 * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16494 * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16495 (rs6000_slow_unaligned_access): New function.
16496 (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16497 (rs6000_rtx_costs): Likewise.
16498 * config/rs6000/rs6000-string.c (expand_block_compare)
16499 (expand_strn_compare): Use targetm.slow_unaligned_access instead
16500 of SLOW_UNALIGNED_ACCESS.
16501 * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16502 * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16503 * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16504 of SLOW_UNALIGNED_ACCESS.
16505 * expmed.c (simple_mem_bitfield_p): Likewise.
16506 * expr.c (alignment_for_piecewise_move): Likewise.
16507 (emit_group_load_1): Likewise.
16508 (emit_group_store): Likewise.
16509 (copy_blkmode_from_reg): Likewise.
16510 (emit_push_insn): Likewise.
16511 (expand_assignment): Likewise.
16512 (store_field): Likewise.
16513 (expand_expr_real_1): Likewise.
16514 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16515 * lra-constraints.c (simplify_operand_subreg): Likewise.
16516 * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16517 * gimple-ssa-store-merging.c: Likewise in block comment at start
16519 * tree-ssa-strlen.c: Include target.h.
16520 (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16521 of SLOW_UNALIGNED_ACCESS.
16522 * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16524 2017-09-12 Richard Sandiford <richard.sandiford@linaro.org>
16526 PR rtl-optimization/82185
16527 * expmed.c (emit_store_flag_int): Only test tem if it has been
16530 2017-09-12 Richard Biener <rguenther@suse.de>
16532 PR middle-end/82149
16533 * match.pd ((FTYPE) N CMP CST): Fix typo.
16535 2017-09-12 Simon Atanasyan <simon.atanasyan@imgtec.com>
16537 * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16539 (mips_near_type_p): Add 'short_call' attribute as a synonym
16541 * doc/extend.texi (short_call): Document new function attribute.
16543 2017-09-12 Jakub Jelinek <jakub@redhat.com>
16546 * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16547 assertion check that in the condition.
16548 (get_atomic_generic_size): Likewise. Before testing if parameter
16549 has pointer type, if it has array type, call for C++
16550 default_conversion to perform array-to-pointer conversion.
16552 2017-09-12 Richard Biener <rguenther@suse.de>
16554 * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16555 for operations we cannot scalarize.
16557 2017-09-12 Aldy Hernandez <aldyh@redhat.com>
16559 * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16560 vectors heap vectors. Clean up comments.
16561 Make visited_bbs a reference.
16562 (profitable_jump_thread_path): Make GC
16563 vectors heap vectors. Clean up comments.
16565 (convert_and_register_jump_thread_path): Make GC vectors heap
16567 (check_subpath_and_update_thread_path): Same. Clean up comments.
16568 Make visited_bbs a reference.
16569 (handle_phi): Abstract common code to to
16570 register_jump_thread_path_if_profitable.
16571 Rename VAR_BB to DEF_BB.
16573 Make GC vectors heap vectors.
16574 Make visited_bbs a reference.
16575 (handle_assignment): Same.
16576 (register_jump_thread_path_if_profitable): New.
16577 (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16579 Make GC vectors heap vectors. Clean up comments.
16580 Make visited_bbs a reference.
16581 (find_jump_threads_backwards): Make visited_bbs live in the stack.
16582 * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16585 2017-09-11 Max Filippov <jcmvbkbc@gmail.com>
16588 * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16589 words of E_DImode object are reachable by xtensa_uimm8x4 access.
16591 2017-09-11 Vidya Praveen <vidyapraveen@arm.com>
16593 Revert r251800 and r251799.
16595 2017-09-11 Martin Jambor <mjambor@suse.cz>
16598 * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16599 arguments in advance.
16600 * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16601 use it to find predecessor edges.
16602 (naive_outof_ssa): Collect vector of predecessors.
16604 2017-09-08 Jason Merrill <jason@redhat.com>
16606 PR c++/70029 - ICE with ref-qualifier and -flto
16607 * langhooks.h (struct lang_hooks_for_types): Add
16608 copy_lang_qualifiers.
16609 * attribs.c (build_type_attribute_qual_variant): Use it.
16610 * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16612 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16613 * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16615 2017-09-08 Eric Botcazou <ebotcazou@adacore.com>
16618 * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16619 (*mulsi3_sp64): New instruction.
16620 (mulsi3): New expander.
16622 2017-09-08 Uros Bizjak <ubizjak@gmail.com>
16624 * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16626 2017-09-08 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16628 * sancov.c: Include memmodel.h.
16630 2017-09-07 Eric Botcazou <ebotcazou@adacore.com>
16633 * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16636 2017-09-07 Carl Love <cel@us.ibm.com>
16638 * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16639 the sldi instruction.
16641 2017-09-07 David Edelsohn <dje.gcc@gmail.com>
16643 * sancov.c: Include tm_p.h.
16645 2017-09-07 Jakub Jelinek <jakub@redhat.com>
16648 * output.h (switch_to_other_text_partition): New declaration.
16649 * varasm.c (switch_to_other_text_partition): New function.
16650 * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16651 NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16652 (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16653 to the other text partition before emitting LCL label and switch back
16654 after emitting the word after it.
16656 2017-09-07 Richard Biener <rguenther@suse.de>
16658 * passes.def (pass_split_crit_edges): Remove instance before PRE.
16659 * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16660 critical edges here, after loop init.
16661 (pass_data_pre): Remove PROP_no_crit_edges flags.
16662 * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16663 for valueization of call args to avoid leaking VN_TOP.
16664 (visit_use): Assert we do not visit default defs.
16665 (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16666 Use error_mark_node to more easily detect leaking VN_TOP.
16667 All default-defs are varying, not VN_TOP. Mark them visited.
16668 (run_scc_vn): Make code match comment.
16670 2017-09-07 Michael Meissner <meissner@linux.vnet.ibm.com>
16672 * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16673 OPTION_MASK_FLOAT128_KEYWORD.
16674 (POWERPC_MASKS): Likewise.
16675 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16676 support for the -mfloat128-type option, and make -mfloat128
16677 default on PowerPC Linux systems. Define or undefine
16678 __FLOAT128__ and __FLOAT128_HARDWARE__ for the current options.
16679 Define __float128 to be __ieee128 if IEEE 128-bit support is
16680 enabled, or undefine it.
16681 (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16682 Delete defining __FLOAT128_TYPE__.
16683 * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16684 -mfloat128-type option and make -mfloat128 default on PowerPC
16686 (TARGET_FLOAT128_TYPE): Likewise.
16687 (-mfloat128-type): Likewise.
16688 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16689 Delete the -mfloat128-type option and make -mfloat128 default on
16690 PowerPC Linux systems. Always use __ieee128 to be the keyword for
16691 the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16692 128-bit floating point is enabled. Change tests from using
16693 -mfloat128-type to -mfloat128.
16694 (rs6000_mangle_type): Use the correct mangling for the __float128
16695 type even if normal long double is restricted to 64-bits.
16696 (floatn_mode): Enable the _Float128 type by default on VSX Linux
16698 * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16699 (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16700 (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16702 * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16703 documentation for -mfloat128.
16705 2017-09-06 Olivier Hainque <hainque@adacore.com>
16707 * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16709 2017-09-06 Wish Wu <wishwu007@gmail.com>
16710 Jakub Jelinek <jakub@redhat.com>
16712 * asan.c (initialize_sanitizer_builtins): Add
16713 BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16714 BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16715 BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16716 BT_FN_VOID_UINT64_PTR variables.
16717 * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16718 (BT_FN_VOID_UINT16_UINT16): Likewise.
16719 (BT_FN_VOID_UINT32_UINT32): Likewise.
16720 (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16721 (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16722 (BT_FN_VOID_UINT64_PTR): Likewise.
16723 * common.opt (flag_sanitize_coverage): New variable.
16724 (fsanitize-coverage=trace-pc): Remove.
16725 (fsanitize-coverage=): Add.
16726 * flag-types.h (enum sanitize_coverage_code): New enum.
16727 * fold-const.c (fold_range_test): Disable non-short-circuit
16728 optimization if flag_sanitize_coverage.
16729 (fold_truth_andor): Likewise.
16730 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16731 * opts.c (COVERAGE_SANITIZER_OPT): Define.
16732 (coverage_sanitizer_opts): New array.
16733 (get_closest_sanitizer_option): Add OPTS argument, handle also
16734 OPT_fsanitize_coverage_.
16735 (parse_sanitizer_options): Adjusted to also handle
16736 OPT_fsanitize_coverage_.
16737 (common_handle_option): Add OPT_fsanitize_coverage_.
16738 * sancov.c (instrument_comparison, instrument_switch): New function.
16739 (sancov_pass): Add trace-cmp support.
16740 * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16741 BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16742 BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16743 BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16744 BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16745 BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16746 BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16747 BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16748 BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16749 * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16751 2017-09-06 Richard Earnshaw <rearnsha@arm.com>
16753 * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16754 error. Only quit immediately if parsing is complete.
16755 (BEGIN): Initialize fatal_err and parse_done.
16756 (begin fpu, end fpu): Check number of arguments.
16757 (begin arch, end arch): Likewise.
16758 (begin cpu, end cpu): Likewise.
16759 (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16760 (optalias): Likewise.
16762 2017-09-06 Richard Earnshaw <rearnsha@arm.com>
16764 * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16765 * config/arm/arm-isa.h: Delete. Move definitions to ...
16766 * arm-cpus.in: ... here. Use new feature and fgroup values.
16767 * config/arm/arm.c (arm_option_override): Use lower case for feature
16769 * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16770 (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16771 * config/arm/parsecpu.awk (END): Add new command 'isa'.
16773 (print_isa_bits_for): New function.
16774 (gen_isa): New function.
16775 (gen_comm_data): Use print_isa_bits_for.
16776 (define feature): New keyword.
16777 (define fgroup): New keyword.
16778 * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16779 (arm-isa.h): Add rule to generate file.
16780 * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16781 case for feature bit names.
16783 2017-09-06 Richard Biener <rguenther@suse.de>
16785 * tree-ssa-pre.c (NECESSARY): Remove.
16786 (create_expression_by_pieces): Do not touch pass-local flags.
16787 (insert_into_preds_of_block): Likewise.
16788 (do_pre_regular_insertion): Likewise.
16789 (eliminate_insert): Likewise.
16790 (eliminate_dom_walker::before_dom_children): Likewise.
16791 (fini_eliminate): Do not look at inserted_exprs.
16792 (mark_operand_necessary): Remove.
16793 (remove_dead_inserted_code): Replace with simple work-list
16794 algorithm based on inserted_exprs and SSA uses.
16795 (pass_pre::execute): Re-order fini_eliminate and
16796 remove_dead_inserted_code.
16798 2017-09-06 Olivier Hainque <hainque@adacore.com>
16800 * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16801 for VxWorks 7. Adjust surrounding comments.
16803 2017-09-06 Richard Biener <rguenther@suse.de>
16805 * gimple-ssa-strength-reduction.c
16806 (find_candidates_dom_walker::before_dom_children): Also allow
16809 2017-09-06 Richard Biener <rguenther@suse.de>
16811 PR tree-optimization/82108
16812 * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16813 for gap in the non-permutation SLP case.
16815 2017-09-06 Martin Jambor <mjambor@suse.cz>
16817 PR tree-optimization/82078
16818 * tree-sra.c (sort_and_splice_var_accesses): Move call to
16819 add_access_to_work_queue...
16820 (build_accesses_from_assign): ...here.
16821 (propagate_all_subaccesses): Make sure racc is the group
16822 representative, if there is one.
16824 2017-09-06 Jakub Jelinek <jakub@redhat.com>
16826 PR middle-end/82095
16827 * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16830 2017-09-06 Richard Biener <rguenther@suse.de>
16832 * gimple-ssa-strength-reduction.c
16833 (find_candidates_dom_walker::before_doom_children): Use a
16834 type and not a mode check.
16836 2017-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16839 * config/arm/predicates.md (arm_general_adddi_operand): Create new
16841 * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16843 2017-09-05 Jeff Law <law@redhat.com>
16845 PR tree-optimization/64910
16846 * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
16847 cases where we have 3 or more operands.
16849 2017-09-05 Jakub Jelinek <jakub@redhat.com>
16851 PR middle-end/81768
16852 * omp-low.c (lower_omp_for): Recompute tree invariant if
16853 gimple_omp_for_initial/final is ADDR_EXPR.
16855 PR middle-end/81768
16856 * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
16857 into gimple val before gimplification fo the COND_EXPR.
16859 2017-09-05 Aldy Hernandez <aldyh@redhat.com>
16861 * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
16862 REGION_COPY argument.
16863 (thread_through_all_blocks): Remove unused argument to
16864 duplicate_thread_path.
16866 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16867 Alan Hayward <alan.hayward@arm.com>
16868 David Sherwood <david.sherwood@arm.com>
16870 * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
16871 Take a scalar_mode rather than a machine_mode.
16872 (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16873 * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
16874 (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16875 (aarch64_gen_adjusted_ldpstp): Likewise.
16876 (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
16878 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16879 Alan Hayward <alan.hayward@arm.com>
16880 David Sherwood <david.sherwood@arm.com>
16882 * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
16883 Take a scalar_int_mode instead of a machine_mode.
16884 (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16885 (aarch64_output_scalar_simd_mov_immediate): Likewise.
16886 (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16887 (aarch64_simd_attr_length_rglist): Delete.
16888 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
16889 a scalar_int_mode instead of a machine_mode.
16890 (aarch64_add_offset): Likewise.
16891 (aarch64_internal_mov_immediate): Likewise
16892 (aarch64_add_constant_internal): Likewise.
16893 (aarch64_add_constant): Likewise.
16894 (aarch64_movw_imm): Likewise.
16895 (aarch64_rtx_arith_op_extract_p): Likewise.
16896 (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16897 (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16898 Remove assert that the mode isn't a vector.
16899 (aarch64_output_scalar_simd_mov_immediate): Likewise.
16900 (aarch64_expand_mov_immediate): Update calls after above changes.
16901 (aarch64_output_casesi): Use as_a <scalar_int_mode>.
16902 (aarch64_and_bitmask_imm): Check for scalar integer modes.
16903 (aarch64_move_imm): Likewise.
16904 (aarch64_can_const_movi_rtx_p): Likewise.
16905 (aarch64_strip_extend): Likewise.
16906 (aarch64_extr_rtx_p): Likewise.
16907 (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
16908 a CONST_INT when the mode parameter is VOIDmode.
16909 (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
16911 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16913 * machmode.h (bitwise_mode_for_mode): Return opt_mode.
16914 * stor-layout.c (bitwise_mode_for_mode): Likewise.
16915 (bitwise_type_for_mode): Update accordingly.
16917 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16919 * stor-layout.h (mode_for_size_tree): Return an opt_mode.
16920 * stor-layout.c (mode_for_size_tree): Likewise.
16921 (mode_for_array): Update accordingly.
16922 (layout_decl): Likewise.
16923 (compute_record_mode): Likewise. Only set the mode once.
16925 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16927 * target.def (get_mask_mode): Change return type to opt_mode.
16929 * doc/tm.texi: Regenerate.
16930 * targhooks.h (default_get_mask_mode): Return an opt_mode.
16931 * targhooks.c (default_get_mask_mode): Likewise.
16932 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16933 * optabs-query.c (can_vec_mask_load_store_p): Update use of
16934 targetm.get_mask_mode.
16935 * tree.c (build_truth_vector_type): Likewise.
16937 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16939 * machmode.h (mode_for_vector): Return an opt_mode.
16940 * stor-layout.c (mode_for_vector): Likewise.
16941 (mode_for_int_vector): Update accordingly.
16942 (layout_type): Likewise.
16943 * config/i386/i386.c (emit_memmov): Likewise.
16944 (ix86_expand_set_or_movmem): Likewise.
16945 (ix86_expand_vector_init): Likewise.
16946 (ix86_get_mask_mode): Likewise.
16947 * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
16949 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
16950 * expmed.c (extract_bit_field_1): Likewise.
16951 * expr.c (expand_expr_real_2): Likewise.
16952 * optabs-query.c (can_vec_perm_p): Likewise.
16953 (can_vec_mask_load_store_p): Likewise.
16954 * optabs.c (expand_vec_perm): Likewise.
16955 * targhooks.c (default_get_mask_mode): Likewise.
16956 * tree-vect-stmts.c (vectorizable_store): Likewise.
16957 (vectorizable_load): Likewise.
16958 (get_vectype_for_scalar_type_and_size): Likewise.
16960 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16962 * machmode.h (mode_for_int_vector): New function.
16963 * stor-layout.c (mode_for_int_vector): Likewise.
16964 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
16965 * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
16966 * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
16967 * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
16968 (s390_expand_vcond): Likewise.
16970 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16972 * machmode.h (opt_machine_mode): New type.
16973 (opt_mode<T>): Allow construction from anything that can be
16975 (is_a, as_a, dyn_cast): Add overloads for opt_mode.
16976 (mode_for_size): Return an opt_machine_mode.
16977 * stor-layout.c (mode_for_size): Likewise.
16978 (mode_for_size_tree): Update call accordingly.
16979 (bitwise_mode_for_mode): Likewise.
16980 (make_fract_type): Likewise.
16981 (make_accum_type): Likewise.
16982 * caller-save.c (replace_reg_with_saved_mem): Update call
16984 * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
16985 * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
16986 * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
16987 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
16988 * expmed.c (extract_bit_field_1): Likewise.
16989 * reload.c (get_secondary_mem): Likewise.
16990 * varasm.c (assemble_integer): Likewise.
16991 * lower-subreg.c (simplify_subreg_concatn): Likewise. Move
16994 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
16996 * machmode.h (decimal_float_mode_for_size): New function.
16997 * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
16998 (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
16999 (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17000 (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17001 (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17002 (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17004 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
17006 * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17007 (get_builtin_sync_mode): Likewise.
17008 (expand_ifn_atomic_compare_exchange): Likewise.
17009 (expand_builtin_atomic_clear): Likewise.
17010 (expand_builtin_atomic_test_and_set): Likewise.
17011 (fold_builtin_atomic_always_lock_free): Likewise.
17012 * calls.c (compute_argument_addresses): Likewise.
17013 (emit_library_call_value_1): Likewise.
17014 (store_one_arg): Likewise.
17015 * combine.c (combine_instructions): Likewise.
17016 * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17017 * config/arm/arm.c (arm_function_value): Likewise.
17018 (aapcs_allocate_return_reg): Likewise.
17019 * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17020 * config/i386/i386.c (construct_container): Likewise.
17021 (ix86_gimplify_va_arg): Likewise.
17022 (ix86_expand_sse_cmp): Likewise.
17023 (emit_memmov): Likewise.
17024 (emit_memset): Likewise.
17025 (expand_small_movmem_or_setmem): Likewise.
17026 (ix86_expand_pextr): Likewise.
17027 (ix86_expand_pinsr): Likewise.
17028 * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17029 * config/microblaze/microblaze.c (microblaze_block_move_straight):
17031 * config/mips/mips.c (mips_function_value_1) Likewise.
17032 (mips_block_move_straight): Likewise.
17033 (mips_expand_ins_as_unaligned_store): Likewise.
17034 * config/powerpcspe/powerpcspe.c
17035 (rs6000_darwin64_record_arg_advance_flush): Likewise.
17036 (rs6000_darwin64_record_arg_flush): Likewise.
17037 * config/rs6000/rs6000.c
17038 (rs6000_darwin64_record_arg_advance_flush): Likewise.
17039 (rs6000_darwin64_record_arg_flush): Likewise.
17040 * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17041 (sparc_function_value_1): Likewise.
17042 * config/spu/spu.c (adjust_operand): Likewise.
17043 (spu_emit_branch_or_set): Likewise.
17044 (arith_immediate_p): Likewise.
17045 * emit-rtl.c (gen_lowpart_common): Likewise.
17046 * expr.c (expand_expr_real_1): Likewise.
17047 * function.c (assign_parm_setup_block): Likewise.
17048 * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17049 * reload1.c (alter_reg): Likewise.
17050 * stor-layout.c (mode_for_vector): Likewise.
17051 (layout_type): Likewise.
17053 2017-09-05 Richard Sandiford <richard.sandiford@linaro.org>
17055 * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17056 (spu_convert_move): Likewise.
17057 * lower-subreg.c (resolve_simple_move): Likewise.
17059 2017-09-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17062 * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17063 define_insn to a define_expand.
17064 (altivec_vsum2sws_direct): New define_insn.
17065 (altivec_vsumsws): Convert from a define_insn to a define_expand.
17067 2017-09-05 Wilco Dijkstra <wdijkstr@arm.com>
17069 * config/arm/arm.c (arm_option_params_internal): Improve setting of
17072 2017-09-05 H.J. Lu <hongjiu.lu@intel.com>
17077 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17078 realign stack if stack alignment needed is less than incoming
17081 2017-09-05 Marek Polacek <polacek@redhat.com>
17084 * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17087 2017-09-05 Wilco Dijkstra <wdijkstr@arm.com>
17089 * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17091 2017-09-05 Richard Biener <rguenther@suse.de>
17093 PR tree-optimization/82084
17094 * fold-const.c (can_native_encode_string_p): Handle wide characters.
17096 2017-09-05 Richard Biener <rguenther@suse.de>
17098 PR tree-optimization/82102
17099 * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17101 2017-09-05 Martin Liska <mliska@suse.cz>
17103 PR tree-optimization/82032
17104 * tree-cfg.c (generate_range_test): New function.
17105 * tree-cfg.h (generate_range_test): Declared here.
17106 * tree-cfgcleanup.c (convert_single_case_switch): New function.
17107 (cleanup_control_expr_graph): Use it.
17108 * tree-switch-conversion.c (try_switch_expansion): Remove
17110 (emit_case_nodes): Use generate_range_test.
17112 2017-09-04 Uros Bizjak <ubizjak@gmail.com>
17115 * config/i386/i386.md (*<btsc><mode>_mask): Add
17116 TARGET_USE_BT to insn constraint.
17117 (*btr<mode>_mask): Ditto.
17119 2017-09-04 Wilco Dijkstra <wdijkstr@arm.com>
17121 * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17122 (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17124 2017-09-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
17127 * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17128 TARGET_NEON and TARGET_IWMMXT.
17129 (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17130 TARGET_NEON and TARGET_IWMMXT.
17131 (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17133 2017-09-04 Uros Bizjak <ubizjak@gmail.com>
17135 * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17136 (ix86_rewrite_tls_address): Ditto.
17137 * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17138 (ix86_rewrite_tls_address_1): Ditto.
17139 (ix86_rewrite_tls_address): Ditto.
17140 * config/i386/predicates.md (tls_address_pattern): New predicate.
17141 * config/i386/i386.md (TLS address splitter): New splitter.
17143 2017-09-04 Richard Biener <rguenther@suse.de>
17145 PR tree-optimization/82084
17146 * fold-const.h (can_native_encode_string_p): Declare.
17147 * fold-const.c (can_native_encode_string_p): Factor out from ...
17148 (native_encode_string): ... here.
17149 * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17150 vectorizing stores from constants we later cannot handle.
17152 2017-09-04 Marek Polacek <polacek@redhat.com>
17155 * doc/invoke.texi: Update -Wtautological-compare documentation.
17157 2017-09-04 Jeff Law <law@redhat.com>
17159 PR tree-optimization/64910
17160 * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17161 swap the first and last operand if the last is a constant.
17163 2017-09-04 Marek Polacek <polacek@redhat.com>
17166 * convert.c (do_narrow): When sanitizing signed integer overflows,
17167 bail out for signed types.
17168 (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17170 2017-09-04 Richard Biener <rguenther@suse.de>
17172 PR tree-optimization/82060
17173 * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17174 Move devirtualization after stmt folding and before EH/AB/noreturn
17175 cleanup to get the stmt refs canonicalized. Use a bool instead
17176 of gimple_modified_p since that doesn't work for NOPs. Schedule
17177 NOPs generated by folding for removal.
17179 2017-09-04 Richard Sandiford <richard.sandiford@linaro.org>
17180 Alan Hayward <alan.hayward@arm.com>
17181 David Sherwood <david.sherwood@arm.com>
17183 * coretypes.h (pad_direction): New enum.
17184 * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17185 (FUNCTION_ARG_PADDING): Likewise.
17186 * target.def (function_arg_padding): New hook.
17187 * targhooks.h (default_function_arg_padding): Declare.
17188 * targhooks.c (default_function_arg_padding): New function.
17189 * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17190 (TARGET_FUNCTION_ARG_PADDING): ...this.
17191 * doc/tm.texi: Regenerate.
17192 * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17193 instead of direction.
17194 (compute_argument_addresses): Likewise.
17195 (load_register_parameters): Likewise.
17196 (emit_library_call_value_1): Likewise.
17197 (store_one_arg): Use targetm.calls.function_arg_padding instead
17198 of FUNCTION_ARG_PADDING.
17199 (must_pass_in_stack_var_size_or_pad): Likewise.
17200 * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17201 (emit_group_store): Likewise.
17202 (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17203 instead of FUNCTION_ARG_PADDING.
17204 (emit_push_insn): Likewise, and propagate enum change throughout
17206 * function.h (direction): Delete.
17207 (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17209 * function.c (assign_parm_find_stack_rtl): Likewise.
17210 (assign_parm_setup_block_p): Likewise.
17211 (assign_parm_setup_block): Likewise.
17212 (gimplify_parameters): Likewise.
17213 (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17214 instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17216 * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17217 (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17218 * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17219 * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17220 (aarch64_function_arg_padding): ...this new function.
17221 (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17222 (TARGET_FUNCTION_ARG_PADDING): Redefine.
17223 * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17224 (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17225 * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17226 * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17227 (arm_pad_arg_upward): Replace with...
17228 (arm_function_arg_padding): ...this new function.
17229 * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17231 * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17232 * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17233 * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17234 (ia64_hpux_function_arg_padding): Replace with...
17235 (ia64_function_arg_padding): ...this new function. Use pad_direction
17236 instead of direction. Check for TARGET_HPUX.
17237 * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17238 * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17239 (iq2000_function_arg_padding): New function.
17240 * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17241 * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17242 (mips_function_arg_padding): ...this new function.
17243 (mips_pad_reg_upward): Update accordingly.
17244 (TARGET_FUNCTION_ARG_PADDING): Redefine.
17245 * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17246 targetm.calls.function_arg_padding.
17247 (FUNCTION_ARG_PADDING): Delete.
17248 (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17249 * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17250 (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17251 * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17252 (nios2_block_reg_padding): Return pad_direction instead of direction.
17253 * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17254 instead of direction.
17255 (nios2_function_arg_padding): Likewise. Make static.
17256 (TARGET_FUNCTION_ARG_PADDING): Redefine.
17257 * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17258 (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17259 * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17260 * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17261 (pa_function_arg_padding): Make static. Return pad_direction instead
17263 * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17264 (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17265 * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17266 instead of direction. Use targetm.calls.function_arg_padding.
17267 * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17268 * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17269 * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17270 * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17271 * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17273 (function_arg_padding): Rename to...
17274 (rs6000_function_arg_padding): ...this. Make static. Return
17275 pad_direction instead of direction.
17276 (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17277 * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17278 (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17279 * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17280 instead of direction. Use targetm.calls.function_arg_padding.
17281 * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17282 * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17283 * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17284 * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17285 * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17286 (function_arg_padding): Rename to...
17287 (rs6000_function_arg_padding): ...this. Make static. Return
17288 pad_direction instead of direction.
17289 (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17290 * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17291 * config/s390/s390.c (s390_function_arg_padding): New function.
17292 (TARGET_FUNCTION_ARG_PADDING): Redefine.
17293 * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17294 * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17295 * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17296 (function_arg_padding): Rename to...
17297 (sparc_function_arg_padding): ...this. Make static. Return
17298 pad_direction instead of direction.
17299 * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17300 * config/spu/spu.c (spu_function_arg_padding): New function.
17301 (TARGET_FUNCTION_ARG_PADDING): Redefine.
17302 * system.h (FUNCTION_ARG_PADDING): Poison.
17304 2017-09-04 Richard Sandiford <richard.sandiford@linaro.org>
17305 Alan Hayward <alan.hayward@arm.com>
17306 David Sherwood <david.sherwood@arm.com>
17308 * target.def (modes_tieable_p): New hook.
17309 * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17310 (TARGET_MODES_TIEABLE_P): ...this.
17311 * doc/tm.texi.in: Regenerate.
17312 * hooks.h (hook_bool_mode_mode_true): Declare.
17313 * hooks.c (hook_bool_mode_mode_true): New function.
17314 * combine.c (subst): Use targetm.modes_tieable_p instead of
17316 * dse.c (find_shift_sequence): Likewise.
17317 * expmed.c (extract_low_bits): Likewise.
17318 * lower-subreg.c: Include target.h.
17319 (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17321 * rtlanal.c (rtx_cost): Likewise.
17322 * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17323 * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17324 * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17325 (TARGET_MODES_TIEABLE_P): Redefine.
17326 * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17327 * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17328 (TARGET_MODES_TIEABLE_P): Redefine.
17329 * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17330 * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17331 (arc_modes_tieable_p): New function.
17332 * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17333 * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17334 * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17335 (arm_modes_tieable_p): Make static.
17336 * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17337 * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17338 * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17339 (TARGET_MODES_TIEABLE_P): Redefine.
17340 * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17341 * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17342 (TARGET_MODES_TIEABLE_P): Redefine.
17343 * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17344 * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17345 (cr16_modes_tieable_p): New function.
17346 * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17347 * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17348 * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17349 (TRULY_NOOP_TRUNCATION): Update comment.
17350 * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17351 (TRULY_NOOP_TRUNCATION): Update comment.
17352 * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17353 (frv_modes_tieable_p): New function.
17354 * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17355 * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17356 * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17357 (TARGET_MODES_TIEABLE_P): Redefine.
17358 * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17359 * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17360 * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17361 (TARGET_MODES_TIEABLE_P): Redefine.
17362 * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17363 * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17364 (ia64_modes_tieable_p): New function.
17365 * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17366 * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17367 (iq2000_modes_tieable_p): New function.
17368 * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17369 * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17370 (lm32_modes_tieable_p): New function.
17371 * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17372 * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17373 * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17374 (TARGET_MODES_TIEABLE_P): Redefine.
17375 * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17376 * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17377 (m32r_modes_tieable_p): New function.
17378 * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17379 * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17380 (m68k_modes_tieable_p): New function.
17381 * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17382 * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17383 (mcore_modes_tieable_p): New function.
17384 * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17385 * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17387 (TARGET_MODES_TIEABLE_P): Redefine.
17388 * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17389 * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17390 * config/mips/mips.c (mips_modes_tieable_p): Make static.
17391 (TARGET_MODES_TIEABLE_P): Redefine.
17392 * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17393 * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17394 * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17395 * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17396 (mn10300_modes_tieable_p): ...this and make static.
17397 (TARGET_MODES_TIEABLE_P): Redefine.
17398 * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17399 * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17400 * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17401 * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17402 (msp430_modes_tieable_p): Make static.
17403 * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17404 * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17405 (TARGET_MODES_TIEABLE_P): Redefine.
17406 * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17407 * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17408 * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17409 (TARGET_MODES_TIEABLE_P): Redefine.
17410 * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17411 * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17412 * config/pa/pa.c (pa_modes_tieable_p): Make static.
17413 (TARGET_MODES_TIEABLE_P): Redefine.
17414 * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17415 * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17416 (pdp11_modes_tieable_p): New function.
17417 * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17418 * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17419 (rs6000_modes_tieable_p): New function.
17420 (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17421 * config/powerpcspe/powerpcspe.md: Update comment.
17422 * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17423 * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17424 (TARGET_MODES_TIEABLE_P): Redefine.
17425 * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17426 * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17427 (rl78_modes_tieable_p): New function.
17428 * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17429 * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17430 (rs6000_modes_tieable_p): New function.
17431 (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17432 * config/rs6000/rs6000.md: Update comment.
17433 * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17434 * config/rx/rx.c (rx_modes_tieable_p): New function.
17435 (TARGET_MODES_TIEABLE_P): Redefine.
17436 * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17437 * config/s390/s390.c (s390_modes_tieable_p): New function.
17438 (TARGET_MODES_TIEABLE_P): Redefine.
17439 * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17440 * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17441 (sh_modes_tieable_p): New function.
17442 * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17443 * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17444 * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17445 (sparc_modes_tieable_p): Make static.
17446 * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17447 * config/spu/spu.c (spu_modes_tieable_p): New function.
17448 (TARGET_MODES_TIEABLE_P): Redefine.
17449 * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17450 * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17451 (TARGET_MODES_TIEABLE_P): Redefine.
17452 * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17453 * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17454 * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17455 * config/v850/v850.c (v850_modes_tieable_p): New function.
17456 (TARGET_MODES_TIEABLE_P): Redefine.
17457 * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17458 * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17459 * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17460 (visium_modes_tieable_p): New function.
17461 * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17462 * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17463 (xtensa_modes_tieable_p): New function.
17464 * system.h (MODES_TIEABLE_P): Poison.
17466 2017-09-04 Richard Sandiford <richard.sandiford@linaro.org>
17467 Alan Hayward <alan.hayward@arm.com>
17468 David Sherwood <david.sherwood@arm.com>
17470 * target.def (hard_regno_mode_ok): New hook.
17471 * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17472 (TARGET_HARD_REGNO_MODE_OK): ...this.
17473 * doc/tm.texi.in: Regenerate.
17474 * hooks.h (hook_bool_uint_mode_true): Declare.
17475 * hooks.c (hook_bool_uint_mode_true): New function.
17476 * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17477 HARD_REGNO_MODE_OK.
17478 * genpreds.c (write_insn_preds_c): Add an include of target.h.
17479 * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17480 instead of HARD_REGNO_MODE_OK.
17481 * caller-save.c: Include target.h.
17482 (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17483 HARD_REGNO_MODE_OK.
17484 * combine.c (can_combine_p): Likewise.
17485 (combinable_i3pat): Likewise.
17486 (can_change_dest_mode): Likewise.
17487 * expr.c (init_expr_target): Likewise.
17488 (convert_move): Likewise.
17489 (convert_modes): Likewise.
17490 * ira.c (setup_prohibited_class_mode_regs): Likewise.
17491 (setup_prohibited_mode_move_regs): Likewise.
17492 * ira.h (target_ira): Likewise.
17493 * lra-assigns.c (find_hard_regno_for_1): Likewise.
17494 * lra-constraints.c (process_alt_operands): Likewise.
17495 (split_reg): Likewise.
17496 * recog.c (peep2_find_free_register): Likewise.
17497 * ree.c (combine_reaching_defs): Likewise.
17498 * regcprop.c (maybe_mode_change): Likewise.
17499 * reginfo.c (init_reg_sets_1): Likewise.
17500 (choose_hard_reg_mode): Likewise.
17501 (simplifiable_subregs): Likewise.
17502 * regrename.c (check_new_reg_p): Likewise.
17503 * reload.c (find_valid_class): Likewise.
17504 (find_valid_class_1): Likewise.
17505 (reload_inner_reg_of_subreg): Likewise.
17506 (push_reload): Likewise.
17507 (combine_reloads): Likewise.
17508 (find_dummy_reload): Likewise.
17509 (find_reloads): Likewise.
17510 * reload1.c (find_reg): Likewise.
17511 (set_reload_reg): Likewise.
17512 (allocate_reload_reg): Likewise.
17513 (choose_reload_regs): Likewise.
17514 (reload_adjust_reg_for_temp): Likewise.
17515 * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17516 (simplify_subreg_regno): Likewise.
17517 * sel-sched.c (init_regs_for_mode): Likewise.
17518 * varasm.c (make_decl_rtl): Likewise.
17519 * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17520 (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17521 HARD_REGNO_MODE_OK.
17522 * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17523 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17524 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17525 * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17526 * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17527 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17528 * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17529 (arc_mode_class): Delete.
17530 (HARD_REGNO_MODE_OK): Delete.
17531 * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17532 (arc_hard_regno_mode_ok): Rename old array to...
17533 (arc_hard_regno_mode_ok_modes): ...this.
17534 (arc_conditional_register_usage): Update accordingly.
17535 (arc_mode_class): Make static.
17536 (arc_hard_regno_mode_ok): New function.
17537 * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17538 * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17539 * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17540 (arm_hard_regno_mode_ok): Make static.
17541 * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17542 HARD_REGNO_MODE_OK.
17543 * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17544 * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17545 * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17547 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17548 * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17549 * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17550 * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17551 (bfin_hard_regno_mode_ok): ...this. Make static and return a bool.
17552 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17553 * config/bfin/predicates.md (valid_reg_operand): Use
17554 targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17555 * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17556 * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17557 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17558 * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17559 * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17560 * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17561 (cr16_hard_regno_mode_ok): Make static and return a bool.
17562 * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17563 * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17564 (cris_hard_regno_mode_ok): New function.
17565 * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17566 (epiphany_mode_class): Delete.
17567 (HARD_REGNO_MODE_OK): Delete.
17568 * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17569 * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17570 (hard_regno_mode_ok): Rename to...
17571 (epiphany_hard_regno_mode_ok): ...this. Make static and return a bool.
17572 * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17573 * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17574 HARD_REGNO_MODE_OK.
17575 * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17576 * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17577 * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17578 (frv_hard_regno_mode_ok): Make static and return a bool.
17579 * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17580 HARD_REGNO_MODE_OK.
17581 * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17582 * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17583 * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17584 * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17586 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17587 * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17588 * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17589 * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17591 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17592 * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17593 * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17594 (ia64_hard_regno_mode_ok): New function.
17595 * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17596 * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17597 (iq2000_hard_regno_mode_ok): New function.
17598 * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17599 * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17600 (lm32_hard_regno_mode_ok): New function.
17601 * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17602 * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17603 * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17604 instead of HARD_REGNO_MODE_OK.
17605 (m32c_hard_regno_ok): Rename to...
17606 (m32c_hard_regno_mode_ok): ...this. Make static and return a bool.
17607 (m32c_cannot_change_mode_class): Update accordingly.
17608 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17609 * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17610 (m32r_mode_class): Delete.
17611 (HARD_REGNO_MODE_OK): Delete.
17612 * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17613 (m32r_hard_regno_mode_ok): Rename to...
17614 (m32r_hard_regno_modes): ...this.
17615 (m32r_mode_class): Make static.
17616 (m32r_hard_regno_mode_ok): New function.
17617 * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17618 * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17619 * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17620 (m68k_hard_regno_mode_ok): Make static.
17621 * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17622 * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17623 (mcore_hard_regno_mode_ok): New function.
17624 * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17625 (HARD_REGNO_MODE_OK): Delete.
17626 * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17628 (microblaze_hard_regno_mode_ok_p): ...this and make static.
17629 (microblaze_hard_regno_mode_ok): New function.
17630 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17631 * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17632 (mips_hard_regno_mode_ok): Delete.
17633 * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17634 (mips_hard_regno_mode_ok_p): ...this and make static.
17635 (mips_hard_regno_mode_ok_p): Rename to...
17636 (mips_hard_regno_mode_ok_uncached): ...this.
17637 (mips_hard_regno_mode_ok): New function.
17638 (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17639 of HARD_REGNO_MODE_OK.
17640 (mips_option_override): Update after above name changes.
17641 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17642 * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17643 * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17644 * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17645 * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17646 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17647 * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17648 * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17649 * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17650 * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17651 (msp430_hard_regno_mode_ok): Make static and return a bool.
17652 * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17653 * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17654 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17656 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17657 * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17658 * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17659 * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17660 * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17661 (PA_HARD_REGNO_MODE_OK): ...this
17662 * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17663 (PA_HARD_REGNO_MODE_OK): ...this.
17664 * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17665 (pa_hard_regno_mode_ok): New function.
17666 * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17667 * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17668 (pdp11_hard_regno_mode_ok): New function.
17669 * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17670 * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17672 * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17674 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17675 (rs6000_hard_regno_mode_ok): Rename to...
17676 (rs6000_hard_regno_mode_ok_uncached): ...this.
17677 (rs6000_init_hard_regno_mode_ok): Update accordingly.
17678 (rs6000_hard_regno_mode_ok): New function.
17679 * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17680 * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17681 * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17682 (riscv_hard_regno_mode_ok): ...this and make static.
17683 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17684 * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17685 * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17686 * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17687 (rl78_hard_regno_mode_ok): Make static and return bool.
17688 * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17689 * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17691 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17692 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17693 (rs6000_hard_regno_mode_ok): Rename to...
17694 (rs6000_hard_regno_mode_ok_uncached): ...this.
17695 (rs6000_init_hard_regno_mode_ok): Update accordingly.
17696 (rs6000_hard_regno_mode_ok): New function.
17697 * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17698 * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17699 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17700 * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17701 * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17702 * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17703 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17704 * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17705 * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17706 * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17707 (sh_hard_regno_mode_ok): Make static.
17708 * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17709 instead of HARD_REGNO_MODE_OK.
17710 * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17711 (sparc_mode_class): Delete.
17712 (HARD_REGNO_MODE_OK): Delete.
17713 * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17714 (hard_regno_mode_classes): Make static.
17715 (sparc_mode_class): Likewise.
17716 (sparc_hard_regno_mode_ok): New function.
17717 * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17718 * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17719 * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17721 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17722 * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17723 * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17724 * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17725 * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17726 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17727 * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17728 * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17729 * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17730 (visium_hard_regno_mode_ok): New function.
17731 * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17732 instead of HARD_REGNO_MODE_OK.
17733 * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17734 (HARD_REGNO_MODE_OK): Delete.
17735 * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17736 (xtensa_hard_regno_mode_ok_p): ...this and make static.
17737 (xtensa_option_override): Update accordingly.
17738 (TARGET_HARD_REGNO_MODE_OK): Redefine.
17739 (xtensa_hard_regno_mode_ok): New function.
17740 * system.h (HARD_REGNO_MODE_OK): Poison.
17742 2017-09-04 Richard Sandiford <richard.sandiford@linaro.org>
17743 Alan Hayward <alan.hayward@arm.com>
17744 David Sherwood <david.sherwood@arm.com>
17746 * target.def (hard_regno_call_part_clobbered): New hook.
17747 * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17748 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17749 * doc/tm.texi: Regenerate.
17750 * hooks.h (hook_bool_uint_mode_false): Declare.
17751 * hooks.c (hook_bool_uint_mode_false): New function.
17752 * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17753 * cselib.c (cselib_process_insn): Use
17754 targetm.hard_regno_call_part_clobbered instead of
17755 HARD_REGNO_CALL_PART_CLOBBERED.
17756 * ira-conflicts.c (ira_build_conflicts): Likewise.
17757 * ira-costs.c (ira_tune_allocno_costs): Likewise.
17758 * lra-constraints.c (need_for_call_save_p): Likewise.
17759 * lra-lives.c: Include target.h.
17760 (check_pseudos_live_through_calls): Use
17761 targetm.hard_regno_call_part_clobbered instead of
17762 HARD_REGNO_CALL_PART_CLOBBERED.
17763 * regcprop.c: Include target.h.
17764 (copyprop_hardreg_forward_1): Use
17765 targetm.hard_regno_call_part_clobbered instead of
17766 HARD_REGNO_CALL_PART_CLOBBERED.
17767 * reginfo.c (choose_hard_reg_mode): Likewise.
17768 * regrename.c (check_new_reg_p): Likewise.
17769 * reload.c (find_equiv_reg): Likewise.
17770 * reload1.c (emit_reload_insns): Likewise.
17771 * sched-deps.c (deps_analyze_insn): Likewise.
17772 * sel-sched.c (init_regs_for_mode): Likewise.
17773 (mark_unavailable_hard_regs): Likewise.
17774 * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17775 * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17776 * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17778 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17779 * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17780 * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17782 * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17784 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17785 * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17786 * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17788 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17789 * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17790 * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17792 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17793 * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17795 * config/powerpcspe/powerpcspe.c
17796 (rs6000_hard_regno_call_part_clobbered): New function.
17797 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17798 * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17799 * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17801 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17802 * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17803 * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17805 (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17806 * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17807 * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17809 2017-09-04 Richard Sandiford <richard.sandiford@linaro.org>
17810 Alan Hayward <alan.hayward@arm.com>
17811 David Sherwood <david.sherwood@arm.com>
17813 * rtl.h (subreg_memory_offset): Declare.
17814 * emit-rtl.c (subreg_memory_offset): New function.
17815 * expmed.c (store_bit_field_1): Use it.
17816 * expr.c (undefined_operand_subword_p): Likewise.
17817 * simplify-rtx.c (simplify_subreg): Likewise.
17819 2017-09-04 Alexander Monakov <amonakov@ispras.ru>
17821 PR rtl-optimization/57448
17824 * optabs.c (expand_atomic_load): Place compiler memory barriers if
17825 using atomic_load pattern.
17826 (expand_atomic_store): Likewise.
17828 2017-09-04 Jakub Jelinek <jakub@redhat.com>
17831 * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17832 and UBSAN_BOUNDS internal calls. Clean up IFN_UBSAN_OBJECT_SIZE
17833 handling. Use replace_call_with_value with NULL instead of
17834 gsi_replace, unlink_stmt_vdef and release_defs.
17836 * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17839 * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17841 2017-09-04 Richard Sandiford <richard.sandiford@linaro.org>
17844 * rtl.h (emit_library_call_value_1): Declare.
17845 (emit_library_call): Replace declaration with a series of overloads.
17846 Remove the parameter count argument.
17847 (emit_library_call_value): Likewise.
17848 * calls.c (emit_library_call_value_1): Make global. Replace varargs
17849 with an "rtx_mode_t *".
17850 (emit_library_call_value): Delete.
17851 (emit_library_call): Likewise.
17852 * asan.c (asan_emit_stack_protection): Update calls accordingly.
17853 (asan_emit_allocas_unpoison): Likewise.
17854 * builtins.c (expand_builtin_powi): Likewise.
17855 (expand_asan_emit_allocas_unpoison): Likewise.
17856 * cfgexpand.c (expand_main_function): Likewise.
17857 * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
17858 * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
17859 * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
17860 * config/arm/arm.c (arm_trampoline_init): Likewise.
17861 (arm_call_tls_get_addr): Likewise.
17862 (arm_expand_divmod_libfunc): Likewise.
17863 * config/bfin/bfin.md (umulsi3_highpart): Likewise.
17864 (smulsi3_highpart): Likewise.
17865 * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
17866 (c6x_expand_compare): Likewise.
17867 (c6x_expand_movmem): Likewise.
17868 * config/frv/frv.c (frv_trampoline_init): Likewise.
17869 * config/i386/i386.c (ix86_trampoline_init): Likewise.
17870 (ix86_expand_divmod_libfunc): Likewise.
17871 * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
17872 (ia64_expand_compare): Likewise.
17873 (ia64_profile_hook): Likewise.
17874 * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
17875 (nonlocal_goto): Likewise.
17876 (restore_stack_nonlocal): Likewise.
17877 * config/m32r/m32r.c (block_move_call): Likewise.
17878 (m32r_trampoline_init): Likewise.
17879 * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
17880 * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
17881 (m68k_call_m68k_read_tp): Likewise.
17882 * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
17883 (microblaze_expand_divide): Likewise.
17884 * config/mips/mips.h (mips_args): Likewise.
17885 * config/mips/sdemtk.h (mips_sync_icache): Likewise.
17886 (MIPS_ICACHE_SYNC): Likewise.
17887 * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
17888 (nios2_trampoline_init): Likewise.
17889 * config/pa/pa.c (hppa_tls_call): Likewise.
17890 (pa_trampoline_init): Likewise.
17891 * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
17892 * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
17893 (expand_strn_compare): Likewise.
17894 (rs6000_generate_compare): Likewise.
17895 (rs6000_expand_float128_convert): Likewise.
17896 (output_profile_hook): Likewise.
17897 (rs6000_trampoline_init): Likewise.
17898 * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
17899 * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
17900 * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
17901 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
17902 (rs6000_generate_compare): Likewise.
17903 (rs6000_expand_float128_convert): Likewise.
17904 (output_profile_hook): Likewise.
17905 (rs6000_trampoline_init): Likewise.
17906 * config/rs6000/rs6000.md (neg<mode>2): Likewise.
17907 * config/sh/sh.c (sh_trampoline_init): Likewise.
17908 * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
17909 (sparc_emit_float_lib_cmp): Likewise.
17910 (sparc32_initialize_trampoline): Likewise.
17911 (sparc64_initialize_trampoline): Likewise.
17912 (sparc_profile_hook): Likewise.
17913 * config/spu/spu.c (ea_load_store): Likewise.
17914 * config/spu/spu.md (floatunssidf2): Likewise.
17915 * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
17916 * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
17917 * config/visium/visium.c (expand_block_move_4): Likewise.
17918 (expand_block_move_2): Likewise.
17919 (expand_block_move_1): Likewise.
17920 (expand_block_set_4): Likewise.
17921 (expand_block_set_2): Likewise.
17922 (expand_block_set_1): Likewise.
17923 (visium_trampoline_init): Likewise.
17924 (visium_profile_hook): Likewise.
17925 * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
17926 (xtensa_setup_frame_addresses): Likewise.
17927 (xtensa_trampoline_init): Likewise.
17928 * except.c (sjlj_emit_function_enter): Likewise.
17929 (sjlj_emit_function_exit): Likewise.
17930 * explow.c (allocate_dynamic_stack_space): Likewise.
17931 (probe_stack_range): Likewise.
17932 * expr.c (convert_mode_scalar): Likewise.
17933 * optabs.c (expand_binop): Likewise.
17934 (expand_twoval_binop_libfunc): Likewise.
17935 (expand_unop): Likewise.
17936 (prepare_cmp_insn): Likewise.
17937 (prepare_float_lib_cmp): Likewise.
17938 (expand_float): Likewise.
17939 (expand_fix): Likewise.
17940 (expand_fixed_convert): Likewise.
17941 (maybe_emit_sync_lock_test_and_set): Likewise.
17942 (expand_atomic_compare_and_swap): Likewise.
17943 (expand_mem_thread_fence): Likewise.
17944 (expand_atomic_fetch_op): Likewise.
17946 2017-09-03 Gerald Pfeifer <gerald@pfeifer.com>
17948 * doc/generic.texi (OpenACC): Adjust URL.
17949 * doc/invoke.texi (C Dialect Options): Ditto.
17951 2017-09-03 Uros Bizjak <ubizjak@gmail.com>
17953 * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
17954 predicate for operand 1. Add (m,<S>) constraint.
17955 (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
17956 Prevent memory operand 1 with register operand 2.
17958 2017-09-01 Segher Boessenkool <segher@kernel.crashing.org>
17960 PR rtl-optimization/82024
17961 * combine.c (try_combine): If the combination result is a PARALLEL,
17962 and we only need to retain the SET in there that would be placed
17963 at I2, check that we can place that at I3 instead, before doing so.
17965 2017-09-01 Jakub Jelinek <jakub@redhat.com>
17968 * config/i386/i386.c (ix86_init_large_pic_reg): Return label
17970 (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
17971 if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
17974 2017-09-01 Joerg Sonnenberger <joerg@bec.de>
17975 Jeff Law <law@redhat.com>
17977 * varasm.c (bss_initializer_p): Do not put constants into .bss
17978 (categorize_decl_for_section): Handle bss_initializer_p returning
17979 false when DECL_INITIAL is NULL.
17981 2017-09-01 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17984 * config/s390/s390.c (s390_can_inline_p): New function.
17986 2017-09-01 Jeff Law <law@redhat.com>
17988 PR tree-optimization/82052
17989 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
17990 Always initialize the returned slot after a hash table miss
17991 when INSERT is true.
17993 2017-09-01 Alexander Monakov <amonakov@ispras.ru>
17995 * config/s390/s390.md (mem_signal_fence): Remove.
17996 * doc/md.texi (mem_signal_fence): Remove.
17997 * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
17999 * target-insns.def (mem_signal_fence): Remove.
18001 2017-09-01 Jakub Jelinek <jakub@redhat.com>
18004 * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18007 * asan.c (create_odr_indicator): Strip name encoding from assembler
18008 name before appending it after __odr_asan_.
18010 2017-09-01 Martin Liska <mliska@suse.cz>
18012 PR tree-optimization/82059
18013 * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18014 frequency only when an edge is redirected.
18016 2017-09-01 Claudiu Zissulescu <claziss@synopsys.com>
18018 * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18019 * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18020 (arc_conditional_register_usage): Remove ARC600 lp_count
18022 (arc_file_start): Emit Tag_ARC_CPU_variation.
18023 (arc_can_use_doloop_p): New conditions to use ZOLs.
18024 (hwloop_fail): New function.
18025 (hwloop_optimize): Likewise.
18026 (hwloop_pattern_reg): Likewise.
18027 (arc_doloop_hooks): New struct, to be used with reorg_loops.
18028 (arc_reorg_loops): New function, calls reorg_loops.
18029 (arc_reorg): Call arc_reorg_loops. Remove old ZOL handling.
18030 (arc600_corereg_hazard): Remove ZOL checking, case handled by
18032 (arc_loop_hazard): Remove function, functionality moved into
18034 (arc_hazard): Remove arc_loop_hazard call.
18035 (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18036 into hwloop_optimize.
18037 (arc_label_align): Remove ZOL handling.
18038 * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18039 * config/arc/arc.md (doloop_begin): Remove pattern.
18040 (doloop_begin_i): Likewise.
18041 (doloop_end_i): Likewise.
18042 (doloop_fallback): Likewise.
18043 (doloop_fallback_m): Likewise.
18044 (doloop_end): Reimplement expand.
18045 (arc_lp): New pattern for LP instruction.
18046 (loop_end): New pattern.
18047 (loop_fail): Likewise.
18048 (decrement_and_branch_until_zero): Likewise.
18049 * config/arc/arc.opt (mlpc-width): New option.
18050 * doc/invoke.texi (mlpc-width): Document option.
18052 2017-09-01 Claudiu Zissulescu <claziss@synopsys.com>
18054 * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18055 (arc_ccfsm_advance): Fix checking for delay slots.
18056 (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18058 2017-09-01 Claudiu Zissulescu <claziss@synopsys.com>
18060 * config/arc/arc.md (movqi_insn): Add stores to save constant long
18062 (movhi_insn): Update store instruction constraint which are saving
18063 6-bit short immediates.
18064 (movsi_insn): Consider also short scaled load operations.
18065 (zero_extendhisi2_i): Use Usd constraint instead of T.
18066 (extendhisi2_i): Add q constraint.
18067 (arc_clzsi2): Add type and length attributes.
18068 (arc_ctzsi2): Likewise.
18069 * config/arc/constraints.md (Usc): Update constraint, the
18070 assembler can parse two relocations for a single instruction.
18072 2017-09-01 Claudiu Zissulescu <claziss@synopsys.com>
18074 * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18075 (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18077 2017-08-31 Olivier Hainque <hainque@adacore.com>
18079 * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18080 match as powerpc-wrs-vxworks*.
18082 2017-08-31 James Greenhalgh <james.greenhalgh@arm.com>
18084 * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18085 register constraint for by-element operand.
18086 (aarch64_mls_elt_merge<mode>): Likewise.
18088 2017-08-31 Claudiu Zissulescu <claziss@synopsys.com>
18090 * config/arc/arc.c (arc_can_follow_jump): Check for short
18093 2017-08-31 Claudiu Zissulescu <claziss@synopsys.com>
18095 * config.gcc: Use g.opt for arc.
18096 * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18097 functionality moved to ...
18098 (legitimate_scaled_address_p): New function, ...here.
18099 (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18100 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18101 (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18103 (arc_override_options): Handle G option.
18104 (arc_output_pic_addr_const): Correct function definition.
18105 (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18106 (arc_decl_anon_ns_mem_p): Delete.
18107 (arc_in_small_data_p): Overhaul this function to take into
18108 consideration the value given via G option.
18109 (arc_rewrite_small_data_1): Renamed and corrected old
18110 arc_rewrite_small_data function.
18111 (arc_rewrite_small_data): New function.
18112 (small_data_pattern): Don't use pic_offset_table_rtx.
18113 * config/arc/arc.h (CC1_SPEC): Recognize G option.
18114 * config/arc/simdext.md (movmisalignv2hi): Use
18115 prepare_move_operands function.
18117 (movmisalign*): Likewise.
18118 * doc/invoke.texi (ARC options): Document -G option.
18120 2017-08-31 Claudiu Zissulescu <claziss@synopsys.com>
18122 * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18124 * config/arc/arc.c (arc_print_operand): Output scalled address for
18125 sdata whenever is possible.
18126 (arc_in_small_data_p): Allow sdata for 64bit datum when double
18127 load/stores are available.
18128 (compact_sda_memory_operand): Check for the alignment required by
18129 code density instructions.
18130 * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18132 * config/arc/constraints.md (Usd): Update constraint.
18133 (Us0): New constraint.
18134 (Usc): Update constraint.
18136 2017-08-31 Richard Biener <rguenther@suse.de>
18138 PR middle-end/82054
18139 * dwarf2out.c (dwarf2out_early_global_decl): Process each
18140 function only once.
18142 2017-08-31 Tamar Christina <tamar.christina@arm.com>
18144 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18145 Resize type_signature.
18147 2017-08-31 Richard Sandiford <richard.sandiford@linaro.org>
18148 Alan Hayward <alan.hayward@arm.com>
18149 David Sherwood <david.sherwood@arm.com>
18151 * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18152 subregs whose inner modes can be stored in GPRs.
18153 (aarch64_classify_index): Likewise.
18155 2017-08-31 Richard Sandiford <richard.sandiford@linaro.org>
18156 Alan Hayward <alan.hayward@arm.com>
18157 David Sherwood <david.sherwood@arm.com>
18159 * config/aarch64/iterators.md (V_cmp_result): Rename to...
18160 (V_INT_EQUIV): ...this.
18161 (v_cmp_result): Rename to...
18162 (v_int_equiv): ...this.
18163 * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18164 * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18165 (copysign<mode>3): Likewise.
18166 (aarch64_simd_bsl<mode>_internal): Likewise.
18167 (aarch64_simd_bsl<mode>): Likewise.
18168 (vec_cmp<mode><mode>): Likewise.
18169 (vcond<mode><mode>): Likewise.
18170 (vcond<v_cmp_mixed><mode>): Likewise.
18171 (vcondu<mode><v_cmp_mixed>): Likewise.
18172 (aarch64_cm<optab><mode>): Likewise.
18173 (aarch64_cmtst<mode>): Likewise.
18174 (aarch64_fac<optab><mode>): Likewise.
18175 (vec_perm_const<mode>): Likewise.
18176 (vcond_mask_<mode><v_cmp_result>): Rename to...
18177 (vcond_mask_<mode><v_int_equiv>): ...this.
18178 (vec_cmp<mode><v_cmp_result>): Rename to...
18179 (vec_cmp<mode><v_int_equiv>): ...this.
18181 2017-08-31 Richard Sandiford <richard.sandiford@linaro.org>
18182 Alan Hayward <alan.hayward@arm.com>
18183 David Sherwood <david.sherwood@arm.com>
18185 * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18187 * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18188 * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18189 (UNSPEC_LD4_DREG): New unspecs.
18190 * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18191 (aarch64_ld2<mode>_dreg_be): Replace with...
18192 (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18194 (aarch64_ld3<mode>_dreg_le)
18195 (aarch64_ld3<mode>_dreg_be): Replace with...
18196 (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18198 (aarch64_ld4<mode>_dreg_le)
18199 (aarch64_ld4<mode>_dreg_be): Replace with...
18200 (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18203 2017-08-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18205 PR tree-optimization/81987
18206 * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18207 insert an initializer in a location not dominated by the stride
18210 2017-08-30 Eric Botcazou <ebotcazou@adacore.com>
18212 * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18213 on the entire header of the finally block in the fallthru case.
18215 2017-08-30 Eric Botcazou <ebotcazou@adacore.com>
18217 * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18219 2017-08-30 Pat Haugen <pthaugen@us.ibm.com>
18221 * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18222 rs6000_emit_move_from_cr and call renamed function.
18223 (rs6000_emit_prologue): Call renamed functions.
18224 * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18225 movesi_from_cr, remove volatile CRs.
18227 2017-08-30 Jon Beniston <jon@beniston.com>
18228 Richard Biener <rguenther@suse.de>
18230 * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18231 of VECTOR_MODE_P check.
18232 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18233 element vector types.
18235 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18237 * df.h (df_read_modify_subreg_p): Remove in favor of...
18238 * rtl.h (read_modify_subreg_p): ...this new function. Take a
18239 const_rtx instead of an rtx.
18240 * cprop.c (local_cprop_find_used_regs): Update accordingly.
18241 * df-problems.c (df_word_lr_mark_ref): Likewise.
18242 * ira-lives.c (mark_pseudo_reg_live): Likewise.
18243 (mark_pseudo_reg_dead): Likewise.
18244 (mark_ref_dead): Likewise.
18245 * reginfo.c (init_subregs_of_mode): Likewise.
18246 * sched-deps.c (sched_analyze_1): Likewise.
18247 * df-scan.c (df_def_record_1): Likewise.
18248 (df_uses_record): Likewise.
18249 (df_read_modify_subreg_p): Remove in favor of...
18250 * rtlanal.c (read_modify_subreg_p): ...this new function. Take a
18251 const_rtx instead of an rtx.
18253 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18254 Alan Hayward <alan.hayward@arm.com>
18255 David Sherwood <david.sherwood@arm.com>
18257 * rtl.h (partial_subreg_p): New function.
18258 * caller-save.c (save_call_clobbered_regs): Use it.
18259 * calls.c (expand_call): Likewise.
18260 * combine.c (combinable_i3pat): Likewise.
18261 (simplify_set): Likewise.
18262 (make_extraction): Likewise.
18263 (make_compound_operation_int): Likewise.
18264 (gen_lowpart_or_truncate): Likewise.
18265 (force_to_mode): Likewise.
18266 (make_field_assignment): Likewise.
18267 (reg_truncated_to_mode): Likewise.
18268 (record_truncated_value): Likewise.
18269 (move_deaths): Likewise.
18270 * cse.c (record_jump_cond): Likewise.
18271 (cse_insn): Likewise.
18272 * cselib.c (cselib_lookup_1): Likewise.
18273 * expmed.c (extract_bit_field_using_extv): Likewise.
18274 * function.c (assign_parm_setup_reg): Likewise.
18275 * ifcvt.c (noce_convert_multiple_sets): Likewise.
18276 * ira-build.c (create_insn_allocnos): Likewise.
18277 * lra-coalesce.c (merge_pseudos): Likewise.
18278 * lra-constraints.c (match_reload): Likewise.
18279 (simplify_operand_subreg): Likewise.
18280 (curr_insn_transform): Likewise.
18281 * lra-lives.c (process_bb_lives): Likewise.
18282 * lra.c (new_insn_reg): Likewise.
18283 (lra_substitute_pseudo): Likewise.
18284 * regcprop.c (mode_change_ok): Likewise.
18285 (maybe_mode_change): Likewise.
18286 (copyprop_hardreg_forward_1): Likewise.
18287 * reload.c (push_reload): Likewise.
18288 (find_reloads): Likewise.
18289 (find_reloads_subreg_address): Likewise.
18290 * reload1.c (alter_reg): Likewise.
18291 (eliminate_regs_1): Likewise.
18292 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18294 2017-08-30 David Edelsohn <dje.gcc@gmail.com>
18296 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18297 back to if statements, including unpack.
18299 2017-08-30 Martin Liska <mliska@suse.cz>
18301 PR inline-asm/82001
18302 * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18304 (func_checker::compare_asm_inputs_outputs): ... this function.
18305 (func_checker::compare_gimple_asm): Use the function to compare
18306 also ASM constrains.
18307 * ipa-icf-gimple.h: Rename the function.
18309 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18310 Alan Hayward <alan.hayward@arm.com>
18311 David Sherwood <david.sherwood@arm.com>
18313 * coretypes.h (complex_mode): New type.
18314 * gdbhooks.py (build_pretty_printer): Handle it.
18315 * machmode.h (complex_mode): New class.
18316 (complex_mode::includes_p): New function.
18317 (is_complex_int_mode): Likewise.
18318 (is_complex_float_mode): Likewise.
18319 * genmodes.c (get_mode_class): Handle complex mode classes.
18320 * function.c (expand_function_end): Use is_complex_int_mode.
18322 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18323 Alan Hayward <alan.hayward@arm.com>
18324 David Sherwood <david.sherwood@arm.com>
18326 * coretypes.h (scalar_mode_pod): New typedef.
18327 * gdbhooks.py (build_pretty_printer): Handle it.
18328 * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18329 * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18330 * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18331 * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18332 as_a <scalar_mode>.
18334 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18335 Alan Hayward <alan.hayward@arm.com>
18336 David Sherwood <david.sherwood@arm.com>
18338 * machmode.h (mode_for_vector): Take a scalar_mode instead
18340 * stor-layout.c (mode_for_vector): Likewise.
18341 * explow.c (promote_mode): Use as_a <scalar_mode>.
18342 * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18344 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18345 Alan Hayward <alan.hayward@arm.com>
18346 David Sherwood <david.sherwood@arm.com>
18348 * target.def (preferred_simd_mode): Take a scalar_mode
18349 instead of a machine_mode.
18350 * targhooks.h (default_preferred_simd_mode): Likewise.
18351 * targhooks.c (default_preferred_simd_mode): Likewise.
18352 * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18353 * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18354 * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18355 * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18356 * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18357 * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18358 * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18359 * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18361 * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18362 * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18363 * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18364 * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18365 (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18366 * doc/tm.texi: Regenerate.
18367 * optabs-query.c (can_vec_mask_load_store_p): Return false for
18370 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18371 Alan Hayward <alan.hayward@arm.com>
18372 David Sherwood <david.sherwood@arm.com>
18374 * target.def (scalar_mode_supported_p): Take a scalar_mode
18375 instead of a machine_mode.
18376 * targhooks.h (default_scalar_mode_supported_p): Likewise.
18377 * targhooks.c (default_scalar_mode_supported_p): Likewise.
18378 * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18379 * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18380 * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18381 * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18382 * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18383 * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18384 * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18385 * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18386 * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18387 * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18388 * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18389 * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18391 * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18392 * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18393 * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18394 * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18395 * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18397 * doc/tm.texi: Regenerate.
18399 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18400 Alan Hayward <alan.hayward@arm.com>
18401 David Sherwood <david.sherwood@arm.com>
18403 * coretypes.h (opt_scalar_mode): New typedef.
18404 * gdbhooks.py (build_pretty_printers): Handle it.
18405 * machmode.h (mode_iterator::get_2xwider): Add overload for
18407 * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18409 * expr.c (convert_mode_scalar): Likewise.
18410 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18411 * optabs.c (expand_float): Likewise.
18412 (expand_fix): Likewise.
18413 * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18415 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18416 Alan Hayward <alan.hayward@arm.com>
18417 David Sherwood <david.sherwood@arm.com>
18419 * optabs.c (expand_float): Explicitly check for scalars before
18420 using a branching expansion.
18421 (expand_fix): Likewise.
18423 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18424 Alan Hayward <alan.hayward@arm.com>
18425 David Sherwood <david.sherwood@arm.com>
18427 * expr.c (convert_mode): Split scalar handling out into...
18428 (convert_mode_scalar): ...this new function. Treat the modes
18431 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18432 Alan Hayward <alan.hayward@arm.com>
18433 David Sherwood <david.sherwood@arm.com>
18435 * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18437 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18439 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18440 Alan Hayward <alan.hayward@arm.com>
18441 David Sherwood <david.sherwood@arm.com>
18443 * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18444 rather than a machine_mode.
18445 (fixed_from_string): Likewise.
18446 (fixed_convert): Likewise.
18447 (fixed_convert_from_int): Likewise.
18448 (fixed_convert_from_real): Likewise.
18449 (real_convert_from_fixed): Likewise.
18450 * fixed-value.c (fixed_from_double_int): Likewise.
18451 (fixed_from_string): Likewise.
18452 (fixed_convert): Likewise.
18453 (fixed_convert_from_int): Likewise.
18454 (fixed_convert_from_real): Likewise.
18455 (real_convert_from_fixed): Likewise.
18456 * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18458 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18459 Alan Hayward <alan.hayward@arm.com>
18460 David Sherwood <david.sherwood@arm.com>
18462 * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18463 of separate mode class checks. Do not allow vector modes here.
18464 (immed_wide_int_const): Use as_a <scalar_mode>.
18465 * explow.c (trunc_int_for_mode): Likewise.
18466 * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18467 (wi::shwi): Likewise.
18468 (wi::min_value): Likewise.
18469 (wi::max_value): Likewise.
18470 * dwarf2out.c (loc_descriptor): Likewise.
18471 * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18472 for CONST_WIDE_INT.
18474 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18475 Alan Hayward <alan.hayward@arm.com>
18476 David Sherwood <david.sherwood@arm.com>
18478 * tree.h (SCALAR_TYPE_MODE): New macro.
18479 * expr.c (expand_expr_addr_expr_1): Use it.
18480 (expand_expr_real_2): Likewise.
18481 * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18482 (fold_convert_const_fixed_from_int): Likewise.
18483 (fold_convert_const_fixed_from_real): Likewise.
18484 (native_encode_fixed): Likewise
18485 (native_encode_complex): Likewise
18486 (native_encode_vector): Likewise.
18487 (native_interpret_fixed): Likewise.
18488 (native_interpret_real): Likewise.
18489 (native_interpret_complex): Likewise.
18490 (native_interpret_vector): Likewise.
18491 * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18492 (simd_clone_adjust_argument_types): Likewise.
18493 (simd_clone_init_simd_arrays): Likewise.
18494 (simd_clone_adjust): Likewise.
18495 * stor-layout.c (layout_type): Likewise.
18496 * tree.c (build_minus_one_cst): Likewise.
18497 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18498 * tree-inline.c (estimate_move_cost): Likewise.
18499 * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18500 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18501 (vectorizable_reduction): Likewise.
18502 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18503 (vect_recog_mixed_size_cond_pattern): Likewise.
18504 (check_bool_pattern): Likewise.
18505 (adjust_bool_pattern): Likewise.
18506 (search_type_for_mask_1): Likewise.
18507 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18508 * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18509 (vectorizable_load): Likewise.
18510 (vectorizable_store): Likewise.
18511 * ubsan.c (ubsan_encode_value): Likewise.
18512 * varasm.c (output_constant): Likewise.
18514 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18515 Alan Hayward <alan.hayward@arm.com>
18516 David Sherwood <david.sherwood@arm.com>
18518 * coretypes.h (scalar_mode): New class.
18519 * machmode.h (scalar_mode): Likewise.
18520 (scalar_mode::includes_p): New function.
18521 (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18522 * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18523 * genmodes.c (get_mode_class): Handle remaining scalar modes.
18524 * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18525 * expmed.c (store_bit_field_1): Likewise.
18526 (extract_bit_field_1): Likewise.
18527 * expr.c (write_complex_part): Likewise.
18528 (read_complex_part): Likewise.
18529 (emit_move_complex_push): Likewise.
18530 (expand_expr_real_2): Likewise.
18531 * function.c (assign_parm_setup_reg): Likewise.
18532 (assign_parms_unsplit_complex): Likewise.
18533 * optabs.c (expand_binop): Likewise.
18534 * rtlanal.c (subreg_get_info): Likewise.
18535 * simplify-rtx.c (simplify_immed_subreg): Likewise.
18536 * varasm.c (output_constant_pool_2): Likewise.
18538 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18539 Alan Hayward <alan.hayward@arm.com>
18540 David Sherwood <david.sherwood@arm.com>
18542 * expmed.c (extract_high_half): Use scalar_int_mode and remove
18544 (expmed_mult_highpart_optab): Likewise.
18545 (expmed_mult_highpart): Likewise.
18547 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18548 Alan Hayward <alan.hayward@arm.com>
18549 David Sherwood <david.sherwood@arm.com>
18551 * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18552 instead of a machine_mode.
18553 (builtin_memset_read_str): Likewise.
18554 * builtins.c (c_readstr): Likewise.
18555 (builtin_memcpy_read_str): Likewise.
18556 (builtin_strncpy_read_str): Likewise.
18557 (builtin_memset_read_str): Likewise.
18558 (builtin_memset_gen_str): Likewise.
18559 (expand_builtin_signbit): Use scalar_int_mode for local variables.
18560 * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18561 instead of a machine_mode.
18562 * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18564 (make_extraction): Likewise.
18565 (try_widen_shift_mode): Take and return scalar_int_modes instead
18567 * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18568 a scalar_int_mode instead of a machine_mode.
18569 * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18570 (avr_addr_space_pointer_mode): Likewise.
18571 * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18572 * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18573 (msp430_unwind_word_mode): Likewise.
18574 * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18575 (spu_addr_space_pointer_mode): Likewise.
18576 (spu_addr_space_address_mode): Likewise.
18577 (spu_libgcc_cmp_return_mode): Likewise.
18578 (spu_libgcc_shift_count_mode): Likewise.
18579 * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18580 (rl78_addr_space_pointer_mode): Likewise.
18581 (fl78_unwind_word_mode): Likewise.
18582 (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18584 * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18585 * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18586 * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18587 (mips_valid_pointer_mode): Likewise.
18588 * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18589 * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18590 (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18592 (ft32_addr_space_address_mode): Likewise.
18593 * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18594 scalar_int_mode instead of a machine_mode.
18595 (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18597 (m32c_addr_space_address_mode): Likewise.
18598 * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18599 (rs6000_eh_return_filter_mode): Likewise.
18600 * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18601 (rs6000_eh_return_filter_mode): Likewise.
18602 * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18603 (s390_libgcc_shift_count_mode): Likewise.
18604 (s390_unwind_word_mode): Likewise.
18605 (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18607 * target.def (mode_rep_extended): Likewise.
18608 (valid_pointer_mode): Likewise.
18609 (addr_space.valid_pointer_mode): Likewise.
18610 (eh_return_filter_mode): Return a scalar_int_mode rather than
18612 (libgcc_cmp_return_mode): Likewise.
18613 (libgcc_shift_count_mode): Likewise.
18614 (unwind_word_mode): Likewise.
18615 (addr_space.pointer_mode): Likewise.
18616 (addr_space.address_mode): Likewise.
18617 * doc/tm.texi: Regenerate.
18618 * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18620 (do_jump): Use scalar_int_mode for local variables.
18621 * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18622 rather than a machine_mode.
18623 * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18624 (scompare_loc_descriptor_wide): Likewise.
18625 (scompare_loc_descriptor_narrow): Likewise.
18626 * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18628 * except.c (sjlj_emit_dispatch_table): Likewise.
18629 (expand_builtin_eh_copy_values): Likewise.
18630 * explow.c (convert_memory_address_addr_space_1): Likewise.
18631 Take a scalar_int_mode rather than a machine_mode.
18632 (convert_memory_address_addr_space): Take a scalar_int_mode rather
18633 than a machine_mode.
18634 (memory_address_addr_space): Use scalar_int_mode for local variables.
18635 * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18636 rather than a machine_mode.
18637 * expmed.c (mask_rtx): Likewise.
18638 (init_expmed_one_conv): Likewise.
18639 (expand_mult_highpart_adjust): Likewise.
18640 (extract_high_half): Likewise.
18641 (expmed_mult_highpart_optab): Likewise.
18642 (expmed_mult_highpart): Likewise.
18643 (expand_smod_pow2): Likewise.
18644 (expand_sdiv_pow2): Likewise.
18645 (emit_store_flag_int): Likewise.
18646 (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18648 (extract_low_bits): Likewise.
18649 * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18651 * expr.c (pieces_addr::adjust): Likewise.
18652 (can_store_by_pieces): Likewise.
18653 (store_by_pieces): Likewise.
18654 (clear_by_pieces_1): Likewise.
18655 (expand_expr_addr_expr_1): Likewise.
18656 (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18657 (expand_expr_real_1): Likewise.
18658 (try_casesi): Likewise.
18659 * final.c (shorten_branches): Likewise.
18660 * fold-const.c (fold_convert_const_int_from_fixed): Change the
18661 type of "mode" to machine_mode.
18662 * internal-fn.c (expand_arith_overflow_result_store): Take a
18663 scalar_int_mode rather than a machine_mode.
18664 (expand_mul_overflow): Use scalar_int_mode for local variables.
18665 * loop-doloop.c (doloop_modify): Likewise.
18666 (doloop_optimize): Likewise.
18667 * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18668 than a machine_mode.
18669 (expand_doubleword_shift_condmove): Likewise.
18670 (expand_doubleword_shift): Likewise.
18671 (expand_doubleword_clz): Likewise.
18672 (expand_doubleword_popcount): Likewise.
18673 (expand_doubleword_parity): Likewise.
18674 (expand_absneg_bit): Use scalar_int_mode for local variables.
18675 (prepare_float_lib_cmp): Likewise.
18676 * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18677 rather than a machine_mode.
18678 (convert_memory_address_addr_space): Likewise.
18679 (get_mode_bounds): Likewise.
18680 (get_address_mode): Return a scalar_int_mode rather than a
18682 * rtlanal.c (get_address_mode): Likewise.
18683 * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18684 than a machine_mode.
18685 * targhooks.c (default_mode_rep_extended): Likewise.
18686 (default_valid_pointer_mode): Likewise.
18687 (default_addr_space_valid_pointer_mode): Likewise.
18688 (default_eh_return_filter_mode): Return a scalar_int_mode rather
18689 than a machine_mode.
18690 (default_libgcc_cmp_return_mode): Likewise.
18691 (default_libgcc_shift_count_mode): Likewise.
18692 (default_unwind_word_mode): Likewise.
18693 (default_addr_space_pointer_mode): Likewise.
18694 (default_addr_space_address_mode): Likewise.
18695 * targhooks.h (default_eh_return_filter_mode): Likewise.
18696 (default_libgcc_cmp_return_mode): Likewise.
18697 (default_libgcc_shift_count_mode): Likewise.
18698 (default_unwind_word_mode): Likewise.
18699 (default_addr_space_pointer_mode): Likewise.
18700 (default_addr_space_address_mode): Likewise.
18701 (default_mode_rep_extended): Take a scalar_int_mode rather than
18703 (default_valid_pointer_mode): Likewise.
18704 (default_addr_space_valid_pointer_mode): Likewise.
18705 * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18707 * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18708 rather than a machine_mode.
18709 * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18710 for local variables.
18711 * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18712 * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18713 than a machine_mode.
18715 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18716 Alan Hayward <alan.hayward@arm.com>
18717 David Sherwood <david.sherwood@arm.com>
18719 * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18720 the mode argument to scalar_int_mode.
18721 (do_jump_by_parts_zero_rtx): Likewise.
18722 (do_jump_by_parts_equality_rtx): Likewise.
18723 (do_jump_by_parts_greater): Take a mode argument.
18724 (do_jump_by_parts_equality): Likewise.
18725 (do_jump_1): Update calls accordingly.
18727 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18728 Alan Hayward <alan.hayward@arm.com>
18729 David Sherwood <david.sherwood@arm.com>
18731 * is-a.h (safe_dyn_cast): New function.
18732 * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18733 (jump_table_for_label): Likewise.
18734 * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18735 instead of an rtx_insn *.
18736 (shorten_branches): Use dyn_cast instead of LABEL_P and
18737 JUMP_TABLE_DATA_P. Use jump_table_for_label and
18738 rtx_jump_table_data::get_data_mode.
18739 (final_scan_insn): Likewise.
18741 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18742 Alan Hayward <alan.hayward@arm.com>
18743 David Sherwood <david.sherwood@arm.com>
18745 * combine.c (try_combine): Use is_a <scalar_int_mode> when
18746 trying to combine a full-register integer set with a subreg
18749 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18750 Alan Hayward <alan.hayward@arm.com>
18751 David Sherwood <david.sherwood@arm.com>
18753 * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18754 that is always either address_mode or pointer_mode.
18756 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18757 Alan Hayward <alan.hayward@arm.com>
18758 David Sherwood <david.sherwood@arm.com>
18760 * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18761 when the two are known to be equal.
18763 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18764 Alan Hayward <alan.hayward@arm.com>
18765 David Sherwood <david.sherwood@arm.com>
18767 * simplify-rtx.c (simplify_const_unary_operation): Use
18768 is_a <scalar_int_mode> instead of checking for a nonzero
18769 precision. Forcibly convert op_mode to a scalar_int_mode
18770 in that case. More clearly differentiate the operand and
18771 result modes and use the former when deciding what the value
18772 of a count-bits operation should be. Use is_int_mode instead
18773 of checking for a MODE_INT. Remove redundant check for whether
18774 this mode has a zero precision.
18776 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18777 Alan Hayward <alan.hayward@arm.com>
18778 David Sherwood <david.sherwood@arm.com>
18780 * optabs.c (widen_leading): Change the type of the mode argument
18781 to scalar_int_mode. Use opt_scalar_int_mode for the mode iterator.
18782 (widen_bswap): Likewise.
18783 (expand_parity): Likewise.
18784 (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18785 (expand_ffs): Likewise.
18786 (epand_unop): Check for scalar integer modes before calling the
18789 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18790 Alan Hayward <alan.hayward@arm.com>
18791 David Sherwood <david.sherwood@arm.com>
18793 * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18795 (expand_expr_real_1): Update call accordingly.
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 * expmed.c (store_bit_field_using_insv): Add op0_mode and
18802 value_mode arguments. Use scalar_int_mode internally.
18803 (store_bit_field_1): Rename the new integer mode from imode
18804 to op0_mode and use it instead of GET_MODE (op0). Update calls
18805 to store_split_bit_field, store_bit_field_using_insv and
18806 store_fixed_bit_field.
18807 (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18808 Use scalar_int_mode internally. Use a bit count rather than a mode
18809 when calculating the largest bit size for get_best_mode.
18810 Update calls to store_split_bit_field and store_fixed_bit_field_1.
18811 (store_fixed_bit_field_1): Add mode and value_mode arguments.
18812 Remove assertion that OP0 has a scalar integer mode.
18813 (store_split_bit_field): Add op0_mode and value_mode arguments.
18814 Update calls to extract_fixed_bit_field.
18815 (extract_bit_field_using_extv): Add an op0_mode argument.
18816 Use scalar_int_mode internally.
18817 (extract_bit_field_1): Rename the new integer mode from imode to
18818 op0_mode and use it instead of GET_MODE (op0). Update calls to
18819 extract_split_bit_field, extract_bit_field_using_extv and
18820 extract_fixed_bit_field.
18821 (extract_fixed_bit_field): Add an op0_mode argument. Update calls
18822 to extract_split_bit_field and extract_fixed_bit_field_1.
18823 (extract_fixed_bit_field_1): Add a mode argument. Remove assertion
18824 that OP0 has a scalar integer mode. Use as_a <scalar_int_mode>
18825 on the target mode.
18826 (extract_split_bit_field): Add an op0_mode argument. Update call
18827 to extract_fixed_bit_field.
18829 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18830 Alan Hayward <alan.hayward@arm.com>
18831 David Sherwood <david.sherwood@arm.com>
18833 * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18834 * explow.c (hard_function_value): Likewise.
18835 * expmed.c (extract_fixed_bit_field_1): Likewise. Move the
18836 convert_to_mode call outside the loop.
18837 * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18838 for the mode iterator. Require the mode specified by max_pieces
18840 (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18842 (copy_blkmode_to_reg): Likewise.
18843 (set_storage_via_setmem): Likewise.
18844 * optabs.c (prepare_cmp_insn): Likewise.
18845 * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
18846 * stor-layout.c (finish_bitfield_representative): Likewise.
18848 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18849 Alan Hayward <alan.hayward@arm.com>
18850 David Sherwood <david.sherwood@arm.com>
18852 * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
18853 * expr.c (convert_move): Use them.
18854 (convert_modes): Likewise.
18855 (store_expr_with_bounds): Likewise.
18857 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18858 Alan Hayward <alan.hayward@arm.com>
18859 David Sherwood <david.sherwood@arm.com>
18861 * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
18862 parameter for the mode of "x". Remove the "known_x", "known_mode"
18863 and "known_ret" arguments. Change the type of the mode argument
18864 to scalar_int_mode.
18865 (rtl_hooks:reg_num_sign_bit_copies): Likewise.
18866 * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
18867 (reg_num_sign_bit_copies_for_combine): Likewise.
18868 * rtlanal.c (nonzero_bits1): Likewise.
18869 (num_sign_bit_copies1): Likewise.
18870 * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
18871 (reg_num_sign_bit_copies_general): Likewise.
18872 * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
18873 (reg_nonzero_bits_general): Likewise.
18875 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18876 Alan Hayward <alan.hayward@arm.com>
18877 David Sherwood <david.sherwood@arm.com>
18879 * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
18880 than in subroutines. Return 1 for non-integer modes.
18881 (cached_num_sign_bit_copies): Change the type of the mode parameter
18882 to scalar_int_mode.
18883 (num_sign_bit_copies1): Likewise. Remove early exit for other mode
18884 classes. Handle CONST_INT_P first and then check whether X also
18885 has a scalar integer mode. Check the same thing for inner registers
18886 of a SUBREG and for values that are being extended or truncated.
18888 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18889 Alan Hayward <alan.hayward@arm.com>
18890 David Sherwood <david.sherwood@arm.com>
18892 * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
18893 in subroutines. Return the mode mask for non-integer modes.
18894 (cached_nonzero_bits): Change the type of the mode parameter
18895 to scalar_int_mode.
18896 (nonzero_bits1): Likewise. Remove early exit for other mode
18897 classes. Handle CONST_INT_P first and then check whether X
18898 also has a scalar integer mode.
18900 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18901 Alan Hayward <alan.hayward@arm.com>
18902 David Sherwood <david.sherwood@arm.com>
18904 * expr.c (widest_int_mode_for_size): Make the comment match the code.
18905 Return a scalar_int_mode and assert that the size is greater than
18907 (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
18908 (op_by_pieces_d::op_by_pieces_d): Likewise.
18909 (op_by_pieces_d::run): Likewise.
18910 (can_store_by_pieces): Likewise.
18912 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18913 Alan Hayward <alan.hayward@arm.com>
18914 David Sherwood <david.sherwood@arm.com>
18916 * combine.c (extract_left_shift): Add a mode argument and update
18918 (make_compound_operation_int): Change the type of the mode parameter
18919 to scalar_int_mode and update the call to extract_left_shift.
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 * combine.c (simplify_and_const_int): Change the type of the mode
18926 parameter to scalar_int_mode.
18927 (simplify_and_const_int_1): Likewise. Update recursive call.
18929 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18930 Alan Hayward <alan.hayward@arm.com>
18931 David Sherwood <david.sherwood@arm.com>
18933 * combine.c (simplify_compare_const): Check that the mode is a
18934 scalar_int_mode (rather than VOIDmode) before testing its
18936 (simplify_comparison): Move COMPARISON_P handling out of the
18937 loop and restrict the latter part of the loop to scalar_int_modes.
18938 Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
18939 and when considering SUBREG_REGs. Use is_int_mode instead of
18940 checking GET_MODE_CLASS against MODE_INT.
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 (try_widen_shift_mode): Move check for equal modes to...
18947 (simplify_shift_const_1): ...here. Use scalar_int_mode for
18948 shift_unit_mode and for modes involved in scalar shifts.
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 (force_int_to_mode): New function, split out from...
18955 (force_to_mode): ...here. Keep xmode up-to-date and use it
18956 instead of GET_MODE (x).
18958 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18959 Alan Hayward <alan.hayward@arm.com>
18960 David Sherwood <david.sherwood@arm.com>
18962 * optabs-query.h (extraction_insn::struct_mode): Change type to
18963 opt_scalar_int_mode and update comment.
18964 (extraction_insn::field_mode): Change type to scalar_int_mode.
18965 (extraction_insn::pos_mode): Likewise.
18966 * combine.c (make_extraction): Update accordingly.
18967 * optabs-query.c (get_traditional_extraction_insn): Likewise.
18968 (get_optab_extraction_insn): Likewise.
18969 * recog.c (simplify_while_replacing): Likewise.
18970 * expmed.c (narrow_bit_field_mem): Change the type of the mode
18971 parameter to opt_scalar_int_mode.
18973 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
18974 Alan Hayward <alan.hayward@arm.com>
18975 David Sherwood <david.sherwood@arm.com>
18977 * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
18978 to a scalar_int_mode instead of a machine_mode.
18979 (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
18980 (get_best_mode): Return a boolean and use a pointer argument to store
18981 the selected mode. Replace the limit mode parameter with a bit limit.
18982 * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
18983 for the values returned by bit_field_mode_iterator::next_mode.
18984 (store_bit_field): Update call to get_best_mode.
18985 (store_fixed_bit_field): Likewise.
18986 (extract_fixed_bit_field): Likewise.
18987 * expr.c (optimize_bitfield_assignment_op): Likewise.
18988 * fold-const.c (optimize_bit_field_compare): Likewise.
18989 (fold_truth_andor_1): Likewise.
18990 * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
18991 Update for new type of m_mode.
18992 (get_best_mode): As above.
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 * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
18999 to scalar_int_mode. Remove check for SCALAR_INT_MODE_P.
19000 (store_bit_field): Check is_a <scalar_int_mode> before calling
19001 strict_volatile_bitfield_p.
19002 (extract_bit_field): Likewise.
19004 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19005 Alan Hayward <alan.hayward@arm.com>
19006 David Sherwood <david.sherwood@arm.com>
19008 * target.def (cstore_mode): Return a scalar_int_mode.
19009 * doc/tm.texi: Regenerate.
19010 * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19011 * targhooks.h (default_cstore_mode): Likewise.
19012 * targhooks.c (default_cstore_mode): Likewise, using a forced
19014 * expmed.c (emit_cstore): Expect the target of the cstore to be
19017 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19018 Alan Hayward <alan.hayward@arm.com>
19019 David Sherwood <david.sherwood@arm.com>
19021 * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19023 (niter_desc): Likewise mode.
19024 (iv_analyze): Add a mode parameter.
19026 (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19027 and change its type to scalar_int_mode.
19028 * loop-iv.c: Update commentary at head of file.
19029 (iv_constant): Pass the mode paraeter before the rtx it describes
19030 and change its type to scalar_int_mode. Remove VOIDmode handling.
19031 (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19032 (iv_extend): Likewise.
19033 (shorten_into_mode): Likewise.
19034 (iv_add): Use scalar_int_mode.
19035 (iv_mult): Likewise.
19036 (iv_shift): Likewise.
19037 (canonicalize_iv_subregs): Likewise.
19038 (get_biv_step_1): Pass the outer_mode parameter before the rtx
19039 it describes and change its mode to scalar_int_mode. Also change
19040 the type of the returned inner_mode to scalar_int_mode.
19041 (get_biv_step): Likewise, turning outer_mode from a pointer
19042 into a direct parameter. Update call to get_biv_step_1.
19043 (iv_analyze_biv): Add an outer_mode parameter. Update calls to
19044 iv_constant and get_biv_step.
19045 (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19046 and change its type to scalar_int_mode. Don't initialise iv->mode
19047 to VOIDmode and remove later checks for its still being VOIDmode.
19048 Update calls to iv_analyze_op and iv_analyze_expr. Check
19049 is_a <scalar_int_mode> when changing the mode under consideration.
19050 (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19051 Update call to iv_analyze_expr.
19052 (iv_analyze_op): Add a mode parameter. Reject subregs whose
19053 inner register is not also a scalar_int_mode. Update call to
19055 (iv_analyze): Add a mode parameter. Update call to iv_analyze_op.
19056 (biv_p): Add a mode parameter. Update call to iv_analyze_biv.
19057 (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19058 separate mode class checks. Update calls to iv_analyze. Remove
19059 fix-up of VOIDmodes after iv_analyze_biv.
19060 * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19061 don't have a scalar_int_mode. Update call to biv_p.
19063 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19064 Alan Hayward <alan.hayward@arm.com>
19065 David Sherwood <david.sherwood@arm.com>
19067 * cfgexpand.c (convert_debug_memory_address): Use
19068 as_a <scalar_int_mode>.
19069 * combine.c (expand_compound_operation): Likewise.
19070 (make_extraction): Likewise.
19071 (change_zero_ext): Likewise.
19072 (simplify_comparison): Likewise.
19073 * cse.c (cse_insn): Likewise.
19074 * dwarf2out.c (minmax_loc_descriptor): Likewise.
19075 (mem_loc_descriptor): Likewise.
19076 (loc_descriptor): Likewise.
19077 * expmed.c (init_expmed_one_mode): Likewise.
19078 (synth_mult): Likewise.
19079 (emit_store_flag_1): Likewise.
19080 (expand_divmod): Likewise. Use HWI_COMPUTABLE_MODE_P instead
19081 of a comparison with size.
19082 * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19083 (reduce_to_bit_field_precision): Likewise.
19084 * function.c (expand_function_end): Likewise.
19085 * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19086 * loop-doloop.c (doloop_modify): Likewise.
19087 * optabs.c (expand_binop): Likewise.
19088 (expand_unop): Likewise.
19089 (expand_copysign_absneg): Likewise.
19090 (prepare_cmp_insn): Likewise.
19091 (maybe_legitimize_operand): Likewise.
19092 * recog.c (const_scalar_int_operand): Likewise.
19093 * rtlanal.c (get_address_mode): Likewise.
19094 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19095 (simplify_cond_clz_ctz): Likewise.
19096 * tree-nested.c (get_nl_goto_field): Likewise.
19097 * tree.c (build_vector_type_for_mode): Likewise.
19098 * var-tracking.c (use_narrower_mode): Likewise.
19100 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19101 Alan Hayward <alan.hayward@arm.com>
19102 David Sherwood <david.sherwood@arm.com>
19104 * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19105 * builtins.c (expand_builtin_signbit): Use it.
19106 * cfgexpand.c (expand_debug_expr): Likewise.
19107 * dojump.c (do_jump): Likewise.
19108 (do_compare_and_jump): Likewise.
19109 * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19110 * expmed.c (make_tree): Likewise.
19111 * expr.c (expand_expr_real_2): Likewise.
19112 (expand_expr_real_1): Likewise.
19113 (try_casesi): Likewise.
19114 * fold-const-call.c (fold_const_call_ss): Likewise.
19115 * fold-const.c (unextend): Likewise.
19116 (extract_muldiv_1): Likewise.
19117 (fold_single_bit_test): Likewise.
19118 (native_encode_int): Likewise.
19119 (native_encode_string): Likewise.
19120 (native_interpret_int): Likewise.
19121 * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19122 * internal-fn.c (expand_addsub_overflow): Likewise.
19123 (expand_neg_overflow): Likewise.
19124 (expand_mul_overflow): Likewise.
19125 (expand_arith_overflow): Likewise.
19126 * match.pd: Likewise.
19127 * stor-layout.c (layout_type): Likewise.
19128 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19129 * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19130 * tree-ssanames.c (get_range_info): Likewise.
19131 * tree-switch-conversion.c (array_value_type) Likewise.
19132 * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19133 (vect_recog_divmod_pattern): Likewise.
19134 (vect_recog_mixed_size_cond_pattern): Likewise.
19135 * tree-vrp.c (extract_range_basic): Likewise.
19136 (simplify_float_conversion_using_ranges): Likewise.
19137 * tree.c (int_fits_type_p): Likewise.
19138 * ubsan.c (instrument_bool_enum_load): Likewise.
19139 * varasm.c (mergeable_string_section): Likewise.
19140 (narrowing_initializer_constant_valid_p): Likewise.
19141 (output_constant): Likewise.
19143 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19144 Alan Hayward <alan.hayward@arm.com>
19145 David Sherwood <david.sherwood@arm.com>
19147 * machmode.h (NARROWEST_INT_MODE): New macro.
19148 * expr.c (alignment_for_piecewise_move): Use it instead of
19149 GET_CLASS_NARROWEST_MODE (MODE_INT).
19150 (push_block): Likewise.
19151 * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19153 * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19155 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19156 Alan Hayward <alan.hayward@arm.com>
19157 David Sherwood <david.sherwood@arm.com>
19159 * postreload.c (move2add_valid_value_p): Change the type of the
19160 mode parameter to scalar_int_mode.
19161 (move2add_use_add2_insn): Add a mode parameter and use it instead
19163 (move2add_use_add3_insn): Likewise.
19164 (reload_cse_move2add): Update accordingly.
19166 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19167 Alan Hayward <alan.hayward@arm.com>
19168 David Sherwood <david.sherwood@arm.com>
19170 * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19172 * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19173 * optabs.c (expand_unop): Likewise.
19175 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19176 Alan Hayward <alan.hayward@arm.com>
19177 David Sherwood <david.sherwood@arm.com>
19179 * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19180 (clz_loc_descriptor): Likewise. Remove SCALAR_INT_MODE_P check.
19181 (popcount_loc_descriptor): Likewise.
19182 (bswap_loc_descriptor): Likewise.
19183 (rotate_loc_descriptor): Likewise.
19184 (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19185 calling the functions above.
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 * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19193 (try_combine): Likewise.
19194 (simplify_if_then_else): Likewise.
19195 * cse.c (cse_insn): Likewise.
19196 * dwarf2out.c (mem_loc_descriptor): Likewise.
19197 * emit-rtl.c (gen_lowpart_common): Likewise.
19198 * simplify-rtx.c (simplify_truncation): Likewise.
19199 (simplify_binary_operation_1): Likewise.
19200 (simplify_const_relational_operation): Likewise.
19201 (simplify_ternary_operation): Likewise.
19202 * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19204 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19205 Alan Hayward <alan.hayward@arm.com>
19206 David Sherwood <david.sherwood@arm.com>
19208 * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19209 * reload.c (push_reload): Likewise.
19210 (find_reloads): Likewise.
19212 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19213 Alan Hayward <alan.hayward@arm.com>
19214 David Sherwood <david.sherwood@arm.com>
19216 * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19217 (make_compound_operation_int): Likewise.
19218 (change_zero_ext): Likewise.
19219 * expr.c (convert_move): Likewise.
19220 (convert_modes): Likewise.
19221 * fwprop.c (forward_propagate_subreg): Likewise.
19222 * loop-iv.c (get_biv_step_1): Likewise.
19223 * optabs.c (widen_operand): Likewise.
19224 * postreload.c (move2add_valid_value_p): Likewise.
19225 * recog.c (simplify_while_replacing): Likewise.
19226 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19227 (simplify_binary_operation_1): Likewise. Remove redundant
19228 mode equality check.
19230 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19231 Alan Hayward <alan.hayward@arm.com>
19232 David Sherwood <david.sherwood@arm.com>
19234 * combine.c (combine_simplify_rtx): Add checks for
19235 is_a <scalar_int_mode>.
19236 (simplify_if_then_else): Likewise.
19237 (make_field_assignment): Likewise.
19238 (simplify_comparison): Likewise.
19239 * ifcvt.c (noce_try_bitop): Likewise.
19240 * loop-invariant.c (canonicalize_address_mult): Likewise.
19241 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19243 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19244 Alan Hayward <alan.hayward@arm.com>
19245 David Sherwood <david.sherwood@arm.com>
19247 * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19248 is_a <scalar_int_mode> instead of != BLKmode.
19250 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19251 Alan Hayward <alan.hayward@arm.com>
19252 David Sherwood <david.sherwood@arm.com>
19254 * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19255 instead of != VOIDmode.
19256 * combine.c (if_then_else_cond): Likewise.
19257 (change_zero_ext): Likewise.
19258 * dwarf2out.c (mem_loc_descriptor): Likewise.
19259 (loc_descriptor): Likewise.
19260 * rtlanal.c (canonicalize_condition): Likewise.
19261 * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19263 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19264 Alan Hayward <alan.hayward@arm.com>
19265 David Sherwood <david.sherwood@arm.com>
19267 * simplify-rtx.c (simplify_binary_operation_1): Use
19268 is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19270 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19271 Alan Hayward <alan.hayward@arm.com>
19272 David Sherwood <david.sherwood@arm.com>
19274 * wide-int.h (int_traits<unsigned char>) New class.
19275 (int_traits<unsigned short>) Likewise.
19276 * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19277 Use GET_MODE_UNIT_PRECISION and remove redundant test for
19279 * combine.c (set_nonzero_bits_and_sign_copies): Use
19280 is_a <scalar_int_mode>.
19281 (find_split_point): Likewise.
19282 (combine_simplify_rtx): Likewise.
19283 (simplify_logical): Likewise.
19284 (expand_compound_operation): Likewise.
19285 (expand_field_assignment): Likewise.
19286 (make_compound_operation): Likewise.
19287 (extended_count): Likewise.
19288 (change_zero_ext): Likewise.
19289 (simplify_comparison): Likewise.
19290 * dwarf2out.c (scompare_loc_descriptor): Likewise.
19291 (ucompare_loc_descriptor): Likewise.
19292 (minmax_loc_descriptor): Likewise.
19293 (mem_loc_descriptor): Likewise.
19294 (loc_descriptor): Likewise.
19295 * expmed.c (init_expmed_one_mode): Likewise.
19296 * lra-constraints.c (lra_constraint_offset): Likewise.
19297 * optabs.c (prepare_libcall_arg): Likewise.
19298 * postreload.c (move2add_note_store): Likewise.
19299 * reload.c (operands_match_p): Likewise.
19300 * rtl.h (load_extend_op): Likewise.
19301 * rtlhooks.c (gen_lowpart_general): Likewise.
19302 * simplify-rtx.c (simplify_truncation): Likewise.
19303 (simplify_unary_operation_1): Likewise.
19304 (simplify_binary_operation_1): Likewise.
19305 (simplify_const_binary_operation): Likewise.
19306 (simplify_const_relational_operation): Likewise.
19307 (simplify_subreg): Likewise.
19308 * stor-layout.c (bitwise_mode_for_mode): Likewise.
19309 * var-tracking.c (adjust_mems): Likewise.
19310 (prepare_call_arguments): Likewise.
19312 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19313 Alan Hayward <alan.hayward@arm.com>
19314 David Sherwood <david.sherwood@arm.com>
19316 * machmode.h (is_int_mode): New fuction.
19317 * combine.c (find_split_point): Use it.
19318 (combine_simplify_rtx): Likewise.
19319 (simplify_if_then_else): Likewise.
19320 (simplify_set): Likewise.
19321 (simplify_shift_const_1): Likewise.
19322 (simplify_comparison): Likewise.
19323 * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19324 * cse.c (notreg_cost): Likewise.
19325 (cse_insn): Likewise.
19326 * cselib.c (cselib_lookup_1): Likewise.
19327 * dojump.c (do_jump_1): Likewise.
19328 (do_compare_rtx_and_jump): Likewise.
19329 * dse.c (get_call_args): Likewise.
19330 * dwarf2out.c (rtl_for_decl_init): Likewise.
19331 (native_encode_initializer): Likewise.
19332 * expmed.c (emit_store_flag_1): Likewise.
19333 (emit_store_flag): Likewise.
19334 * expr.c (convert_modes): Likewise.
19335 (store_field): Likewise.
19336 (expand_expr_real_1): Likewise.
19337 * fold-const.c (fold_read_from_constant_string): Likewise.
19338 * gimple-ssa-sprintf.c (get_format_string): Likewise.
19339 * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19340 * optabs.c (expand_binop): Likewise.
19341 (expand_unop): Likewise.
19342 (expand_abs_nojump): Likewise.
19343 (expand_one_cmpl_abs_nojump): Likewise.
19344 * simplify-rtx.c (mode_signbit_p): Likewise.
19345 (val_signbit_p): Likewise.
19346 (val_signbit_known_set_p): Likewise.
19347 (val_signbit_known_clear_p): Likewise.
19348 (simplify_relational_operation_1): Likewise.
19349 * tree.c (vector_type_mode): Likewise.
19351 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19352 Alan Hayward <alan.hayward@arm.com>
19353 David Sherwood <david.sherwood@arm.com>
19355 * machmode.h (smallest_mode_for_size): Fix formatting.
19356 (smallest_int_mode_for_size): New function.
19357 * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19358 instead of smallest_mode_for_size.
19359 * combine.c (make_extraction): Likewise.
19360 * config/arc/arc.c (arc_expand_movmem): Likewise.
19361 * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19362 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19363 * config/s390/s390.c (s390_expand_insv): Likewise.
19364 * config/sparc/sparc.c (assign_int_registers): Likewise.
19365 * config/spu/spu.c (spu_function_value): Likewise.
19366 (spu_function_arg): Likewise.
19367 * coverage.c (get_gcov_type): Likewise.
19368 (get_gcov_unsigned_t): Likewise.
19369 * dse.c (find_shift_sequence): Likewise.
19370 * expmed.c (store_bit_field_1): Likewise.
19371 * expr.c (convert_move): Likewise.
19372 (store_field): Likewise.
19373 * internal-fn.c (expand_arith_overflow): Likewise.
19374 * optabs-query.c (get_best_extraction_insn): Likewise.
19375 * optabs.c (expand_twoval_binop_libfunc): Likewise.
19376 * stor-layout.c (layout_type): Likewise.
19377 (initialize_sizetypes): Likewise.
19378 * targhooks.c (default_get_mask_mode): Likewise.
19379 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19381 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19382 Alan Hayward <alan.hayward@arm.com>
19383 David Sherwood <david.sherwood@arm.com>
19385 * machmode.h (opt_mode::else_blk): New function.
19386 (int_mode_for_mode): Declare.
19387 * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19388 * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19390 * cfgexpand.c (expand_debug_expr): Likewise.
19391 * combine.c (gen_lowpart_or_truncate): Likewise.
19392 (gen_lowpart_for_combine): Likewise.
19393 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19394 * config/avr/avr.c (avr_to_int_mode): Likewise.
19395 (avr_out_plus_1): Likewise.
19396 (avr_out_plus): Likewise.
19397 (avr_out_round): Likewise.
19398 * config/i386/i386.c (ix86_split_to_parts): Likewise.
19399 * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19400 (s390_expand_vcond): Likewise.
19401 * config/spu/spu.c (spu_split_immediate): Likewise.
19402 (spu_expand_mov): Likewise.
19403 * dse.c (get_stored_val): Likewise.
19404 * expmed.c (store_bit_field_1): Likewise.
19405 (convert_extracted_bit_field): Use int_mode_for_mode instead of
19407 (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19408 (extract_low_bits): Likewise.
19409 * expr.c (emit_group_load_1): Likewise. Separate out the BLKmode
19410 handling rather than repeating the check.
19411 (emit_group_store): Likewise.
19412 (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19413 * optabs.c (expand_absneg_bit): Likewise.
19414 (expand_copysign_absneg): Likewise.
19415 (expand_copysign_bit): Likewise.
19416 * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19417 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19418 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19419 * var-tracking.c (prepare_call_arguments): Likewise.
19420 * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19421 int_mode_for_mode instead of mode_for_size.
19422 * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19424 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19425 Alan Hayward <alan.hayward@arm.com>
19426 David Sherwood <david.sherwood@arm.com>
19428 * machmode.h (int_mode_for_size): New function.
19429 * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19430 instead of mode_for_size.
19431 * calls.c (save_fixed_argument_area): Likewise. Make use of BLKmode
19433 * combine.c (expand_field_assignment): Use int_mode_for_size
19434 instead of mode_for_size.
19435 (make_extraction): Likewise.
19436 (simplify_shift_const_1): Likewise.
19437 (simplify_comparison): Likewise.
19438 * dojump.c (do_jump): Likewise.
19439 * dwarf2out.c (mem_loc_descriptor): Likewise.
19440 * emit-rtl.c (init_derived_machine_modes): Likewise.
19441 * expmed.c (flip_storage_order): Likewise.
19442 (convert_extracted_bit_field): Likewise.
19443 * expr.c (copy_blkmode_from_reg): Likewise.
19444 * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19445 * internal-fn.c (expand_mul_overflow): Likewise.
19446 * lower-subreg.c (simple_move): Likewise.
19447 * optabs-libfuncs.c (init_optabs): Likewise.
19448 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19449 * tree.c (vector_type_mode): Likewise.
19450 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19451 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19452 * tree-vect-generic.c (expand_vector_parallel): Likewise.
19453 * tree-vect-stmts.c (vectorizable_load): Likewise.
19454 (vectorizable_store): Likewise.
19456 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19457 Alan Hayward <alan.hayward@arm.com>
19458 David Sherwood <david.sherwood@arm.com>
19460 * coretypes.h (pod_mode): New type.
19461 (scalar_int_mode_pod): New typedef.
19462 * machmode.h (pod_mode): New class.
19463 (int_n_data_t::m): Change type to scalar_int_mode_pod.
19464 * genmodes.c (emit_mode_int_n): Update accordingly.
19465 * lower-subreg.h (target_lower_subreg): Change type to
19466 scalar_int_mode_pod.
19467 * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19468 scalar_int_mode_pod.
19470 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19471 Alan Hayward <alan.hayward@arm.com>
19472 David Sherwood <david.sherwood@arm.com>
19474 * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19475 machine_mode to scalar_int_mode.
19476 * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19477 (rs6000_option_override_internal): Remove cast to int.
19478 * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19479 machine_mode to scalar_int_mode.
19480 * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19481 (rs6000_option_override_internal): Remove cast to int.
19482 * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19483 * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19485 * config/s390/s390.c (s390_expand_builtin): Likewise.
19486 * coretypes.h (scalar_int_mode): New type.
19487 (opt_scalar_int_mode): New typedef.
19488 * machmode.h (scalar_int_mode): New class.
19489 (scalar_int_mode::includes_p): New function.
19490 (byte_mode): Change type to scalar_int_mode.
19491 (word_mode): Likewise.
19492 (ptr_mode): Likewise.
19493 * emit-rtl.c (byte_mode): Likewise.
19494 (word_mode): Likewise.
19495 (ptr_mode): Likewise.
19496 (init_derived_machine_modes): Update accordingly.
19497 * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19498 and MODE_PARTIAL_INT.
19499 * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19500 opt_scalar_int_mode.
19502 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19503 Alan Hayward <alan.hayward@arm.com>
19504 David Sherwood <david.sherwood@arm.com>
19506 * target.def (libgcc_floating_mode_supported_p): Take a
19508 * doc/tm.texi: Regenerate.
19509 * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19511 * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19512 * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19515 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19516 Alan Hayward <alan.hayward@arm.com>
19517 David Sherwood <david.sherwood@arm.com>
19519 * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19520 * doc/tm.texi: Regenerate.
19521 * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19522 * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19523 * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19524 * targhooks.h (default_floatn_mode): Likewise.
19525 * targhooks.c (default_floatn_mode): Likewise.
19526 * tree.c (build_common_tree_nodes): Update accordingly.
19528 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19529 Alan Hayward <alan.hayward@arm.com>
19530 David Sherwood <david.sherwood@arm.com>
19532 * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19533 (mode_iterator::iterate_p): Likewise.
19534 (mode_iterator::get_wider): Likewise.
19535 * expr.c (init_expr_target): Use opt_scalar_float_mode.
19537 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19538 Alan Hayward <alan.hayward@arm.com>
19539 David Sherwood <david.sherwood@arm.com>
19541 * coretypes.h (opt_scalar_float_mode): New typedef.
19542 * machmode.h (float_mode_for_size): New function.
19543 * emit-rtl.c (double_mode): Delete.
19544 (init_emit_once): Use float_mode_for_size.
19545 * stor-layout.c (layout_type): Likewise.
19546 * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19548 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19549 Alan Hayward <alan.hayward@arm.com>
19550 David Sherwood <david.sherwood@arm.com>
19552 * output.h (assemble_real): Take a scalar_float_mode.
19553 * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19554 * config/arm/arm.md (consttable_4): Likewise.
19555 (consttable_8): Likewise.
19556 (consttable_16): Likewise.
19557 * config/mips/mips.md (consttable_float): Likewise.
19558 * config/s390/s390.c (s390_output_pool_entry): Likewise.
19559 * varasm.c (assemble_real): Take a scalar_float_mode.
19560 (output_constant_pool_2): Update accordingly.
19561 (output_constant): Likewise.
19563 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19564 Alan Hayward <alan.hayward@arm.com>
19565 David Sherwood <david.sherwood@arm.com>
19567 * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19568 * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19569 * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19570 (native_encode_real): Likewise.
19571 (native_interpret_real): Likewise.
19572 * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19573 * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19575 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19576 Alan Hayward <alan.hayward@arm.com>
19577 David Sherwood <david.sherwood@arm.com>
19579 * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19580 <scalar_float_mode>. Simplify.
19581 (gen_extend_conv_libfunc): Likewise.
19583 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19584 Alan Hayward <alan.hayward@arm.com>
19585 David Sherwood <david.sherwood@arm.com>
19587 * coretypes.h (scalar_float_mode): New type.
19588 * machmode.h (mode_traits::from_int): Use machine_mode if
19589 USE_ENUM_MODES is defined.
19590 (is_a): New function.
19592 (dyn_cast): Likewise.
19593 (scalar_float_mode): New class.
19594 (scalar_float_mode::includes_p): New function.
19595 (is_float_mode): Likewise.
19596 * gdbhooks.py (MachineModePrinter): New class.
19597 (build_pretty_printer): Use it for scalar_float_mode.
19598 * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19599 (format_helper::format_helper): Turn into a template.
19600 * genmodes.c (get_mode_class): New function.
19601 (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19602 or machine_mode if none.
19603 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19604 as_a <scalar_float_mode>.
19605 * dwarf2out.c (mem_loc_descriptor): Likewise.
19606 (insert_float): Likewise.
19607 (add_const_value_attribute): Likewise.
19608 * simplify-rtx.c (simplify_immed_subreg): Likewise.
19609 * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19610 (expand_unop): Update accordingly.
19611 (expand_abs_nojump): Likewise.
19612 (expand_copysign_absneg): Take a scalar_float_mode.
19613 (expand_copysign_bit): Likewise.
19614 (expand_copysign): Update accordingly.
19616 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19617 Alan Hayward <alan.hayward@arm.com>
19618 David Sherwood <david.sherwood@arm.com>
19620 * coretypes.h (opt_mode): New class.
19621 * machmode.h (opt_mode): Likewise.
19622 (opt_mode::else_void): New function.
19623 (opt_mode::require): Likewise.
19624 (opt_mode::exists): Likewise.
19625 (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19626 (GET_MODE_2XWIDER_MODE): Likewise.
19627 (mode_iterator::get_wider): Update accordingly.
19628 (mode_iterator::get_2xwider): Likewise.
19629 (mode_iterator::get_known_wider): Likewise, turning into a template.
19630 * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19631 forcing a wider mode to exist.
19632 * config/cr16/cr16.h (LONG_REG_P): Likewise.
19633 * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19634 * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19635 GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19636 * lower-subreg.c (init_lower_subreg): Likewise.
19637 * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19638 on the final iteration.
19639 * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19640 a wider mode exists before asking for a move pattern.
19641 (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19642 forcing a wider mode to exist.
19643 (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19644 returning false if no such mode exists.
19645 * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19646 * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19647 * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19648 Avoid checking for a MODE_INT if we already know the mode is not a
19650 (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19651 forcing a wider mode to exist.
19652 (expmed_mult_highpart_optab): Likewise.
19653 (expmed_mult_highpart): Likewise.
19654 * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19656 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19657 * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19658 * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19659 * internal-fn.c (expand_mul_overflow): Update use of
19660 GET_MODE_2XWIDER_MODE.
19661 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19662 * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19663 GET_MODE_WIDER_MODE.
19664 (convert_plusminus_to_widen): Likewise.
19665 * tree-switch-conversion.c (array_value_type): Likewise.
19666 * var-tracking.c (emit_note_insn_var_location): Likewise.
19667 * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19668 Return false inside rather than outside the loop if no wider mode
19670 * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19671 and GET_MODE_2XWIDER_MODE
19672 (can_compare_p): Use else_void.
19673 * gdbhooks.py (OptMachineModePrinter): New class.
19674 (build_pretty_printer): Use it for opt_mode.
19676 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19677 Alan Hayward <alan.hayward@arm.com>
19678 David Sherwood <david.sherwood@arm.com>
19680 * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19681 once. Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19683 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19684 Alan Hayward <alan.hayward@arm.com>
19685 David Sherwood <david.sherwood@arm.com>
19687 * machmode.h (mode_traits): New structure.
19688 (get_narrowest_mode): New function.
19689 (mode_iterator::start): Likewise.
19690 (mode_iterator::iterate_p): Likewise.
19691 (mode_iterator::get_wider): Likewise.
19692 (mode_iterator::get_known_wider): Likewise.
19693 (mode_iterator::get_2xwider): Likewise.
19694 (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19695 (FOR_EACH_MODE): Likewise.
19696 (FOR_EACH_MODE_FROM): Likewise.
19697 (FOR_EACH_MODE_UNTIL): Likewise.
19698 (FOR_EACH_WIDER_MODE): Likewise.
19699 (FOR_EACH_2XWIDER_MODE): Likewise.
19700 * builtins.c (expand_builtin_strlen): Use new mode iterators.
19701 * combine.c (simplify_comparison): Likewise
19702 * config/i386/i386.c (type_natural_mode): Likewise.
19703 * cse.c (cse_insn): Likewise.
19704 * dse.c (find_shift_sequence): Likewise.
19705 * emit-rtl.c (init_derived_machine_modes): Likewise.
19706 (init_emit_once): Likewise.
19707 * explow.c (hard_function_value): Likewise.
19708 * expmed.c (extract_fixed_bit_field_1): Likewise.
19709 (extract_bit_field_1): Likewise.
19710 (expand_divmod): Likewise.
19711 (emit_store_flag_1): Likewise.
19712 * expr.c (init_expr_target): Likewise.
19713 (convert_move): Likewise.
19714 (alignment_for_piecewise_move): Likewise.
19715 (widest_int_mode_for_size): Likewise.
19716 (emit_block_move_via_movmem): Likewise.
19717 (copy_blkmode_to_reg): Likewise.
19718 (set_storage_via_setmem): Likewise.
19719 (compress_float_constant): Likewise.
19720 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19721 * optabs-query.c (get_best_extraction_insn): Likewise.
19722 * optabs.c (expand_binop): Likewise.
19723 (expand_twoval_unop): Likewise.
19724 (expand_twoval_binop): Likewise.
19725 (widen_leading): Likewise.
19726 (widen_bswap): Likewise.
19727 (expand_parity): Likewise.
19728 (expand_unop): Likewise.
19729 (prepare_cmp_insn): Likewise.
19730 (prepare_float_lib_cmp): Likewise.
19731 (expand_float): Likewise.
19732 (expand_fix): Likewise.
19733 (expand_sfix_optab): Likewise.
19734 * postreload.c (move2add_use_add2_insn): Likewise.
19735 * reg-stack.c (reg_to_stack): Likewise.
19736 * reginfo.c (choose_hard_reg_mode): Likewise.
19737 * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19738 * stor-layout.c (mode_for_size): Likewise.
19739 (smallest_mode_for_size): Likewise.
19740 (mode_for_vector): Likewise.
19741 (finish_bitfield_representative): Likewise.
19742 * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19743 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19744 * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19745 * var-tracking.c (prepare_call_arguments): Likewise.
19747 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19748 Alan Hayward <alan.hayward@arm.com>
19749 David Sherwood <david.sherwood@arm.com>
19751 * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19752 * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19753 USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19754 * machmode.h (mode_size): Move earlier in file.
19755 (mode_precision): Likewise.
19756 (mode_inner): Likewise.
19757 (mode_nunits): Likewise.
19758 (mode_unit_size): Likewise.
19759 (unit_unit_precision): Likewise.
19760 (mode_wider): Likewise.
19761 (mode_2xwider): Likewise.
19762 (machine_mode): New class.
19763 (mode_to_bytes): New function.
19764 (mode_to_bits): Likewise.
19765 (mode_to_precision): Likewise.
19766 (mode_to_inner): Likewise.
19767 (mode_to_unit_size): Likewise.
19768 (mode_to_unit_precision): Likewise.
19769 (mode_to_nunits): Likewise.
19770 (GET_MODE_SIZE): Use mode_to_bytes.
19771 (GET_MODE_BITSIZE): Use mode_to_bits.
19772 (GET_MODE_PRECISION): Use mode_to_precision.
19773 (GET_MODE_INNER): Use mode_to_inner.
19774 (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19775 (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19776 (GET_MODE_NUNITS): Use mode_to_nunits.
19777 * system.h (ALWAYS_INLINE): New macro.
19778 * config/powerpcspe/powerpcspe-c.c
19779 (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19780 int for arg1_mode and arg2_mode.
19782 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
19783 Alan Hayward <alan.hayward@arm.com>
19784 David Sherwood <david.sherwood@arm.com>
19786 * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19787 Prefix mode names with E_ in case statements.
19788 * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19789 * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19790 (aarch64_split_simd_move): Likewise.
19791 (aarch64_gen_storewb_pair): Likewise.
19792 (aarch64_gen_loadwb_pair): Likewise.
19793 (aarch64_gen_store_pair): Likewise.
19794 (aarch64_gen_load_pair): Likewise.
19795 (aarch64_get_condition_code_1): Likewise.
19796 (aarch64_constant_pool_reload_icode): Likewise.
19797 (get_rsqrte_type): Likewise.
19798 (get_rsqrts_type): Likewise.
19799 (get_recpe_type): Likewise.
19800 (get_recps_type): Likewise.
19801 (aarch64_gimplify_va_arg_expr): Likewise.
19802 (aarch64_simd_container_mode): Likewise.
19803 (aarch64_emit_load_exclusive): Likewise.
19804 (aarch64_emit_store_exclusive): Likewise.
19805 (aarch64_expand_compare_and_swap): Likewise.
19806 (aarch64_gen_atomic_cas): Likewise.
19807 (aarch64_emit_bic): Likewise.
19808 (aarch64_emit_atomic_swap): Likewise.
19809 (aarch64_emit_atomic_load_op): Likewise.
19810 (aarch64_evpc_trn): Likewise.
19811 (aarch64_evpc_uzp): Likewise.
19812 (aarch64_evpc_zip): Likewise.
19813 (aarch64_evpc_ext): Likewise.
19814 (aarch64_evpc_rev): Likewise.
19815 (aarch64_evpc_dup): Likewise.
19816 (aarch64_gen_ccmp_first): Likewise.
19817 (aarch64_gen_ccmp_next): Likewise.
19818 * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19819 (alpha_emit_xfloating_libcall): Likewise.
19820 (emit_insxl): Likewise.
19821 (alpha_arg_type): Likewise.
19822 * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19823 (arc_preferred_simd_mode): Likewise.
19824 (arc_secondary_reload): Likewise.
19825 (get_arc_condition_code): Likewise.
19826 (arc_print_operand): Likewise.
19827 (arc_legitimate_constant_p): Likewise.
19828 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19829 * config/arc/arc.md (casesi_load): Likewise.
19830 (casesi_compact_jump): Likewise.
19831 * config/arc/predicates.md (proper_comparison_operator): Likewise.
19832 (cc_use_register): Likewise.
19833 * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19834 * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19835 (arm_init_iwmmxt_builtins): Likewise.
19836 * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19837 (neon_expand_vector_init): Likewise.
19838 (arm_attr_length_move_neon): Likewise.
19839 (maybe_get_arm_condition_code): Likewise.
19840 (arm_emit_vector_const): Likewise.
19841 (arm_preferred_simd_mode): Likewise.
19842 (arm_output_iwmmxt_tinsr): Likewise.
19843 (thumb1_output_casesi): Likewise.
19844 (thumb2_output_casesi): Likewise.
19845 (arm_emit_load_exclusive): Likewise.
19846 (arm_emit_store_exclusive): Likewise.
19847 (arm_expand_compare_and_swap): Likewise.
19848 (arm_evpc_neon_vuzp): Likewise.
19849 (arm_evpc_neon_vzip): Likewise.
19850 (arm_evpc_neon_vrev): Likewise.
19851 (arm_evpc_neon_vtrn): Likewise.
19852 (arm_evpc_neon_vext): Likewise.
19853 (arm_validize_comparison): Likewise.
19854 * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
19855 * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
19856 * config/avr/avr.c (avr_rtx_costs_1): Likewise.
19857 * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
19858 (c6x_preferred_simd_mode): Likewise.
19859 * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
19860 (epiphany_rtx_costs): Likewise.
19861 * config/epiphany/predicates.md (proper_comparison_operator):
19863 * config/frv/frv.c (condexec_memory_operand): Likewise.
19864 (frv_emit_move): Likewise.
19865 (output_move_single): Likewise.
19866 (output_condmove_single): Likewise.
19867 (frv_hard_regno_mode_ok): Likewise.
19868 (frv_matching_accg_mode): Likewise.
19869 * config/h8300/h8300.c (split_adds_subs): Likewise.
19870 (h8300_rtx_costs): Likewise.
19871 (h8300_print_operand): Likewise.
19872 (compute_mov_length): Likewise.
19873 (output_logical_op): Likewise.
19874 (compute_logical_op_length): Likewise.
19875 (compute_logical_op_cc): Likewise.
19876 (h8300_shift_needs_scratch_p): Likewise.
19877 (output_a_shift): Likewise.
19878 (compute_a_shift_length): Likewise.
19879 (compute_a_shift_cc): Likewise.
19880 (expand_a_rotate): Likewise.
19881 (output_a_rotate): Likewise.
19882 * config/i386/i386.c (classify_argument): Likewise.
19883 (function_arg_advance_32): Likewise.
19884 (function_arg_32): Likewise.
19885 (function_arg_64): Likewise.
19886 (function_value_64): Likewise.
19887 (ix86_gimplify_va_arg): Likewise.
19888 (ix86_legitimate_constant_p): Likewise.
19889 (put_condition_code): Likewise.
19890 (split_double_mode): Likewise.
19891 (ix86_avx256_split_vector_move_misalign): Likewise.
19892 (ix86_expand_vector_logical_operator): Likewise.
19893 (ix86_split_idivmod): Likewise.
19894 (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
19895 (ix86_build_const_vector): Likewise.
19896 (ix86_build_signbit_mask): Likewise.
19897 (ix86_match_ccmode): Likewise.
19898 (ix86_cc_modes_compatible): Likewise.
19899 (ix86_expand_branch): Likewise.
19900 (ix86_expand_sse_cmp): Likewise.
19901 (ix86_expand_sse_movcc): Likewise.
19902 (ix86_expand_int_sse_cmp): Likewise.
19903 (ix86_expand_vec_perm_vpermi2): Likewise.
19904 (ix86_expand_vec_perm): Likewise.
19905 (ix86_expand_sse_unpack): Likewise.
19906 (ix86_expand_int_addcc): Likewise.
19907 (ix86_split_to_parts): Likewise.
19908 (ix86_vectorize_builtin_gather): Likewise.
19909 (ix86_vectorize_builtin_scatter): Likewise.
19910 (avx_vpermilp_parallel): Likewise.
19911 (inline_memory_move_cost): Likewise.
19912 (ix86_tieable_integer_mode_p): Likewise.
19913 (x86_maybe_negate_const_int): Likewise.
19914 (ix86_expand_vector_init_duplicate): Likewise.
19915 (ix86_expand_vector_init_one_nonzero): Likewise.
19916 (ix86_expand_vector_init_one_var): Likewise.
19917 (ix86_expand_vector_init_concat): Likewise.
19918 (ix86_expand_vector_init_interleave): Likewise.
19919 (ix86_expand_vector_init_general): Likewise.
19920 (ix86_expand_vector_set): Likewise.
19921 (ix86_expand_vector_extract): Likewise.
19922 (emit_reduc_half): Likewise.
19923 (ix86_emit_i387_round): Likewise.
19924 (ix86_mangle_type): Likewise.
19925 (ix86_expand_round_sse4): Likewise.
19926 (expand_vec_perm_blend): Likewise.
19927 (canonicalize_vector_int_perm): Likewise.
19928 (ix86_expand_vec_one_operand_perm_avx512): Likewise.
19929 (expand_vec_perm_1): Likewise.
19930 (expand_vec_perm_interleave3): Likewise.
19931 (expand_vec_perm_even_odd_pack): Likewise.
19932 (expand_vec_perm_even_odd_1): Likewise.
19933 (expand_vec_perm_broadcast_1): Likewise.
19934 (ix86_vectorize_vec_perm_const_ok): Likewise.
19935 (ix86_expand_vecop_qihi): Likewise.
19936 (ix86_expand_mul_widen_hilo): Likewise.
19937 (ix86_expand_sse2_abs): Likewise.
19938 (ix86_expand_pextr): Likewise.
19939 (ix86_expand_pinsr): Likewise.
19940 (ix86_preferred_simd_mode): Likewise.
19941 (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
19942 * config/i386/sse.md (*andnot<mode>3): Likewise.
19943 (<mask_codefor><code><mode>3<mask_name>): Likewise.
19944 (*<code><mode>3): Likewise.
19945 * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
19946 (ia64_expand_atomic_op): Likewise.
19947 (ia64_arg_type): Likewise.
19948 (ia64_mode_to_int): Likewise.
19949 (ia64_scalar_mode_supported_p): Likewise.
19950 (ia64_vector_mode_supported_p): Likewise.
19951 (expand_vec_perm_broadcast): Likewise.
19952 * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
19953 (iq2000_function_arg_advance): Likewise.
19954 (iq2000_function_arg): Likewise.
19955 * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
19956 * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
19957 (m68k_libcall_value): Likewise.
19958 (m68k_function_value): Likewise.
19959 (sched_attr_op_type): Likewise.
19960 * config/mcore/mcore.c (mcore_output_move): Likewise.
19961 * config/microblaze/microblaze.c (microblaze_function_arg_advance):
19963 (microblaze_function_arg): Likewise.
19964 * config/mips/mips.c (mips16_build_call_stub): Likewise.
19965 (mips_print_operand): Likewise.
19966 (mips_mode_ok_for_mov_fmt_p): Likewise.
19967 (mips_vector_mode_supported_p): Likewise.
19968 (mips_preferred_simd_mode): Likewise.
19969 (mips_expand_vpc_loongson_even_odd): Likewise.
19970 (mips_expand_vec_unpack): Likewise.
19971 (mips_expand_vi_broadcast): Likewise.
19972 (mips_expand_vector_init): Likewise.
19973 (mips_expand_vec_reduc): Likewise.
19974 (mips_expand_msa_cmp): Likewise.
19975 * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
19976 * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
19977 (cc_flags_for_mode): Likewise.
19978 * config/msp430/msp430.c (msp430_print_operand): Likewise.
19979 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
19980 (nds32_output_casesi_pc_relative): Likewise.
19981 * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19982 * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
19983 (nvptx_gen_unpack): Likewise.
19984 (nvptx_gen_pack): Likewise.
19985 (nvptx_gen_shuffle): Likewise.
19986 (nvptx_gen_wcast): Likewise.
19987 (nvptx_preferred_simd_mode): Likewise.
19988 * config/pa/pa.c (pa_secondary_reload): Likewise.
19989 * config/pa/predicates.md (base14_operand): Likewise.
19990 * config/powerpcspe/powerpcspe-c.c
19991 (altivec_resolve_overloaded_builtin): Likewise.
19992 * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
19994 (rs6000_preferred_simd_mode): Likewise.
19995 (output_vec_const_move): Likewise.
19996 (rs6000_expand_vector_extract): Likewise.
19997 (rs6000_split_vec_extract_var): Likewise.
19998 (reg_offset_addressing_ok_p): Likewise.
19999 (rs6000_legitimate_offset_address_p): Likewise.
20000 (rs6000_legitimize_address): Likewise.
20001 (rs6000_emit_set_const): Likewise.
20002 (rs6000_const_vec): Likewise.
20003 (rs6000_emit_move): Likewise.
20004 (spe_build_register_parallel): Likewise.
20005 (rs6000_darwin64_record_arg_recurse): Likewise.
20006 (swap_selector_for_mode): Likewise.
20007 (spe_init_builtins): Likewise.
20008 (paired_init_builtins): Likewise.
20009 (altivec_init_builtins): Likewise.
20010 (do_load_for_compare): Likewise.
20011 (rs6000_generate_compare): Likewise.
20012 (rs6000_expand_float128_convert): Likewise.
20013 (emit_load_locked): Likewise.
20014 (emit_store_conditional): Likewise.
20015 (rs6000_output_function_epilogue): Likewise.
20016 (rs6000_handle_altivec_attribute): Likewise.
20017 (rs6000_function_value): Likewise.
20018 (emit_fusion_gpr_load): Likewise.
20019 (emit_fusion_p9_load): Likewise.
20020 (emit_fusion_p9_store): Likewise.
20021 * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20022 (fusion_gpr_mem_load): Likewise.
20023 (fusion_addis_mem_combo_load): Likewise.
20024 (fusion_addis_mem_combo_store): Likewise.
20025 * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20026 (fusion_gpr_mem_load): Likewise.
20027 (fusion_addis_mem_combo_load): Likewise.
20028 (fusion_addis_mem_combo_store): Likewise.
20029 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20031 * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20032 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20033 (rs6000_preferred_simd_mode): Likewise.
20034 (output_vec_const_move): Likewise.
20035 (rs6000_expand_vector_extract): Likewise.
20036 (rs6000_split_vec_extract_var): Likewise.
20037 (reg_offset_addressing_ok_p): Likewise.
20038 (rs6000_legitimate_offset_address_p): Likewise.
20039 (rs6000_legitimize_address): Likewise.
20040 (rs6000_emit_set_const): Likewise.
20041 (rs6000_const_vec): Likewise.
20042 (rs6000_emit_move): Likewise.
20043 (rs6000_darwin64_record_arg_recurse): Likewise.
20044 (swap_selector_for_mode): Likewise.
20045 (paired_init_builtins): Likewise.
20046 (altivec_init_builtins): Likewise.
20047 (rs6000_expand_float128_convert): Likewise.
20048 (emit_load_locked): Likewise.
20049 (emit_store_conditional): Likewise.
20050 (rs6000_output_function_epilogue): Likewise.
20051 (rs6000_handle_altivec_attribute): Likewise.
20052 (rs6000_function_value): Likewise.
20053 (emit_fusion_gpr_load): Likewise.
20054 (emit_fusion_p9_load): Likewise.
20055 (emit_fusion_p9_store): Likewise.
20056 * config/rx/rx.c (rx_gen_move_template): Likewise.
20057 (flags_from_mode): Likewise.
20058 * config/s390/predicates.md (s390_alc_comparison): Likewise.
20059 (s390_slb_comparison): Likewise.
20060 * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20061 (s390_vector_mode_supported_p): Likewise.
20062 (s390_cc_modes_compatible): Likewise.
20063 (s390_match_ccmode_set): Likewise.
20064 (s390_canonicalize_comparison): Likewise.
20065 (s390_emit_compare_and_swap): Likewise.
20066 (s390_branch_condition_mask): Likewise.
20067 (s390_rtx_costs): Likewise.
20068 (s390_secondary_reload): Likewise.
20069 (__SECONDARY_RELOAD_CASE): Likewise.
20070 (s390_expand_cs): Likewise.
20071 (s390_preferred_simd_mode): Likewise.
20072 * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20073 * config/sh/sh.c (sh_print_operand): Likewise.
20074 (dump_table): Likewise.
20075 (sh_secondary_reload): Likewise.
20076 * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20077 * config/sh/sh.md (casesi_worker_1): Likewise.
20078 (casesi_worker_2): Likewise.
20079 * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20080 (fcc_comparison_operator): Likewise.
20081 * config/sparc/sparc.c (sparc_expand_move): Likewise.
20082 (emit_soft_tfmode_cvt): Likewise.
20083 (sparc_preferred_simd_mode): Likewise.
20084 (output_cbranch): Likewise.
20085 (sparc_print_operand): Likewise.
20086 (sparc_expand_vec_perm_bmask): Likewise.
20087 (vector_init_bshuffle): Likewise.
20088 * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20089 (spu_vector_mode_supported_p): Likewise.
20090 (spu_expand_insv): Likewise.
20091 (spu_emit_branch_or_set): Likewise.
20092 (spu_handle_vector_attribute): Likewise.
20093 (spu_builtin_splats): Likewise.
20094 (spu_builtin_extract): Likewise.
20095 (spu_builtin_promote): Likewise.
20096 (spu_expand_sign_extend): Likewise.
20097 * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20098 (tilegx_simd_int): Likewise.
20099 * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20100 (tilepro_simd_int): Likewise.
20101 * config/v850/v850.c (const_double_split): Likewise.
20102 (v850_print_operand): Likewise.
20103 (ep_memory_offset): Likewise.
20104 * config/vax/vax.c (vax_rtx_costs): Likewise.
20105 (vax_output_int_move): Likewise.
20106 (vax_output_int_add): Likewise.
20107 (vax_output_int_subtract): Likewise.
20108 * config/visium/predicates.md (visium_branch_operator): Likewise.
20109 * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20110 (visium_print_operand_address): Likewise.
20111 * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20112 * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20113 (xtensa_expand_conditional_branch): Likewise.
20114 (xtensa_copy_incoming_a7): Likewise.
20115 (xtensa_output_literal): Likewise.
20116 * dfp.c (decimal_real_maxval): Likewise.
20117 * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20119 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
20120 Alan Hayward <alan.hayward@arm.com>
20121 David Sherwood <david.sherwood@arm.com>
20123 * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20124 (mode_nunits_inline): Likewise.
20125 (mode_inner_inline): Likewise.
20126 (mode_unit_size_inline): Likewise.
20127 (mode_unit_precision_inline): Likewise.
20128 (emit_insn_modes_h): Likewise. Also emit a #define of the
20130 (emit_mode_wider): Add an E_ prefix to mode names.
20131 (emit_mode_complex): Likewise.
20132 (emit_mode_inner): Likewise.
20133 (emit_mode_adjustments): Likewise.
20134 (emit_mode_int_n): Likewise.
20135 * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20136 (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20137 (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20138 (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20139 (CRC32_BUILTIN, ENTRY): Likewise.
20140 * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20141 (aarch64_pop_regs): Likewise.
20142 (aarch64_process_components): Likewise.
20143 * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20144 * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20145 (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20146 (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20147 * config/arm/arm.c (arm_init_libfuncs): Likewise.
20148 * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20150 * config/i386/i386-builtin.def (pcmpestr): Likewise.
20151 (pcmpistr): Likewise.
20152 * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20153 * config/mmix/mmix.c (mmix_output_condition): Likewise.
20154 * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20156 * config/rl78/rl78.c (mduc_regs): Likewise.
20157 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20158 (htm_expand_builtin): Likewise.
20159 * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20160 * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20161 * config/xtensa/xtensa.c (print_operand): Likewise.
20162 * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20163 (NUM_MODE_VECTOR_INT): Likewise.
20164 * genoutput.c (null_operand): Likewise.
20165 (output_operand_data): Likewise.
20166 * genrecog.c (print_parameter_value): Likewise.
20167 * lra.c (debug_operand_data): Likewise.
20169 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
20171 * dwarf2out.c (scompare_loc_descriptor_wide)
20172 (scompare_loc_descriptor_narrow): New functions, split out from...
20173 (scompare_loc_descriptor): ...here.
20174 * expmed.c (emit_store_flag_int): New function, split out from...
20175 (emit_store_flag): ...here.
20177 2017-08-30 Richard Biener <rguenther@suse.de>
20179 * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20180 (dwarf2out_early_finish): Move setting of AT_pubnames from
20181 early debug output to early finish.
20183 2017-08-30 Jozef Lawrynowicz <jozef.l@somniumtech.com>
20185 * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20186 and -mdata-region to the assembler.
20188 2017-08-30 Richard Biener <rguenther@suse.de>
20190 * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20192 (gen_subprogram_die): Add DW_AT_object_pointer only early.
20193 (dwarf2out_early_global_decl): Only generate a DIE for the
20194 abstract origin if it doesn't already exist or is a declaration DIE.
20195 (resolve_addr): Do not add the linkage name twice when
20196 generating a stub DIE for the DW_TAG_GNU_call_site target.
20198 2017-08-30 Richard Sandiford <richard.sandiford@linaro.org>
20200 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20201 Use machine_mode rather than int for arg1_mode.
20203 2017-08-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20206 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20207 that the second argument of the built-in functions to unpack
20208 128-bit scalar types to 64-bit values is 0 or 1. Change to use a
20209 switch statement instead a lot of if statements.
20210 * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20211 Allow 64-bit values to be in Altivec registers as well as
20212 traditional floating point registers.
20213 (pack<mode>, FMOVE128_VSX iterator): Likewise.
20215 2017-08-29 Alexander Monakov <amonakov@ispras.ru>
20217 * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20218 MAX_REGS_PER_ADDRESS == 1.
20220 2017-08-29 Uros Bizjak <ubizjak@gmail.com>
20222 * config/i386/i386.opt (flag_fentry): Do not init to -1.
20223 * config/i386/i386.c (ix86_option_override_internal): Simplify
20224 setting of opts->x_flag_entry.
20226 2017-08-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20227 Jakub Jelinek <jakub@redhat.com>
20228 Richard Biener <rguenther@suse.de>
20230 PR tree-optimization/81503
20231 * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20232 folded constant fits in the target type; reorder tests for clarity.
20234 2017-08-29 Martin Liska <mliska@suse.cz>
20236 * passes.def: Include pass_lower_switch.
20237 * stmt.c (dump_case_nodes): Remove and move to
20238 tree-switch-conversion.
20239 (case_values_threshold): Likewise.
20240 (expand_switch_as_decision_tree_p): Likewise.
20241 (emit_case_decision_tree): Likewise.
20242 (expand_case): Likewise.
20243 (balance_case_nodes): Likewise.
20244 (node_has_low_bound): Likewise.
20245 (node_has_high_bound): Likewise.
20246 (node_is_bounded): Likewise.
20247 (emit_case_nodes): Likewise.
20248 (struct simple_case_node): New struct.
20249 (add_case_node): Remove.
20250 (emit_case_dispatch_table): Use vector instead of case_list.
20251 (reset_out_edges_aux): Remove.
20252 (compute_cases_per_edge): Likewise.
20253 (expand_case): Build list of simple_case_node.
20254 (expand_sjlj_dispatch_table): Use it.
20255 * tree-switch-conversion.c (struct case_node): Moved from
20256 stmt.c and adjusted.
20257 (emit_case_nodes): Likewise.
20258 (node_has_low_bound): Likewise.
20259 (node_has_high_bound): Likewise.
20260 (node_is_bounded): Likewise.
20261 (case_values_threshold): Likewise.
20262 (reset_out_edges_aux): Likewise.
20263 (compute_cases_per_edge): Likewise.
20264 (add_case_node): Likewise.
20265 (dump_case_nodes): Likewise.
20266 (balance_case_nodes): Likewise.
20267 (expand_switch_as_decision_tree_p): Likewise.
20268 (emit_jump): Likewise.
20269 (emit_case_decision_tree): Likewise.
20270 (try_switch_expansion): Likewise.
20271 (do_jump_if_equal): Likewise.
20272 (emit_cmp_and_jump_insns): Likewise.
20273 (fix_phi_operands_for_edge): New function.
20274 (record_phi_operand_mapping): Likewise.
20275 (class pass_lower_switch): New pass.
20276 (pass_lower_switch::execute): New function.
20277 (make_pass_lower_switch): Likewise.
20278 (conditional_probability):
20279 * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20280 * tree-pass.h: Add make_pass_lower_switch.
20282 2017-08-29 Jozef Lawrynowicz <jozef.l@somniumtech.com>
20285 * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20288 2017-08-29 Richard Biener <rguenther@suse.de>
20290 * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20291 we do not add a DW_AT_inline attribute twice.
20292 (gen_subprogram_die): Remove code setting DW_AT_inline on
20293 DECL_ABSTRACT_P nodes.
20295 2017-08-29 Richard Sandiford <richard.sandiford@linaro.org>
20297 * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20298 calls to internal functions.
20299 (gimplify_modify_expr): Likewise.
20300 * tree-call-cdce.c (use_internal_fn): Likewise.
20301 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20302 (convert_to_divmod): Set the nothrow flag.
20303 * tree-if-conv.c (predicate_mem_writes): Likewise.
20304 * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20305 (vectorizable_call): Likewise.
20306 (vectorizable_store): Likewise.
20307 (vectorizable_load): Likewise.
20308 * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20309 (vect_recog_mask_conversion_pattern): Likewise.
20311 2017-08-29 Martin Liska <mliska@suse.cz>
20314 * gcc.c (driver_handle_option): Add new argument.
20315 * opts-common.c (handle_option): Pass
20316 target_option_override_hook.
20317 * opts-global.c (lang_handle_option): Add new option.
20318 (set_default_handlers): Add new argument.
20319 (decode_options): Likewise.
20320 * opts.c (target_handle_option): Likewise.
20321 (common_handle_option): Call target_option_override_hook.
20322 * opts.h (struct cl_option_handler_func): Add hook for
20323 target option override.
20324 (struct cl_option_handlers): Likewise.
20325 (set_default_handlers): Add new argument.
20326 (decode_options): Likewise.
20327 (common_handle_option): Likewise.
20328 (target_handle_option): Likewise.
20329 * toplev.c (toplev::main): Pass targetm.target_option.override
20332 2017-08-29 Richard Biener <rguenther@suse.de>
20333 Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20335 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20336 life to the active subtree.
20338 2017-08-28 Jeff Law <law@redhat.com>
20340 * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20341 derive_equivalences.
20342 (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20343 Code moved into....
20344 (edge_info::derive_equivalences): New private member function
20346 * tree-ssa-dom.c (class edge_info): Changed from a struct
20347 to a class. Add ctor/dtor, methods and data members.
20348 (edge_info::edge_info): Renamed from allocate_edge_info.
20349 Initialize additional members.
20350 (edge_info::~edge_info): New.
20351 (free_dom_edge_info): Delete the edge info.
20352 (record_edge_info): Use new class & associated member functions.
20353 Tighten forms for testing for edge equivalences.
20354 (record_temporary_equivalences): Iterate over the simple
20355 equivalences rather than assuming there's only one per edge.
20356 (cprop_into_successor_phis): Iterate over the simple
20357 equivalences rather than assuming there's only one per edge.
20358 (optimize_stmt): Use operand_equal_p rather than pointer
20359 equality for mini-DSE code.
20361 2017-08-28 Nathan Sidwell <nathan@acm.org>
20363 * gcc.c (execute): Fold SIGPIPE handling into switch
20364 statement. Adjust internal error message.
20366 2017-08-28 Richard Biener <rguenther@suse.de>
20369 * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20370 Do nothing for removed DIEs.
20372 2017-08-28 Richard Biener <rguenther@suse.de>
20374 PR tree-optimization/81977
20375 * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20378 2017-08-28 Alexander Monakov <amonakov@ispras.ru>
20381 * doc/md.texi (mem_thread_fence): Remove mention of mode. Rewrite.
20382 * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20383 using targetm.gen_mem_thread_fence.
20385 2017-08-27 Uros Bizjak <ubizjak@gmail.com>
20388 * config/i386/i386.md (*<btsc><mode>): Change operand 2
20389 predicate to register_operand. Reorder operands.
20390 (*btr<mode>): Ditto.
20391 (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20392 (*btr<mode>_mask): Ditto.
20394 2017-08-25 Steven Munroe <munroesj@gcc.gnu.org>
20396 * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20397 * config/rs6000/xmmintrin.h: New file.
20398 * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20400 2017-08-25 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20403 * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20404 parameter and_insn and return it.
20405 (recombine_lvx_pattern): Insert a copy to ensure availability of
20406 the base register of the copied masking operation at the point of
20407 the instruction replacement.
20408 (recombine_stvx_pattern): Likewise.
20410 2017-08-25 Michael Meissner <meissner@linux.vnet.ibm.com>
20412 * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20413 undocumented switches.
20414 (-mpower9-dform-vector): Likewise.
20415 (-mpower9-dform): Likewise.
20416 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20417 comments to delete references to -mpower9-dform* switches.
20418 * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20419 Delete reference to -mpower9-dform* switches, test for
20420 -mpower9-vector instead.
20421 * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20422 (OTHER_P9_VECTOR_MASKS): Likewise.
20423 (POWERPC_MASKS): Likewise.
20424 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20425 tests against -mpower9-dform* to -mpower9-vector. Delete code
20426 that checked for -mpower9-dform* consistancy with other options.
20427 Add test for -mpower9-misc to enable other power9 switches.
20428 (rs6000_init_hard_regno_mode_ok): Likewise.
20429 (rs6000_option_override_internal): Likewise.
20430 (rs6000_emit_prologue): Likewise.
20431 (rs6000_emit_epilogue): Likewise.
20432 (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20433 (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20434 (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20436 (emit_fusion_p9_store): Likewise.
20437 * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20438 resetting these macros if the assembler does not support ISA 3.0
20440 (TARGET_P9_DFORM_VECTOR): Likewise.
20441 * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20442 Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20444 2017-08-25 Alan Modra <amodra@gmail.com>
20446 PR rtl-optimization/81747
20447 * cse.c (cse_extended_basic_block): Don't attempt to record
20448 equivalences for degenerate conditional branches that branch
20449 to their fall-through.
20451 2017-08-24 Martin Sebor <msebor@redhat.com>
20453 PR middle-end/81908
20454 * gimple-fold.c (size_must_be_zero_p): New function.
20455 (gimple_fold_builtin_memory_op): Call it.
20457 2017-08-24 Steven Munroe <munroesj@gcc.gnu.org>
20459 * config/rs6000/mm_malloc.h: New file.
20461 2017-08-24 Bin Cheng <bin.cheng@arm.com>
20463 PR tree-optimization/81913
20464 * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20465 analysis when either IVs in condition can wrap.
20467 2017-08-24 Uros Bizjak <ubizjak@gmail.com>
20469 * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20470 * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20472 2017-08-24 Richard Biener <rguenther@suse.de>
20475 * targhooks.c (default_target_can_inline_p): Properly
20476 use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20477 is present and always compare.
20478 * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20479 infer -mfpmath=sse from TARGET_SSE_P.
20480 (ix86_can_inline_p): Properly use target_option_default_node when
20481 no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20483 2017-08-24 Richard Biener <rguenther@suse.de>
20486 * dwarf2out.c (output_die): Handle flag_generate_offload like
20488 (output_comp_unit): Likewise.
20489 (gen_array_type_die): Likewise.
20490 (dwarf2out_early_finish): Likewise.
20491 (note_variable_value_in_expr): Likewise.
20492 (dwarf2out_finish): Likewise. Adjust assert.
20493 * cgraphunit.c (symbol_table::compile): Move setting of
20494 flag_generate_offload earlier ...
20495 (symbol_table::finalize_compilation_unit): ... here, before
20496 early debug finalization.
20498 2017-08-24 Richard Biener <rguenther@suse.de>
20500 * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20501 and ipa-fnsummary.h.
20502 (ix86_can_inline_p): When ix86_fpmath flags do not match
20503 check whether the callee uses FP math at all.
20505 2017-08-24 Aldy Hernandez <aldyh@redhat.com>
20507 PR middle-end/81931
20508 * tree-ssanames.c (get_nonzero_bits): Use element_precision
20509 instead of TYPE_PRECISION.
20511 2017-08-24 Richard Sandiford <richard.sandiford@linaro.org>
20512 Alan Hayward <alan.hayward@arm.com>
20513 David Sherwood <david.sherwood@arm.com>
20515 * combine.c (make_extraction): Use subreg_offset_from_lsb.
20517 2017-08-23 Daniel Santos <daniel.santos@pobox.com>
20519 * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20521 (ix86_frame::stack_realign_allocate): New field.
20522 (struct machine_frame_state): Modify comments.
20523 (machine_frame_state::sp_realigned_fp_end): New field.
20524 * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20525 layout calculation.
20526 (sp_valid_at): Add assertion to assure no attempt to access invalid
20527 offset of a realigned stack.
20528 (fp_valid_at): Likewise.
20529 (choose_baseaddr): Modify comments.
20530 (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20531 ix86_expand_prologue.
20532 (ix86_expand_prologue): Modify stack realignment and allocation.
20533 (ix86_expand_epilogue): Modify comments.
20534 * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20535 avx2_runtime, avx512f, and avx512f_runtime.
20537 2017-08-23 Uros Bizjak <ubizjak@gmail.com>
20539 * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20540 (mstackrealign): Do not init to -1.
20541 * config/i386/i386.c (ix86_option_override_internal):
20542 Check opts_set, not opts when setting default value of
20543 opts->x_ix86_force_align_arg_pointer.
20545 2017-08-23 Richard Biener <rguenther@suse.de>
20547 * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20548 lang_hooks.decl_printable_name.
20549 * print-rtl-function.c (print_rtx_function): Likewise.
20550 * tree-pretty-print.c (dump_function_header): Likewise.
20552 2017-08-23 Richard Biener <rguenther@suse.de>
20555 * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20556 -g0 at compile-time.
20558 2017-08-23 Tamar Christina <tamar.christina@arm.com>
20560 PR middle-end/19706
20561 * doc/sourcebuild.texi (Other hardware attributes):
20564 2017-08-23 Tamar Christina <tamar.christina@arm.com>
20566 PR middle-end/19706
20567 * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20568 Fix single-use check.
20570 2017-08-23 Thomas Preud'homme <thomas.preudhomme@arm.com>
20572 * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20574 2017-08-22 Daniel Santos <daniel.santos@pobox.com>
20576 * doc/install.texi: Modify to add more details on running selected
20579 2017-08-22 Daniel Santos <daniel.santos@pobox.com>
20581 * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20582 is combined with -mabi=ms.
20583 (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20586 2017-08-22 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20588 PR tree-optimization/81488
20589 * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20590 and cached_basis fields.
20591 (MAX_SPREAD): New constant.
20592 (alloc_cand_and_find_basis): Initialize new fields.
20593 (clear_visited): New function.
20594 (create_phi_basis_1): Rename from create_phi_basis, set visited
20595 and cached_basis fields.
20596 (create_phi_basis): New wrapper function.
20597 (phi_add_costs_1): Rename from phi_add_costs, add spread
20598 parameter, set visited field, short-circuit when limits reached.
20599 (phi_add_costs): New wrapper function.
20600 (record_phi_increments_1): Rename from record_phi_increments, set
20602 (record_phi_increments): New wrapper function.
20603 (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20604 (phi_incr_cost): New wrapper function.
20605 (all_phi_incrs_profitable_1): Rename from
20606 all_phi_incrs_profitable, set visited field.
20607 (all_phi_incrs_profitable): New wrapper function.
20609 2017-08-22 Richard Sandiford <richard.sandiford@linaro.org>
20610 Alan Hayward <alan.hayward@arm.com>
20611 David Sherwood <david.sherwood@arm.com>
20613 * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20614 that takes the outer and inner modes.
20615 * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20616 comparison as the canonical test for a paradoxical subreg.
20617 * combine.c (simplify_set): Use paradoxical_subreg_p.
20618 (make_extraction): Likewise.
20619 (force_to_mode): Likewise.
20620 (rtx_equal_for_field_assignment_p): Likewise.
20621 (gen_lowpart_for_combine): Likewise.
20622 (simplify_comparison): Likewise.
20623 * cse.c (equiv_constant): Likewise.
20624 * expmed.c (store_bit_field_1): Likewise.
20625 * final.c (alter_subreg): Likewise.
20626 * fwprop.c (propagate_rtx): Likewise.
20627 (forward_propagate_subreg): Likewise.
20628 * ira-conflicts.c (ira_build_conflicts): Likewise.
20629 * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20630 * lra-constraints.c (curr_insn_transform): Likewise.
20631 (split_reg): Likewise.
20632 * lra-eliminations.c (move_plus_up): Likewise.
20633 (lra_eliminate_regs_1): Likewise.
20634 * recog.c (general_operand): Likewise.
20635 * ree.c (combine_reaching_defs): Likewise.
20636 * reload.c (push_reload): Likewise.
20637 (find_reloads): Likewise.
20638 * reload1.c (elimination_effects): Likewise.
20639 (compute_reload_subreg_offset): Likewise.
20640 (choose_reload_regs): Likewise.
20641 * rtlanal.c (subreg_lsb_1): Likewise.
20642 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20643 (simplify_subreg): Likewise.
20644 * var-tracking.c (track_loc_p): Likewise.
20645 * emit-rtl.c (byte_lowpart_offset): Likewise.
20646 (paradoxical_subreg_p): Delete out-of-line definition.
20648 2017-08-22 Jeff Law <law@redhat.com>
20650 PR tree-optimization/81741
20651 PR tree-optimization/71947
20652 * tree-ssa-dom.c: Include tree-inline.h.
20653 (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20654 equivalences if one is more expensive to compute than the other.
20655 * tree-ssa-scopedtables.h (class const_or_copies): Make
20656 record_const_or_copy_raw method private.
20657 (class avail_exprs_stack): New method simplify_binary_operation.
20658 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20659 avail_exprs_stack::simplify_binary_operation as needed.
20660 (avail_exprs_stack::simplify_binary_operation): New function.
20662 2017-08-22 Sebastian Huber <sebastian.huber@embedded-brains.de>
20664 * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20665 * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20666 (DOT_SYMBOLS): Likewise.
20667 (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20668 (RELOCATABLE_NEEDS_FIXUP): Likewise.
20669 (RS6000_ABI_NAME): Likewise.
20670 (TARGET_CMODEL): Likewise.
20671 (TOC_SECTION_ASM_OP): Likewise.
20672 (SET_CMODEL): New macro.
20673 (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20675 2017-08-22 Richard Biener <rguenther@suse.de>
20677 * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20678 to allow for free-lang-data replacements similar to verify_type_variant.
20680 2017-08-22 Richard Sandiford <richard.sandiford@linaro.org>
20681 Alan Hayward <alan.hayward@arm.com>
20682 David Sherwood <david.sherwood@arm.com>
20684 * config/aarch64/aarch64.md (casesi): Use DImode rather than
20685 VOIDmode for the LABEL_REF.
20687 2017-08-22 Richard Biener <rguenther@suse.de>
20689 * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20691 2017-08-22 Richard Biener <rguenther@suse.de>
20693 * common.opt (feliminate-dwarf2-dups): Ignore.
20694 * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20695 * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20696 same_die_p_wrap, compute_section_prefix,
20697 is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20698 (comdat_symbol_id, comdat_symbol_number): Likewise.
20699 (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20701 (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20702 (output_die): Mark unreachable path unreachable.
20703 (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20704 (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20705 (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20706 (dwarf2out_early_finish): Likewise.
20708 2017-08-22 Aldy Hernandez <aldyh@redhat.com>
20710 * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20712 2017-08-22 Georg-Johann Lay <avr@gjlay.de>
20715 * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20716 not VAR_P. Filter attribute warnings with OPT_Wattributes.
20717 (avr_attribute_table) <io, io_low, address>: Initialize
20718 .decl_required with true.
20720 2017-08-21 Michael Meissner <meissner@linux.vnet.ibm.com>
20722 * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20723 undocumented debugging options.
20724 (-mvsx-scalar-double): Likewise.
20725 (-mallow-df-permute): Likewise.
20726 (-mvectorize-builtins): Likewise.
20727 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20728 (rs6000_builtin_vectorized_function): Likewise.
20729 (rs6000_builtin_md_vectorized_function): Likewise.
20730 (rs6000_opt_vars): Likewise.
20732 2017-08-21 Uros Bizjak <ubizjak@gmail.com>
20735 * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20736 (*btrq_imm): Rename from *btrq.
20737 (*btcq_imm): Rename from *btcq.
20738 (btsc): New code attribute.
20739 (*<btsc><mode>): New insn pattern.
20740 (*btr<mode>): Ditto.
20741 (*<btsc><mode>_mask): New insn_and_split pattern.
20742 (*btr<mode>_mask): Ditto.
20744 2017-08-21 Richard Sandiford <richard.sandiford@linaro.org>
20745 Alan Hayward <alan.hayward@arm.com>
20746 David Sherwood <david.sherwood@arm.com>
20748 * function.c (pad_below): Simplify padding calculation.
20750 2017-08-21 Richard Sandiford <richard.sandiford@linaro.org>
20751 Alan Hayward <alan.hayward@arm.com>
20752 David Sherwood <david.sherwood@arm.com>
20754 * target.def (function_prologue): Remove frame size argument.
20755 (function_epilogue): Likewise.
20756 * doc/tm.texi: Regenerate.
20757 * final.c (final_start_function): Update call to function_prologue.
20758 (final_end_function): Update call to function_epilogue.
20759 (default_function_pro_epilogue): Remove frame size argument.
20760 * output.h (default_function_pro_epilogue): Likewise.
20761 * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20762 (arm_output_function_prologue): Likewise.
20763 * config/frv/frv.c (frv_function_prologue): Likewise.
20764 (frv_function_epilogue): Likewise.
20765 * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20766 * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20767 (ia64_output_function_epilogue): Likewise.
20768 * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20769 (m32r_output_function_epilogue): Likewise.
20770 * config/microblaze/microblaze.c (microblaze_function_prologue)
20771 (microblaze_function_epilogue): Likewise.
20772 * config/mips/mips.c (mips_output_function_prologue): Likewise.
20773 (mips_output_function_epilogue): Likewise.
20774 * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20775 (mmix_target_asm_function_epilogue): Likewise.
20776 * config/msp430/msp430.c (msp430_start_function): Likewise.
20777 * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20778 (nds32_asm_function_epilogue): Likewise.
20779 * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20780 * config/pa/pa.c (pa_output_function_prologue): Likewise.
20781 (pa_output_function_epilogue): Likewise.
20782 * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20783 (rs6000_output_function_epilogue): Likewise.
20784 * config/rl78/rl78.c (rl78_start_function): Likewise.
20785 * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20786 (rs6000_output_function_epilogue): Likewise.
20787 * config/rx/rx.c (rx_output_function_prologue): Likewise.
20788 * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20789 * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20790 (sparc_asm_function_epilogue): Likewise.
20792 2017-08-21 Richard Sandiford <richard.sandiford@linaro.org>
20794 * tree.h (type_has_mode_precision_p): New function.
20795 * convert.c (convert_to_integer_1): Use it.
20796 * expr.c (expand_expr_real_2): Likewise.
20797 (expand_expr_real_1): Likewise.
20798 * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20799 * match.pd: Likewise.
20800 * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20801 * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20802 * tree-tailcall.c (process_assignment): Likewise.
20803 * tree-vect-loop.c (vectorizable_reduction): Likewise.
20804 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20805 (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20806 * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20807 (vectorizable_assignment): Likewise.
20808 (vectorizable_shift): Likewise.
20809 (vectorizable_operation): Likewise.
20810 * tree-vrp.c (register_edge_assert_for_2): Likewise.
20812 2017-08-21 Wilco Dijkstra <wdijkstr@arm.com>
20814 * match.pd: Add pow (C, x) simplification.
20816 2017-08-21 Richard Biener <rguenther@suse.de>
20818 PR tree-optimization/81900
20819 * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20820 for blocks with abnormal predecessors.
20821 (compute_antic): Do not set visited flag prematurely.
20823 2017-08-21 Georg-Johann Lay <avr@gjlay.de>
20826 * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20828 2017-08-21 Richard Sandiford <richard.sandiford@linaro.org>
20830 * stor-layout.h (vector_type_mode): Move to...
20831 * tree.h (vector_type_mode): ...here.
20832 * stor-layout.c (vector_type_mode): Move to...
20833 * tree.c (vector_type_mode): ...here. Include rtl.h and regs.h.
20835 2017-08-21 Richard Biener <rguenther@suse.de>
20837 * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20838 register_external_die hooks.
20839 (debug_false_tree_charstarstar_uhwistar): Declare.
20840 (debug_nothing_tree_charstar_uhwi): Likewise.
20841 * debug.c (do_nothing_debug_hooks): Adjust.
20842 (debug_false_tree_charstarstar_uhwistar): New do nothing.
20843 (debug_nothing_tree_charstar_uhwi): Likewise.
20844 * dbxout.c (dbx_debug_hooks): Adjust.
20845 (xcoff_debug_hooks): Likewise.
20846 * sdbout.c (sdb_debug_hooks): Likewise.
20847 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
20848 * dwarf2out.c (macinfo_label_base): New global.
20849 (dwarf2out_register_external_die): New function for the
20850 register_external_die hook.
20851 (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
20852 (dwarf2_debug_hooks): Use them.
20853 (dwarf2_lineno_debug_hooks): Adjust.
20854 (struct die_struct): Add with_offset flag.
20855 (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
20856 DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
20857 DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
20858 DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
20859 DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
20860 DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
20861 defining section names for the early LTO debug variants.
20862 (reset_indirect_string): New helper.
20863 (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
20864 (print_dw_val): Add support for offsetted symbol references.
20865 (get_ultimate_context): Split out from is_cxx.
20866 (is_cxx): Use get_ultimate_context.
20867 (is_fortran): Add decl overload.
20868 (compute_comp_unit_symbol): Split out worker from
20869 compute_section_prefix.
20870 (compute_section_prefix): Call compute_comp_unit_symbol and
20871 set comdat_type_p here.
20872 (output_die): Skip DIE symbol output for the LTO added one.
20873 Handle DIE symbol references with offset.
20874 (output_comp_unit): Guard section name mangling properly.
20875 For LTO debug sections emit a symbol at the section beginning
20876 which we use to refer to its DIEs.
20877 (add_abstract_origin_attribute): For DIEs registered via
20878 dwarf2out_register_external_die directly refer to the early
20879 DIE rather than indirectly through the shadow one we created.
20880 Remove obsolete call to dwarf2out_abstract_function for
20881 non-function/block origins.
20882 (gen_array_type_die): When generating early LTO debug do
20883 not emit DW_AT_string_length.
20884 (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
20885 late when in LTO. As suggested place a gcc_unreachable for
20886 the DECL_ABSTRACT_P case.
20887 (gen_subprogram_die): Avoid another specification DIE
20888 for early built declarations/definitions for the late LTO case.
20889 (gen_variable_die): Add type references for late duplicated VLA dies
20891 (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
20892 we have the abstract instance already.
20893 (process_scope_var): Adjust decl DIE contexts in LTO which
20894 first puts them in limbo.
20895 (gen_decl_die): Do not generate type DIEs late apart from
20896 types for VLAs or for decls we do not yet have a DIE. Do not
20897 call dwarf2out_abstract_function late.
20898 (dwarf2out_early_global_decl): Make sure to create DIEs
20899 for abstract instances of a decl first.
20900 (dwarf2out_late_global_decl): Adjust comment.
20901 (output_macinfo_op): With multiple macro sections use
20902 macinfo_label_base to distinguish labels.
20903 (output_macinfo): Likewise. Update macinfo_label_base.
20904 Pass in the line info label.
20905 (note_variable_value_in_expr): When generating LTO resolve
20906 all variable values here by generating DIEs as needed.
20907 (init_sections_and_labels): Add early LTO debug flag parameter
20908 and generate different sections and names if set. Add generation
20909 counter for the labels so we can have multiple of them.
20910 (reset_dies): Helper to allow DIEs to be output multiple times.
20911 (dwarf2out_finish): When outputting DIEs to the fat part of an
20912 LTO object first reset DIEs.
20913 (dwarf2out_early_finish): Output early DIEs when generating LTO.
20914 (modified_type_die): Check for decl_ultimate_origin being self
20916 (gen_type_die_with_usage): Likewise.
20917 (gen_typedef_die): Allow decl_ultimate_origin being self.
20918 (set_decl_abstract_flags): Remove.
20919 (set_block_abstract_flags): Likewise.
20920 (dwarf2out_abstract_function): Treat the early generated DIEs
20921 as the abstract copy and only add DW_AT_inline and
20922 DW_AT_artificial here and call set_decl_origin_self.
20923 If the DIE has an abstract origin don't do anything.
20924 * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
20925 if we have none yet (Go fails to build one, PR78628).
20926 (variably_modified_type_p): Prevent endless recursion for Ada
20927 cyclic pointer types.
20928 * lto-streamer-in.c: Include debug.h.
20929 (dref_queue): New global.
20930 (lto_read_tree_1): Stream in DIE references.
20931 (lto_input_tree): Register DIE references.
20932 (input_function): Stream DECL_DEBUG_ARGS.
20933 * lto-streamer-out.c: Include debug.h.
20934 (lto_write_tree_1): Output DIE references.
20935 (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
20936 Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
20937 (output_function): Stream DECL_DEBUG_ARGS.
20938 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20939 Stream DECL_ABSTRACT_ORIGIN.
20940 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20941 (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
20942 DECL_CONTEXT for file-scope decls.
20943 * lto-streamer.h (struct dref_entry): Declare.
20944 (dref_queue): Likewise.
20945 * cfgexpand.c (pass_expand::execute): Do not call the
20946 outlining_inline_function hook here.
20947 * lto-wrapper.c (debug_obj): New global.
20948 (tool_cleanup): Unlink it if required.
20949 (debug_objcopy): New function.
20950 (run_gcc): Handle early debug sections in the IL files by
20951 extracting them to separate files, partially linkin them and
20952 feeding the result back as result to the linker.
20953 * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
20954 DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
20955 DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
20956 sections into a separate segment.
20957 * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
20959 (darwin_asm_dwarf_section): Likewise.
20960 (darwin_asm_output_dwarf_offset): Likewise.
20961 * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
20963 2017-08-21 Richard Sandiford <richard.sandiford@linaro.org>
20964 Alan Hayward <alan.hayward@arm.com>
20965 David Sherwood <david.sherwood@arm.com>
20967 * read-md.h (md_reader::record_potential_iterator_use): Replace
20968 pointer argument with an rtx and an index.
20969 * read-rtl.c (iterator_group::apply_iterator): Likewise.
20970 (apply_mode_iterator): Likewise.
20971 (apply_code_iterator): Likewise.
20972 (apply_int_iterator): Likewise.
20973 (apply_subst_iterator): Likewise.
20974 (record_iterator_use): Likewise.
20975 (record_attribute_use): Likewise.
20976 (md_reader::record_potential_iterator_use): Likewise. Update calls
20977 to record_iterator_use and apply_iterator.
20978 (iterator_use): Replace ptr with x and index.
20979 (attribute_use): Likewise.
20980 (apply_attribute_uses): Update calls to apply_iterator.
20981 (apply_iterators): Likewise. Update initialization of iterator_use.
20982 (rtx_reader::read_rtx_code): Update calls to record_iterator_use
20983 and record_potential_iterator_use.
20984 (rtx_reader::read_rtx_operand): Likewise.
20986 2017-08-21 Richard Sandiford <richard.sandiford@linaro.org>
20987 Alan Hayward <alan.hayward@arm.com>
20988 David Sherwood <david.sherwood@arm.com>
20990 * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
20993 2017-08-21 Richard Biener <rguenther@suse.de>
20995 PR middle-end/81884
20996 * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
20997 at struct end conservatively when comparing common bases.
20999 2017-08-21 Richard Biener <rguenther@suse.de>
21001 * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21002 (mem_ref_in_stmt): Remove.
21003 (determine_max_movement): Use ref index to get at the reference.
21004 (invariantness_dom_walker::before_dom_children): Deal with
21005 lim data already initialized.
21006 (gather_mem_refs_stmt): Initialize lim data and record ref index.
21008 2017-08-19 Uros Bizjak <ubizjak@gmail.com>
21010 * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21011 (TARGET_ISA_ROUND): Ditto.
21012 (TARGET_ROUND): Ditto.
21013 * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21014 * config/i386/i386.md: Ditto.
21015 * config/i386/sse.md: Ditto.
21016 * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21017 with OPTION_MASK_ISA_SSE4_1.
21019 2017-08-19 Uros Bizjak <ubizjak@gmail.com>
21022 * doc/extend.texi (x86 Built-in Functions): Correct the name of
21023 __builtin_ia32_lzcnt_u16.
21025 2017-08-18 Peter Bergner <bergner@vnet.ibm.com>
21028 * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21029 (rs6000_set_current_function): Rewrite function to use it.
21031 2017-08-18 H.J. Lu <hongjiu.lu@intel.com>
21034 * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21035 and TYPE_WARN_IF_NOT_ALIGN.
21036 * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21037 (handle_warn_if_not_align): New.
21038 (place_union_field): Call handle_warn_if_not_align.
21039 (place_field): Call handle_warn_if_not_align.
21040 Copy TYPE_WARN_IF_NOT_ALIGN.
21041 (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21042 (layout_type): Likewise.
21043 * tree-core.h (tree_type_common): Add warn_if_not_align. Set
21045 (tree_decl_common): Add warn_if_not_align.
21046 * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21047 * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21048 (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21049 (DECL_WARN_IF_NOT_ALIGN): Likewise.
21050 (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21051 * doc/extend.texi: Document warn_if_not_aligned attribute.
21052 * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21054 2017-08-17 Martin Liska <mliska@suse.cz>
21057 * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21058 (get_data_dependence): Use it as pointer type.
21059 (distribute_loop): Likewise.
21061 2017-08-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21063 * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21064 (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21065 (p8_vmrgow_<mode>_direct): New define_insn.
21066 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21067 handle endianness for vmrgew and vmrgow permute patterns.
21069 2017-08-17 Peter Bergner <bergner@vnet.ibm.com>
21071 * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21072 * config/rs6000/rs6000-cpus.def: Remove comment.
21073 (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21074 (POWERPC_MASKS): Likewise.
21075 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21076 use of TARGET_VSX_TIMODE.
21077 (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21078 (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21079 TARGET_VSX_TIMODE. Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21080 (rs6000_option_override_internal): Remove dead code.
21081 (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21082 (rs6000_legitimize_reload_address): Likewise.
21083 (rs6000_legitimate_address_p): Likewise.
21084 (rs6000_opt_masks): Delete "vsx-timode".
21085 (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21086 from function comment.
21087 * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21088 * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21089 (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21091 * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21092 * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21093 * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21094 (VSX_TI): Likewise.
21096 (define_peephole2): Likewise.
21098 2017-08-17 Martin Sebor <msebor@redhat.com>
21101 * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21102 past the end of an array.
21103 (test_pp_format): Add test cases.
21105 2017-08-17 Richard Sandiford <richard.sandiford@linaro.org>
21107 * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21108 missing ECF_NOTHROW flags.
21110 2017-08-17 Peter Bergner <bergner@vnet.ibm.com>
21113 * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21114 operands residing in integer registers.
21115 (*vsx_le_perm_load_<mode>): Likewise.
21116 (*vsx_le_perm_store_<mode>): Likewise.
21117 (define_peephole2): Add peepholes to optimize the above.
21119 2017-08-17 Marek Polacek <polacek@redhat.com>
21121 PR middle-end/81814
21122 * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21123 to mimic what shorten_compare did. Change the return type to bool.
21124 (fold_cond_expr_with_comparison): Update call to
21125 operand_equal_for_comparison_p.
21126 (fold_ternary_loc): Likewise.
21128 2017-08-17 Jackson Woodruff <jackson.woodruff@arm.com>
21130 * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21132 (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21134 2017-08-17 Richard Biener <rguenther@suse.de>
21136 * tree-ssa-structalias.c (solve_graph): When propagating
21137 to successors update the graphs succ edges and avoid duplicate work.
21139 2017-08-17 Maxim Ostapenko <m.ostapenko@samsung.com>
21142 * config/i386/i386.c (ix86_option_override_internal): Save target
21143 specific options after ix86_stack_protector_guard_reg was changed.
21145 2017-08-17 Richard Biener <rguenther@suse.de>
21147 PR tree-optimization/81827
21148 * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21149 (new_var_info): Initialize it conservatively.
21150 (get_call_vi): Mark register vars.
21151 (new_scalar_tmp_constraint_exp): Likewise.
21152 (handle_rhs_call): Likewise.
21153 (handle_const_call): Likewise.
21154 (create_function_info_for): Likewise.
21155 (solve_constraints): Sort varinfos to separate register from
21156 non-register vars to pack points-to solution bitmaps during
21159 2017-08-17 Marek Polacek <polacek@redhat.com>
21161 * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21163 2017-08-17 Richard Biener <rguenther@suse.de>
21165 * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21166 to true when overflow is undefined and we saturated the result.
21168 2017-08-17 Alan Modra <amodra@gmail.com>
21171 * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21172 Don't use store multiple if only one reg needs saving.
21173 (interesting_frame_related_regno): New function.
21174 (rs6000_frame_related): Don't emit frame info for regs that
21176 (rs6000_emit_epilogue): Likewise.
21178 2017-08-16 Nathan Sidwell <nathan@acm.org>
21180 * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21181 * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21182 (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21183 * tree.c (free_lang_data_in_type): Use else-if chain. Always
21184 clear TYPE_LANG_1. Remove obsolete member-function stripping.
21185 (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21186 (verify_type): Adjust for TYPE_BINFO move.
21187 * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21188 process TYPE_BINFO directly.
21189 (hash_tree): Likewise.
21190 * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21192 * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21195 2017-08-16 David Malcolm <dmalcolm@redhat.com>
21197 * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21199 2017-08-16 Uros Bizjak <ubizjak@gmail.com>
21202 * config/i386/i386.md (*anddi_1_btr): Change predicates of
21203 operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21204 Add ix86_binary_operator_ok to insn constraint.
21205 (*iordi_1_bts): Ditto.
21206 (*xordi_1_btc): Ditto.
21207 (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21208 Update corresponding peephole2 pattern.
21212 2017-08-16 Bin Cheng <bin.cheng@arm.com>
21214 PR tree-optimization/81832
21215 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21216 copy loop header which has IFN_LOOP_DIST_ALIAS call.
21218 2017-08-16 Marek Polacek <polacek@redhat.com>
21221 * fold-const.c (fold_indirect_ref_1): Restore original behavior
21222 regarding size_zero_node.
21224 2017-08-16 Martin Liska <mliska@suse.cz>
21227 * config.gcc: Respect previously set extra_objs in case
21230 2017-08-16 Richard Sandiford <richard.sandiford@linaro.org>
21232 PR tree-optimization/81835
21233 * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21234 the phi SSA_NAME. Check that the condition in a COND_EXPR does
21235 not depend on the phi.
21237 2017-08-16 Alan Modra <amodra@gmail.com>
21239 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21242 2017-08-16 Alan Modra <amodra@gmail.com>
21244 * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21245 (save_reg_p): ..into this. Update all callers.
21246 (first_reg_to_save): Simplify.
21248 2017-08-16 Alan Modra <amodra@gmail.com>
21250 * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21253 2017-08-15 Joseph Myers <joseph@codesourcery.com>
21257 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21258 constant count if that count is less than 32.
21260 2017-08-15 Nathan Sidwell <nathan@acm.org>
21262 * gcc.c (execute): Emit friendlier message if inferior is killed
21263 by an external cause.
21265 2017-08-15 Richard Biener <rguenther@suse.de>
21267 PR tree-optimization/81790
21268 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21269 CONSTRUCTORs from simplifying and VN.
21271 2017-08-14 Martin Sebor <msebor@redhat.com>
21273 * builtin-attrs.def: Add comments.
21275 2017-08-14 Martin Sebor <msebor@redhat.com>
21278 * doc/extend.texi (attribute nonstring): Document new attribute.
21280 2017-08-14 Martin Sebor <msebor@redhat.com>
21283 * tree-diagnostic.c (default_tree_printer): Handle %G.
21284 * gimple-pretty-print.h (percent_G_format): Declare new function.
21285 * gimple-pretty-print.c (percent_G_format): Define.
21286 * tree-pretty-print.c (percent_K_format): Add argument.
21288 2017-08-14 Martin Sebor <msebor@redhat.com>
21290 PR translation/79998
21291 * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21292 Remove a stray space.
21294 2017-08-14 Uros Bizjak <ubizjak@gmail.com>
21297 * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21298 (*iordi_1_bts): Ditto.
21299 (*xordi_1_btc): Ditto.
21301 2017-08-14 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21304 * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21305 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21307 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21308 quoted strings, and make more translator-friendly.
21309 (darwin_rs6000_override_options): Likewise.
21310 (rs6000_option_override_internal): Likewise.
21311 (rs6000_return_in_memory): Fix overlong line.
21312 (init_cmulative_args): Use quoted strings, and make more
21313 translator-friendly.
21314 (rs6000_pass_by_reference): Fix overlong line.
21315 (def_builtin): Use quoted strings.
21316 (altivec_expand_predicate_builtin): Use quoted strings, and make
21317 more translator-friendly.
21318 (htm_expand_builtin): Use quoted strings.
21319 (cpu_expand_builtin): Use quoted strings, and make more
21320 translator-friendly.
21321 (altivec_expand_builtin): Likewise.
21322 (paired_expand_predicate_builtin): Likewise.
21323 (rs6000_invalid_builtin): Likewise.
21324 (builtin_function_type): Use quoted strings.
21325 (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21326 more translator-friendly.
21327 (rs6000_trampoline_init): Likewise.
21328 (rs6000_handle_altivec_attribute): Likewise.
21329 (rs6000_inner_target_options): Use quoted strings.
21330 (rs6000_disable_incompatible_switches): Likewise.
21331 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21332 strings, and make more translator-friendly.
21333 (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21335 2017-08-14 Bin Cheng <bin.cheng@arm.com>
21337 PR tree-optimization/81799
21338 * tree-loop-distribution.c (version_loop_by_alias_check): Force
21339 cond_expr to simple gimple operand.
21341 2017-08-14 Wilco Dijkstra <wdijkstr@arm.com>
21343 PR middle-end/46932
21344 * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21346 2017-08-14 Georg-Johann Lay <avr@gjlay.de>
21350 * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21351 * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21352 TARGET_GASISR_PROLOGUES.
21353 * config/avr/avr.c (avr_option_override): Same.
21354 (avr_pass_pre_proep::execute): Same.
21356 2017-08-13 H.J. Lu <hongjiu.lu@intel.com>
21359 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21360 frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21362 2017-08-13 Uros Bizjak <ubizjak@gmail.com>
21364 * config/i386/i386.md (*load_tp_<mode>): Redefine as
21365 define_insn_and_split. Split to a memory load from 0 in
21366 DEFAULT_TLS_SEG_REG address space. Merge with *load_tp_x32
21367 using PTR mode iterator.
21368 (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21369 Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21370 (*add_tp_<mode>): Redefine as define_insn_and_split.
21371 Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21372 address space. Merge with *add_tp_x32 using PTR mode iterator.
21373 (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21374 Split to an add with a memory load from 0 in
21375 DEFAULT_TLS_SEG_REG address space.
21377 2017-08-12 Andrew Pinski <apinski@cavium.com>
21379 * config/aarch64/aarch64-option-extensions.def (rdma):
21380 Fix feature string to what Linux prints out in /proc/cpuinfo.
21382 2017-08-12 Pierre-Marie de Rodat <derodat@adacore.com>
21385 * dwarf2out.c (modified_type_die): For C typedef types that have
21386 an ultimate origin, process the ultimate origin instead of the
21388 (gen_typedef_die): Assert that input DECLs have no ultimate
21390 (gen_type_die_with_usage): For typedef variants that have an
21391 ultimate origin, just call gen_decl_die on the original DECL.
21392 (process_scope_var): Avoid creating DIEs for local typedefs and
21393 concrete static variables.
21395 2017-08-12 Alan Modra <amodra@gmail.com>
21399 * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21400 match gnu-user.h startfile.
21401 (ENDFILE_LINUX_SPEC): Similarly.
21403 2017-08-11 Thomas Schwinge <thomas@codesourcery.com>
21406 * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21408 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21410 2017-08-11 Tamar Christina <tamar.christina@arm.com>
21411 * config/aarch64/aarch64.md (mov<mode>): Change.
21412 (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21413 aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21414 * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21416 2017-08-11 Eric Botcazou <ebotcazou@adacore.com>
21418 * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21419 for storage order barriers.
21421 2017-08-11 Martin Liska <mliska@suse.cz>
21424 * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21425 variables of void type.
21427 2017-08-11 Martin Liska <mliska@suse.cz>
21429 * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21430 TARGET_SUPPORTS_ALIASES.
21431 * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21432 * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21433 (optimize_weakref): Likewise.
21434 * symtab.c (symtab_node::noninterposable_alias): Likewise.
21435 * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21436 * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21438 2017-08-11 Martin Liska <mliska@suse.cz>
21441 * config/i386/i386.c (make_resolver_func): Do complete
21442 refactoring of the function.
21444 2017-08-10 Uros Bizjak <ubizjak@gmail.com>
21447 * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21448 * config/i386/i386.c (ix86_stack_protect_guard): Use
21449 ix86_stack_protect_guard_symbol_str to generate varible declaration.
21450 * doc/invoke.texi (x86 Options): Document
21451 -mstack-protector-guard-symbol= option.
21453 2017-08-10 Uros Bizjak <ubizjak@gmail.com>
21455 * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21456 * config/i386/i386.c (ix86_split_stack_guard): New function.
21457 (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21458 (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21459 (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21460 (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21461 * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21462 (split_stack_space_check): Call ix86_split_stack_guard.
21464 2017-08-10 Martin Sebor <msebor@redhat.com>
21466 * print-tree.c (print_node): Print location using the established
21468 Replace spaces with colons.
21469 (debug_raw, debug): Ditto.
21471 2017-08-10 Martin Sebor <msebor@redhat.com>
21474 * pretty-print.c (pp_format): Correct the handling of %s precision.
21476 2017-08-10 H.J. Lu <hongjiu.lu@intel.com>
21479 * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21481 (ix86_finalize_stack_frame_flags): This. Also clear
21482 frame_pointer_needed if -fno-omit-frame-pointer is used without
21484 (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21485 with ix86_finalize_stack_frame_flags.
21486 (ix86_expand_epilogue): Likewise.
21487 (ix86_expand_split_stack_prologue): Likewise.
21488 * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21490 2017-08-10 Martin Liska <mliska@suse.cz>
21493 * c-attribs.c (handle_target_attribute):
21494 Report warning for an empty string argument of target attribute.
21496 2017-08-09 Jakub Jelinek <jakub@redhat.com>
21499 * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21501 * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21502 or DECL_NONLOCAL labels.
21503 (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21504 or DECL_NONLOCAL labels here.
21506 2017-08-09 Will Schmidt <will_schmidt@vnet.ibm.com>
21508 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21509 to indicate when early gimple folding has been disabled.
21510 (rs6000_gimple_fold_builtin): Add debug content.
21511 (rs6000_invalid_builtin): Fix whitespace.
21512 (rs6000_expand_builtin): Fix whitespace.
21513 * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21515 2017-08-09 Segher Boessenkool <segher@kernel.crashing.org>
21518 * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21519 SAVE_MULTIPLE if not all the registers that saves, should be saved.
21521 2017-08-09 Jim Wilson <jim.wilson@linaro.org>
21523 * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21524 (qdf24xx): Likewise.
21525 * config/aarch64/aarch64.md: Include falkor.md.
21526 * config/aarch64/falkor.md: New.
21528 2017-08-09 Marek Polacek <polacek@redhat.com>
21531 * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21532 * diagnostic.c (emit_diagnostic): Add a comment.
21533 (emit_diagnostic_valist): New function.
21535 2017-08-09 Marek Polacek <polacek@redhat.com>
21538 * input.c (make_location): New overload.
21539 * input.h (make_location): Declare.
21541 2017-08-08 Alan Modra <amodra@gmail.com>
21542 H.J. Lu <hongjiu.lu@intel.com>
21545 * gcc.c (NO_PIE_SPEC): Delete.
21546 (PIE_SPEC): Define as !no-pie/pie. Move static|shared|r
21548 (LINK_PIE_SPEC): ..to here.
21549 (LINK_COMMAND_SPEC): Support -no-pie.
21550 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21551 chain of crtbegin*.o selection, update for PIE_SPEC changes and
21553 (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21554 * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21555 (ENDFILE_CRTEND_SPEC): Similarly.
21557 2017-08-08 Uros Bizjak <ubizjak@gmail.com>
21560 * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21561 (mstack-protector-guard-offset=): Ditto.
21562 * config/i386/i386.c (ix86_option_override): Handle
21563 -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21565 (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21566 ix86_stack_protect_guard_offset variables.
21567 (TARGET_STACK_PROTECT_GUARD): Always define.
21568 * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21569 and -mstack-protector-guard-offset= options.
21571 2017-08-08 Bin Cheng <bin.cheng@arm.com>
21573 * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21574 boundary case for the last candidate.
21576 2017-08-08 Bin Cheng <bin.cheng@arm.com>
21578 * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21579 * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21581 2017-08-08 Tamar Christina <tamar.christina@arm.com>
21583 PR middle-end/19706
21584 * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21585 * config/aarch64/aarch64-builtins.c
21586 (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21587 * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21588 * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21590 2017-08-08 Tamar Christina <tamar.christina@arm.com>
21591 Andrew Pinski <pinskia@gmail.com>
21593 PR middle-end/19706
21594 * internal-fn.def (XORSIGN): New.
21595 * optabs.def (xorsign_optab): New.
21596 * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21597 (convert_expand_mult_copysign): New.
21598 (pass_optimize_widening_mul::execute): Call
21599 convert_expand_mult_copysign.
21601 2017-08-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21603 PR tree-optimization/81354
21604 * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21605 Insert on edges rather than explicitly creating landing pads.
21606 (analyze_candidates_and_replace): Commit edge inserts.
21608 2017-08-08 Richard Biener <rguenther@suse.de>
21610 PR middle-end/81719
21611 * tree-ssa-loop-niter.c: Include tree-dfa.h.
21612 (expand_simple_operations): Also look through ADDR_EXPRs with
21613 MEM_REF bases treating them as POINTER_PLUS_EXPR.
21615 2017-08-08 Richard Biener <rguenther@suse.de>
21617 PR tree-optimization/81723
21618 * tree-vect-slp.c (struct bst_traits): New hash traits.
21619 (bst_fail): New global.
21620 (vect_build_slp_tree_2): New worker, split out from ...
21621 (vect_build_slp_tree): ... this now wrapping it with using
21622 bst_fail set to cache SLP tree build fails. Properly handle
21624 (vect_analyze_slp_instance): Allocate and free bst_fail.
21626 2017-08-08 Martin Liska <mliska@suse.cz>
21629 * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21630 LABEL_DECLs that can be from a different function.
21632 2017-08-08 Bin Cheng <bin.cheng@arm.com>
21634 PR tree-optimization/81744
21635 * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21636 loop's number of iterations.
21638 2017-08-08 Martin Liska <mliska@suse.cz>
21640 * asan.c: Include header files.
21641 * attribs.c (build_decl_attribute_variant): New function moved
21643 (build_type_attribute_qual_variant): Likewise.
21644 (cmp_attrib_identifiers): Likewise.
21645 (simple_cst_list_equal): Likewise.
21646 (omp_declare_simd_clauses_equal): Likewise.
21647 (attribute_value_equal): Likewise.
21648 (comp_type_attributes): Likewise.
21649 (build_type_attribute_variant): Likewise.
21650 (lookup_ident_attribute): Likewise.
21651 (remove_attribute): Likewise.
21652 (merge_attributes): Likewise.
21653 (merge_type_attributes): Likewise.
21654 (merge_decl_attributes): Likewise.
21655 (merge_dllimport_decl_attributes): Likewise.
21656 (handle_dll_attribute): Likewise.
21657 (attribute_list_equal): Likewise.
21658 (attribute_list_contained): Likewise.
21659 * attribs.h (lookup_attribute): New function moved from tree.[ch].
21660 (lookup_attribute_by_prefix): Likewise.
21661 * bb-reorder.c: Include header files.
21662 * builtins.c: Likewise.
21663 * calls.c: Likewise.
21664 * cfgexpand.c: Likewise.
21665 * cgraph.c: Likewise.
21666 * cgraphunit.c: Likewise.
21667 * convert.c: Likewise.
21668 * dwarf2out.c: Likewise.
21669 * final.c: Likewise.
21670 * fold-const.c: Likewise.
21671 * function.c: Likewise.
21672 * gimple-expr.c: Likewise.
21673 * gimple-fold.c: Likewise.
21674 * gimple-pretty-print.c: Likewise.
21675 * gimple.c: Likewise.
21676 * gimplify.c: Likewise.
21677 * hsa-common.c: Likewise.
21678 * hsa-gen.c: Likewise.
21679 * internal-fn.c: Likewise.
21680 * ipa-chkp.c: Likewise.
21681 * ipa-cp.c: Likewise.
21682 * ipa-devirt.c: Likewise.
21683 * ipa-fnsummary.c: Likewise.
21684 * ipa-inline.c: Likewise.
21685 * ipa-visibility.c: Likewise.
21687 * lto-cgraph.c: Likewise.
21688 * omp-expand.c: Likewise.
21689 * omp-general.c: Likewise.
21690 * omp-low.c: Likewise.
21691 * omp-offload.c: Likewise.
21692 * omp-simd-clone.c: Likewise.
21693 * opts-global.c: Likewise.
21694 * passes.c: Likewise.
21695 * predict.c: Likewise.
21696 * sancov.c: Likewise.
21697 * sanopt.c: Likewise.
21698 * symtab.c: Likewise.
21699 * toplev.c: Likewise.
21700 * trans-mem.c: Likewise.
21701 * tree-chkp.c: Likewise.
21702 * tree-eh.c: Likewise.
21703 * tree-into-ssa.c: Likewise.
21704 * tree-object-size.c: Likewise.
21705 * tree-parloops.c: Likewise.
21706 * tree-profile.c: Likewise.
21707 * tree-ssa-ccp.c: Likewise.
21708 * tree-ssa-live.c: Likewise.
21709 * tree-ssa-loop.c: Likewise.
21710 * tree-ssa-sccvn.c: Likewise.
21711 * tree-ssa-structalias.c: Likewise.
21712 * tree-ssa.c: Likewise.
21713 * tree-streamer-in.c: Likewise.
21714 * tree-vectorizer.c: Likewise.
21715 * tree-vrp.c: Likewise.
21716 * tsan.c: Likewise.
21717 * ubsan.c: Likewise.
21718 * varasm.c: Likewise.
21719 * varpool.c: Likewise.
21720 * tree.c: Remove functions moved to attribs.[ch].
21721 * tree.h: Likewise.
21722 * config/aarch64/aarch64.c: Add attrs.h header file.
21723 * config/alpha/alpha.c: Likewise.
21724 * config/arc/arc.c: Likewise.
21725 * config/arm/arm.c: Likewise.
21726 * config/avr/avr.c: Likewise.
21727 * config/bfin/bfin.c: Likewise.
21728 * config/c6x/c6x.c: Likewise.
21729 * config/cr16/cr16.c: Likewise.
21730 * config/cris/cris.c: Likewise.
21731 * config/darwin.c: Likewise.
21732 * config/epiphany/epiphany.c: Likewise.
21733 * config/fr30/fr30.c: Likewise.
21734 * config/frv/frv.c: Likewise.
21735 * config/ft32/ft32.c: Likewise.
21736 * config/h8300/h8300.c: Likewise.
21737 * config/i386/winnt.c: Likewise.
21738 * config/ia64/ia64.c: Likewise.
21739 * config/iq2000/iq2000.c: Likewise.
21740 * config/lm32/lm32.c: Likewise.
21741 * config/m32c/m32c.c: Likewise.
21742 * config/m32r/m32r.c: Likewise.
21743 * config/m68k/m68k.c: Likewise.
21744 * config/mcore/mcore.c: Likewise.
21745 * config/microblaze/microblaze.c: Likewise.
21746 * config/mips/mips.c: Likewise.
21747 * config/mmix/mmix.c: Likewise.
21748 * config/mn10300/mn10300.c: Likewise.
21749 * config/moxie/moxie.c: Likewise.
21750 * config/msp430/msp430.c: Likewise.
21751 * config/nds32/nds32-isr.c: Likewise.
21752 * config/nds32/nds32.c: Likewise.
21753 * config/nios2/nios2.c: Likewise.
21754 * config/nvptx/nvptx.c: Likewise.
21755 * config/pa/pa.c: Likewise.
21756 * config/pdp11/pdp11.c: Likewise.
21757 * config/powerpcspe/powerpcspe.c: Likewise.
21758 * config/riscv/riscv.c: Likewise.
21759 * config/rl78/rl78.c: Likewise.
21760 * config/rx/rx.c: Likewise.
21761 * config/s390/s390.c: Likewise.
21762 * config/sh/sh.c: Likewise.
21763 * config/sol2.c: Likewise.
21764 * config/sparc/sparc.c: Likewise.
21765 * config/spu/spu.c: Likewise.
21766 * config/stormy16/stormy16.c: Likewise.
21767 * config/tilegx/tilegx.c: Likewise.
21768 * config/tilepro/tilepro.c: Likewise.
21769 * config/v850/v850.c: Likewise.
21770 * config/vax/vax.c: Likewise.
21771 * config/visium/visium.c: Likewise.
21772 * config/xtensa/xtensa.c: Likewise.
21774 2017-08-07 Michael Meissner <meissner@linux.vnet.ibm.com>
21777 * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21778 constraints since the -mupper-regs-* switches have been
21780 (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21781 into a vector from a double word element that was extracted from
21782 another vector, and eliminate extra XXPERMDI instructions.
21783 (vsx_concat_<mode>_2): Likewise.
21784 (vsx_concat_<mode>_3): Likewise.
21785 (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21786 concat to allow optimizing inserts from previous extracts.
21788 2017-08-07 Uros Bizjak <ubizjak@gmail.com>
21790 * config/i386/i386.c (ix86_stack_protect_guard): Generate
21791 memory reference to a SSP offset in TLS address space.
21792 (ix86_print_operand) <case '@'>: Remove.
21793 (ix86_print_operand_punct_valid_p): Remove '@' code.
21794 * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21795 UNSPEC_SP_TLS_TEST.
21796 (stack_tls_protect_set_<mode>): Remove.
21797 (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21798 (stack_tls_protect_test_<mode>): Remove.
21799 (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21801 2017-08-07 Olivier Hainque <hainque@adacore.com>
21804 * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21806 2017-08-07 Douglas Rupp <rupp@adacore.com>
21808 * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21809 variable was referenced as multidir in command.
21811 2017-08-07 Jakub Jelinek <jakub@redhat.com>
21814 * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21817 2017-08-07 Martin Liska <mliska@suse.cz>
21819 * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21820 * config/rl78/rl78.c: Add include of attribs.h.
21821 * config/sh/sh.c: Likewise.
21822 * config/v850/v850.c: Likewise.
21824 2017-08-07 Tom de Vries <tom@codesourcery.com>
21826 PR middle-end/78266
21827 * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21829 2017-08-07 Martin Liska <mliska@suse.cz>
21831 * config/mips/mips.c: Include attribs.h.
21833 2017-08-07 Thomas Koenig <tkoenig@gcc.gnu.org>
21836 * doc/invoke.texi: Document change in behvaior for -Ofast for
21839 2017-08-07 Wilco Dijkstra <wdijkstr@arm.com>
21841 * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21843 (aarch64_pop_regs): Likewise.
21844 (aarch64_gen_load_pair): Likewise.
21845 (aarch64_save_callee_saves): Likewise.
21846 (aarch64_restore_callee_saves): Likewise.
21848 2017-08-07 H.J. Lu <hongjiu.lu@intel.com>
21850 * config/i386/i386.c: Revert the last change.
21852 2017-08-07 H.J. Lu <hongjiu.lu@intel.com>
21855 * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21857 (ix86_finalize_stack_frame_flags): This. Also clear
21858 frame_pointer_needed if -fno-omit-frame-pointer is used without
21860 (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21861 with ix86_finalize_stack_frame_flags.
21862 (ix86_expand_epilogue): Likewise.
21863 (ix86_expand_split_stack_prologue): Likewise.
21865 2017-08-07 H.J. Lu <hongjiu.lu@intel.com>
21868 * config/i386/i386.c (get_builtin_code_for_version): Set priority
21869 to P_AES for Westmere.
21871 2017-08-07 Jonathan Yong <10walls@gmail.com>
21873 * config/i386/mingw.opt (fset-stack-executable): Removed.
21874 * config/i386/cygming.opt (fset-stack-executable): Moved
21876 * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
21878 2017-08-07 Segher Boessenkool <segher@kernel.crashing.org>
21880 * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
21882 2017-08-07 Marek Polacek <polacek@redhat.com>
21884 PR middle-end/81737
21885 * fold-const.c (fold_indirect_ref_1): Check type_domain.
21887 2017-08-07 Martin Liska <mliska@suse.cz>
21889 * attribs.h (canonicalize_attr_name): New function.
21890 (cmp_attribs): Move from c-format.c and adjusted.
21891 (is_attribute_p): Moved from tree.h.
21892 * tree-inline.c: Add new includes.
21893 * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
21894 (private_is_attribute_p): Remove.
21895 (private_lookup_attribute): Likewise.
21896 (private_lookup_attribute_by_prefix): Simplify.
21897 (remove_attribute): Use is_attribute_p.
21898 * tree.h: Remove removed declarations.
21900 2017-08-07 Jakub Jelinek <jakub@redhat.com>
21902 PR middle-end/81698
21903 * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
21904 instead of computing it in the function. Formatting fix.
21905 (expand_case): Don't rely on default_edge being the first edge,
21906 clear it if removing it, pass default_edge to
21907 emit_case_dispatch_table.
21908 (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
21911 2017-08-06 Uros Bizjak <ubizjak@gmail.com>
21913 * config/alpha/alpha.c (alpha_reorg): If trap is the last active
21914 insn in the function, emit NOP after the insn.
21916 2017-08-06 Tom de Vries <tom@codesourcery.com>
21918 * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
21921 2017-08-06 Tom de Vries <tom@codesourcery.com>
21923 * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
21926 2017-08-04 Yury Gribov <tetra2005@gmail.com>
21928 PR tree-optimization/57371
21929 * match.pd: New pattern.
21931 2017-08-04 Marek Polacek <polacek@redhat.com>
21933 PR middle-end/81695
21934 * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
21935 perform the computation in offset_int.
21937 2017-08-04 Richard Sandiford <richard.sandiford@linaro.org>
21939 PR tree-optimization/81136
21940 * tree-vectorizer.h: Include tree-hash-traits.h.
21941 (vec_base_alignments): New typedef.
21942 (vec_info): Add a base_alignments field.
21943 (vect_record_base_alignments): Declare.
21944 * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
21946 (DR_IS_CONDITIONAL_IN_STMT): New macro.
21947 (create_data_ref): Add an is_conditional_in_stmt argument.
21948 * tree-data-ref.c (create_data_ref): Likewise. Use it to initialize
21949 the is_conditional_in_stmt field.
21950 (data_ref_loc): Add an is_conditional_in_stmt field.
21951 (get_references_in_stmt): Set the is_conditional_in_stmt field.
21952 (find_data_references_in_stmt): Update call to create_data_ref.
21953 (graphite_find_data_references_in_stmt): Likewise.
21954 * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
21955 * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
21956 (vect_record_base_alignment): New function.
21957 (vect_record_base_alignments): Likewise.
21958 (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
21959 for nested statements even if we fail to compute a misalignment.
21960 Use pooled base alignments for unconditional references.
21961 (vect_find_same_alignment_drs): Compare base addresses instead
21963 (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
21964 * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
21966 2017-08-04 Richard Sandiford <richard.sandiford@linaro.org>
21968 * tree-vectorizer.h (vec_info): Add a constructor and destructor.
21969 Add an explicit name for the enum. Use auto_vec for slp_instances
21970 and grouped_stores.
21971 (_loop_vec_info): Add a constructor and destructor. Use auto_vec
21973 (_bb_vec_info): Add a constructor and destructor.
21974 (vinfo_for_stmt): Return NULL for uids of -1 as well.
21975 (destroy_loop_vec_info): Delete.
21976 (vect_destroy_datarefs): Likewise.
21977 * tree-vectorizer.c (vect_destroy_datarefs): Delete.
21978 (vec_info::vec_info): New function.
21979 (vec_info::~vec_info): Likewise.
21980 (vectorize_loops): Use delete instead of destroy_loop_vec_info.
21981 * tree-parloops.c (gather_scalar_reductions): Use delete instead of
21982 destroy_loop_vec_info.
21983 * tree-vect-loop.c (new_loop_vec_info): Replace with...
21984 (_loop_vec_info::_loop_vec_info): ...this.
21985 (destroy_loop_vec_info): Replace with...
21986 (_loop_vec_info::~_loop_vec_info): ...this. Unconditionally delete
21987 the stmt_vec_infos. Leave handling of vec_info information to its
21988 destructor. Remove explicit vector releases.
21989 (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
21990 (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
21991 * tree-vect-slp.c (new_bb_vec_info): Replace with...
21992 (_bb_vec_info::_bb_vec_info): ...this. Don't reserve space in
21993 BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
21994 (destroy_bb_vec_info): Replace with...
21995 (_bb_vec_info::~_bb_vec_info): ...this. Leave handling of vec_info
21996 information to its destructor.
21997 (vect_slp_analyze_bb_1): Use new and delete instead of
21998 new_bb_vec_info and destroy_bb_vec_info.
21999 (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22002 2017-08-04 Richard Sandiford <richard.sandiford@linaro.org>
22004 * tree-data-ref.h (subscript): Add access_fn field.
22005 (data_dependence_relation): Add could_be_independent_p.
22006 (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22007 (same_access_functions): Move to tree-data-ref.c.
22008 * tree-data-ref.c (ref_contains_union_access_p): New function.
22009 (access_fn_component_p): Likewise.
22010 (access_fn_components_comparable_p): Likewise.
22011 (dr_analyze_indices): Add a reference to access_fn_component_p.
22012 (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22014 (constant_access_functions): Likewise.
22015 (add_other_self_distances): Likewise.
22016 (same_access_functions): Likewise. (Moved from tree-data-ref.h.)
22017 (initialize_data_dependence_relation): Use XCNEW and remove
22018 explicit zeroing of DDR_REVERSED_P. Look for a subsequence
22019 of access functions that have the same type. Allow the
22020 subsequence to end with different bases in some circumstances.
22021 Record the chosen access functions in SUB_ACCESS_FN.
22022 (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22023 a_index and b_index. Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22024 (subscript_dependence_tester_1): Likewise dra and drb.
22025 (build_classic_dist_vector): Update calls accordingly.
22026 (subscript_dependence_tester): Likewise.
22027 * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22028 DDR_COULD_BE_INDEPENDENT_P.
22029 * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22030 comp_alias_ddrs instead of may_alias_ddrs.
22031 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22033 (vect_analyze_data_ref_dependence): Use it if
22034 DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22035 distance vectors if that fails.
22036 (dependence_distance_ge_vf): New function.
22037 (vect_prune_runtime_alias_test_list): Use it. Don't clear
22038 LOOP_VINFO_MAY_ALIAS_DDRS.
22040 2017-08-04 Richard Biener <rguenther@suse.de>
22042 PR middle-end/81705
22043 * fold-const.c (fold_binary_loc): Properly restrict
22044 minus_var0 && minus_var1 case when associating undefined overflow
22047 2017-08-04 Tom de Vries <tom@codesourcery.com>
22049 * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22051 2017-08-03 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22053 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22054 Don't start diagnostic messages with a capital letter.
22055 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22057 (rs6000_invalid_builtin): Likewise.
22058 (rs6000_trampoline_init): Likewise.
22060 2017-08-03 Jakub Jelinek <jakub@redhat.com>
22063 * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22064 after setting changeable df flags.
22066 2017-08-03 Richard Biener <rguenther@suse.de>
22068 * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22069 up if the use is in USE - X.
22071 2017-08-03 Alexander Monakov <amonakov@ispras.ru>
22073 * toplev.c (dumpfile.h): New include.
22074 (internal_error_reentered): New static function. Use it...
22075 (internal_error_function): ...here to handle reentered internal_error.
22077 2017-08-03 Richard Biener <rguenther@suse.de>
22079 PR middle-end/81148
22080 * fold-const.c (split_tree): Add minus_var and minus_con
22081 arguments, remove unused loc arg. Never generate NEGATE_EXPRs
22082 here but always use minus_*.
22083 (associate_trees): Assert we never associate with MINUS_EXPR
22084 and NULL first operand. Do not recurse for PLUS_EXPR operands
22085 when associating as MINUS_EXPR either.
22086 (fold_binary_loc): Track minus_var and minus_con.
22088 2017-08-03 Tom de Vries <tom@codesourcery.com>
22091 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22092 ACCEL_COMPILER, apply finish_options on
22093 DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22095 2017-08-03 Tom de Vries <tom@codesourcery.com>
22098 * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22099 function_entry_patch_area_size > 0.
22101 2017-08-03 Jakub Jelinek <jakub@redhat.com>
22104 * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22105 instead of 10??LU, perform unit multiplication in wide_int,
22106 don't change alloc_object_size_limit if the limit is larger
22109 PR tree-optimization/81655
22110 PR tree-optimization/81588
22111 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22112 the case when ranges[i].low and high are 1 for unsigned type with
22115 PR middle-end/81052
22116 * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22117 (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22119 2017-08-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22121 * tree-vrp.h: Add include guard.
22123 2017-08-02 Uros Bizjak <ubizjak@gmail.com>
22126 * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22127 (ud2): New insn pattern.
22128 * config/i386/i386.c (ix86_expand_epilogue):
22129 For naked functions, generate ud2 instead of trap insn.
22131 2017-08-02 Marek Polacek <polacek@redhat.com>
22134 * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22136 2017-08-02 Tom de Vries <tom@codesourcery.com>
22137 Cesar Philippidis <cesar@codesourcery.com>
22139 * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22140 Add missing edge probabilities.
22142 2017-08-02 Tamar Christina <tamar.christina@arm.com>
22144 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22145 Correct endianness.
22147 2017-08-02 Jakub Jelinek <jakub@redhat.com>
22149 PR middle-end/79499
22150 * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22151 for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22152 of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22153 prologue_seq sequences - if any.
22155 2017-08-02 Richard Biener <rguenther@suse.de>
22157 * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22158 via vectors if supported, integer extracts via punning if supported
22159 or otherwise vector extracts.
22161 2017-08-02 Richard Biener <rguenther@suse.de>
22163 * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22165 (bitmap_insert_into_set): ... this.
22167 2017-08-02 Richard Biener <rguenther@suse.de>
22169 PR tree-optimization/81633
22171 2015-08-17 Alan Hayward <alan.hayward@arm.com>
22173 PR tree-optimization/71752
22174 * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22176 2017-08-01 Daniel Santos <daniel.santos@pobox.com>
22178 * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22179 (machine_function::call_ms2sysv_pad_out): Remove field.
22180 * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22181 (ix86_compute_frame_layout): Likewise.
22183 2017-08-01 H.J. Lu <hongjiu.lu@intel.com>
22186 * config/i386/i386.c (ix86_set_func_type): Disallow naked
22187 attribute with interrupt attribute.
22189 2017-08-01 Andrew Pinski <apinski@cavium.com>
22191 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22192 BIT_INSERT_EXPR's operand 1
22193 to see if the types precision matches.
22195 2017-08-01 Martin Liska <mliska@suse.cz>
22197 PR middle-end/70140
22198 * builtins.c (expand_builtin_memcpy_args): Remove.
22199 (expand_builtin_memcpy): Call newly added function
22200 expand_builtin_memory_copy_args.
22201 (expand_builtin_memcpy_with_bounds): Likewise.
22202 (expand_builtin_mempcpy): Remove last argument.
22203 (expand_builtin_mempcpy_with_bounds): Likewise.
22204 (expand_builtin_memory_copy_args): New function created from
22205 expand_builtin_mempcpy_args with small modifications.
22206 (expand_builtin_mempcpy_args): Remove.
22207 (expand_builtin_stpcpy): Remove unused argument.
22208 (expand_builtin): Likewise.
22209 (expand_builtin_with_bounds): Likewise.
22211 2017-08-01 Martin Liska <mliska@suse.cz>
22214 Make mempcpy more optimal (PR middle-end/70140).
22216 2017-08-01 Jakub Jelinek <jakub@redhat.com>
22219 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22220 __builtin_vec_cmpne verify both arguments are compatible vectors
22221 before looking at TYPE_MODE on the element type. For __builtin_vec_ld
22222 verify arg1_type is a pointer or array type. For __builtin_vec_st,
22223 move computation of aligned to after checking the argument types.
22227 * config/rs6000/vsx.md (vextract_fp_from_shorth,
22228 vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22231 2017-08-01 Jerome Lambourg <lambourg@adacore.com>
22232 Doug Rupp <rupp@adacore.com>
22233 Olivier Hainque <hainque@adacore.com>
22235 * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22236 well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22238 * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22239 for TARGET_OS_CPP_BUILTIN.
22240 (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22241 refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22243 (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22244 passing required abi options to the assembler for EABI configurations.
22245 (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22246 of .text.hot and .text.unlikely sections for kernel modules when
22247 using ARM style exceptions.
22248 (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22249 options. Add EXTRA_CC1_SPEC.
22250 (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22252 (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22254 (ARM_TARGET2_DWARF_FORMAT): Define.
22255 * config/arm/t-vxworks: Adjust multilib control to removal of the
22256 Diab command line options.
22258 2017-08-01 Martin Liska <mliska@suse.cz>
22260 PR gcov-profile/81561
22261 * gcov.c (unblock): Make unblocking safe as we need to preserve
22262 index correspondence of blocks and block_lists.
22264 2017-08-01 Richard Biener <rguenther@suse.de>
22266 PR tree-optimization/81181
22267 * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22268 (compute_antic): ... end of iteration here.
22270 2017-08-01 James Greenhalgh <james.greenhalgh@arm.com>
22272 * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22273 (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22274 (ftree-slp-vectorize): Likewise.
22275 * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22276 can no longer be set independent of flag_tree_loop_vectorize.
22277 * omp-general.c (emp_max_vf): Likewise.
22278 * opts.c (enable_fdo_optimizations): Remove references to
22279 flag_tree_vectorize, these are now implicit.
22280 (common_handle_option): Remove handling for OPT_ftree_vectorize,
22281 and leave it for the options machinery.
22283 2017-08-01 Martin Liska <mliska@suse.cz>
22285 PR middle-end/70140
22286 * builtins.c (expand_builtin_memcpy_args): Remove.
22287 (expand_builtin_memcpy): Call newly added function
22288 expand_builtin_memory_copy_args.
22289 (expand_builtin_memcpy_with_bounds): Likewise.
22290 (expand_builtin_mempcpy): Remove last argument.
22291 (expand_builtin_mempcpy_with_bounds): Likewise.
22292 (expand_builtin_memory_copy_args): New function created from
22293 expand_builtin_mempcpy_args with small modifications.
22294 (expand_builtin_mempcpy_args): Remove.
22295 (expand_builtin_stpcpy): Remove unused argument.
22296 (expand_builtin): Likewise.
22297 (expand_builtin_with_bounds): Likewise.
22299 2017-08-01 Uros Bizjak <ubizjak@gmail.com>
22302 * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22303 print "ds:" only for immediates in generic address space.
22305 2017-08-01 Uros Bizjak <ubizjak@gmail.com>
22308 * config/i386/i386.c (ix86_funciton_naked): New prototype.
22309 (ix86_function_ok_for_sibcall): Return false for naked functions.
22311 2017-08-01 Richard Biener <rguenther@suse.de>
22313 * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22314 (compute_antic): Seed worklist with exit block predecessors.
22315 * cfganal.c (dfs_find_deadend): For a cycle return the source
22316 of the edge closing it.
22318 2017-08-01 Tamar Christina <tamar.christina@arm.com>
22320 * config/aarch64/aarch64.c
22321 (aarch64_can_const_movi_rtx_p): Move 0 check.
22323 2017-08-01 Bin Cheng <bin.cheng@arm.com>
22325 * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22326 * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22328 * match.pd: Ditto in address comparison pattern.
22330 2017-08-01 Bin Cheng <bin.cheng@arm.com>
22332 PR tree-optimization/81627
22333 * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22334 closed ssa form for store-store chain.
22336 2017-08-01 Bin Cheng <bin.cheng@arm.com>
22338 PR tree-optimization/81620
22339 * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22340 for store-store chain.
22342 2017-08-01 Jakub Jelinek <jakub@redhat.com>
22344 PR tree-optimization/81588
22345 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22346 ranges[i].in_p, invert comparison code ccode. For >/>=,
22347 swap rhs1 and rhs2 and comparison code unconditionally,
22348 for </<= don't do that. Don't swap rhs1/rhs2 again if
22349 ranges[i].in_p, instead invert comparison code ccode if
22350 opcode or oe->rank is BIT_IOR_EXPR.
22353 * optabs.def (vec_extract_optab, vec_init_optab): Change from
22354 a direct optab to conversion optab.
22355 * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22356 with GET_MODE_INNER as last argument instead of optab_handler.
22357 * expmed.c (extract_bit_field_1): Likewise. Use vector from
22358 vector extraction if possible and optab is available.
22359 * expr.c (store_constructor): Use convert_optab_handler instead
22360 of optab_handler. Use vector initialization from smaller
22361 vectors if possible and optab is available.
22362 * tree-vect-stmts.c (vectorizable_load): Likewise.
22363 * doc/md.texi (vec_extract, vec_init): Document that the optabs
22364 now have two modes.
22365 * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22366 of vec_init from half-sized vectors with the same element mode.
22367 * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22368 (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22369 (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22370 reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22371 reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22372 after mode in gen_vec_extract* calls.
22373 (vec_extract<mode>): Renamed to ...
22374 (vec_extract<mode><ssescalarmodelower>): ... this.
22375 (vec_extract<mode><ssehalfvecmodelower>): New expander.
22376 (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22377 element mode after mode in gen_vec_init* calls.
22378 (VEC_INIT_HALF_MODE): New mode iterator.
22379 (vec_init<mode>): Renamed to ...
22380 (vec_init<mode><ssescalarmodelower>): ... this.
22381 (vec_init<mode><ssehalfvecmodelower>): New expander.
22382 * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22383 (vec_extractv2sfsf): ... this.
22384 (vec_initv2sf): Renamed to ...
22385 (vec_initv2sfsf): ... this.
22386 (vec_extractv2si): Renamed to ...
22387 (vec_extractv2sisi): ... this.
22388 (vec_initv2si): Renamed to ...
22389 (vec_initv2sisi): ... this.
22390 (vec_extractv4hi): Renamed to ...
22391 (vec_extractv4hihi): ... this.
22392 (vec_initv4hi): Renamed to ...
22393 (vec_initv4hihi): ... this.
22394 (vec_extractv8qi): Renamed to ...
22395 (vec_extractv8qiqi): ... this.
22396 (vec_initv8qi): Renamed to ...
22397 (vec_initv8qiqi): ... this.
22398 * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22399 (vec_init<mode>): Renamed to ...
22400 (vec_init<mode><VEC_base_l>): ... this.
22401 (vec_extract<mode>): Renamed to ...
22402 (vec_extract<mode><VEC_base_l>): ... this.
22403 * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22404 (vec_initv2sfsf): ... this.
22405 * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22406 vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22407 vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22408 element mode after mode in gen_vec_init* calls.
22409 * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22410 (vec_init<mode><Vel>): ... this.
22411 (vec_extract<mode>): Renamed to ...
22412 (vec_extract<mode><Vel>): ... this.
22413 * config/aarch64/iterators.md (Vel): New mode attribute.
22414 * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22415 Add element mode after mode in gen_vec_extract* calls.
22416 * config/s390/vector.md (non_vec_l): New mode attribute.
22417 (vec_extract<mode>): Renamed to ...
22418 (vec_extract<mode><non_vec_l>): ... this.
22419 (vec_init<mode>): Renamed to ...
22420 (vec_init<mode><non_vec_l>): ... this.
22421 * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22422 s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22424 * config/arm/iterators.md (V_elem_l): New mode attribute.
22425 * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22426 (vec_extract<mode><V_elem_l>): ... this.
22427 (vec_extractv2di): Renamed to ...
22428 (vec_extractv2didi): ... this.
22429 (vec_init<mode>): Renamed to ...
22430 (vec_init<mode><V_elem_l>): ... this.
22431 (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22432 reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22433 reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22434 Add element mode after gen_vec_extract* calls.
22435 * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22436 (vec_init<mode><unitmode>): ... this.
22437 (vec_extract<mode>): Renamed to ...
22438 (vec_extract<mode><unitmode>): ... this.
22439 * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22440 (vec_init<mode><unitmode>): ... this.
22441 * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22442 (vec_initv2sfsf): ... this.
22443 (vec_extractv2sf): Renamed to ...
22444 (vec_extractv2sfsf): ... this.
22445 (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22446 Add element mode after gen_vec_extract* calls.
22447 * config/mips/mips.md (unitmode): New mode iterator.
22448 * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22449 spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22450 * config/spu/spu.md (inner_l): New mode attribute.
22451 (vec_init<mode>): Renamed to ...
22452 (vec_init<mode><inner_l>): ... this.
22453 (vec_extract<mode>): Renamed to ...
22454 (vec_extract<mode><inner_l>): ... this.
22455 * config/sparc/sparc.md (veltmode): New mode iterator.
22456 (vec_init<VMALL:mode>): Renamed to ...
22457 (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22458 * config/ia64/vect.md (vec_initv2si): Renamed to ...
22459 (vec_initv2sisi): ... this.
22460 (vec_initv2sf): Renamed to ...
22461 (vec_initv2sfsf): ... this.
22462 (vec_extractv2sf): Renamed to ...
22463 (vec_extractv2sfsf): ... this.
22464 * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22465 (vec_init<mode>): Renamed to ...
22466 (vec_init<mode><VEC_base_l>): ... this.
22467 (vec_extract<mode>): Renamed to ...
22468 (vec_extract<mode><VEC_base_l>): ... this.
22469 * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22470 (vec_initv2sfsf): ... this.
22471 * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22472 vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22473 vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22474 gen_vec_init* calls.
22476 2017-08-01 Richard Biener <rguenther@suse.de>
22478 PR tree-optimization/81297
22479 * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22480 TREE_OVERFLOW from INTEGER_CSTs.
22482 2017-07-31 Segher Boessenkool <segher@kernel.crashing.org>
22484 * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22486 2017-07-31 Carl Love <cel@us.ibm.com>
22488 * config/rs6000/rs6000-c: Add support for built-in functions
22489 vector signed char vec_xl_be (signed long long, signed char *);
22490 vector unsigned char vec_xl_be (signed long long, unsigned char *);
22491 vector signed int vec_xl_be (signed long long, signed int *);
22492 vector unsigned int vec_xl_be (signed long long, unsigned int *);
22493 vector signed long long vec_xl_be (signed long long, signed long long *);
22494 vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22495 vector signed short vec_xl_be (signed long long, signed short *);
22496 vector unsigned short vec_xl_be (signed long long, unsigned short *);
22497 vector double vec_xl_be (signed long long, double *);
22498 vector float vec_xl_be (signed long long, float *);
22499 * config/rs6000/altivec.h (vec_xl_be): Add #define.
22500 * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22501 XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22503 * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22504 (altivec_expand_builtin): Add switch statement to call
22505 altivec_expand_xl_be for each builtin.
22506 (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22507 __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22508 __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22509 __builtin_vsx_le_be_v16qi.
22510 * doc/extend.texi: Update the built-in documentation file for the
22511 new built-in functions.
22513 2017-07-31 Uros Bizjak <ubizjak@gmail.com>
22516 * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22518 (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22520 2017-07-31 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22522 * config.gcc: Add z14.
22523 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22524 CPU model numbers for z13s and z14.
22525 * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22527 * config/s390/s390-opts.h (enum processor_type): Rename
22528 PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22529 * config/s390/s390.c (processor_table): Add field for CPU name to
22530 be passed to Binutils.
22531 (s390_asm_output_machine_for_arch): Use the new field in
22532 processor_table for Binutils.
22533 (s390_expand_builtin): Replace arch12 with z14.
22534 (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22535 (s390_get_sched_attrmask): Likewise.
22536 (s390_get_unit_mask): Likewise.
22537 * config/s390/s390.opt: Add z14 to processor_type enum.
22539 2017-07-31 Martin Jambor <mjambor@suse.cz>
22542 * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22543 regardless of optimization level.
22545 2017-07-31 Jan Hubicka <hubicka@ucw.cz>
22546 Martin Liska <mliska@suse.cz>
22548 * predict.def: Remove old comment and adjust probability.
22549 * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22550 PREDICT statements.
22552 2017-07-31 Uros Bizjak <ubizjak@gmail.com>
22555 * config/i386/i386.c (ix86_function_naked): New function.
22556 (ix86_can_use_return_insn_p): Return false for naked functions.
22557 (ix86_expand_prologue): Skip prologue for naked functions.
22558 (ix86_expand_epilogue): Skip epilogue for naked functions
22559 and emit trap instruction.
22560 (ix86_warn_func_return): New function.
22561 (ix86_attribute_table): Add "naked" attribute specification.
22562 (TARGET_WARN_FUNC_RETURN): Define.
22563 * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22565 2017-07-31 Martin Liska <mliska@suse.cz>
22567 * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22568 (dump_gimple_bb_header): Always dump BB info.
22569 (pp_cfg_jump): Do not append info about BB when dumping a jump.
22571 2017-07-31 Martin Liska <mliska@suse.cz>
22574 * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22575 also with current_function_decl non-null equality.
22577 2017-07-31 Jakub Jelinek <jakub@redhat.com>
22580 * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22581 change type to the element type, instead add eltype variable and
22582 use it where we are interested in the element type.
22584 PR tree-optimization/81603
22585 * ipa-polymorphic-call.c
22586 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22587 offset arithmetic in offset_int, bail out if the resulting bit offset
22588 doesn't fit into shwi.
22590 2017-07-31 Martin Liska <mliska@suse.cz>
22592 * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22593 (gimplify_save_expr): Fix comment.
22595 2017-07-30 H.J. Lu <hongjiu.lu@intel.com>
22598 * config/i386/i386.c (ix86_function_arg): Update arguments for
22600 (ix86_compute_frame_layout): Set the initial stack offset to
22601 INCOMING_FRAME_SP_OFFSET. Update red-zone offset with
22602 INCOMING_FRAME_SP_OFFSET.
22603 (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22604 stack before exception handler returns.
22605 * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22606 the 'ERROR_CODE' for exception handler.
22608 2017-07-30 Uros Bizjak <ubizjak@gmail.com>
22610 * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22611 (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22612 (ASM_OUTPUT_REG_POP): Ditto.
22613 * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22614 instead of asm_fprintf to output pure string.
22616 2017-07-29 Jakub Jelinek <jakub@redhat.com>
22618 * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22619 to imported_module_or_decl hook.
22620 (debug_nothing_tree_tree_tree_bool): Remove.
22621 (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22622 * debug.c (do_nothing_debug_hooks): Use
22623 debug_nothing_tree_tree_tree_bool_bool instead of
22624 debug_nothing_tree_tree_tree_bool.
22625 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22626 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22627 * sdbout.c (sdb_debug_hooks): Likewise.
22628 * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22629 (gen_namespace_die): Add DW_AT_export_symbols attribute if
22631 (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22632 if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22633 attribute, don't add anything.
22635 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22637 * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22638 (fold_build2_stat_loc): Likewise.
22639 (fold_build3_stat_loc): Likewise.
22640 * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22641 (fold_build1_loc): Remove macro.
22642 (fold_build2_loc): Likewise.
22643 (fold_build3_loc): Likewise.
22645 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22647 * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22648 (gimple_build_debug_bind_source_stat): Likewise.
22649 * gimple.h (gimple_build_debug_bind): Remove macro.
22650 (gimple_build_debug_bind_source): Likewise.
22652 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22654 * bitmap.c (bitmap_alloc): Adjust.
22655 (bitmap_gc_alloc): Likewise.
22656 * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22658 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22660 * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22661 (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22662 * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22663 (bitmap_gc_alloc_stat): Likewise.
22664 (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22666 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22668 * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22669 * rtl.h (shallow_copy_rtx): Remove macro.
22671 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22673 * emit-rtl.c (gen_raw_REG): Adjust.
22674 * gengenrtl.c (gendef): Likewise.
22675 * rtl.c (rtx_alloc_stat): Remove _stat from name.
22676 * rtl.h (rtx_alloc): Remove macro.
22678 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22680 * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22681 (build_tree_list_stat): Likewise.
22682 * tree.h (build_tree_list): Remove macro.
22683 (build_tree_list_vec): Likewise.
22685 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22687 * tree.c (make_vector_stat): Remove _stat from name.
22688 (build_vector_stat): Likewise.
22689 * tree.h (make_vector_stat): Remove macro.
22690 (build_vector_stat): Likewise.
22692 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22694 * tree.h (build_var_debug_value): Remove prototype.
22696 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22698 * tree.c (tree_cons_stat): Remove _stat from name.
22699 * tree.h (tree_cons): Remove macro.
22701 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22703 * tree.c (build_vl_exp_stat): Remove _stat from name.
22704 * tree.h (build_vl_exp): Remove macro.
22706 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22708 * tree.c (build_decl_stat): Remove _stat from name.
22709 * tree.h (build_decl): Remove macro.
22711 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22713 * gimple.c (gimple_build_with_ops_stat): Adjust.
22714 (gimple_alloc_stat): Remove _stat from name.
22715 * gimple.h (gimple_alloc): Remove macro.
22717 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22719 * tree.c (make_tree_vec_stat): Remove _stat from name.
22720 (grow_tree_vec_stat): Likewise.
22721 * tree.h (make_tree_vec_stat): Adjust prototype.
22722 (grow_tree_vec_stat): Likewise.
22723 (make_tree_vec): Remove macro.
22724 (grow_tree_vec): Likewise.
22726 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22728 * fold-const.c (fold_build1_stat_loc): Adjust.
22729 (fold_build2_stat_loc): Likewise.
22730 (fold_build3_stat_loc): Likewise.
22731 * tree.c (build0_stat): Remove _stat from name.
22732 (build1_stat): Likewise.
22733 (build2_stat): Likewise.
22734 (build3_stat): Likewise.
22735 (build4_stat): Likewise.
22736 (build5_stat): Likewise.
22737 * tree.h (build1_loc): Remove macro, and rename _stat function
22739 (build2_loc): Likewise.
22740 (build3_loc): Likewise.
22741 (build4_loc): Likewise.
22742 (build5_loc): Likewise.
22744 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22746 * tree.c (make_int_cst_stat): Remove _stat from name.
22747 * tree.h (make_int_cst_stat): Adjust prototype.
22748 (make_int_cst): Remove macro.
22750 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22752 * tree.c (make_tre_binfo_stat): Remove _stat from name.
22753 * tree.h (make_tree_binfo_stat): Adjust prototype.
22754 (make_tree_binfo): Remove.
22756 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22758 * tree.c (copy_node_stat): Rename to copy_node.
22759 (build_distinct_type_copy): Adjust.
22760 * tree.h (copy_node_stat): Adjust prototype.
22761 (copy_node): Remove macro.
22763 2017-07-28 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
22765 * tree.c (make_node_stat): rename to make_node.
22766 (build_tree_list_stat): Adjust.
22767 (build0_stat): Likewise.
22768 (build2_stat): Likewise.
22769 (build3_stat): Likewise.
22770 (build4_stat): Likewise.
22771 (build5_stat): Likewise.
22772 (build_decl_stat): Likewise.
22773 * tree.h (make_node_stat): Adjust prototype.
22774 (make_node): remove macro.
22776 2017-07-28 Peter Bergner <bergner@vnet.ibm.com>
22778 * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22779 (PPC_FEATURE2_SCV): Likewise.
22780 * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22782 2017-07-28 Tamar Christina <tamar.christina@arm.com>
22784 * config/aarch64/aarch64.c
22785 (aarch64_internal_mov_immediate): Add new special pattern.
22786 * config/aarch64/aarch64.md (*movdi_aarch64):
22787 Add reg/32bit const mov case.
22789 2017-07-28 Tamar Christina <tamar.christina@arm.com>
22790 Richard Sandiford <richard.sandiford@linaro.org>
22792 * config/aarch64/aarch64.md (mov<mode>): Generalize.
22793 (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22794 Add integer and movi cases.
22795 (movi-split-hf-df-sf split, fp16): New.
22796 (enabled): Added TARGET_FP_F16INST.
22797 * config/aarch64/iterators.md (GPF_HF): New.
22798 * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22800 2017-07-28 Tamar Christina <tamar.christina@arm.com>
22802 * config/aarch64/aarch64.c
22803 (aarch64_simd_container_mode): Add prototype.
22804 (aarch64_expand_mov_immediate): Add HI support.
22805 (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22806 (aarch64_can_const_movi_rtx_p): New.
22807 (aarch64_preferred_reload_class):
22808 Remove restrictions of using FP registers for certain SIMD operations.
22809 (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22810 (aarch64_valid_floating_const): Add integer move validation.
22811 (aarch64_simd_imm_scalar_p): Remove.
22812 (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22813 (aarch64_legitimate_constant_p): Expand list of supported cases.
22814 * config/aarch64/aarch64-protos.h
22815 (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22816 (aarch64_reinterpret_float_as_int): New.
22817 (aarch64_simd_imm_scalar_p): Remove.
22818 * config/aarch64/constraints.md (Uvi): New.
22819 (Dd): Split into Ds and new Dd.
22820 * config/aarch64/aarch64.md (*movsi_aarch64):
22822 (*movdi_aarch64): Add SIMD mov case.
22824 2017-07-28 Bin Cheng <bin.cheng@arm.com>
22826 * tree-predcom.c: (struct chain): Handle store-store chain in which
22827 stores for elimination only store loop invariant values.
22828 (execute_pred_commoning_chain): Ditto.
22829 (prepare_initializers_chain_store_elim): Ditto.
22830 (prepare_finalizers): Ditto.
22831 (is_inv_store_elimination_chain): New function.
22832 (initialize_root_vars_store_elim_1): New function.
22834 2017-07-28 Bin Cheng <bin.cheng@arm.com>
22836 * tree-predcom.c: Revise general description of the pass.
22837 (enum chain_type): New enum type for store elimination.
22838 (struct chain): New field supporting store elimination.
22839 (struct component): Ditto.
22840 (dump_chain): Dump store-stores chain.
22841 (release_chain): Release resources.
22842 (split_data_refs_to_components): Compute and create component
22843 contains only stores for elimination.
22844 (get_chain_last_ref_at): New function.
22845 (make_invariant_chain): Initialization.
22846 (make_rooted_chain): Specify chain type in parameter and record it.
22847 (add_looparound_copies): Skip for store-stores chain.
22848 (determine_roots_comp): Compute type of chain and pass it to
22850 (initialize_root_vars_store_elim_2): New function.
22851 (finalize_eliminated_stores): New function.
22852 (remove_stmt): Handle store for elimination.
22853 (execute_pred_commoning_chain): Execute predictive commoning on
22854 store-store chains.
22855 (determine_unroll_factor): Skip unroll for store-stores chain.
22856 (prepare_initializers_chain_store_elim): New function.
22857 (prepare_initializers_chain): Hanlde store-store chain.
22858 (prepare_finalizers_chain, prepare_finalizers): New function.
22859 (tree_predictive_commoning_loop): Return integer value indicating
22860 if loop is unrolled or lcssa form is corrupted.
22861 (tree_predictive_commoning): Rewrite for lcssa form if necessary.
22863 2017-07-28 Bin Cheng <bin.cheng@arm.com>
22865 * tree-predcom.c (initialize_root): Delete.
22866 (execute_pred_commoning_chain): Initialize root vars and replace
22867 reference of non-combined chain directly, rather than call above
22870 2017-07-28 Bin Cheng <bin.cheng@arm.com>
22872 * tree-predcom.c (ref_at_iteration): Add parameter NITERS. Compute
22873 memory reference to DR at (NITERS + ITERS)-th iteration of loop.
22875 2017-07-28 Bin Cheng <bin.cheng@arm.com>
22877 * tree-predcom.c (struct chain): New field init_seq.
22878 (release_chain): Release init_seq.
22879 (prepare_initializers_chain): Record intialization stmts in above
22881 (insert_init_seqs): New function.
22882 (tree_predictive_commoning_loop): Call insert_init_seqs.
22884 2017-07-28 Bin Cheng <bin.cheng@arm.com>
22886 * tree-predcom.c (determine_roots_comp): Skip trivial components.
22888 2017-07-28 Richard Biener <rguenther@suse.de>
22890 * match.pd: Remove superfluous :c.
22891 * genmatch.c (simplify::id): Add member.
22892 (lower_commutative, lower_opt_convert, lower_cond, lower_for):
22894 (current_id): New global.
22895 (dt_node::parent): Move from ...
22896 (dt_operand::parent): ... here. Add for_id member.
22897 (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
22898 (decision_tree::find_node): Relax order requirement when
22899 merging DT_TRUE nodes to ones inbetween the current simplify
22900 and the one we try to merge with. Add diagnostic whenever
22901 we need to enforce pattern order by not merging.
22902 (decision_tree::insert): Set current_id.
22903 (decision_tree::print_node): Dump parent node and for_id.
22904 (parser::last_id): Add member.
22905 (parser::push_simplify): Assign unique id.
22906 (parser::parser): Initialize last_id.
22908 2017-07-28 Martin Liska <mliska@suse.cz>
22911 * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
22912 gimple_build_debug_bind.
22914 2017-07-28 Richard Biener <rguenther@suse.de>
22916 PR tree-optimization/81502
22917 * match.pd: Add pattern combining BIT_INSERT_EXPR with
22919 * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
22921 (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
22922 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
22923 for BIT_FIELD_REF args.
22924 * fold-const.c (make_bit_field_ref): Likewise.
22925 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22927 2017-07-28 Jakub Jelinek <jakub@redhat.com>
22930 * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
22931 * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
22932 * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
22933 Or it into SANITIZER_UNDEFINED.
22934 * ubsan.c: Include gimple-fold.h and varasm.h.
22935 (ubsan_expand_ptr_ifn): New function.
22936 (instrument_pointer_overflow): New function.
22937 (maybe_instrument_pointer_overflow): New function.
22938 (instrument_object_size): Formatting fix.
22939 (pass_ubsan::execute): Call instrument_pointer_overflow
22940 and maybe_instrument_pointer_overflow.
22941 * internal-fn.c (expand_UBSAN_PTR): New function.
22942 * ubsan.h (ubsan_expand_ptr_ifn): Declare.
22943 * sanitizer.def (__ubsan_handle_pointer_overflow,
22944 __ubsan_handle_pointer_overflow_abort): New builtins.
22945 * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
22946 * internal-fn.def (UBSAN_PTR): New internal function.
22947 * opts.c (sanitizer_opts): Add pointer-overflow.
22948 * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
22949 * fold-const.c (build_range_check): Compute pointer range check in
22950 integral type if pointer arithmetics would be needed. Formatting
22953 2017-07-28 Martin Liska <mliska@suse.cz>
22956 * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
22957 parameters that are of a variable-length.
22959 2017-07-28 Sebastian Huber <sebastian.huber@embedded-brains.de>
22961 * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h. Add
22963 * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
22964 (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
22965 (CRT_CALL_STATIC_FUNCTION): Likewise.
22966 (ASM_DEFAULT_SPEC): New define.
22967 (ASM_SPEC32): Likewise.
22968 (ASM_SPEC64): Likewise.
22969 (ASM_SPEC_COMMON): Likewise.
22970 (ASM_SPEC): Likewise.
22971 (INVALID_64BIT): Likewise.
22972 (LINK_OS_DEFAULT_SPEC): Likewise.
22973 (LINK_OS_SPEC32): Likewise.
22974 (LINK_OS_SPEC64): Likewise.
22975 (POWERPC_LINUX): Likewise.
22976 (PTRDIFF_TYPE): Likewise.
22977 (RESTORE_FP_PREFIX): Likewise.
22978 (RESTORE_FP_SUFFIX): Likewise.
22979 (SAVE_FP_PREFIX): Likewise.
22980 (SAVE_FP_SUFFIX): Likewise.
22981 (SIZE_TYPE): Likewise.
22982 (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
22983 (TARGET_64BIT): Likewise.
22984 (TARGET_64BIT): Likewise.
22985 (TARGET_AIX): Likewise.
22986 (WCHAR_TYPE_SIZE): Likewise.
22987 (WCHAR_TYPE): Undefine.
22988 (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
22989 (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
22990 (CPP_OS_RTEMS_SPEC): Delete.
22991 (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems. Add
22992 asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
22994 * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
22996 2017-07-28 Jakub Jelinek <jakub@redhat.com>
22998 PR tree-optimization/81578
22999 * tree-parloops.c (build_new_reduction): Bail out if
23000 reduction_code isn't one of the standard OpenMP reductions.
23001 Move the details printing after that decision.
23003 2017-07-27 Peter Bergner <bergner@vnet.ibm.com>
23005 * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23006 related to reload_in_progress.
23007 (splat_input_operand): Likewise.
23008 * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23010 * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23012 (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23013 (TARGET_INSTANTIATE_DECLS): Likewise.
23014 (legitimate_indexed_address_p): Delete reload_in_progress code.
23015 (rs6000_debug_legitimate_address_p): Likewise.
23016 (rs6000_eliminate_indexed_memrefs): Likewise.
23017 (rs6000_emit_le_vsx_store): Likewise.
23018 (rs6000_emit_move_si_sf_subreg): Likewise.
23019 (rs6000_emit_move): Likewise.
23020 (register_to_reg_type): Likewise.
23021 (rs6000_pre_atomic_barrier): Likewise.
23022 (rs6000_machopic_legitimize_pic_address): Likewise.
23023 (rs6000_allocate_stack_temp): Likewise.
23024 (rs6000_address_for_fpconvert): Likewise.
23025 (rs6000_address_for_altivec): Likewise.
23026 (rs6000_secondary_memory_needed_rtx): Delete function.
23027 (rs6000_check_sdmode): Likewise.
23028 (rs6000_alloc_sdmode_stack_slot): Likewise.
23029 (rs6000_instantiate_decls): Likewise.
23030 * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23031 * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23032 Delete reload_in_progress.
23033 (*vec_reload_and_plus_<mptrsize>): Likewise.
23034 * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23035 (vsx_div_v2di): Likewise.
23036 (vsx_udiv_v2di): Likewise.
23038 2017-07-27 Peter Bergner <bergner@vnet.ibm.com>
23040 * config/rs6000/rs6000.opt (mlra): Replace with stub.
23041 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23042 * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23043 (rs6000_debug_reg_global): Delete print of LRA status.
23044 (rs6000_option_override_internal): Delete dead LRA related code.
23045 (rs6000_lra_p): Delete function.
23046 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23048 2017-07-27 Sebastian Huber <sebastian.huber@embedded-brains.de>
23050 * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23051 * config/riscv/rtems.h: New file.
23053 2017-07-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23054 Sudakshina Das <sudi.das@arm.com>
23056 * config/aarch64/aarch64.md
23057 (define_split for and<mode>3nr_compare): Move
23058 non aarch64_logical_operand to a register.
23059 (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23060 register immediate operand to a register.
23061 * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23063 2017-07-27 Peter Bergner <bergner@vnet.ibm.com>
23065 PR middle-end/81564
23066 * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23068 2017-07-27 Richard Biener <rguenther@suse.de>
23070 PR tree-optimization/81573
23071 PR tree-optimization/81494
23072 * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23073 multi defuse cycle case.
23075 2017-07-27 Richard Biener <rguenther@suse.de>
23077 PR tree-optimization/81571
23078 * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23081 2017-07-27 Eric Botcazou <ebotcazou@adacore.com>
23083 * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23084 earlier and only if MASK_FPU is set. Adjust formatting.
23086 2017-07-27 Martin Liska <mliska@suse.cz>
23088 * opt-functions.awk: Add validation of value of Init.
23089 * optc-gen.awk: Pass new argument.
23091 2017-07-27 Martin Liska <mliska@suse.cz>
23093 * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23094 Fix wrong condition.
23096 2017-07-27 Martin Liska <mliska@suse.cz>
23098 * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23099 BBs and edges seen by autoFDO.
23101 2017-07-27 Richard Biener <rguenther@suse.de>
23103 PR tree-optimization/81502
23104 * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23105 with incompatible but same sized type.
23106 (execute_update_addresses_taken): Likewise.
23108 2017-07-27 James Greenhalgh <james.greenhalgh@arm.com>
23110 * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23111 flag_tree_loop_vectorize rather than flag_tree_vectorize.
23113 2017-07-27 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23116 * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23117 ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23118 Change s_operand to memory_operand.
23120 2017-07-27 Richard Sandiford <richard.sandiford@linaro.org>
23122 * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23123 * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23124 (rs6000_emit_le_vsx_permute): ...this. Take the destination as input.
23125 Emit instructions rather than returning an expression. Handle TFmode
23126 and KFmode by casting to TImode.
23127 (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23128 (rs6000_emit_le_vsx_store): Likewise.
23129 * config/rs6000/vsx.md (VSX_TI): New iterator.
23130 (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23131 (*vsx_le_undo_permute_<mode>): Likewise.
23132 (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23133 emit the split sequence.
23134 (*vsx_le_perm_store_<mode>): Likewise.
23136 2017-07-27 Jakub Jelinek <jakub@redhat.com>
23138 PR tree-optimization/81555
23139 PR tree-optimization/81556
23140 * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23141 if true, force CHANGED for the recursive invocation.
23142 (reassociate_bb): Remember original length of ops array, pass
23143 len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23145 * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23146 attributes for noipa attribute. For naked attribute use
23147 lookup_attribute first before lookup_attribute_spec.
23148 * final.c (rest_of_handle_final): Disable IPA RA for functions with
23150 * ipa-visibility.c (non_local_p): Fix comment typos. Return true
23151 for functions with noipa attribute.
23152 (cgraph_externally_visible_p): Return true for functions with noipa
23154 * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23155 for functions with noipa attribute.
23156 * doc/extend.texi: Document noipa function attribute.
23157 * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23158 also for functions with noipa attribute.
23159 (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23161 2017-07-26 Andrew Pinski <apinski@cavium.com>
23163 * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23164 vec_unalign_load_cost and vec_unalign_store_cost.
23166 2017-07-26 Michael Meissner <meissner@linux.vnet.ibm.com>
23168 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23169 -mvsx-small-integer option.
23170 (ISA_3_0_MASKS_IEEE): Likewise.
23171 (OTHER_VSX_VECTOR_MASKS): Likewise.
23172 (POWERPC_MASKS): Likewise.
23173 * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23174 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23175 code, only testing for DImode being allowed in non-VSX floating
23177 (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23178 to TARGET_P8_VECTOR test. Remove redundant VSX test inside of
23180 (rs6000_option_override_internal): Delete -mvsx-small-integer.
23181 (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23182 TARGET_P8_VECTOR test.
23183 (rs6000_secondary_reload_simple_move): Likewise.
23184 (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23185 since TARGET_P9_VECTOR was already tested.
23186 (rs6000_opt_masks): Remove -mvsx-small-integer.
23187 * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23188 TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23190 (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23191 test for TARGET_VEXTRACTUB was used, and that uses
23193 (p9 extract splitter): Likewise.
23194 (vsx_extract_<mode>_di_p9): Likewise.
23195 (vsx_extract_<mode>_store_p9): Likewise.
23196 (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23197 for TARGET_P9_VECTOR was used. Delete code that is now dead with
23198 the elimination of TARGET_VSX_SMALL_INTEGER.
23199 (vsx_extract_<mode>_p8): Likewise.
23200 (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23201 (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23202 (vsx_set_<mode>_p9): Likewise.
23203 (vsx_set_v4sf_p9): Likewise.
23204 (vsx_set_v4sf_p9_zero): Likewise.
23205 (vsx_insert_extract_v4sf_p9): Likewise.
23206 (vsx_insert_extract_v4sf_p9_2): Likewise.
23207 * config/rs6000/rs6000.md (sign extend splitter): Change
23208 TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23209 (floatsi<mode>2_lfiwax_mem): Likewise.
23210 (floatunssi<mode>2_lfiwzx_mem): Likewise.
23211 (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23212 since a test for TARGET_P9_VECTOR was used.
23213 (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23214 (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23215 (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23216 (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23217 TARGET_P8_VECTOR test.
23218 (fix_trunc<mode>si2_stfiwx): Likewise.
23219 (fix_trunc<mode>si2_internal): Likewise.
23220 (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23221 TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23223 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23224 (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23225 TARGET_P8_VECTOR test.
23226 (fixuns_trunc<mode>si2_stfiwx): Likewise.
23227 (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23228 TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23230 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23231 (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23232 since a test for TARGET_P9_VECTOR was used.
23233 (splitter for loading small constants): Likewise.
23235 2017-07-26 Andrew Pinski <apinski@cavium.com>
23237 * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23240 2017-07-26 H.J. Lu <hongjiu.lu@intel.com>
23243 * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23244 (fp_valid_at): Likewise.
23246 2017-07-26 James Greenhalgh <james.greenhalgh@arm.com>
23248 * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23249 (qdf24xx_addrcost_table): Likewise.
23250 (cortexa57_tunings): Update to use generic_branch_cost.
23251 (cortexa72_tunings): Likewise.
23252 (cortexa73_tunings): Likewise.
23253 (qdf24xx_tunings): Likewise.
23255 2017-07-26 James Greenhalgh <james.greenhalgh@arm.com>
23257 * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23258 (thunderx2t99_branch_cost): Likewise.
23259 (cortexa35_tunings): Update to use generic_branch_cost.
23260 (cortexa53_tunings): Likewise.
23261 (cortexa57_tunings): Likewise.
23262 (cortexa72_tunings): Likewise.
23263 (cortexa73_tunings): Likewise.
23264 (thunderx2t99_tunings): Likewise.
23266 2017-07-26 Sebastian Huber <sebastian.huber@embedded-brains.de>
23268 * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23269 (sparc_option_override): Honour MASK_FSMULD.
23270 * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23271 * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23272 * config/sparc/sparc.opt (mfsmuld): New option.
23273 * doc/invoke.texi (mfsmuld): Document option.
23275 2017-07-26 Marek Polacek <polacek@redhat.com>
23277 PR middle-end/70992
23278 * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23280 2017-07-26 Richard Biener <rguenther@suse.de>
23282 * gimple-match-head.c (do_valueize): Return OP if valueize
23284 (get_def): New helper to get at the def stmt of a SSA name
23285 if valueize allows.
23286 * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23287 do_valueize to get at the def stmt.
23288 (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23290 2017-07-26 Wilco Dijkstra <wdijkstr@arm.com>
23292 PR middle-end/46932
23293 * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23295 2017-07-26 Martin Liska <mliska@suse.cz>
23298 * function.c (expand_function_start): Make expansion of
23299 nonlocal_goto_save_area after parm_birth_insn.
23301 2017-07-26 Sebastian Huber <sebastian.huber@embedded-brains.de>
23303 * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23304 from all CPU target flags enable members.
23306 2017-07-26 Richard Biener <rguenther@suse.de>
23308 * genmatch.c (dt_simplify::gen): Make iterator vars const.
23309 (decision_tree::gen): Make 'type' const.
23310 (write_predicate): Likewise.
23312 2017-07-24 Michael Meissner <meissner@linux.vnet.ibm.com>
23314 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23315 Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23316 (rs6000_option_override_internal): Likewise.
23317 (rs6000_expand_vector_set): Likewise.
23318 * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23319 (TARGET_UPPER_REGS_SF): Likewise.
23320 (TARGET_UPPER_REGS_DI): Likewise.
23321 (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23322 (TARGET_DIRECT_MOVE_64BIT): Likewise.
23323 * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23324 (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23325 (Splitters for DI constants in Altivec registers): Likewise.
23326 * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23327 (vsx_set_v4sf_p9): Likewise.
23328 (vsx_set_v4sf_p9_zero): Likewise.
23329 (vsx_insert_extract_v4sf_p9): Likewise.
23330 (vsx_insert_extract_v4sf_p9_2): Likewise.
23332 2017-07-25 Carl Love <cel@us.ibm.com>
23334 * doc/extend.texi: Update the built-in documentation file for the
23335 existing built-in functions
23336 vector signed char vec_cnttz (vector signed char);
23337 vector unsigned char vec_cnttz (vector unsigned char);
23338 vector signed short vec_cnttz (vector signed short);
23339 vector unsigned short vec_cnttz (vector unsigned short);
23340 vector signed int vec_cnttz (vector signed int);
23341 vector unsigned int vec_cnttz (vector unsigned int);
23342 vector signed long long vec_cnttz (vector signed long long);
23343 vector unsigned long long vec_cnttz (vector unsigned long long);
23345 2017-07-25 Andrew Pinski <apinski@cavium.com>
23347 * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23348 accesses where the use is for the first operand of a BIT_INSERT.
23350 2017-07-25 Jim Wilson <jim.wilson@linaro.org>
23353 * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23354 for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23356 2017-07-25 Jim Wilson <jim.wilson@linaro.org>
23358 * config/i386/gstabs.h: Delete.
23359 * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23361 2017-07-25 Uros Bizjak <ubizjak@gmail.com>
23363 * config/i386/i386.c (ix86_decompose_address): Do not check for
23364 register RTX when looking at index_reg or base_reg.
23365 * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23367 2017-07-25 Eric Botcazou <ebotcazou@adacore.com>
23369 * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23370 to update EH info here.
23372 2017-07-25 Alexander Monakov <amonakov@ispras.ru>
23374 * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23376 2017-07-25 Alexander Monakov <amonakov@ispras.ru>
23378 * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23380 2017-07-25 Torsten Duwe <duwe@suse.de>
23382 * common.opt: Introduce -fpatchable-function-entry
23383 command line option, and its variables function_entry_patch_area_size
23384 and function_entry_patch_area_start.
23385 * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23386 including a two-value parser.
23387 * target.def (print_patchable_function_entry): New target hook.
23388 * targhooks.h (default_print_patchable_function_entry): New function.
23389 * targhooks.c (default_print_patchable_function_entry): Likewise.
23390 * toplev.c (process_options): Switch off IPA-RA if
23391 patchable function entries are being generated.
23392 * varasm.c (assemble_start_function): Look at the
23393 patchable-function-entry command line switch and current
23394 function attributes and maybe generate NOP instructions by
23395 calling the print_patchable_function_entry hook.
23396 * doc/extend.texi: Document patchable_function_entry attribute.
23397 * doc/invoke.texi: Document -fpatchable_function_entry
23398 command line option.
23399 * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23401 * doc/tm.texi: Re-generate.
23403 2017-07-25 Jakub Jelinek <jakub@redhat.com>
23406 * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23407 TARGET_AVX512DQ rather than TARGET_AVX512BW.
23409 2017-07-25 Tamar Christina <tamar.christina@arm.com>
23411 * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23413 2017-07-25 Richard Biener <rguenther@suse.de>
23415 PR tree-optimization/81455
23416 * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23417 not walk in cycles when looking for guards.
23419 2017-07-25 Richard Biener <rguenther@suse.de>
23421 PR tree-optimization/81529
23422 * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23423 when optimizing backedge uses.
23425 2017-07-25 David Edelsohn <dje.gcc@gmail.com>
23427 * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23429 * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23430 to dl_section_ref. On AIX, append an expression to subtract
23431 the size of the section length to dl_section_ref.
23433 2017-07-25 Segher Boessenkool <segher@kernel.crashing.org>
23435 * configure.ac: If any of the config.* scripts fail, exit 1.
23436 * configure: Regenerate.
23438 2017-07-25 Richard Biener <rguenther@suse.de>
23440 PR middle-end/81546
23441 * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23442 of immediate uses, be more verbose on errors.
23444 2017-07-25 Richard Biener <rguenther@suse.de>
23446 PR tree-optimization/81510
23447 * tree-vect-loop.c (vect_is_simple_reduction): When the
23448 reduction stmt is not inside the loop bail out.
23450 2017-07-25 Richard Biener <rguenther@suse.de>
23452 PR tree-optimization/81303
23453 * tree-vect-loop-manip.c (vect_loop_versioning): Build
23454 profitability check against LOOP_VINFO_NITERSM1.
23456 2017-07-25 Alexander Monakov <amonakov@ispras.ru>
23458 * domwalk.c (cmp_bb_postorder): Simplify.
23459 (sort_bbs_postorder): New function. Use it...
23460 (dom_walker::walk): ...here to optimize common cases.
23462 2017-07-25 Martin Liska <mliska@suse.cz>
23465 * ipa-visibility.c (function_and_variable_visibility): Make the
23466 redirection just on target that supports aliasing.
23467 Fix GNU coding style.
23469 2017-07-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23472 * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23474 * config/arm/rtems.h: Likewise.
23475 * config/bfin/rtems.h: Likewise.
23476 * config/i386/rtemself.h: Likewise.
23477 * config/lm32/rtems.h: Likewise.
23478 * config/m32c/rtems.h: Likewise.
23479 * config/m68k/rtemself.h: Likewise.
23480 * config/microblaze/rtems.h: Likewise.
23481 * config/mips/rtems.h: Likewise.
23482 * config/moxie/rtems.h: Likewise.
23483 * config/nios2/rtems.h: Likewise.
23484 * config/powerpcspe/rtems.h: Likewise.
23485 * config/rs6000/rtems.h: Likewise.
23486 * config/rtems.h: Likewise.
23487 * config/sh/rtems.h: Likewise.
23488 * config/sh/rtemself.h: Likewise.
23489 * config/sparc/rtemself.h: Likewise.
23491 2017-07-25 Georg-Johann Lay <avr@gjlay.de>
23494 * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23495 * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23496 * tree-ssa-structalias.c (alias_get_name): Same.
23498 2017-07-25 Bin Cheng <bin.cheng@arm.com>
23501 * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23502 instructions if no du chain is found.
23504 2017-07-25 Georg-Johann Lay <avr@gjlay.de>
23506 * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23508 2017-07-25 Richard Biener <rguenther@suse.de>
23510 PR middle-end/81505
23511 * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23514 2017-07-24 Michael Meissner <meissner@linux.vnet.ibm.com>
23516 * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23517 upper-regs options.
23518 (ISA_2_7_MASKS_SERVER): Likewise.
23519 (ISA_3_0_MASKS_IEEE): Likewise.
23520 (OTHER_P8_VECTOR_MASKS): Likewise.
23521 (OTHER_VSX_VECTOR_MASKS): Likewise.
23522 (POWERPC_MASKS): Likewise.
23523 (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23524 duplicate list of options.
23525 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23526 explicit -mupper-regs options.
23527 * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23528 -mupper-regs* options. Delete -mvsx-scalar-memory, which was an
23529 alias for -mupper-regs-df.
23530 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23531 (rs6000_init_hard_regno_mode_ok): Likewise.
23532 (rs6000_option_override_internal): Likewise.
23533 (rs6000_opt_masks): Likewise.
23534 * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23535 options in terms of whether -mvsx or -mpower8-vector was used.
23536 (TARGET_UPPER_REGS_DI): Likewise.
23537 (TARGET_UPPER_REGS_SF): Likewise.
23538 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23539 -mupper-regs-* options.
23541 2017-07-24 Segher Boessenkool <segher@kernel.crashing.org>
23543 * passes.c (emergency_dump_function): Print some empty lines and a
23544 header before the RTL dump.
23546 2017-07-24 Segher Boessenkool <segher@kernel.crashing.org>
23548 * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23550 2017-07-24 Wilco Dijkstra <wdijkstr@arm.com>
23553 * config/aarch64/aarch64.c (aarch64_classify_symbol):
23554 Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23556 2017-07-24 Carl Love <cel@us.ibm.com>
23558 * config/rs6000/rs6000-c.c: Add support for built-in functions
23559 vector float vec_extract_fp32_from_shorth (vector unsigned short);
23560 vector float vec_extract_fp32_from_shortl (vector unsigned short);
23561 * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23562 vec_extract_fp_from_shortl): Add defines for the two builtins.
23563 * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23564 VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23566 * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23567 (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23568 * doc/extend.texi: Update the built-in documentation file for the
23569 new built-in function.
23571 2017-07-24 Jakub Jelinek <jakub@redhat.com>
23574 * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23576 * doc/generic.texi: Likewise.
23577 * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23578 for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23580 2017-07-24 Jackson Woodruff <jackson.woodruff@arm.com>
23582 * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23583 (aarch64_mls_elt_merge<mode>): Likewise.
23585 2017-07-23 Krister Walfridsson <krister.walfridsson@gmail.com>
23587 * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23588 having __cxa_atexit.
23590 2017-07-23 Michael Collison <michael.collison@arm.com>
23592 * config/arm/arm.c (arm_option_override): Deprecate
23593 use of -mstructure-size-boundary.
23594 * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23595 * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23597 2017-07-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23600 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23601 Reduce cost estimate for direct moves.
23603 2017-07-23 Uros Bizjak <ubizjak@gmail.com>
23606 * config/i386/i386.c (ix86_option_override_internal): Disable
23607 BMI, BMI2 and TBM instructions for -m16.
23609 2017-07-21 Carl Love <cel@us.ibm.com>
23611 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23612 ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23613 ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23614 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23615 builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23616 * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23617 VMULOSW): New enum "unspec" values.
23618 (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23619 altivec_vmulosw): New patterns.
23620 * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23621 VMULOSW): Add definitions.
23623 2017-07-21 Jim Wilson <jim.wilson@linaro.org>
23625 * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23626 (qdf24xx): Likewise.
23627 * config/aarch64/aarch64-options-extensions.def (rdma); New.
23628 * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23629 (AARCH64_FL_V8_1): Renumber.
23630 (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23631 (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23632 * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23633 * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs. Add
23634 rdma to feature modifiers list.
23636 2017-07-21 Yury Gribov <tetra2005@gmail.com>
23638 PR middle-end/56727
23639 * ipa-visibility (function_and_variable_visibility): Convert
23640 recursive PLT call to direct call if appropriate.
23642 2017-07-21 Andrew Pinski <apinski@cavium.com>
23644 * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23645 operand 1 to see if the types precision matches.
23646 * fold-const.c (operand_equal_p): Likewise.
23648 2017-07-21 Richard Biener <rguenther@suse.de>
23650 PR tree-optimization/81303
23651 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23652 in datarefs vector. Allow NULL dr0 for no peeling cost estimate.
23653 (vect_peeling_hash_get_lowest_cost): Adjust.
23654 (vect_enhance_data_refs_alignment): Likewise. Use
23655 vect_get_peeling_costs_all_drs to compute the penalty for no
23656 peeling to match up costs.
23658 2017-07-21 Richard Biener <rguenther@suse.de>
23660 PR tree-optimization/81500
23661 * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23662 we didn't identify a reduction path.
23664 2017-07-21 Tom de Vries <tom@codesourcery.com>
23665 Cesar Philippidis <cesar@codesourcery.com>
23667 PR gcov-profile/81442
23668 * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23671 2017-07-21 Tom de Vries <tom@codesourcery.com>
23674 * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23676 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23677 nvptx_override_options_after_change.
23679 2017-07-21 Ulrich Drepper <drepper@redhat.com>
23681 * dwarf2out.c (output_file_names): Avoid double testing for
23682 dwarf_version >= 5.
23684 2017-07-21 Georg-Johann Lay <avr@gjlay.de>
23686 * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23688 2016-07-21 Jan Hubicka <hubicka@ucw.cz>
23690 * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23692 (try_crossjump_to_edge): Do not punt on partitioned functions.
23694 2016-07-21 Jan Hubicka <hubicka@ucw.cz>
23696 * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23697 Put all BBs reachable only via paths crossing cold region to cold
23699 * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23701 2016-07-21 Richard Biener <rguenther@suse.de>
23703 PR tree-optimization/81303
23704 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23705 into account prologue and epilogue iterations when raising
23706 min_profitable_iters to sth at least covering one vector iteration.
23708 2017-07-21 Tamar Christina <tamar.christina@arm.com>
23710 * config/arm/arm.c (arm_test_cpu_arch_dat):
23713 2017-07-20 Nathan Sidwell <nathan@acm.org>
23715 Remove TYPE_METHODS.
23716 * tree.h (TYPE_METHODS): Delete.
23717 * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23718 * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23719 (dbxout_type_methods): Scan TYPE_FIELDS.
23720 (dbxout_type): Don't check TYPE_METHODS here.
23721 * function.c (use_register_for_decl): Always ignore register for
23722 class types when not optimizing.
23723 * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23724 * tree.c (free_lang_data_in_type): Stitch out member functions and
23725 templates from TYPE_FIELDS.
23726 (build_distinct_type_copy, verify_type_variant,
23727 verify_type): Member fns are on TYPE_FIELDS.
23728 * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23729 * tree-pretty-print.c (dump_generic_node): Likewise.
23731 2017-07-20 Jakub Jelinek <jakub@redhat.com>
23734 * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23735 V2TImode and V4TImode.
23736 (ix86_expand_vector_extract): Likewise.
23737 * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23738 TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23739 (ssescalarmode): Handle V4TImode and V2TImode.
23740 (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23741 (*vec_extractv2ti, *vec_extractv4ti): New insns.
23742 (VEXTRACTI128_MODE): New mode iterator.
23743 (splitter for *vec_extractv?ti first element): New.
23744 (VEC_INIT_MODE): New mode iterator.
23745 (vec_init<mode>): Consolidate 3 expanders into one using
23746 VEC_INIT_MODE mode iterator.
23748 2017-07-20 Alexander Monakov <amonakov@ispras.ru>
23750 * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23751 non_spilled_static_chain_regno_p.
23753 2017-07-20 Alexander Monakov <amonakov@ispras.ru>
23755 * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23757 2017-07-20 Jan Hubicka <hubicka@ucw.cz>
23759 * bb-reorder.c (connect_traces): Allow copying of blocks within
23762 2017-07-20 Richard Biener <rguenther@suse.de>
23764 * gimple.h (gimple_phi_result): Add gphi * overload.
23765 (gimple_phi_result_ptr): Likewise.
23766 (gimple_phi_arg): Likewise. Adjust index assert to only
23767 allow actual argument accesses rather than all slots available
23769 (gimple_phi_arg_def): Add gphi * overload.
23770 * tree-phinodes.c (make_phi_node): Initialize only actual
23772 (resize_phi_node): Clear memory not covered by old node,
23773 do not initialize excess argument slots.
23774 (reserve_phi_args_for_new_edge): Initialize new argument slot
23777 2017-07-20 Bin Cheng <bin.cheng@arm.com>
23779 PR tree-optimization/81388
23781 2016-07-21 Bin Cheng <bin.cheng@arm.com>
23783 * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23784 by removing computation of may_be_zero.
23786 2017-07-18 Jan Hubicka <hubicka@ucw.cz>
23787 Tom de Vries <tom@codesourcery.com>
23789 PR middle-end/81030
23790 * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23791 when gimple level profile disagrees with what RTL expander did.
23793 2017-07-20 Richard Biener <rguenther@suse.de>
23795 PR tree-optimization/61171
23796 * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23797 (vect_analyze_stmt): Add slp instance parameter.
23798 (vectorizable_reduction): Likewise.
23799 * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23800 (vect_is_simple_reduction): Deal with chains not detected
23801 as SLP reduction chain, specifically not properly associated
23802 chains containing a mix of plus/minus.
23803 (get_reduction_op): Remove.
23804 (get_initial_defs_for_reduction): Simplify, pass in whether
23805 this is a reduction chain, pass in the SLP node for the PHIs.
23806 (vect_create_epilog_for_reduction): Get the SLP instance as
23808 (vectorizable_reduction): Get the SLP instance as arg.
23809 During analysis remember the SLP node with the PHIs in the
23810 instance. Simplify getting at the vectorized reduction PHIs.
23811 * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23812 through SLP instance.
23813 (vect_slp_analyze_operations): Likewise.
23814 * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23815 (vect_transform_stmt): Likewise.
23817 2017-07-20 Tom de Vries <tom@codesourcery.com>
23819 PR tree-optimization/81489
23820 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23821 read of phi arg location to before loop that modifies phi.
23823 2017-07-20 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
23825 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23828 2017-07-19 Jan Hubicka <hubicka@ucw.cz>
23830 PR middle-end/81331
23831 * except.c (execute): Fix ordering issue.
23833 2018-07-19 Segher Boessenkool <segher@kernel.crashing.org>
23835 PR rtl-optimization/81423
23836 * combine.c (make_compound_operation_int): Don't try to optimize
23837 the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23839 2017-07-19 Segher Boessenkool <segher@kernel.crashing.org>
23841 PR rtl-optimization/81423
23842 * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23843 with a constant that is -1 in the truncated to mode.
23845 2017-07-19 Jan Hubicka <hubicka@ucw.cz>
23847 * predict.c (propagate_unlikely_bbs_forward): Break out from ...
23848 (determine_unlikely_bbs): ... here.
23849 * predict.h (propagate_unlikely_bbs_forward): Declare.
23850 * cfgexpand.c (pass_expand::execute): Use it.
23851 * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
23853 (find_rarely_executed_basic_blocks_and_crossing_edges): Use
23854 propagate_unlikely_bbs_forward.
23856 2017-07-19 Jan Hubicka <hubicka@ucw.cz>
23858 PR middle-end/81331
23859 * except.c (maybe_add_nop_after_section_switch): New function.
23862 2017-07-19 Tom de Vries <tom@codesourcery.com>
23864 * gimple.h (gimple_phi_set_arg): Make assert more strict.
23866 2017-07-19 Tom de Vries <tom@codesourcery.com>
23868 * gimple.h (gimple_phi_arg): Make assert more strict.
23870 2017-07-19 Steven Munroe <munroesj@gcc.gnu.org>
23872 * config.gcc (powerpc*-*-*): Add mmintrin.h.
23873 * config/rs6000/mmintrin.h: New file.
23874 * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
23876 2017-07-19 Jakub Jelinek <jakub@redhat.com>
23878 PR tree-optimization/81346
23879 * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
23881 2017-07-19 Tom de Vries <tom@codesourcery.com>
23883 * config/nvptx/nvptx.md (VECIM): Add V2DI.
23885 2017-07-19 Tom de Vries <tom@codesourcery.com>
23887 * config/nvptx/nvptx-modes.def: Add V2DImode.
23888 * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
23889 * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
23890 (nvptx_output_mov_insn): Handle lack of mov.b128.
23891 (nvptx_print_operand): Handle 'H' and 'L' codes.
23892 (nvptx_vector_mode_supported): Allow V2DImode.
23893 (nvptx_preferred_simd_mode): New function.
23894 (nvptx_data_alignment): New function.
23895 (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
23896 nvptx_preferred_simd_mode.
23897 * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
23899 (DATA_ALIGNMENT): Define. Set to nvptx_data_alignment.
23901 2017-07-19 Tom de Vries <tom@codesourcery.com>
23903 * config/nvptx/nvptx-modes.def: New file. Add V2SImode.
23904 * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
23905 (nvptx_vector_mode_supported): New function. Allow V2SImode.
23906 (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
23907 * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
23908 (mov<VECIM>_insn): New define_insn.
23909 (define_expand "mov<VECIM>): New define_expand.
23911 2017-07-19 Tom de Vries <tom@codesourcery.com>
23913 * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
23915 2017-07-19 Jakub Jelinek <jakub@redhat.com>
23917 PR tree-optimization/81346
23918 * fold-const.h (fold_div_compare, range_check_type): Declare.
23919 * fold-const.c (range_check_type): New function.
23920 (build_range_check): Use range_check_type.
23921 (fold_div_compare): No longer static, rewritten into
23922 a match.pd helper function.
23923 (fold_comparison): Don't call fold_div_compare here.
23924 * match.pd (X / C1 op C2): New optimization using fold_div_compare
23925 as helper function.
23927 2017-07-19 Nathan Sidwell <nathan@acm.org>
23929 * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
23930 (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
23931 * tree.c (find_decls_types_r, verify_type): Use
23932 TYPE_{MIN,MAX}_VALUE_RAW.
23933 * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
23934 (hash_tree): Likewise.
23935 * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
23937 * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
23940 2017-07-18 Tom de Vries <tom@codesourcery.com>
23942 PR middle-end/81464
23943 * omp-expand.c (expand_omp_for_static_chunk): Handle
23944 equal-argument loop exit phi.
23946 2017-07-18 Uros Bizjak <ubizjak@gmail.com>
23949 * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
23950 (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
23951 operand 2 predicate.
23952 (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
23953 operand 2 predicate.
23954 (ror,rol -> rorx splitters): Use const_int_operand as
23955 operand 2 predicate.
23957 2017-06-18 Richard Biener <rguenther@suse.de>
23959 PR tree-optimization/81410
23960 * tree-vect-stmts.c (vectorizable_load): Properly adjust for
23961 the gap in the ! slp_perm SLP case after each group.
23963 2017-07-18 Jan Hubicka <hubicka@ucw.cz>
23965 PR middle-end/81463
23966 * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
23969 2017-07-18 Jan Hubicka <hubicka@ucw.cz>
23971 PR middle-end/81462
23972 * predict.c (set_even_probabilities): Cleanup; do not affect
23973 probabilities that are already known.
23974 (combine_predictions_for_bb): Call even when count is set.
23976 2017-07-18 Nathan Sidwell <nathan@acm.org>
23978 * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
23981 2017-07-18 Bin Cheng <bin.cheng@arm.com>
23984 * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
23985 optimization for loop niter analysis.
23987 2017-07-18 Georg-Johann Lay <avr@gjlay.de>
23990 * config/avr/avr.c (avr_optimize_casesi): Don't use
23991 INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
23993 2017-07-18 Robin Dapp <rdapp@linux.vnet.ibm.com>
23995 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
23996 body_cost_vec from _vect_peel_extended_info.
23997 (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
23998 (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24001 2017-07-18 Bin Cheng <bin.cheng@arm.com>
24003 * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24005 2017-07-18 Richard Biener <rguenther@suse.de>
24007 PR tree-optimization/80620
24008 PR tree-optimization/81403
24009 * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24010 info when re-using a VN table entry.
24012 2017-07-18 Richard Biener <rguenther@suse.de>
24014 PR tree-optimization/81418
24015 * tree-vect-loop.c (vectorizable_reduction): Properly compute
24016 vectype_in. Verify that with lane-reducing reduction operations
24017 we have a single def-use cycle.
24019 2017-07-17 Carl Love <cel@us.ibm.com>
24021 Revert commit r249424 2017-06-20 Carl Love <cel@us.ibm.com>
24023 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24024 ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24025 ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24026 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24027 builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24028 * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24029 VMULOSW): New enum "unspec" values.
24030 (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24031 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24032 altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24033 altivec_vmulosw): New patterns.
24034 * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24035 VMULOSW): Add definitions.
24037 2017-07-17 Uros Bizjak <ubizjak@gmail.com>
24039 * config/alpha/alpha.c: Include predict.h.
24041 2017-07-17 Yury Gribov <tetra2005@gmail.com>
24043 * tree-vrp.c (compare_assert_loc): Fix comparison function
24044 to return predictable results.
24046 2017-07-17 Claudiu Zissulescu <claziss@synopsys.com>
24048 * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24050 (subdi3): Likewise.
24051 * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24052 * doc/invoke.texi (mexpand-adddi): Update text.
24054 2017-07-17 Claudiu Zissulescu <claziss@synopsys.com>
24056 * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24057 that also clobbers the CC register. The old expand code is moved
24059 (*arc_clzsi2): ... here.
24060 (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24061 the CC register. The old expand code is moved to ...
24062 (arc_ctzsi2): ... here.
24064 2017-07-17 Claudiu Zissulescu <claziss@synopsys.com>
24066 * config/arc/arc.opt (mindexed-loads): Use initial value
24067 TARGET_INDEXED_LOADS_DEFAULT.
24068 (mauto-modify-reg): Use initial value
24069 TARGET_AUTO_MODIFY_REG_DEFAULT.
24070 * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24071 (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24072 * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24073 (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24075 2017-07-17 Martin Liska <mliska@suse.cz>
24078 * opts.c (finish_options): Do not allow -fgnu-tm
24079 w/ -fsanitize={kernel-,}address. Say sorry.
24081 2017-07-17 Bin Cheng <bin.cheng@arm.com>
24084 * tree-loop-distribution.c (classify_partition): Only assert on
24085 numer of iterations.
24086 (merge_dep_scc_partitions): Delete prameter. Update function call.
24087 (distribute_loop): Remove code handling loop with unknown niters.
24088 (pass_loop_distribution::execute): Skip loop with unknown niters.
24090 2017-07-17 Bin Cheng <bin.cheng@arm.com>
24093 * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24094 function sort_partitions_by_post_order.
24096 2017-07-17 Bin Cheng <bin.cheng@arm.com>
24098 PR tree-optimization/81374
24099 * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24100 the max index of basic blocks, rather than number of basic blocks.
24102 2017-07-17 Claudiu Zissulescu <claziss@synopsys.com>
24104 * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24106 (arc_legitimate_pic_operand_p): Likewise.
24107 * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24109 (arc_needs_pcl_p): Likewise.
24110 (arc_legitimate_pc_offset_p): Likewise.
24111 (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24112 function is also used in constrains.md.
24113 (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24114 validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24115 PLUS. Only return true/false in known cases, otherwise assert.
24116 (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24117 is already called in arc_legitimate_constant_p.
24118 * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24120 (LEGITIMATE_PIC_OPERAND_P): Use
24121 arc_raw_symbolic_reference_mentioned_p function.
24122 * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24127 2017-07-17 Claudiu Zissulescu <claziss@synopsys.com>
24128 Andrew Burgess <andrew.burgess@embecosm.com>
24130 * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24131 (arc_return_address_register): New function.
24132 * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24133 (arc_handle_fndecl_attribute): Add naked attribute.
24134 (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24135 (TARGET_WARN_FUNC_RETURN): Likewise.
24136 (arc_allocate_stack_slots_for_args): New function.
24137 (arc_warn_func_return): Likewise.
24138 (machine_function): Change type fn_type.
24139 (arc_compute_function_type): Consider new naked function type,
24140 change function return type.
24141 (arc_must_save_register): Adapt to handle new
24142 arc_compute_function_type's return type.
24143 (arc_expand_prologue): Likewise.
24144 (arc_expand_epilogue): Likewise.
24145 (arc_return_address_regs): Delete.
24146 (arc_return_address_register): New function.
24147 (arc_epilogue_uses): Use above function.
24148 * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24149 (arc_function_type): Change encoding, add naked type.
24150 (ARC_INTERRUPT_P): Change to handle the new encoding.
24151 (ARC_FAST_INTERRUPT_P): Likewise.
24152 (ARC_NORMAL_P): Define.
24153 (ARC_NAKED_P): Likewise.
24154 (arc_compute_function_type): Delete prototype.
24155 * config/arc/arc.md (in_ret_delay_slot): Use
24156 arc_return_address_register function.
24157 (simple_return): Likewise.
24158 (p_return_i): Likewise.
24160 2017-07-17 Jakub Jelinek <jakub@redhat.com>
24162 PR tree-optimization/81428
24163 * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24164 can't be built for those types.
24166 2017-07-17 Georg-Johann Lay <avr@gjlay.de>
24168 Remove stuff dead since r239246.
24170 * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24171 * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24172 (avr_inform_devices): Remove dead stuff.
24174 2017-07-17 Tamar Christina <tamar.christina@arm.com>
24176 * config/arm/arm_neon.h: Fix softp typo.
24178 2017-07-17 Jakub Jelinek <jakub@redhat.com>
24180 PR tree-optimization/81365
24181 * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24182 aggregate moves onto bb predecessor edges, make sure there are no
24183 loads that could alias the lhs in between the start of bb and the
24186 2017-07-17 Georg-Johann Lay <avr@gjlay.de>
24189 * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24190 (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24191 [LSHIFTRT, outer_code = TRUNCATE]: Same.
24193 2017-07-17 Jakub Jelinek <jakub@redhat.com>
24195 PR tree-optimization/81396
24196 * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24197 (init_symbolic_number): Initialize it to 1.
24198 (perform_symbolic_merge): Add n_ops from both operands into the new
24200 (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24201 without base_addr as useless if they need more than one operation.
24202 (bswap_replace): Handle !bswap case for NULL base_addr.
24204 2017-07-17 Tom de Vries <tom@codesourcery.com>
24207 * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24210 2017-07-17 Sebastian Huber <sebastian.huber@embedded-brains.de>
24212 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24213 conditional builtin define __FIX_LEON3FT_B2BST.
24215 2017-07-17 Daniel Cederman <cederman@gaisler.com>
24217 * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24218 MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24221 2017-07-16 Eric Botcazou <ebotcazou@adacore.com>
24223 PR rtl-optimization/81424
24224 * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24225 to remove potential trapping from operands if -fnon-call-exceptions.
24227 2017-07-16 Jan Hubicka <hubicka@ucw.cz>
24229 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24230 profile_proability for scalling.
24231 * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24233 2017-07-16 Jan Hubicka <hubicka@ucw.cz>
24235 * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24237 2017-07-16 Jan Hubicka <hubicka@ucw.cz>
24239 * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24240 fixpoint arithmetics.
24242 2017-07-16 Jan Hubicka <hubicka@ucw.cz>
24244 * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24245 fixpoint arithmetics.
24247 2017-07-16 Jan Hubicka <hubicka@ucw.cz>
24249 * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24250 fixpoint arithmetics.
24252 2017-07-16 Jan Hubicka <hubicka@ucw.cz>
24254 * profile-count.h (profile_probability::from_reg_br_prob_note,
24255 profile_probability::to_reg_br_prob_note): New functions.
24256 * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24257 * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24258 * predict.c (probability_reliable_p): Update.
24259 (edge_probability_reliable_p): Update.
24260 (br_prob_note_reliable_p): Update.
24261 (invert_br_probabilities): Update.
24262 (add_reg_br_prob_note): New function.
24263 (combine_predictions_for_insn): Update.
24264 * asan.c (asan_clear_shadow): Update.
24265 * cfgbuild.c (compute_outgoing_frequencies): Update.
24266 * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24267 (update_br_prob_note): Update.
24268 (rtl_verify_edges): Update.
24269 (purge_dead_edges): Update.
24270 (fixup_reorder_chain): Update.
24271 * emit-rtl.c (try_split): Update.
24272 * ifcvt.c (cond_exec_process_insns): Update.
24273 (cond_exec_process_if_block): Update.
24274 (dead_or_predicable): Update.
24275 * internal-fn.c (expand_addsub_overflow): Update.
24276 (expand_neg_overflow): Update.
24277 (expand_mul_overflow): Update.
24278 * loop-doloop.c (doloop_modify): Update.
24279 * loop-unroll.c (compare_and_jump_seq): Update.
24280 * optabs.c (emit_cmp_and_jump_insn_1): Update.
24281 * predict.h: Update.
24282 * reorg.c (mostly_true_jump): Update.
24284 * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24285 * config/alpha/alpha.c (emit_unlikely_jump): Update.
24286 * config/arc/arc.c: (emit_unlikely_jump): Update.
24287 * config/arm/arm.c: (emit_unlikely_jump): Update.
24288 * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24289 * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24290 * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24291 (ix86_print_operand): Update.
24292 (ix86_split_fp_branch): Update.
24293 (predict_jump): Update.
24294 * config/ia64/ia64.c (ia64_print_operand): Update.
24295 * config/mmix/mmix.c (mmix_print_operand): Update.
24296 * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24297 (rs6000_expand_split_stack_prologue): Update.
24298 * config/rs6000/rs6000.c: Update.
24299 * config/s390/s390.c (s390_expand_vec_strlen): Update.
24300 (s390_expand_vec_movstr): Update.
24301 (s390_expand_cs_tdsi): Update.
24302 (s390_expand_split_stack_prologue): Update.
24303 * config/sh/sh.c (sh_print_operand): Update.
24304 (expand_cbranchsi4): Update.
24305 (expand_cbranchdi4): Update.
24306 * config/sparc/sparc.c (output_v9branch): Update.
24307 * config/spu/spu.c (get_branch_target): Update.
24308 (ea_load_store_inline): Update.
24309 * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24310 * config/tilepro/tilepro.c: Update.
24312 2017-07-16 Eric Botcazou <ebotcazou@adacore.com>
24314 * gimplify.c (mostly_copy_tree_r): Revert latest change.
24315 (gimplify_save_expr): Likewise.
24317 2017-07-07 Jan Hubicka <hubicka@ucw.cz>
24319 * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24321 2017-07-07 Jan Hubicka <hubicka@ucw.cz>
24323 * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24325 * timevar.def (TV_IPA_FNSUMMARY): Define.
24327 2017-07-16 Daniel Cederman <cederman@gaisler.com>
24329 * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24330 to back store errata sensitive sequence from being generated.
24331 (sqrtdf2_fix): Likewise.
24333 2017-07-07 Jan Hubicka <hubicka@ucw.cz>
24335 * tree-ssa-threadupdate.c (compute_path_counts,
24336 update_joiner_offpath_counts): Use profile_probability.
24338 2017-07-15 Thomas Preud'homme <thomas.preudhomme@arm.com>
24341 2017-07-14 Thomas Preud'homme <thomas.preudhomme@arm.com>
24343 * config/arm/arm-c.c (arm_cpu_builtins): Define
24344 __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24346 2017-07-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
24348 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24349 array entries to represent __ieee128 versions of the
24350 scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24351 scalar_extract_sig, and scalar_insert_exp built-in functions.
24352 (altivec_resolve_overloaded_builtin): Add special case handling
24353 for the __builtin_scalar_insert_exp function, as represented by
24354 the P9V_BUILTIN_VEC_VSIEDP constant.
24355 * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24356 exponent support for __ieee128 argument.
24357 (VSESQP): Add scalar extract signature support for __ieee128
24359 (VSTDCNQP): Add scalar test negative support for __ieee128
24361 (VSIEQP): Add scalar insert exponent support for __int128 argument
24362 with __ieee128 result.
24363 (VSIEQPF): Add scalar insert exponent support for __ieee128
24364 argument with __ieee128 result.
24365 (VSTDCQP): Add scalar test data class support for __ieee128
24367 (VSTDCNQP): Add overload support for scalar test negative with
24368 __ieee128 argument.
24369 (VSTDCQP): Add overload support for scalar test data class
24370 __ieee128 argument.
24371 * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24372 UNSPEC_VSX_SXSIGDP.
24373 (UNSPEC_VSX_SIEXPQP): New constant.
24374 (xsxexpqp): New insn for VSX scalar extract exponent quad
24376 (xsxsigqp): New insn for VSX scalar extract significand quad
24378 (xsiexpqpf): New insn for VSX scalar insert exponent quad
24379 precision with floating point argument.
24380 (xststdcqp): New expand for VSX scalar test data class quad
24382 (xststdcnegqp): New expand for VSX scalar test negative quad
24384 (xststdcqp): New insn to match expansions for VSX scalar test data
24385 class quad precision and VSX scalar test negative quad precision.
24386 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24387 special case operand checking to enforce that second operand of
24388 VSX scalar test data class with quad precision argument is a 7-bit
24390 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24391 prototypes and descriptions of __ieee128 versions of
24392 scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24393 scalar_test_data_class, and scalar_test_neg built-in functions.
24395 2016-07-14 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24397 PR tree-optimization/81162
24398 * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24399 replace a negate with an add.
24401 2017-07-14 James Greenhalgh <james.greenhalgh@arm.com>
24403 * doc/invoke.texi (arm/-mcpu): Document +crypto.
24405 2017-07-14 Thomas Preud'homme <thomas.preudhomme@arm.com>
24407 * config/arm/arm-c.c (arm_cpu_builtins): Define
24408 __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24410 2017-07-14 Thomas Preud'homme <thomas.preudhomme@arm.com>
24412 * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24413 (armv8-r): Set ARM Cortex-R52 as default CPU.
24414 * config/arm/arm-tables.opt: Regenerate.
24415 * config/arm/arm-tune.md: Regenerate.
24416 * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24418 * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24419 extension for -mcpu=cortex-r52.
24421 2017-07-14 Thomas Preud'homme <thomas.preudhomme@arm.com>
24423 * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24424 (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24425 * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24426 (fp-armv8): Define it as FP_ARMv8 only.
24427 config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24428 (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24430 config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24431 TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24432 * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24433 first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24434 than TARGET_FPU_ARMV8.
24435 * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24436 __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24437 * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24439 * config/arm/neon.md (neon_vrint): Likewise.
24440 (neon_vcvt): Likewise.
24441 (neon_<fmaxmin_op><mode>): Likewise.
24442 (<fmaxmin><mode>3): Likewise.
24443 * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24444 * config/arm/predicates.md (arm_cond_move_operator): Check against
24445 TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24447 2017-07-14 Jackson Woodruff <jackson.woodruff@arm.com>
24449 * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24450 to top of function.
24452 2017-07-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24454 * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24455 loop in comment with memset.
24457 2017-07-14 Martin Liska <mliska@suse.cz>
24459 * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24460 * dwarf2out.c (is_java): Remove the function.
24461 (output_pubname): Remove usage of the function.
24462 (lower_bound_default): Remove usage of DW_LANG_Java.
24463 (gen_compile_unit_die): Likewise.
24464 * gcc.c: Remove compiler defaults for .java and .zip files.
24465 * gimple-expr.c (remove_suffix): Change as there's no longer
24466 extension than 4-letter one.
24467 * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24468 (gimplify_save_expr): Likewise.
24469 * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24470 as it's possible even for other languages than Java.
24471 * langhooks.h (struct lang_hooks): Remove Java from a comment.
24472 * lto-opts.c (lto_write_options): Remove reference to Java.
24473 * opts.c (strip_off_ending): Update file extension handling.
24474 * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24475 * tree-eh.c (lower_resx): Likewise.
24476 * tree.c (free_lang_data_in_type): Remove dead code.
24477 (find_decls_types_r): Likewise.
24478 (build_common_builtin_nodes): Remove Java from a comment.
24479 (verify_type): Remove dead code.
24480 * varasm.c (assemble_external): Remove Java from a comment.
24482 2017-07-14 Martin Liska <mliska@suse.cz>
24484 * opts.c (finish_options): Add quotes.
24485 (common_handle_option): Likewise.
24487 2017-07-14 Martin Liska <mliska@suse.cz>
24489 * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24490 * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24491 Remove N_SO_PASCAL.
24492 * dwarf2out.c (lower_bound_default): Do not handle
24494 (gen_compile_unit_die): Likewise.
24495 * gcc.c: Remove default extension binding for GNU Pascal.
24496 * stmt.c: Remove Pascal language from a comment.
24497 * xcoffout.c: Likewise.
24499 2017-07-13 David Malcolm <dmalcolm@redhat.com>
24502 * diagnostic-show-locus.c (fixit_cmp): New function.
24503 (layout::layout): Sort m_fixit_hints.
24504 (column_range::column_range): Assert that the values are valid.
24505 (struct char_span): New struct.
24506 (correction::overwrite): New method.
24507 (struct source_line): New struct.
24508 (line_corrections::add_hint): Add assertions. Reimplement memcpy
24509 calls in terms of classes source_line and char_span, and
24510 correction::overwrite.
24511 (selftest::test_overlapped_fixit_printing_2): New function.
24512 (selftest::diagnostic_show_locus_c_tests): Call it.
24514 2017-07-13 Will Schmidt <will_schmidt@vnet.ibm.com>
24516 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24517 early if there is no lhs.
24519 2017-07-13 Martin Liska <mliska@suse.cz>
24521 * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24522 (gen_reference_type_die): Likewise.
24523 * stor-layout.c: Remove Pascal-related comment.
24525 2017-07-13 Martin Liska <mliska@suse.cz>
24527 * opts.c (finish_options): Add quotes to error messages.
24528 (parse_sanitizer_options): Likewise.
24530 2017-07-13 Thomas Preud'homme <thomas.preudhomme@arm.com>
24532 * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24534 2017-07-13 Richard Earnshaw <rearnsha@arm.com>
24536 * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24538 2017-07-13 Maxim Ostapenko <m.ostapenko@samsung.com>
24540 * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24542 * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24544 2017-07-12 Michael Meissner <meissner@linux.vnet.ibm.com>
24547 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24548 provides the hardware capability bits, define the macro
24549 __BUILTIN_CPU_SUPPORTS__.
24550 * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24551 if GLIBC does not provide the hardware capability bits. Add a
24552 gcc_unreachable call if the built-in cpu function is neither
24553 __builtin_cpu_is nor __builtin_cpu_supports.
24554 (rs6000_get_function_versions_dispatcher): Change the warning
24555 that an old GLIBC is used which does not export the capability
24556 bits to be an error.
24557 * doc/extend.texi (target_clones attribute): Document the
24558 restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24559 (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24560 needed by __builtin_cpu_is and __builtin_cpu_supports. Document
24561 the macros defined by GCC if the newer GLIBC is available.
24563 2017-07-12 Jeff Law <law@redhat.com>
24565 * config/riscv/riscv.c: Remove unnecessary includes. Reorder
24566 remaining includes slightly.
24567 * config/riscv/riscv-builtins.c: Include profile-count.h.
24569 2017-07-12 Georg-Johann Lay <avr@gjlay.de>
24572 * config/avr/avr.c (avr_set_current_function): In diagnostic
24573 messages: Quote keywords and (parts of) identifiers.
24574 [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24577 2017-07-12 Carl Love <cel@us.ibm.com>
24579 * config/rs6000/rs6000-c.c: Add support for built-in functions
24580 vector bool char vec_revb (vector bool char);
24581 vector bool short vec_revb (vector short char);
24582 vector bool int vec_revb (vector bool int);
24583 vector bool long long vec_revb (vector bool long long);
24584 * doc/extend.texi: Update the built-in documentation file for the
24585 new built-in functions.
24587 2017-07-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
24589 * config/s390/s390.md: Remove movcc splitter.
24591 2017-07-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
24593 * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24594 load/store on condition.
24596 2017-07-12 Georg-Johann Lay <avr@gjlay.de>
24599 * config/avr/avr.c (avr_encode_section_info)
24600 [progmem && !TREE_READONLY]: Error if progmem object needs
24603 2017-07-11 Michael Collison <michael.collison@arm.com>
24605 * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24608 2017-07-11 Carl Love <cel@us.ibm.com>
24610 * config/rs6000/rs6000-c.c: Add support for builtins
24611 vector unsigned int vec_parity_lsbb (vector signed int);
24612 vector unsigned int vec_parity_lsbb (vector unsigned int);
24613 vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24614 vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24615 vector unsigned long long vec_parity_lsbb (vector signed long long);
24616 vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24617 * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24618 * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24619 * doc/extend.texi: Update the built-in documentation file for the
24620 new built-in functions.
24622 2017-07-11 David Malcolm <dmalcolm@redhat.com>
24624 * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24625 (layout::m_primary_loc): New field.
24626 (layout::layout): Initialize new field. Move location filtering
24627 logic from here to...
24628 (layout::maybe_add_location_range): ...this new method. Add
24629 support for filtering to just the lines already specified by other
24631 (layout::will_show_line_p): New method.
24632 (gcc_rich_location::add_location_if_nearby): New method.
24633 (selftest::test_add_location_if_nearby): New test function.
24634 (selftest::diagnostic_show_locus_c_tests): Call it.
24635 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24638 2017-07-11 Tom de Vries <tom@codesourcery.com>
24640 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24641 (bb_first_real_insn): New function.
24642 (nvptx_single): Add extra initialization of broadcasted condition
24645 2017-07-11 Nathan Sidwell <nathan@acm.org>
24647 * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24649 2017-07-11 Georg-Johann Lay <avr@gjlay.de>
24651 * doc/extend.texi (AVR Function Attributes): Remove weblink to
24652 Binutils doc as TEXI will mess them up.
24653 * doc/invoke.texi (AVR Options): Same here.
24655 2017-07-11 Daniel Cederman <cederman@gaisler.com>
24657 * config/sparc/sparc.opt (mfix-ut700): New option.
24658 (mfix-gr712rc): Likewise.
24659 (sparc_fix_b2bst): New variable.
24660 * doc/invoke.texi (SPARC options): Document them.
24661 (ARM options): Fix warnings.
24662 * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24663 instructions to prevent sequences that can trigger the store-store
24664 errata for certain LEON3FT processors.
24665 (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24666 (sparc_option_override): Set sparc_fix_b2bst appropriately.
24667 * config/sparc/sparc.md (fix_b2bst): New attribute.
24668 (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24670 2017-07-10 Uros Bizjak <ubizjak@gmail.com>
24673 * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24675 (*rsqrtsf2_sse): Ditto.
24677 (div<mode>3): Macroize insn from divdf3 and divsf3
24678 using MODEF mode iterator.
24680 2017-07-10 Martin Sebor <msebor@redhat.com>
24682 PR tree-optimization/80397
24683 * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24684 instead of testing for equality to INTEGER_TYPE.
24686 2017-07-10 Vineet Gupta <vgupta@synopsys.com>
24688 * config.gcc: Remove uclibc from arc target spec.
24690 2017-07-10 Claudiu Zissulescu <claziss@synopsys.com>
24692 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24694 2017-07-07 Jan Hubicka <hubicka@ucw.cz>
24697 * lto-wrapper.c (remove_option): New function.
24698 (merge_and_complain): Merge PIC/PIE options more realistically.
24700 2017-07-10 Georg-Johann Lay <avr@gjlay.de>
24702 Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24706 * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24707 (HAVE_AS_AVR_MGCCISR_OPTION): If so, AC_DEFINE it.
24708 * config.in: Regenerate.
24709 * configure: Regenerate.
24710 * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24711 * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24712 * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24713 (TARGET_GASISR_PROLOGUES): ...target mask.
24714 * common/config/avr/avr-common.c
24715 (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24716 Set -mgas-isr-prologues.
24717 * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24718 INSERT_PASS_BEFORE for it.
24719 * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24720 * config/avr/avr.c (avr_option_override)
24721 [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24722 (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24723 (avr_attribute_table) <no_gccisr>: Add new function attribute.
24724 (avr_set_current_function) <is_no_gccisr>: Init machine field.
24725 (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24727 (make_avr_pass_pre_proep): New function.
24728 (emit_push_sfr) <treg>: Add argument to function and use it
24729 instead of TMP_REG.
24730 (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24731 and set machine->gasisr.yes.
24732 (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24733 (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24734 __gcc_isr.n_pushed to .L__stack_usage.
24735 (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24736 (avr_asm_final_postscan_insn): ...this new static function.
24737 * config/avr/avr.h (machine_function)
24738 <is_no_gccisr, use_L__stack_usage>: New fields.
24739 <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24740 * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24741 (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24742 (gasisr, *gasisr): New expander and insn.
24743 * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24744 [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24745 * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24747 2017-07-10 Richard Earnshaw <rearnsha@arm.com>
24749 * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24752 2017-07-10 Georg-Johann Lay <avr@gjlay.de>
24754 Move jump-tables out of .text again.
24757 * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24758 (ASM_OUTPUT_ADDR_VEC): New function.
24759 (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24760 (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24761 INSN_ADDRESSes as asm comment.
24762 * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24763 (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24764 (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24765 * config/avr/avr.md (*tablejump): Adjust comment.
24766 * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24767 * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24769 * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24770 (avr_output_addr_vec): New proto.
24771 (avr_log_t) <insn_addresses>: New field.
24773 2017-07-09 H.J. Lu <hongjiu.lu@intel.com>
24776 * config/i386/i386.c (ix86_function_arg_advance): Set
24777 outgoing_args_on_stack to true if there are outgoing arguments
24779 (ix86_function_arg): Likewise.
24780 (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24781 arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24782 * config/i386/i386.h (machine_function): Add
24783 outgoing_args_on_stack.
24785 2017-07-09 Krister Walfridsson <krister.walfridsson@gmail.com>
24787 * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24788 supporting pthreds.
24789 * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24791 2017-07-08 Richard Sandiford <richard.sandiford@linaro.org>
24793 * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24794 (REAL_H): Remove $(MACHMODE_H).
24795 (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24797 (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24798 $(MACHMODE_H) and double-int.h.
24799 (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24801 (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24804 2017-07-07 Andrew Pinski <apinski@cavium.com>
24806 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24807 prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24809 2017-07-07 Michael Meissner <meissner@linux.vnet.ibm.com>
24811 * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24812 Add warning if GCC was not configured to link against a GLIBC that
24813 exports the hardware capability bits.
24814 (make_resolver_func): Make resolver function private and not a
24815 COMDAT function. Create the name with clone_function_name instead
24816 of make_unique_name.
24819 * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24820 correct operand in doing the split.
24822 2017-07-07 Carl Love <cel@us.ibm.com>
24824 * config/rs6000/rs6000-c: Add support for built-in function
24825 vector unsigned short vec_pack_to_short_fp32 (vector float,
24827 * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24828 BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24829 * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24830 * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24831 (convert_4f32_8i16): Add define_expand.
24832 * doc/extend.texi: Update the built-in documentation file for the
24833 new built-in function.
24835 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
24837 * config/sparc/m8.md: New file.
24838 * config/sparc/sparc.md: Include m8.md.
24840 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
24842 * config/sparc/sparc.opt: New option -mvis4b.
24843 * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
24844 (sparc_option_override): Handle VIS4B.
24845 (enum sparc_builtins): Define
24846 SPARC_BUILTIN_DICTUNPACK{8,16,32},
24847 SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
24848 SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
24849 SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
24850 SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
24851 (check_constant_argument): New function.
24852 (sparc_vis_init_builtins): Define builtins
24853 __builtin_vis_dictunpack{8,16,32},
24854 __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
24855 __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
24856 __builtin_vis_fpcmpde{8,16,32}shl and
24857 __builtin_vis_fpcmpur{8,16,32}shl.
24858 (sparc_expand_builtin): Check that the constant operands to
24859 __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
24860 constant and in range.
24861 * config/sparc/sparc-c.c (sparc_target_macros): Handle
24863 * config/sparc/sparc.h (SPARC_IMM2_P): Define.
24864 (SPARC_IMM5_P): Likewise.
24865 * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
24866 (enabled): Handle vis4b.
24867 (UNSPEC_DICTUNPACK): New unspec.
24868 (UNSPEC_FPCMPSHL): Likewise.
24869 (UNSPEC_FPUCMPSHL): Likewise.
24870 (UNSPEC_FPCMPDESHL): Likewise.
24871 (UNSPEC_FPCMPURSHL): Likewise.
24872 (cpu_feature): New CPU feature `vis4b'.
24873 (dictunpack{8,16,32}): New insns.
24874 (FPCSMODE): New mode iterator.
24875 (fpcscond): New code iterator.
24876 (fpcsucond): Likewise.
24877 (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
24878 (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
24879 (fpcmpde{8,16,32}{si,di}shl): Likewise.
24880 (fpcmpur{8,16,32}{si,di}shl): Likewise.
24881 * config/sparc/constraints.md: Define constraints `q' for unsigned
24882 2-bit integer constants and `t' for unsigned 5-bit integer
24884 * config/sparc/predicates.md (imm5_operand_dictunpack8): New
24886 (imm5_operand_dictunpack16): Likewise.
24887 (imm5_operand_dictunpack32): Likewise.
24888 (imm2_operand): Likewise.
24889 * doc/invoke.texi (SPARC Options): Document -mvis4b.
24890 * doc/extend.texi (SPARC VIS Built-in Functions): Document the
24891 ditunpack* and fpcmp*shl builtins.
24893 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
24895 * config.gcc: Handle m8 in --with-{cpu,tune} options.
24896 * config.in: Add HAVE_AS_SPARC6 define.
24897 * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
24899 * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
24901 (ASM_CPU32_DEFAUILT_SPEC): Likewise.
24902 (CPP_CPU_SPEC): Handle m8.
24903 (ASM_CPU_SPEC): Likewise.
24904 * config/sparc/sparc-opts.h (enum processor_type): Add
24906 * config/sparc/sparc.c (m8_costs): New struct.
24907 (sparc_option_override): Handle TARGET_CPU_m8.
24908 (sparc32_initialize_trampoline): Likewise.
24909 (sparc64_initialize_trampoline): Likewise.
24910 (sparc_issue_rate): Likewise.
24911 (sparc_register_move_cost): Likewise.
24912 * config/sparc/sparc.h (TARGET_CPU_m8): Define.
24913 (CPP_CPU64_DEFAULT_SPEC): Define for M8.
24914 (ASM_CPU64_DEFAULT_SPEC): Likewise.
24915 (CPP_CPU_SPEC): Handle M8.
24916 (ASM_CPU_SPEC): Likewise.
24917 (AS_M8_FLAG): Define.
24918 * config/sparc/sparc.md: Add m8 to the cpu attribute.
24919 * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
24920 * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
24922 * configure: Regenerate.
24923 * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
24926 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
24928 * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
24930 * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
24931 ("*movdi_insn_sp32"): Do not set v3pipe.
24932 ("*movsi_insn"): Likewise.
24933 ("*movdi_insn_sp64"): Likewise.
24934 ("*movsf_insn"): Likewise.
24935 ("*movdf_insn_sp32"): Likewise.
24936 ("*movdf_insn_sp64"): Likewise.
24937 ("*zero_extendsidi2_insn_sp64"): Likewise.
24938 ("*sign_extendsidi2_insn"): Likewise.
24939 ("*mov<VM32:mode>_insn"): Likewise.
24940 ("*mov<VM64:mode>_insn_sp64"): Likewise.
24941 ("*mov<VM64:mode>_insn_sp32"): Likewise.
24942 ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
24943 ("<vlop:code><VL:mode>3"): Likewise.
24944 ("*not_<vlop:code><VL:mode>3"): Likewise.
24945 ("*nand<VL:mode>_vis"): Likewise.
24946 ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
24947 ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
24948 ("one_cmpl<VL:mode>2"): Likewise.
24949 ("faligndata<VM64:mode>_vis"): Likewise.
24950 ("alignaddrsi_vis"): Likewise.
24951 ("alignaddrdi_vis"): Likweise.
24952 ("alignaddrlsi_vis"): Likewise.
24953 ("alignaddrldi_vis"): Likewise.
24954 ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
24955 ("bmaskdi_vis"): Likewise.
24956 ("bmasksi_vis"): Likewise.
24957 ("bshuffle<VM64:mode>_vis"): Likewise.
24958 ("cmask8<P:mode>_vis"): Likewise.
24959 ("cmask16<P:mode>_vis"): Likewise.
24960 ("cmask32<P:mode>_vis"): Likewise.
24961 ("pdistn<P:mode>_vis"): Likewise.
24962 ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
24964 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
24966 * config/sparc/sparc.md ("subtype"): New insn attribute.
24967 ("*wrgsr_sp64"): Set insn subtype.
24968 ("*rdgsr_sp64"): Likewise.
24969 ("alignaddrsi_vis"): Likewise.
24970 ("alignaddrdi_vis"): Likewise.
24971 ("alignaddrlsi_vis"): Likewise.
24972 ("alignaddrldi_vis"): Likewise.
24973 ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
24974 ("fexpand_vis"): Likewise.
24975 ("fpmerge_vis"): Likewise.
24976 ("faligndata<VM64:mode>_vis"): Likewise.
24977 ("bshuffle<VM64:mode>_vis"): Likewise.
24978 ("cmask8<P:mode>_vis"): Likewise.
24979 ("cmask16<P:mode>_vis"): Likewise.
24980 ("cmask32<P:mode>_vis"): Likewise.
24981 ("fchksm16_vis"): Likewise.
24982 ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
24983 ("fmean16_vis"): Likewise.
24984 ("fp<plusminus_insn>64_vis"): Likewise.
24985 ("<plusminus_insn>v8qi3"): Likewise.
24986 ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
24987 ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
24988 ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
24989 ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
24990 ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
24991 ("*movqi_insn"): Likewise.
24992 ("*movhi_insn"): Likewise.
24993 ("*movsi_insn"): Likewise.
24994 ("movsi_pic_gotdata_op"): Likewise.
24995 ("*movdi_insn_sp32"): Likewise.
24996 ("*movdi_insn_sp64"): Likewise.
24997 ("movdi_pic_gotdata_op"): Likewise.
24998 ("*movsf_insn"): Likewise.
24999 ("*movdf_insn_sp32"): Likewise.
25000 ("*movdf_insn_sp64"): Likewise.
25001 ("*zero_extendhisi2_insn"): Likewise.
25002 ("*zero_extendqihi2_insn"): Likewise.
25003 ("*zero_extendqisi2_insn"): Likewise.
25004 ("*zero_extendqidi2_insn"): Likewise.
25005 ("*zero_extendhidi2_insn"): Likewise.
25006 ("*zero_extendsidi2_insn_sp64"): Likewise.
25007 ("ldfsr"): Likewise.
25008 ("prefetch_64"): Likewise.
25009 ("prefetch_32"): Likewise.
25010 ("tie_ld32"): Likewise.
25011 ("tie_ld64"): Likewise.
25012 ("*tldo_ldub_sp32"): Likewise.
25013 ("*tldo_ldub1_sp32"): Likewise.
25014 ("*tldo_ldub2_sp32"): Likewise.
25015 ("*tldo_ldub_sp64"): Likewise.
25016 ("*tldo_ldub1_sp64"): Likewise.
25017 ("*tldo_ldub2_sp64"): Likewise.
25018 ("*tldo_ldub3_sp64"): Likewise.
25019 ("*tldo_lduh_sp32"): Likewise.
25020 ("*tldo_lduh1_sp32"): Likewise.
25021 ("*tldo_lduh_sp64"): Likewise.
25022 ("*tldo_lduh1_sp64"): Likewise.
25023 ("*tldo_lduh2_sp64"): Likewise.
25024 ("*tldo_lduw_sp32"): Likewise.
25025 ("*tldo_lduw_sp64"): Likewise.
25026 ("*tldo_lduw1_sp64"): Likewise.
25027 ("*tldo_ldx_sp64"): Likewise.
25028 ("*mov<VM32:mode>_insn"): Likewise.
25029 ("*mov<VM64:mode>_insn_sp64"): Likewise.
25030 ("*mov<VM64:mode>_insn_sp32"): Likewise.
25032 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
25034 * config/sparc/sparc.md ("type"): New insn type viscmp.
25035 ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25037 ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25038 ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25039 ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25040 * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25042 ("n7_vis_logical_11cycle"): Likewise.
25043 * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25044 * config/sparc/niagara2.md ("niag3_vis": Likewise.
25045 * config/sparc/niagara.md ("niag_vis"): Likewise.
25046 * config/sparc/ultra3.md ("us3_fga"): Likewise.
25047 * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25049 2017-07-07 Jose E. Marchesi <jose.marchesi@oracle.com>
25051 * config/sparc/sparc.md: New instruction type `bmask'.
25052 (bmaskdi_vis): Use the `bmask' type.
25053 (bmasksi_vis): Likewise.
25054 * config/sparc/ultra3.md (us3_array): Likewise.
25055 * config/sparc/niagara7.md (n7_array): Likewise.
25056 * config/sparc/niagara4.md (n4_array): Likewise.
25057 * config/sparc/niagara2.md (niag2_vis): Likewise.
25058 (niag3_vis): Likewise.
25059 * config/sparc/niagara.md (niag_vis): Likewise.
25061 2017-07-06 Jan Hubicka <hubicka@ucw.cz>
25063 * ipa-comdats.c: Remove optimize check from gate.
25064 * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25065 for functions not optimized.
25066 (ipa_fn_summary_read): Skip optimize check.
25067 (ipa_fn_summary_write): Likewise.
25068 * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25070 * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25072 (can_inline_edge_p): Check flag_pcc_struct_return for match.
25073 (check_callers): Give up on caller which is not optimized.
25074 (inline_small_functions): Likewise.
25075 (ipa_inline): Do not give up when not optimizing.
25076 * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25077 away unoptimizes cdtors.
25078 (whole_program_function_and_variable_visibility): Do
25079 ipa_discover_readonly_nonaddressable_vars in LTO mode.
25080 * ipa.c (process_references): Do not check optimize.
25081 (symbol_table::remove_unreachable_nodes): Update optimize check.
25082 (set_writeonly_bit): Update optimize check.
25083 (pass_ipa_cdtor_merge::gate): Do not check optimize.
25084 (pass_ipa_single_use::gate): Remove.
25086 2017-07-06 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25088 * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25089 insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25090 rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25091 mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25092 permute_load, permute_store, adjust_extract, adjust_splat,
25093 adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25094 replace_swap_with_copy, dump_swap_insn_table,
25095 alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25096 recombine_lvx_pattern, recombine_stvx_pattern,
25097 recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25098 make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25099 to file rs6000-p8swap.c.
25100 * config/rs6000/rs6000-p8swap.c: New file.
25101 * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25102 * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25103 and rs6000*-*-* targets.
25105 2017-07-06 David Malcolm <dmalcolm@redhat.com>
25107 * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25109 2017-07-06 Jan Hubicka <hubicka@ucw.cz>
25111 * lto-wrapper.c (merge_and_complain): Do not merge
25112 fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25113 fsigned_zeros, ftrapping_math, fwrapv.
25114 (append_compiler_options): Do not track these options.
25115 (append_linker_options): Likewie
25117 2017-07-06 Jan Hubicka <hubicka@ucw.cz>
25119 * cgraphunit.c (cgraph_node::finalize_function): When
25120 !flag_toplevel_reorde set no_reorder flag.
25121 (varpool_node::finalize_decl): Likewise.
25122 (symbol_table::compile): Drop no toplevel reorder path.
25124 2017-07-06 Jan Hubicka <hubicka@ucw.cz>
25126 * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25127 edges; zero probability is not better than uninitialized.
25129 2017-07-06 Maxim Ostapenko <m.ostapenko@samsung.com>
25131 * asan.h (asan_sanitize_allocas_p): Declare.
25132 * asan.c (asan_sanitize_allocas_p): New function.
25133 (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25134 (handle_builtin_alloca): Likewise.
25135 * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25136 if !asan_sanitize_allocas_p.
25137 * params.def (asan-instrument-allocas): Add new option.
25138 * params.h (ASAN_PROTECT_ALLOCAS): Define.
25139 * opts.c (common_handle_option): Disable allocas sanitization for
25142 2017-07-06 Maxim Ostapenko <m.ostapenko@samsung.com>
25144 * asan.c: Include gimple-fold.h.
25145 (get_last_alloca_addr): New function.
25146 (handle_builtin_stackrestore): Likewise.
25147 (handle_builtin_alloca): Likewise.
25148 (asan_emit_allocas_unpoison): Likewise.
25149 (get_mem_refs_of_builtin_call): Add new parameter, remove const
25150 quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25151 BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25152 (instrument_builtin_call): Pass gimple iterator to
25153 get_mem_refs_of_builtin_call.
25154 (last_alloca_addr): New global.
25155 * asan.h (asan_emit_allocas_unpoison): Declare.
25156 * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25157 (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25158 * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25159 if function calls alloca.
25160 * gimple-fold.c (replace_call_with_value): Remove static keyword.
25161 * gimple-fold.h (replace_call_with_value): Declare.
25162 * internal-fn.c: Include asan.h.
25163 * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25164 BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25166 2017-07-06 David Malcolm <dmalcolm@redhat.com>
25168 * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25169 (C_SELFTEST_FLAGS): New.
25170 (CPP_SELFTEST_FLAGS): New.
25171 (SELFTEST_DEPS): New, from deps of s-selftest.
25172 (C_SELFTEST_DEPS): New, from deps of s-selftest.
25173 (CPP_SELFTEST_DEPS): New.
25174 (selftest): Add dependency on s-selftest-c++.
25175 (s-selftest): Rename to...
25176 (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25177 and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25178 than SELFTEST_FLAGS.
25179 (selftest-gdb): Rename to...
25180 (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25182 (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25183 (selftest-valgrind): Rename to...
25184 (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25186 (selftest-valgrind): Reintroduce as an alias for
25187 selftest-c-valgrind.
25188 (s-selftest-c++): New.
25189 (selftest-c++-gdb): New.
25190 (selftest-c++-valgrind): New.
25192 2017-07-06 Olivier Hainque <hainque@adacore.com>
25194 * gcc.c (process_command): When deciding if undefined variables
25195 should be ignored when processing specs, accept "gcc -v" as well.
25197 2017-07-06 Jan Hubicka <hubicka@ucw.cz>
25199 * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25200 afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25202 2017-07-06 Thomas Preud'homme <thomas.preudhomme@arm.com>
25204 * config/arm/arm-cpus.in (armv8-r): Add new entry.
25205 * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25206 * config/arm/arm-tables.opt: Regenerate.
25207 * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25209 * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25211 2017-07-06 Carl Love <cel@us.ibm.com>
25213 * ChangeLog: Clean up from mid air collision
25215 2017-07-06 Carl Love <cel@us.ibm.com>
25217 * config/rs6000/rs6000-c.c: Add support for built-in functions
25218 vector signed int vec_subc (vector signed int, vector signed int);
25219 vector signed __int128 vec_subc (vector signed __int128,
25220 vector signed __int128);
25221 vector unsigned __int128 vec_subc (vector unsigned __int128,
25222 vector unsigned __int128);
25223 vector signed int vec_sube (vector signed int, vector signed int,
25224 vector signed int);
25225 vector unsigned int vec_sube (vector unsigned int,
25226 vector unsigned int,
25227 vector unsigned int);
25228 vector signed __int128 vec_sube (vector signed __int128,
25229 vector signed __int128,
25230 vector signed__int128);
25231 vector unsigned __int128 vec_sube (vector unsigned __int128,
25232 vector unsigned __int128,
25233 vector unsigned __int128);
25234 vector signed int vec_subec (vector signed int, vector signed int,
25235 vector signed int);
25236 vector unsigned int vec_subec (vector unsigned int,
25237 vector unsigned int,
25238 vector unsigned int);
25239 vector signed __int128 vec_subec (vector signed __int128,
25240 vector signed __int128,
25241 vector signed__int128);
25242 vector unsigned __int128 vec_subec (vector unsigned __int128,
25243 vector unsigned __int128,
25244 vector unsigned __int128);
25245 * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25246 ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25247 * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25248 BU_ALTIVEC_OVERLOAD_X definitions.
25249 * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25250 * doc/extend.texi: Update the built-in documentation file for the new
25251 built-in functions.
25253 2017-07-06 David Malcolm <dmalcolm@redhat.com>
25256 * diagnostic-show-locus.c (layout::layout): Use start and finish
25257 spelling location for the start and finish of each range.
25258 * genmatch.c (linemap_client_expand_location_to_spelling_point):
25259 Add unused aspect param.
25260 * input.c (expand_location_1): Add "aspect" param, and use it
25261 to access the correct part of the location.
25262 (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25264 (expand_location_to_spelling_point): Likewise.
25265 (linemap_client_expand_location_to_spelling_point): Add "aspect"
25266 param, and pass it to expand_location_1.
25268 2017-07-06 Sebastian Peryt <sebastian.peryt@intel.com>
25270 * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25271 _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25272 _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25273 _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25274 _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25275 _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25276 _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25277 _mm_maskz_getmant_ss): New intrinsics.
25278 (__builtin_ia32_getexpss128_mask): Changed to ...
25279 __builtin_ia32_getexpss128_round ... this.
25280 (__builtin_ia32_getexpsd128_mask): Changed to ...
25281 __builtin_ia32_getexpsd128_round ... this.
25282 * config/i386/i386-builtin-types.def
25283 ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25284 (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25285 * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25286 __builtin_ia32_getexpss_mask_round, __builtin_ia32_getmantsd_mask_round,
25287 __builtin_ia32_getmantss_mask_round): New builtins.
25288 * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25289 V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25290 (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25291 CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25292 * config/i386/sse.md
25293 (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25294 avx512f_sgetexp<mode><mask_scalar_name>
25295 <round_saeonly_scalar_name> ... this.
25296 (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25297 %0, %1, %2<round_saeonly_op3>}): Changed to ...
25298 vgetexp<ssescalarmodesuffix>
25299 \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25300 %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25301 (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25302 avx512f_vgetmant<mode><mask_scalar_name>
25303 <round_saeonly_scalar_name> ... this.
25304 (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25305 %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25306 vgetmant<ssescalarmodesuffix>
25307 \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25308 %0<mask_scalar_operand4>, %1, %2
25309 <round_saeonly_scalar_mask_op4>, %3} ... this.
25310 * config/i386/subst.md (mask_scalar_operand4,
25311 round_saeonly_scalar_mask_operand4, round_saeonly_scalar_mask_op4,
25312 round_saeonly_scalar_nimm_predicate): New subst attributes.
25314 2017-07-06 Julia Koval <julia.koval@intel.com>
25316 * config/i386/i386.c (ix86_erase_embedded_rounding):
25317 Remove code for old rounding pattern.
25319 2017-07-06 Richard Earnshaw <rearnsha@arm.com>
25321 * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25323 2017-07-06 Christophe Lyon <christophe.lyon@linaro.org>
25325 * doc/sourcebuild.texi (Test Directives, Variants of
25326 dg-require-support): Add documentation for dg-require-stack-check.
25328 2017-07-05 Sebastian Peryt <sebastian.peryt@intel.com>
25330 * config/i386/subst.md (mask_scalar, round_scalar,
25331 round_saeonly_scalar): New meta-templates.
25332 (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25333 round_scalar_mask_operand3, round_scalar_mask_op3,
25334 round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25335 round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25336 round_saeonly_scalar_constraint,
25337 round_saeonly_scalar_prefix): New subst attribute.
25338 * config/i386/sse.md
25339 (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25340 <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25341 <round_scalar_name> ... this.
25342 (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25343 <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25344 <round_scalar_name> ... this.
25345 (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25346 <sse>_vm<code><mode>3<mask_scalar_name>
25347 <round_saeonly_scalar_name> ... this.
25348 (v<plusminus_mnemonic><ssescalarmodesuffix>
25349 \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25350 %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25351 v<plusminus_mnemonic><ssescalarmodesuffix>
25352 \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25353 %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25354 (v<multdiv_mnemonic><ssescalarmodesuffix>
25355 \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25356 %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25357 v<multdiv_mnemonic><ssescalarmodesuffix>
25358 \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25359 %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25360 (v<maxmin_float><ssescalarmodesuffix>
25361 \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25362 %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25363 v<maxmin_float><ssescalarmodesuffix>
25364 \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25365 %0<mask_scalar_operand3>, %1, %<iptr>2
25366 <round_saeonly_scalar_mask_op3>} ... this.
25368 2017-07-05 Richard Earnshaw <rearnsha@arm.com>
25370 * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25371 (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25373 2017-07-05 Richard Sandiford <richard.sandiford@linaro.org>
25374 Alan Hayward <alan.hayward@arm.com>
25375 David Sherwood <david.sherwood@arm.com>
25377 * combine.c (simplify_if_then_else): Remove "enum" before
25379 * compare-elim.c (can_eliminate_compare): Likewise.
25380 * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25382 (aarch64_lookup_simd_builtin_type): Likewise.
25383 (aarch64_simd_builtin_type): Likewise.
25384 (aarch64_init_simd_builtin_types): Likewise.
25385 (aarch64_simd_expand_args): Likewise.
25386 * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25388 (aarch64_reverse_mask): Likewise.
25389 (aarch64_simd_emit_reg_reg_move): Likewise.
25390 (aarch64_gen_adjusted_ldpstp): Likewise.
25391 (aarch64_ccmp_mode_to_code): Likewise.
25392 (aarch64_operands_ok_for_ldpstp): Likewise.
25393 (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25394 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25396 (aarch64_min_divisions_for_recip_mul): Likewise.
25397 (aarch64_reassociation_width): Likewise.
25398 (aarch64_get_condition_code_1): Likewise.
25399 (aarch64_simd_emit_reg_reg_move): Likewise.
25400 (aarch64_simd_attr_length_rglist): Likewise.
25401 (aarch64_reverse_mask): Likewise.
25402 (aarch64_operands_ok_for_ldpstp): Likewise.
25403 (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25404 (aarch64_gen_adjusted_ldpstp): Likewise.
25405 * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25407 * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25408 * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25409 (arm_lookup_simd_builtin_type): Likewise.
25410 (arm_simd_builtin_type): Likewise.
25411 (arm_init_simd_builtin_types): Likewise.
25412 (arm_expand_builtin_args): Likewise.
25413 * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25414 * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25415 (ft32_setup_incoming_varargs): Likewise.
25416 (ft32_function_arg): Likewise.
25417 (ft32_function_arg_advance): Likewise.
25418 (ft32_pass_by_reference): Likewise.
25419 (ft32_arg_partial_bytes): Likewise.
25420 (ft32_valid_pointer_mode): Likewise.
25421 (ft32_addr_space_pointer_mode): Likewise.
25422 (ft32_addr_space_legitimate_address_p): Likewise.
25423 * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25425 * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25426 (ix86_emit_outlined_ms2sysv_restore): Likewise.
25427 (iamcu_alignment): Likewise.
25428 (canonicalize_vector_int_perm): Likewise.
25429 (ix86_noce_conversion_profitable_p): Likewise.
25430 (ix86_mpx_bound_mode): Likewise.
25431 (ix86_operands_ok_for_move_multiple): Likewise.
25432 * config/microblaze/microblaze-protos.h
25433 (microblaze_expand_conditional_branch_reg): Likewise.
25434 * config/microblaze/microblaze.c
25435 (microblaze_expand_conditional_branch_reg): Likewise.
25436 * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25438 (rs6000_reassociation_width): Likewise.
25439 (rs6000_invalid_binary_op): Likewise.
25440 (fusion_p9_p): Likewise.
25441 (emit_fusion_p9_load): Likewise.
25442 (emit_fusion_p9_store): Likewise.
25443 * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25445 (riscv_hard_regno_mode_ok_p): Likewise.
25446 (riscv_address_insns): Likewise.
25447 (riscv_split_symbol): Likewise.
25448 (riscv_legitimize_move): Likewise.
25449 (riscv_function_value): Likewise.
25450 (riscv_hard_regno_nregs): Likewise.
25451 (riscv_expand_builtin): Likewise.
25452 * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25453 (riscv_build_integer): Likewise.
25454 (riscv_split_integer): Likewise.
25455 (riscv_legitimate_constant_p): Likewise.
25456 (riscv_cannot_force_const_mem): Likewise.
25457 (riscv_regno_mode_ok_for_base_p): Likewise.
25458 (riscv_valid_base_register_p): Likewise.
25459 (riscv_valid_offset_p): Likewise.
25460 (riscv_valid_lo_sum_p): Likewise.
25461 (riscv_classify_address): Likewise.
25462 (riscv_legitimate_address_p): Likewise.
25463 (riscv_address_insns): Likewise.
25464 (riscv_load_store_insns): Likewise.
25465 (riscv_force_binary): Likewise.
25466 (riscv_split_symbol): Likewise.
25467 (riscv_force_address): Likewise.
25468 (riscv_legitimize_address): Likewise.
25469 (riscv_move_integer): Likewise.
25470 (riscv_legitimize_const_move): Likewise.
25471 (riscv_legitimize_move): Likewise.
25472 (riscv_address_cost): Likewise.
25473 (riscv_subword): Likewise.
25474 (riscv_output_move): Likewise.
25475 (riscv_canonicalize_int_order_test): Likewise.
25476 (riscv_emit_int_order_test): Likewise.
25477 (riscv_function_arg_boundary): Likewise.
25478 (riscv_pass_mode_in_fpr_p): Likewise.
25479 (riscv_pass_fpr_single): Likewise.
25480 (riscv_pass_fpr_pair): Likewise.
25481 (riscv_get_arg_info): Likewise.
25482 (riscv_function_arg): Likewise.
25483 (riscv_function_arg_advance): Likewise.
25484 (riscv_arg_partial_bytes): Likewise.
25485 (riscv_function_value): Likewise.
25486 (riscv_pass_by_reference): Likewise.
25487 (riscv_setup_incoming_varargs): Likewise.
25488 (riscv_print_operand): Likewise.
25489 (riscv_elf_select_rtx_section): Likewise.
25490 (riscv_save_restore_reg): Likewise.
25491 (riscv_for_each_saved_reg): Likewise.
25492 (riscv_register_move_cost): Likewise.
25493 (riscv_hard_regno_mode_ok_p): Likewise.
25494 (riscv_hard_regno_nregs): Likewise.
25495 (riscv_class_max_nregs): Likewise.
25496 (riscv_memory_move_cost): Likewise.
25497 * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25498 * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25499 (rl78_addr_space_address_mode): Likewise.
25500 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25502 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25503 (rs6000_reassociation_width): Likewise.
25504 (rs6000_invalid_binary_op): Likewise.
25505 (fusion_p9_p): Likewise.
25506 (emit_fusion_p9_load): Likewise.
25507 (emit_fusion_p9_store): Likewise.
25508 * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25509 (ok_for_simple_move_operands): Likewise.
25510 (ok_for_simple_move_strict_operands): Likewise.
25511 (ok_for_simple_arith_logic_operands): Likewise.
25512 (visium_legitimize_reload_address): Likewise.
25513 (visium_select_cc_mode): Likewise.
25514 (output_cbranch): Likewise.
25515 (visium_split_double_move): Likewise.
25516 (visium_expand_copysign): Likewise.
25517 (visium_expand_int_cstore): Likewise.
25518 (visium_expand_fp_cstore): Likewise.
25519 * config/visium/visium.c (visium_pass_by_reference): Likewise.
25520 (visium_function_arg): Likewise.
25521 (visium_function_arg_advance): Likewise.
25522 (visium_libcall_value): Likewise.
25523 (visium_setup_incoming_varargs): Likewise.
25524 (visium_legitimate_constant_p): Likewise.
25525 (visium_legitimate_address_p): Likewise.
25526 (visium_legitimize_address): Likewise.
25527 (visium_secondary_reload): Likewise.
25528 (visium_register_move_cost): Likewise.
25529 (visium_memory_move_cost): Likewise.
25530 (prepare_move_operands): Likewise.
25531 (ok_for_simple_move_operands): Likewise.
25532 (ok_for_simple_move_strict_operands): Likewise.
25533 (ok_for_simple_arith_logic_operands): Likewise.
25534 (visium_function_value_1): Likewise.
25535 (rtx_ok_for_offset_p): Likewise.
25536 (visium_legitimize_reload_address): Likewise.
25537 (visium_split_double_move): Likewise.
25538 (visium_expand_copysign): Likewise.
25539 (visium_expand_int_cstore): Likewise.
25540 (visium_expand_fp_cstore): Likewise.
25541 (visium_split_cstore): Likewise.
25542 (visium_select_cc_mode): Likewise.
25543 (visium_split_cbranch): Likewise.
25544 (output_cbranch): Likewise.
25545 (visium_print_operand_address): Likewise.
25546 * expmed.c (flip_storage_order): Likewise.
25547 * expmed.h (emit_cstore): Likewise.
25548 (flip_storage_order): Likewise.
25549 * genrecog.c (validate_pattern): Likewise.
25550 * hsa-gen.c (gen_hsa_addr): Likewise.
25551 * internal-fn.c (expand_arith_overflow): Likewise.
25552 * ira-color.c (allocno_copy_cost_saving): Likewise.
25553 * lra-assigns.c (find_hard_regno_for_1): Likewise.
25554 * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25555 (process_invariant_for_inheritance): Likewise.
25556 * lra-eliminations.c (move_plus_up): Likewise.
25557 * omp-low.c (lower_oacc_reductions): Likewise.
25558 * simplify-rtx.c (simplify_subreg): Likewise.
25559 * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25560 (TARGET_CHKP_BOUND_MODE): Likewise..
25561 * targhooks.c (default_chkp_bound_mode): Likewise.
25562 (default_setup_incoming_vararg_bounds): Likewise.
25563 * targhooks.h (default_chkp_bound_mode): Likewise.
25564 (default_setup_incoming_vararg_bounds): Likewise.
25565 * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25566 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25567 (have_whole_vector_shift): Likewise.
25568 * tree-vect-stmts.c (vectorizable_load): Likewise.
25569 * doc/tm.texi: Regenerate.
25571 2017-07-05 Georg-Johann Lay <avr@gjlay.de>
25573 Graceful degrade if Binutils PR21472 is not available.
25576 * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25577 .rodata in flash test fails.
25578 (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25579 * confgure: Regenerate.
25580 * config.in: Regenerate.
25581 * config/avr/avr.c (avr_asm_named_section)
25582 [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25583 __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25584 (avr_asm_init_sections): Same.
25586 2017-07-05 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
25588 * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25589 (fma<VH:mode>4_intrinsic): Likewise.
25590 (*fmsub<VCVTF:mode>4): Likewise.
25591 (*fmsub<VH:mode>4_intrinsic): Likewise.
25593 2017-07-05 Georg-Johann Lay <avr@gjlay.de>
25596 * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25597 Don't depend on "optimize > 0".
25598 (out_movhi_r_mr, out_movqi_mr_r): Same.
25599 (out_movhi_mr_r, out_movqi_r_mr): Same.
25600 (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25601 io_address_operand on "optimize > 0".
25603 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25605 * tree-loop-distribution.c: Add general explanantion on the pass.
25606 (generate_loops_for_partition): Mark distributed loop.
25607 (pg_add_dependence_edges): New parameter. Handle alias data
25608 dependence specially and record it in the parameter if asked.
25609 (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25610 (init_partition_graph_vertices, add_partition_graph_edge): New.
25611 (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25612 (free_partition_graph_vdata, build_partition_graph): New.
25613 (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25614 (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25615 (data_ref_segment_size, latch_dominated_by_data_ref): New.
25616 (compute_alias_check_pairs, version_loop_by_alias_check): New.
25617 (version_for_distribution_p, finalize_partitions): New.
25618 (distribute_loop): Handle alias data dependence specially. Factor
25619 out loop fusion code as functions and call these functions.
25621 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25623 * tree-loop-distribution.c (classify_partition): New parameter and
25624 better handle reduction statement.
25625 (rdg_build_partitions): Revise comment.
25626 (distribute_loop): Compute statements in all partitions and pass it
25627 to classify_partition.
25629 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25631 * tree-loop-distribution.c (enum partition_type): New.
25632 (struct partition): New field type.
25633 (partition_merge_into): Add parameter. Update partition type.
25634 (data_dep_in_cycle_p, update_type_for_merge): New functions.
25635 (build_rdg_partition_for_vertex): Compute partition type.
25636 (rdg_build_partitions): Dump partition type.
25637 (distribute_loop): Update calls to partition_merge_into.
25639 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25641 * tree-loop-distribution.c (struct ddr_hasher): New.
25642 (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25644 (classify_partition): Call get_data_dependence.
25645 (pg_add_dependence_edges): Ditto.
25646 (distribute_loop): Release data dependence hash table.
25648 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25650 * tree-loop-distribution.c (ref_base_address): Delete.
25651 (similar_memory_accesses): Rename ...
25652 (share_memory_accesses): ... to this. Check if partitions access
25653 the same memory reference.
25654 (distribute_loop): Call share_memory_accesses.
25656 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25658 * tree-loop-distribution.c (struct partition): New field recording
25659 its data reference.
25660 (partition_alloc, partition_free): Init and release data refs.
25661 (partition_merge_into): Merge data refs.
25662 (build_rdg_partition_for_vertex): Collect data refs for partition.
25663 (pg_add_dependence_edges): Change parameters from vector to bitmap.
25665 (distribute_loop): Remve data refs from vertice data of partition
25668 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25670 * tree-loop-distribution.c (params.h): Include header file.
25671 (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25672 (datarefs_vec): New global var.
25673 (create_rdg_vertices): Use datarefs_vec directly.
25674 (free_rdg): Don't free data references.
25675 (build_rdg): Update use. Don't free data references.
25676 (distribute_loop): Compute global variable for data references.
25677 Bail out if there are too many data references.
25679 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25681 * tree-loop-distribution.c (loop_nest): New global var.
25682 (build_rdg): Use loop directly, rather than loop nest.
25683 (pg_add_dependence_edges): Remove loop nest parameter. Use global
25685 (distribute_loop): Compute global variable loop nest. Update use.
25687 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25689 * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25690 (partition_merge_into): New parameter. Dump reason for fusion.
25691 (distribute_loop): Update use of partition_merge_into.
25693 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25695 * tree-loop-distribution.c (bb_top_order_index): New.
25696 (bb_top_order_index_size, bb_top_order_cmp): New.
25697 (stmts_from_loop): Use topological order.
25698 (pass_loop_distribution::execute): Compute and release topological
25699 order for basic blocks.
25701 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25703 * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25706 2017-07-05 Bin Cheng <bin.cheng@arm.com>
25708 * cfgloop.h (struct loop): Add comment. New field orig_loop_num.
25709 * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25710 * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25711 * internal-fn.def (LOOP_DIST_ALIAS): New.
25712 * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25713 (fold_loop_internal_call): ... this.
25714 (vect_loop_dist_alias_call): New function.
25715 (set_uid_loop_bbs): Call fold_loop_internal_call.
25716 (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25719 2017-07-04 Uros Bizjak <ubizjak@gmail.com>
25722 * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25723 Require dead FLAGS_REG at the beginning of a peephole.
25725 2017-07-04 Uros Bizjak <ubizjak@gmail.com>
25728 * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25729 arguments in the call to __builtin_ia32_sbb_u32.
25730 (_subborrow_u64): Swap _X and _Y arguments in the call to
25731 __builtin_ia32_sbb_u64.
25733 2017-07-04 Jakub Jelinek <jakub@redhat.com>
25736 * tree-vrp.c (compare_assert_loc): Turn into a function template
25737 with stable template parameter. Only test if a->e is NULL,
25738 !a->e == !b->e has been verified already. Use e == NULL or
25739 e != NULL instead of e or ! e tests. If stable is true, don't use
25740 iterative_hash_expr, on the other side allow a or b or both NULL
25741 and sort the NULLs last.
25742 (process_assert_insertions): Sort using compare_assert_loc<false>
25743 instead of compare_assert_loc, later sort using
25744 compare_assert_loc<true> before calling process_assert_insertions_for
25745 in a loop. Use break instead of continue once seen NULL pointer.
25747 2017-07-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
25749 * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25750 Cortex-R7 and Cortex-R8 processors.
25752 2017-07-04 Jan Hubicka <hubicka@ucw.cz>
25754 * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25755 uninitialized while src is not.
25757 2017-07-04 Richard Earnshaw <rearnsha@arm.com>
25759 * common/config/arm/arm-common.c: Adjust include path for
25761 * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25762 (arm-cpu.h): Create in build directory. Adjust dependency rules.
25763 (arm-cpu-data.h): Likewise.
25764 (arm-cpu-cdata.h): Likewise.
25765 * config/arm/arm-cpu.h: Delete.
25766 * config/arm/arm-cpu-cdata.h: Delete.
25767 * config/arm/arm-cpu-data.h: Delete.
25769 2017-07-04 James Greenhalgh <james.greenhalgh@arm.com>
25771 * config/arm/arm-cpus.in (cortex-a55): New.
25772 (cortex-a75): Likewise.
25773 (cortex-a75.cortex-a55): Likewise.
25774 * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25776 * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25777 * config/arm/arm-cpu-cdata.h: Regenerate.
25778 * config/arm/arm-cpu-data.h: Regenerate.
25779 * config/arm/arm-cpu.h: Regenerate.
25780 * config/arm/arm-tables.opt: Regenerate.
25781 * config/arm/arm-tune.md: Regenerate.
25783 2017-07-04 Jan Hubicka <hubicka@ucw.cz>
25785 * haifa-sched.c (sched_create_recovery_edges): Update profile.
25787 2017-07-04 Jan Hubicka <hubicka@ucw.cz>
25789 * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25792 2017-07-04 Richard Sandiford <richard.sandiford@linaro.org>
25794 PR tree-optimization/81292
25795 * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25796 full_string_p, also call adjust_related_strinfos if the adjustment
25797 is simple, otherwise invalidate related strinfos.
25799 2017-07-04 Martin Liska <mliska@suse.cz>
25802 * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25803 newly created variable as DECL_IGNORED_P.
25805 2017-07-04 Martin Liska <mliska@suse.cz>
25808 * ipa-inline.c (inline_small_functions):
25809 Use xstrdup_for_dump.
25811 2017-07-04 Tom de Vries <tom@codesourcery.com>
25813 * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25815 2017-07-03 Dominique d'Humieres <dominiq@lps.ens.fr>
25818 * config/darwin.c (darwin_function_switched_text_sections):
25821 2017-07-03 Jan Hubicka <hubicka@ucw.cz>
25823 * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25825 2017-07-03 Richard Earnshaw <rearnsha@arm.com>
25827 * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25829 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25831 * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25832 min_profitable_iters, and th as inclusive lower bounds.
25833 Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25834 (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25835 for min_profitable_iters and min_profitable_estimate.
25836 (vect_transform_loop): Treat th as an inclusive lower bound.
25837 * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25839 2017-07-03 Dominique d'Humieres <dominiq@lps.ens.fr>
25842 * config/darwin.c (darwin_function_switched_text_sections):
25843 Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
25844 in two pieces, and suppress the use of buf.
25846 2017-07-03 Nathan Sidwell <nathan@acm.org>
25848 * hash-table.h (hash_table_mod1): Fix indentation.
25850 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
25852 PR middle-end/81290
25853 * predict.c (force_edge_cold): Be more careful about propagation
25855 * profile-count.h (profile_probability::guessed,
25856 profile_probability::fdo, profile_count::guessed, profile_count::fdo):
25858 * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
25860 2017-07-03 James Greenhalgh <james.greenhalgh@arm.com>
25862 * doc/invoke.texi (rcpc architecture extension): Document it.
25864 2017-07-03 Richard Biener <rguenther@suse.de>
25866 PR tree-optimization/60510
25867 * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
25868 the scalar reduction PHI and use it.
25869 (vectorizable_reduction): Properly guard the single_defuse_cycle
25870 path for non-SLP reduction chains where we cannot use it.
25871 Rework reduc_def/index and vector type deduction. Rework
25872 vector operand gathering during reduction op code-gen.
25873 * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
25874 chains dissolve the chain and leave it to non-SLP reduction
25877 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25879 * tree-data-ref.h (dr_alignment): Declare.
25880 * tree-data-ref.c (dr_alignment): New function.
25881 * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
25882 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25884 * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
25886 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25888 * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
25889 and base_misalignment fields.
25890 (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
25891 * tree-data-ref.c: Include builtins.h.
25892 (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
25893 * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
25894 (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
25895 * tree-vect-data-refs.c: Include tree-cfg.h.
25896 (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
25897 fields instead of calculating an alignment here.
25898 (vect_analyze_data_refs): Use dr_analyze_innermost. Dump the new
25899 innermost_loop_behavior fields.
25901 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25903 * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
25905 (DR_STEP_ALIGNMENT): New macro.
25906 * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
25907 * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
25908 (create_data_ref): Print it.
25909 * tree-vect-stmts.c (vectorizable_load): Use the step alignment
25910 to tell whether the step preserves vector (mis)alignment.
25911 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
25912 Move the check for an integer step and generalise to all INTEGER_CST.
25913 (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
25914 Print the outer step alignment.
25916 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25918 * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
25919 with offset_alignment.
25920 (DR_ALIGNED_TO): Delete.
25921 (DR_OFFSET_ALIGNMENT): New macro.
25922 * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
25923 (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
25924 * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
25925 (create_data_ref): Likewise.
25926 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
25927 (vect_analyze_data_refs): Likewise.
25928 * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
25929 creating dummy innermost behavior.
25931 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25933 * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
25934 with a "innermost_loop_behavior *" and refeence tree.
25935 * tree-data-ref.c (dr_analyze_innermost): Likewise.
25936 (create_data_ref): Update call accordingly.
25937 * tree-predcom.c (find_looparound_phi): Likewise.
25939 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
25941 * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
25942 fields with dr_wrt_vec_loop.
25943 (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
25944 (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
25945 (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
25946 (vect_dr_behavior): New function.
25947 (vect_create_addr_base_for_vector_ref): Remove loop parameter.
25948 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
25949 vect_dr_behavior. Use a step_preserves_misalignment_p boolean to
25950 track whether the step preserves the misalignment.
25951 (vect_create_addr_base_for_vector_ref): Remove loop parameter.
25952 Use vect_dr_behavior.
25953 (vect_setup_realignment): Update call accordingly.
25954 (vect_create_data_ref_ptr): Likewise. Use vect_dr_behavior.
25955 * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
25956 call to vect_create_addr_base_for_vector_ref.
25957 (vect_create_cond_for_align_checks): Likewise.
25958 * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
25959 STMT_VINFO_DR_WRT_VEC_LOOP as a block.
25960 (vect_recog_mask_conversion_pattern): Likewise.
25961 * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
25962 (new_stmt_vec_info): Remove redundant zeroing.
25964 2017-07-03 Richard Earnshaw <rearnsha@arm.com>
25966 * common/config/arm/arm-common.c (arm_be8_option): New function.
25967 * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
25968 (ISA_ARMv6): Add isa_bit_be8.
25969 * config/arm/arm.h (arm_be8_option): Add prototype.
25970 (BE8_SPEC_FUNCTION): New define.
25971 (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
25972 * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
25973 (mlittle-endian): Similarly.
25974 (mbe8, mbe32): New options.
25975 * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
25976 * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
25978 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
25980 * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
25982 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
25984 * tree-cfgcleanup.c (want_merge_blocks_p): New function.
25985 (cleanup_tree_cfg_bb): Use it.
25986 * profile-count.h (profile_count::of_for_merging, profile_count::merge):
25988 * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
25990 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
25993 * loop-doloop.c (add_test): Update profile.
25995 2017-07-03 Martin Liska <mliska@suse.cz>
25998 * sanopt.c (rewrite_usage_of_param): New function.
25999 (sanitize_rewrite_addressable_params): Likewise.
26000 (pass_sanopt::execute): Call rewrite_usage_of_param.
26002 2017-07-03 Richard Biener <rguenther@suse.de>
26004 * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26005 back to using VIEW_CONVERT_EXPR.
26007 2017-07-03 Martin Liska <mliska@suse.cz>
26010 * doc/extend.texi: Document when a resolver function is
26011 generated for target_clones.
26013 2017-07-03 Martin Liska <mliska@suse.cz>
26015 * asan.c (asan_emit_stack_protection): Unpoison just red zones
26016 and shadow memory of auto variables which are subject of
26017 use-after-scope sanitization.
26018 (asan_expand_mark_ifn): Add do set only when is_poison.
26020 2016-07-03 Richard Biener <rguenther@suse.de>
26022 * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26024 (vect_force_simple_reduction): Record reduction def -> phi mapping.
26025 (vectorizable_reduction): Perform reduction PHI creation when
26026 visiting a reduction PHI and adjust and simplify code generation
26027 phase of the reduction op. Cache dts, use fold_binary, not fold_build2.
26028 (vect_transform_loop): Visit reduction PHIs.
26029 * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26030 defs into the SLP tree.
26031 (vect_build_slp_tree): Reduction defs terminate the recursion.
26032 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26034 (vect_get_vec_defs_for_stmt_copy): Export.
26035 (vect_get_vec_defs): Likewise.
26036 * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26038 (vect_get_vec_defs_for_stmt_copy): Declare.
26039 (vect_get_vec_defs): Likewise.
26041 2017-07-03 Richard Sandiford <richard.sandiford@linaro.org>
26043 * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26044 parameter with a "loop" parameter and use it instead of the
26045 loop containing DR_STMT. Don't check simple_iv when doing
26046 BB analysis. Describe the two analysis modes in the comment.
26048 2017-07-03 Tom de Vries <tom@codesourcery.com>
26050 PR tree-optimization/69468
26051 * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26052 (find_same_succ_bb): Handle ignore_edge_flags.
26054 2017-07-03 Tom de Vries <tom@codesourcery.com>
26056 PR tree-optimization/81192
26057 * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26059 (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26061 (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26063 2017-07-03 Tom de Vries <tom@codesourcery.com>
26065 PR tree-optimization/81192
26066 * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26067 BB_SAME_SUCC (bb) == NULL.
26069 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
26071 * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26074 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
26076 * dumpfile.c: Include profile-count.h
26077 * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26079 (insert_cond_bb): Update profile.
26080 * tree-cfg.h (insert_cond_bb): Update prototype.
26081 * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26082 * tree-dump.c: Do not include tree-cfg.
26084 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
26086 * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26088 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
26090 * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26093 2017-07-02 Jan Hubicka <hubicka@ucw.cz>
26095 * tree-complex.c (expand_complex_div_wide): update profile.
26097 2017-07-02 Richard Sandiford <richard.sandiford@linaro.org>
26098 Alan Hayward <alan.hayward@arm.com>
26099 David Sherwood <david.sherwood@arm.com>
26101 * Makefile.in (MACHMODE_H): Remove insn-modes.h
26102 (CORETYPES_H): New define.
26103 (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26104 (insn-modes-inline.h, s-modes-inline-h): New rules.
26105 (generated_files): Add insn-modes-inline.h.
26106 (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26107 (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26108 (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26109 (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26110 (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26111 (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26112 (build/gencodes.o, build/genconditions.o): Likewise.
26113 (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26114 (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26115 (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26116 (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26117 (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26118 (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26119 * coretypes.h: Include everything up to real.h for generators.
26120 Include insn-modes.h first. Include wide-int-print.h after
26121 wide-int.h. Include insn-modes-inline.h and then machmode.h.
26122 * machmode.h: Don't include insn-modes.h here.
26123 * function-tests.c: Remove includes of signop.h, machmode.h,
26124 double-int.h and wide-int.h.
26126 * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26128 * optc-save-gen.awk: Likewise.
26129 * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26130 * godump.c: Remove include of wide-int-print.h.
26131 * pretty-print.h: Likewise.
26132 * wide-int-print.cc: Likewise.
26133 * wide-int.cc: Likewise.
26134 * hash-map-tests.c: Remove include of signop.h.
26135 * hash-set-tests.c: Likewise.
26136 * rtl-tests.c: Likewise.
26137 * mkconfig.sh: Remove include of machmode.h.
26138 * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26140 (emit_insn_modes_inline_h): ...this new function. Emit the code
26141 into an insn-modes-inline.h header file, adding appropriate
26142 include guards and end comments.
26143 (emit_insn_modes_c_header): Remove include of machmode.h.
26144 (emit_min_insn_modes_c_header): Include coretypes.h rather than
26146 (main): Handle -i flag and call emit_insn_modes_inline_h when
26149 2017-07-02 Richard Sandiford <richard.sandiford@linaro.org>
26151 * tree-ssa-strlen.c (strinfo): Rename the length field to
26152 nonzero_chars. Add a full_string_p field.
26153 (compare_nonzero_chars, zero_length_string_p): New functions.
26154 (get_addr_stridx): Add an offset_out parameter.
26155 Use compare_nonzero_chars.
26156 (get_stridx): Update accordingly. Use compare_nonzero_chars.
26157 (new_strinfo): Update after above changes to strinfo.
26158 (set_endptr_and_length): Set full_string_p.
26159 (get_string_length): Update after above changes to strinfo.
26160 (unshare_strinfo): Update call to new_strinfo.
26161 (maybe_invalidate): Likewise.
26162 (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26163 Use compare_nonzero_chars and zero_string_p. Treat nonzero_chars
26164 as a uhwi instead of an shwi. Update after above changes to
26165 strinfo and new_strinfo.
26166 (zero_length_string): Assert that chainsi contains full strings.
26167 Use zero_length_string_p. Update call to new_strinfo.
26168 (adjust_related_strinfos): Update after above changes to strinfo.
26169 Copy full_string_p from origsi.
26170 (adjust_last_stmt): Use zero_length_string_p.
26171 (handle_builtin_strlen): Update after above changes to strinfo and
26172 new_strinfo. Install the lhs as the string length if the previous
26173 entry didn't describe a full string.
26174 (handle_builtin_strchr): Update after above changes to strinfo
26176 (handle_builtin_strcpy): Likewise.
26177 (handle_builtin_strcat): Likewise.
26178 (handle_builtin_malloc): Likewise.
26179 (handle_pointer_plus): Likewise.
26180 (handle_builtin_memcpy): Likewise. Track nonzero characters
26181 that aren't necessarily followed by a nul terminator.
26182 (handle_char_store): Likewise.
26184 2017-07-02 Richard Sandiford <richard.sandiford@linaro.org>
26186 PR tree-optimization/80769
26187 * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26188 for malloc and calloc. Document the new invariant that all related
26189 strinfos have delayed lengths or none do.
26190 (verify_related_strinfos): Move earlier in file.
26191 (set_endptr_and_length): New function, split out from...
26192 (get_string_length): ...here. Also set the lengths of related
26194 (zero_length_string): Assert that chainsi has known (rather than
26196 (adjust_related_strinfos): Likewise.
26198 2017-07-02 Richard Sandiford <richard.sandiford@linaro.org>
26200 PR tree-optimization/81136
26201 * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26202 assert that two references with the same misalignment have the same
26203 compile-time misalignment if those compile-time misalignments
26206 2017-07-01 Andi Kleen <ak@linux.intel.com>
26208 * print-tree.c (print_node): Print all attributes.
26210 2017-07-01 Jan Hubicka <hubicka@ucw.cz>
26212 * cfg.c (scale_bbs_frequencies): New function.
26213 * cfg.h (scale_bbs_frequencies): Declare it.
26214 * cfgloopanal.c (single_likely_exit): Cleanup.
26215 * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26217 (scale_loop_profile): Likewise.
26218 (loop_version): Likewise.
26219 (create_empty_loop_on_edge): Update.
26220 * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26221 scale_loop_frequencies, scale_loop_profile, loopify,
26222 loop_version): Update prototypes.
26223 * modulo-sched.c (sms_schedule): Update.
26224 * predict.c (unlikely_executed_edge_p): Also check probability.
26225 (probably_never_executed_edge_p): Fix typo.
26226 * tree-if-conv.c (version_loop_for_if_conversion): Update.
26227 * tree-parloops.c (gen_parallel_loop): Update.
26228 * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26229 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26230 * tree-ssa-loop-split.c (split_loop): Update.
26231 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26232 * tree-vect-loop-manip.c (vect_do_peeling): Update.
26233 (vect_loop_versioning): Update.
26234 * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26236 2017-07-01 Jan Hubicka <hubicka@ucw.cz>
26238 * trans-mem.c (split_bb_make_tm_edge): Update profile.
26240 2017-07-01 Jan Hubicka <hubicka@ucw.cz>
26242 * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26243 to keep profile consistent.
26245 2017-07-01 Jan Hubicka <hubicka@ucw.cz>
26247 * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26248 * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26249 * profile-count.h (max_safe_multiplier): Make unsigned.
26250 (profile_count::guessed_zero): New.
26252 2017-07-01 Jan Hubicka <hubicka@ucw.cz>
26254 * bb-reorder.c (fix_up_crossing_landing_pad,
26255 fix_crossing_conditional_branches): Use make_single_succ_edge
26256 to keep profile consistent.
26258 2017-07-01 Jan Hubicka <hubicka@ucw.cz>
26260 * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26263 2017-07-01 Jakub Jelinek <jakub@redhat.com>
26266 * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26267 the right scopes, make sure cond_jump isn't preserved between multiple
26268 iterations. Search for fallthru edge whenever there are 3+ edges and
26269 use find_fallthru_edge for it.
26271 2017-06-29 Jan Hubicka <hubicka@ucw.cz>
26273 Patch by Alexander Monakov <amonakov@ispras.ru>
26274 * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26275 probabilities consistently.
26277 2017-06-29 Jan Hubicka <hubicka@ucw.cz>
26279 * pa.c (pa_expand_compare_and_swap_loop): Update call of
26280 emit_cmp_and_jump_insns.
26282 2017-06-29 Jan Hubicka <hubicka@ucw.cz>
26285 * tree-inline.c (expand_call_inline): Combine profile statuses.
26287 2017-06-30 Andrew Pinski <apinski@cavium.com>
26289 * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26290 fold_stmt returned true.
26292 2017-06-30 Nathan Sidwell <nathan@acm.org>
26294 * ggc.h (empty_string): Delete.
26295 * cfgexpand.c (expand_asm_stmt): Use plain "".
26296 * optabs.c (expand_asm_memory_barrier): Likewise.
26297 * stringpool.c (empty_string): Delete.
26298 (digit_vector, digit_string): Delete.
26299 (ggc_alloc_string): Use plain "", don't optimize single digit
26300 strings. Use ggc_alloc_atomic.
26302 2017-06-30 Richard Earnshaw <rearnsha@arm.com>
26304 * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26305 comparison set and one other set, use the cost of the non-comparison
26308 2017-06-30 Nathan Sidwell <nathan@acm.org>
26310 * ggc.h: Replace all 'static inline' with plain 'inline'. Fix
26313 2017-06-30 Peter Bergner <bergner@vnet.ibm.com>
26315 * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26316 loops. Remove now unneeded calls to gimple_switch_set_label() that
26317 just set removed labels to NULL_TREE.
26319 2017-06-30 Aldy Hernandez <aldyh@redhat.com>
26321 * tree-ssanames.c (set_range_info_raw): Abstract from ...
26322 (set_range_info): ...here. Only call set_range_info_raw if domain
26324 (set_nonzero_bits): Call set_range_info_raw.
26325 * tree-ssanames.h (set_range_info_raw): New.
26327 2017-06-30 Jakub Jelinek <jakub@redhat.com>
26330 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26331 V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26332 of nonimmediate_operand and <store_mask_constraint> instead of m
26333 for the input operand. For V8FI iterator, always split if input
26334 is a MEM. For V16FI and V8SF_256 iterators, don't test if both
26335 operands are MEM if <mask_applied>. For VI4F_256 iterator, use
26336 <store_mask_predicate> instead of register_operand and
26337 <store_mask_constraint> instead of v for the input operand. Make
26338 sure both operands aren't MEMs for if not <mask_applied>.
26340 2017-06-30 Sylvestre Ledru <sylvestre@debian.org>
26342 * lto-wrapper.c (copy_file) Close both file descriptors before
26345 2017-06-30 Martin Liska <mliska@suse.cz>
26348 * multiple_target.c (create_dispatcher_calls): Make ifunc
26349 also for function that don't have calls or are not referenced.
26351 2017-06-30 Richard Biener <rguenther@suse.de>
26353 * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26354 analyze the first scalar stmt. Move vector type computation
26355 for the BB case here from ...
26356 * tree-vect-stmts.c (vect_analyze_stmt): ... here. Guard
26357 live operation processing in the SLP case properly.
26359 2017-06-30 Richard Biener <rguenther@suse.de>
26361 * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26363 2017-06-30 Martin Liska <mliska@suse.cz>
26366 * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26367 before BUILT_IN_UNWIND_RESUME when ASAN is used.
26369 2017-06-30 Yvan Roux <yvan.roux@linaro.org>
26371 * doc/invoke.texi (AArch64): Add missing options and remove redundant
26374 2017-06-30 Richard Biener <rguenther@suse.de>
26376 PR tree-optimization/81249
26377 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26378 condition reduction result to original scalar type.
26380 2017-06-30 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26382 * profile-count.h (enum profile_quality): Fix typos and whitespace
26385 2017-06-30 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26387 * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26388 type for branch probabilities.
26390 2017-06-29 Julian Brown <julian@codesourcery.com>
26391 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
26393 * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26394 * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26395 (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26396 (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26398 2017-06-29 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
26400 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26401 check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26402 * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26403 CC usage from generic code to here.
26404 * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26405 CC usage into the target macros.
26407 2017-06-29 Maya Rashish <coypu@sdf.org>
26409 * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26412 2017-06-29 Jan Hubicka <hubicka@ucw.cz>
26414 * arm/arm-builtins.c: Include profile-count.h
26415 * except.c (sjlj_emit_function_enter): Use
26416 profile_probability::unlikely.
26418 2017-06-29 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26420 * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26421 and tocrel_offset be pointer args rather than implicitly using
26423 (legitimate_constant_pool_address_p, rs6000_emit_move,
26424 const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26425 tocrel_offset and use in toc_relative_expr_p call.
26426 (print_operand, print_operand_address): Use static tocrel_base_oac
26427 and tocrel_offset_oac.
26428 (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26431 2017-06-29 Maya Rashish <coypu@sdf.org>
26433 * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26435 2017-06-29 Eric Botcazou <ebotcazou@adacore.com>
26437 * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26438 objects, take into account only the alignment of 'op0' and 'mode1' if
26441 2017-06-29 Steve Ellcey <sellcey@cavium.com>
26443 * ccmp.c (ccmp_tree_comparison_p): New function.
26444 (ccmp_candidate_p): Update to use above function.
26445 (get_compare_parts): New function.
26446 (expand_ccmp_next): Update to use new functions.
26447 (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26449 (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26450 take mode as argument.
26451 * ccmp.h (expand_ccmp_expr): Add mode as argument.
26452 * expr.c (expand_expr_real_1): Pass mode as argument.
26454 2017-06-29 Segher Boessenkool <segher@kernel.crashing.org>
26456 * combine.c (combine_instructions): Print insns to dump_file, together
26459 2017-06-29 Jan Hubicka <hubicka@ucw.cz>
26461 * asan.c (asan_emit_stack_protection): Update.
26462 (create_cond_insert_point): Update.
26463 * auto-profile.c (afdo_propagate_circuit): Update.
26464 * basic-block.h (struct edge_def): Turn probability to
26465 profile_probability.
26466 (EDGE_FREQUENCY): Update.
26467 * bb-reorder.c (find_traces_1_round): Update.
26468 (better_edge_p): Update.
26469 (sanitize_hot_paths): Update.
26470 * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26471 (make_single_succ_edge): Update.
26472 (check_bb_profile): Update.
26473 (dump_edge_info): Update.
26474 (update_bb_profile_for_threading): Update.
26475 * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26476 probabilitycount to 0.
26477 * cfgbuild.c (compute_outgoing_frequencies): Update.
26478 * cfgcleanup.c (try_forward_edges): Update.
26479 (outgoing_edges_match): Update.
26480 (try_crossjump_to_edge): Update.
26481 * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26482 (expand_gimple_tailcall): Update.
26483 (construct_init_block): Use make_single_succ_edge.
26484 (construct_exit_block): Use make_single_succ_edge.
26485 * cfghooks.c (verify_flow_info): Update.
26486 (redirect_edge_succ_nodup): Update.
26487 (split_edge): Update.
26488 (account_profile_record): Update.
26489 * cfgloopanal.c (single_likely_exit): Update.
26490 * cfgloopmanip.c (scale_loop_profile): Update.
26491 (set_zero_probability): Remove.
26492 (duplicate_loop_to_header_edge): Update.
26493 * cfgloopmanip.h (loop_version): Update prototype.
26494 * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26495 (force_nonfallthru_and_redirect): Update.
26496 (update_br_prob_note): Update.
26497 (rtl_verify_edges): Update.
26498 (purge_dead_edges): Update.
26499 (rtl_lv_add_condition_to_bb): Update.
26500 * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26501 * cgraphunit.c (init_lowered_empty_function): Update.
26502 (cgraph_node::expand_thunk): Update.
26503 * cilk-common.c: Include profile-count.h
26504 * dojump.c (inv): Remove.
26505 (jumpifnot): Update.
26506 (jumpifnot_1): Update.
26507 (do_jump_1): Update.
26509 (do_jump_by_parts_greater_rtx): Update.
26510 (do_compare_rtx_and_jump): Update.
26511 * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26512 do_jump_1. do_compare_rtx_and_jump): Update prototype.
26513 * dwarf2cfi.c: Include profile-count.h
26514 * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26515 (sjlj_emit_dispatch_table): Likewise.
26516 * explow.c: Include profile-count.h
26517 * expmed.c (emit_store_flag_force): Update.
26518 (do_cmp_and_jump): Update.
26519 * expr.c (compare_by_pieces_d::generate): Update.
26520 (compare_by_pieces_d::finish_mode): Update.
26521 (emit_block_move_via_loop): Update.
26522 (store_expr_with_bounds): Update.
26523 (store_constructor): Update.
26524 (expand_expr_real_2): Update.
26525 (expand_expr_real_1): Update.
26526 * expr.h (try_casesi, try_tablejump): Update prototypes.
26527 * gimple-pretty-print.c (dump_probability): Update.
26528 (dump_profile): New.
26529 (dump_gimple_label): Update.
26530 (dump_gimple_bb_header): Update.
26531 * graph.c (draw_cfg_node_succ_edges): Update.
26532 * hsa-gen.c (convert_switch_statements): Update.
26533 * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26534 (find_if_case_1): Update.
26535 (find_if_case_2): Update.
26536 * internal-fn.c (expand_arith_overflow_result_store): Update.
26537 (expand_addsub_overflow): Update.
26538 (expand_neg_overflow): Update.
26539 (expand_mul_overflow): Update.
26540 (expand_vector_ubsan_overflow): Update.
26541 * ipa-cp.c (good_cloning_opportunity_p): Update.
26542 * ipa-split.c (split_function): Use make_single_succ_edge.
26543 * ipa-utils.c (ipa_merge_profiles): Update.
26544 * loop-doloop.c (add_test): Update.
26545 (doloop_modify): Update.
26546 * loop-unroll.c (compare_and_jump_seq): Update.
26547 (unroll_loop_runtime_iterations): Update.
26548 * lra-constraints.c (lra_inheritance): Update.
26549 * lto-streamer-in.c (input_cfg): Update.
26550 * lto-streamer-out.c (output_cfg): Update.
26551 * mcf.c (adjust_cfg_counts): Update.
26552 * modulo-sched.c (sms_schedule): Update.
26553 * omp-expand.c (expand_omp_for_init_counts): Update.
26554 (extract_omp_for_update_vars): Update.
26555 (expand_omp_ordered_sink): Update.
26556 (expand_omp_for_ordered_loops): Update.
26557 (expand_omp_for_generic): Update.
26558 (expand_omp_for_static_nochunk): Update.
26559 (expand_omp_for_static_chunk): Update.
26560 (expand_cilk_for): Update.
26561 (expand_omp_simd): Update.
26562 (expand_omp_taskloop_for_outer): Update.
26563 (expand_omp_taskloop_for_inner): Update.
26564 * omp-simd-clone.c (simd_clone_adjust): Update.
26565 * optabs.c (expand_doubleword_shift): Update.
26566 (expand_abs): Update.
26567 (emit_cmp_and_jump_insn_1): Update.
26568 (expand_compare_and_swap_loop): Update.
26569 * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26570 * predict.c (predictable_edge_p): Update.
26571 (edge_probability_reliable_p): Update.
26572 (set_even_probabilities): Update.
26573 (combine_predictions_for_insn): Update.
26574 (combine_predictions_for_bb): Update.
26575 (propagate_freq): Update.
26576 (estimate_bb_frequencies): Update.
26577 (force_edge_cold): Update.
26578 * profile-count.c (profile_count::dump): Add missing space into dump.
26579 (profile_count::debug): Add newline.
26580 (profile_count::differs_from_p): Explicitly convert to unsigned.
26581 (profile_count::stream_in): Update.
26582 (profile_probability::dump): New member function.
26583 (profile_probability::debug): New member function.
26584 (profile_probability::differs_from_p): New member function.
26585 (profile_probability::differs_lot_from_p): New member function.
26586 (profile_probability::stream_in): New member function.
26587 (profile_probability::stream_out): New member function.
26588 * profile-count.h (profile_count_quality): Rename to ...
26589 (profile_quality): ... this one.
26590 (profile_probability): New.
26591 (profile_count): Update.
26592 * profile.c (compute_branch_probabilities): Update.
26593 * recog.c (peep2_attempt): Update.
26594 * sched-ebb.c (schedule_ebbs): Update.
26595 * sched-rgn.c (find_single_block_region): Update.
26596 (compute_dom_prob_ps): Update.
26597 (schedule_region): Update.
26598 * sel-sched-ir.c (compute_succs_info): Update.
26599 * stmt.c (struct case_node): Update.
26600 (do_jump_if_equal): Update.
26601 (get_outgoing_edge_probs): Update.
26602 (conditional_probability): Update.
26603 (emit_case_dispatch_table): Update.
26604 (expand_case): Update.
26605 (expand_sjlj_dispatch_table): Update.
26606 (emit_case_nodes): Update.
26607 * targhooks.c: Update.
26608 * tracer.c (better_p): Update.
26609 (find_best_successor): Update.
26610 * trans-mem.c (expand_transaction): Update.
26611 * tree-call-cdce.c: Update.
26612 * tree-cfg.c (gimple_split_edge): Upate.
26613 (move_sese_region_to_fn): Upate.
26614 * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26615 * tree-eh.c (lower_resx): Upate.
26616 (cleanup_empty_eh_move_lp): Upate.
26617 * tree-if-conv.c (version_loop_for_if_conversion): Update.
26618 * tree-inline.c (copy_edges_for_bb): Update.
26619 (copy_cfg_body): Update.
26620 * tree-parloops.c (gen_parallel_loop): Update.
26621 * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26622 (gimple_gen_time_profiler): Update.
26623 * tree-ssa-dce.c (remove_dead_stmt): Update.
26624 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26625 * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26626 * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26627 (unloop_loops): Update.
26628 (try_peel_loop): Update.
26629 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26630 * tree-ssa-loop-split.c (connect_loops): Update.
26631 (split_loop): Update.
26632 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26633 (hoist_guard): Update.
26634 * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26635 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26636 (value_replacement): Update.
26637 * tree-ssa-reassoc.c (branch_fixup): Update.
26638 * tree-ssa-tail-merge.c (replace_block_by): Update.
26639 * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26640 (create_edge_and_update_destination_phis): Update.
26641 (compute_path_counts): Update.
26642 (recompute_probabilities): Update.
26643 (update_joiner_offpath_counts): Update.
26644 (freqs_to_counts_path): Update.
26645 (duplicate_thread_path): Update.
26646 * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26647 (struct switch_conv_info): Update.
26648 (gen_inbound_check): Update.
26649 * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26650 (vect_do_peeling): Update.
26651 (vect_loop_versioning): Update.
26652 * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26653 (optimize_mask_stores): Update.
26654 * ubsan.c (ubsan_expand_null_ifn): Update.
26655 * value-prof.c (gimple_divmod_fixed_value): Update.
26656 (gimple_divmod_fixed_value_transform): Update.
26657 (gimple_mod_pow2): Update.
26658 (gimple_mod_pow2_value_transform): Update.
26659 (gimple_mod_subtract): Update.
26660 (gimple_mod_subtract_transform): Update.
26661 (gimple_ic): Update.
26662 (gimple_stringop_fixed_value): Update.
26663 (gimple_stringops_transform): Update.
26664 * value-prof.h: Update.
26666 2017-06-29 Carl Love <cel@us.ibm.com>
26668 * config/rs6000/rs6000-c.c: Add support for built-in functions
26669 vector signed int vec_signed (vector float);
26670 vector signed long long vec_signed (vector double);
26671 vector signed int vec_signed2 (vector double, vector double);
26672 vector signed int vec_signede (vector double);
26673 vector signed int vec_signedo (vector double);
26674 * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26675 instruction generator.
26676 * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26677 UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26678 (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26680 (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26681 vunsignede_v2df): Add define_expands.
26682 * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26683 VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26684 VEC_UNSIGNEDO): Add definitions.
26685 * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26686 UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26687 (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26688 (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26689 vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26690 * config/rs6000/altivec.h (vec_signed, vec_signed2,
26691 vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26692 vec_unsignede, vec_unsignedo): Add builtin defines.
26693 * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26695 * doc/extend.texi: Update the built-in documentation file for the
26696 new built-in functions.
26698 2017-06-29 Richard Biener <rguenther@suse.de>
26700 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26701 reduction chains to LOOP_VINFO_REDUCTIONS.
26702 * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26703 SLP reductions after processing reduction chains.
26705 2017-06-29 Nathan Sidwell <nathan@acm.org>
26707 * builtins.c (fold_builtin_FUNCTION): Use
26708 lang_hooks.decl_printable_name.
26710 2017-06-29 Peter Bergner <bergner@vnet.ibm.com>
26712 PR middle-end/81194
26713 * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26714 with only one label.
26715 * stmt.c (expand_case): Assert NCASES is greater than one.
26717 2017-06-29 Richard Biener <rguenther@suse.de>
26719 * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26721 (group_case_labels): Likewise.
26722 (find_taken_edge): Push sanity checking on val to workers...
26723 (find_taken_edge_cond_expr): ... here
26724 (find_taken_edge_switch_expr): ... and here, handle cases
26725 with just a default label.
26726 * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26727 (group_case_labels): Likewise.
26728 * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26729 group_case_labels does anything cleanup the CFG again.
26731 2017-06-29 Bin Cheng <bin.cheng@arm.com>
26733 PR tree-optimization/81196
26734 * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26735 exit condition comparing two IVs.
26737 2017-06-29 Richard Earnshaw <rearnsha@arm.com>
26739 * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26740 profile to the dummy entry at the end of the list of architectures.
26741 * config/arm/arm-cpu-cdata.h: Regenerated.
26743 2017-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26744 Michael Collison <michael.collison@arm.com>
26747 * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26749 (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26750 (*aarch64_reg_<mode>3_minus_mask): New pattern.
26751 (*aarch64_<optab>_reg_di3_mask2): New pattern.
26752 * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26753 of shift when the shift amount is masked with constant equal to
26754 the size of the mode.
26755 * config/aarch64/predicates.md (subreg_lowpart_operator): New
26758 2017-06-29 Martin Liska <mliska@suse.cz>
26760 * config/i386/i386.opt: Change range from [1,5] to [0,5].
26762 2017-06-29 Yury Gribov <tetra2005@gmail.com>
26765 * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26768 (edge_growth_cache_entry::edge_growth_cache_entry): New
26770 (reset_edge_growth_cache): Update to use constructor.
26772 2017-06-28 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
26774 * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26775 (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26776 (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26778 2017-06-28 Sebastian Peryt <sebastian.peryt@intel.com>
26780 * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26781 (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26783 2017-06-28 Szabolcs Nagy <szabolcs.nagy@arm.com>
26785 * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26786 (*-linux-uclibc*): Add t-uclibc tmake_file.
26787 * config/t-musl: New.
26788 * config/t-uclibc: New.
26790 2017-06-28 Richard Earnshaw <rearnsha@arm.com>
26792 * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26794 (gen_comm_data): Emit architectural setting of arch_prof.
26795 * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26797 (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26798 (armv8-m.base, armv8-m.main): Likewise.
26799 * arm-protos.h (arm_build_target): Add profile field.
26800 (arch_option): Likewise.
26801 * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26803 * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26804 arm_active_target.profile.
26806 2017-06-28 Richard Biener <rguenther@suse.de>
26808 PR middle-end/81227
26809 * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26810 TYPE_OVERFLOW_WRAPS.
26811 * match.pd (negate_expr_p): Likewise.
26812 * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26813 fold_build2, not fold_binary.
26815 2017-06-28 Wilco Dijkstra <wdijkstr@arm.com>
26817 * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26818 Convert memory address to Pmode.
26819 (aarch64_print_operand): Assert MEM operands are always Pmode.
26821 2017-06-28 Wilco Dijkstra <wdijkstr@arm.com>
26824 * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26825 Remove redundant if.
26826 (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26827 * config/arm/aarch-common-protos.h
26828 (aarch_forward_to_shift_is_not_shifted_re): Remove.
26829 * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26831 2017-06-28 Michael Meissner <meissner@linux.vnet.ibm.com>
26834 * multiple_target.c (create_dispatcher_calls): Set the default
26835 clone to be static, not public.
26837 2017-06-28 Richard Biener <rguenther@suse.de>
26839 * tree-vect-loop.c (vectorizable_reduction): Move special
26840 cond reduction IV var creation ...
26841 (vect_create_epilog_for_reduction): ... here. Remove induction_index
26842 parameter. Use STMT_VINFO_VECTYPE.
26843 * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
26846 2017-06-28 Martin Liska <mliska@suse.cz>
26849 * ipa-visibility.c (non_local_p): Handle visibility.
26851 2017-06-28 Martin Liska <mliska@suse.cz>
26854 * common.opt: Add IntegerRange to various options.
26855 * opt-functions.awk (integer_range_info): New function.
26856 * optc-gen.awk: Add integer_range_info to cl_options struct.
26857 * opts-common.c (decode_cmdline_option): Handle
26858 CL_ERR_INT_RANGE_ARG.
26859 (cmdline_handle_error): Likewise.
26860 * opts.c (print_filtered_help): Show valid interval in
26861 when --help is provided.
26862 * opts.h (struct cl_option): Add range_min and range_max fields.
26863 * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
26865 2017-06-28 Marc Glisse <marc.glisse@inria.fr>
26867 * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
26868 (x * C EQ/NE y * C): New transformation.
26870 2017-06-28 Christophe Lyon <christophe.lyon@linaro.org>
26872 * genmultilib (combination_space): Accept '+' in option names.
26874 2017-06-28 Martin Liska <mliska@suse.cz>
26877 * asan.c (instrument_derefs): Bail out inner references
26878 that are hard register variables.
26880 2017-06-28 Jakub Jelinek <jakub@redhat.com>
26883 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
26884 rather than def_builtin_pure for __builtin_ia32_gatherpf*.
26886 2017-06-28 Richard Biener <rguenther@suse.de>
26888 * tree-vectorizer.h (vect_get_vec_defs): Remove.
26889 (vect_get_slp_defs): Adjust.
26890 * tree-vect-loop.c (get_initial_defs_for_reduction): Split
26892 * tree-vect-slp.c (vect_get_constant_vectors): ... here and
26894 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
26895 get_initial_defs_for_reduction instead of vect_get_vec_defs.
26896 (vectorizable_reduction): Adjust.
26897 * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
26899 (vect_get_slp_defs): Likewise.
26900 * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
26901 (vectorizable_bswap): Adjust.
26902 (vectorizable_call): Likewise.
26903 (vectorizable_conversion): Likewise.
26904 (vectorizable_assignment): Likewise.
26905 (vectorizable_shift): Likewise.
26906 (vectorizable_operation): Likewise.
26907 (vectorizable_store): Likewise.
26908 (vectorizable_condition): Likewise.
26909 (vectorizable_comparison): Likewise.
26911 2017-06-28 Michael Collison <michael.collison@arm.com>
26914 * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
26916 (arm_gen_movmemqi): Removed unused variable 'i'.
26917 Convert 'for' loop into 'while' loop.
26918 (arm_expand_prologue): Remove last unnecessary set of insn.
26919 (thumb_pop): Remove unused variable 'pushed_words'.
26920 (thumb_exit): Remove last unnecessary set of regs_to_pop.
26922 2017-06-28 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26924 * config/s390/predicates.md: Use s390_rel_address_ok_p.
26925 * config/s390/s390-protos.h: Add prototype of
26926 s390_rel_address_ok_p.
26927 * config/s390/s390.c (s390_got_symbol): New function.
26928 (s390_rel_address_ok_p): New function.
26929 (legitimize_pic_address): Use s390_rel_address_ok_p.
26930 (s390_load_got): Use s390_got_symbol.
26931 (s390_option_override): Issue error if
26932 -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
26933 * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
26935 * config/s390/s390.opt: New option mpic-data-is-text-relative.
26937 2017-06-27 Andrew Pinski <apinski@cavium.com>
26939 * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
26940 (X * copysign (1.0, X)): New pattern.
26941 (X * copysign (1.0, -X)): New pattern.
26942 (copysign (-1.0, CST)): New pattern.
26944 2017-06-27 Joseph Myers <joseph@codesourcery.com>
26946 * genmultilib (combination_space): Remove variable.
26947 Validate reuse rules against regular expression for any sequence
26948 of multilib options in any order.
26950 2017-06-27 Michael Collison <michael.collison@arm.com>
26952 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
26953 call aarch64_split_simd_combine.
26954 * (aarch64_combine_internal<mode>): Delete pattern.
26955 * config/aarch64/aarch64.c (aarch64_split_simd_combine):
26956 Allow register and subreg operands.
26958 2017-06-27 Jerome Lambourg <lambourg@adacore.com>
26960 * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
26961 specific need, just fallback on defaults.
26962 (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
26964 2017-06-27 Jerome Lambourg <lambourg@adacore.com>
26965 Olivier Hainque <hainque@adacore.com>
26967 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
26969 (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
26970 targets. Pick a default if no particular attempt applied.
26971 (STACK_CHECK_PROTECT): Double for 64bit targets, which have
26974 2017-06-27 Jerome Lambourg <lambourg@adacore.com>
26976 * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
26977 (x86_64-wrs-vxworks7): Likewise.
26979 2017-06-27 Marek Polacek <polacek@redhat.com>
26982 * ubsan.c (instrument_null): Check get_base_address's result for null.
26984 2017-06-27 Marc Glisse <marc.glisse@inria.fr>
26986 * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
26988 2017-06-27 Marc Glisse <marc.glisse@inria.fr>
26990 * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
26991 BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
26992 (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
26993 BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
26994 New function types.
26995 * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
26996 BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
26997 BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
26998 BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
26999 BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27000 BUILT_IN_FEUPDATEENV): New builtins.
27001 * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27002 TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27003 * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27004 fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27006 (builtin_structptr_types): Adjust size.
27007 * tree.c (builtin_structptr_types): Add four entries.
27009 2017-06-27 Jerome Lambourg <lambourg@adacore.com>
27010 Olivier Hainque <hainque@adacore.com>
27012 * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27013 (TLS_SYM): New local macro, forcing reference to __tls__ on
27014 link command lines for VxWorks 7 RTPs, triggering initialization
27016 (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27017 OS features TLS support, true for RTPs on VxWorks 7.
27018 * config/vxworks.c (vxworks_override_options): Setup emutls
27021 2017-06-27 Jakub Jelinek <jakub@redhat.com>
27023 * predict.c (test_prediction_value_range): Use -1U instead of -1
27024 to avoid narrowing conversion warning.
27025 * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27026 to avoid narrowing conversion warning.
27027 * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27029 * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27031 2017-06-27 Jerome Lambourg <lambourg@adacore.com>
27033 * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27034 64bit configurations.
27035 (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27036 (SIZE_TYPE): Likewise.
27037 * config/vxworks.c (vxworks_emutls_var_fields): Use
27038 long_unsigned_type_node instead of unsigned_type_node as the offset
27039 field type, which is "pointer" mode in emutls.c.
27041 2017-06-27 Jakub Jelinek <jakub@redhat.com>
27044 * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27046 PR middle-end/81207
27047 * gimple-fold.c (replace_call_with_call_and_fold): Handle
27048 gimple_vuse copying separately from gimple_vdef copying.
27050 2017-06-27 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27052 * value-prof.c (free_hist): Remove call to memset and the enclosing if
27055 2017-06-26 Jerome Lambourg <lambourg@adacore.com>
27056 Olivier Hainque <hainque@adacore.com>
27058 * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27059 for all vxworks7 targets.
27060 * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27061 (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27062 (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27063 variations for VX6/VX7 and 32/64bits later on in ...
27064 (VXWORKS_LIB_SPEC): Leverage new macros.
27065 (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27066 as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27068 2017-06-26 Jerome Lambourg <lambourg@adacore.com>
27070 * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27071 _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27073 2017-06-26 Carl Love <cel@us.ibm.com>
27075 * config/rs6000/rs6000-c.c: Add support for built-in functions
27076 vector bool char vec_reve (vector bool char);
27077 vector signed char vec_reve (vector signed char);
27078 vector unsigned char vec_reve (vector unsigned char);
27079 vector bool int vec_reve (vector bool int);
27080 vector signed int vec_reve (vector signed int);
27081 vector unsigned int vec_reve (vector unsigned int);
27082 vector bool long long vec_reve (vector bool long long);
27083 vector signed long long vec_reve (vector signed long long);
27084 vector unsigned long long vec_reve (vector unsigned long long);
27085 vector bool short vec_reve (vector bool short);
27086 vector signed short vec_reve (vector signed short);
27087 vector double vec_reve (vector double);
27088 vector float vec_reve (vector float);
27089 * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27090 VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27091 * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27092 (altivec_vreve): New pattern.
27093 * config/rs6000/altivec.h (vec_reve): New define.
27094 * doc/extend.texi (vec_rev): Update the built-in documentation file
27095 for the new built-in functions.
27097 2016-06-26 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27099 PR tree-optimization/71815
27100 * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27102 (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27104 (slsr_process_phi): Likewise.
27105 (replace_uncond_cands_and_profitable_phis): Don't replace a
27106 multiply candidate with a stride of 1 (copy or cast).
27107 (phi_incr_cost): Call uses_consumed_by_stmt rather than
27109 (lowest_cost_path): Likewise.
27110 (total_savings): Likewise.
27112 2017-06-26 Richard Biener <rguenther@suse.de>
27115 * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27116 Use def_builtin_pure for all gather builtins.
27118 2017-06-26 Richard Biener <rguenther@suse.de>
27120 PR tree-optimization/81203
27121 * tree-tailcall.c (find_tail_calls): Do not move stmts into
27122 non-dominating BBs.
27124 2017-06-26 Marek Polacek <polacek@redhat.com>
27127 * doc/invoke.texi: Document -Wmultistatement-macros.
27129 2017-06-26 Christophe Lyon <christophe.lyon@linaro.org>
27131 * doc/sourcebuild.texi (ARM-specific attributes): Document new
27132 arm_neon_ok_no_float_abi effective target.
27134 2017-06-26 Richard Biener <rguenther@suse.de>
27136 PR tree-optimization/80928
27137 * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27138 (copy_bbs): Set BB_DUPLICATED flag early.
27139 (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27141 (execute_on_shrinking_pred): Likewise.
27142 * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27143 BB_DUPLICATED blocks.
27144 * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27145 iterate over all PHIs considering removal of *gsi.
27147 2017-06-23 Jim Wilson <jim.wilson@linaro.org>
27149 * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27152 2017-06-23 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27154 * config/rs6000/rs6000-string.c: (expand_block_clear,
27155 do_load_for_compare, select_block_compare_mode,
27156 compute_current_alignment, expand_block_compare,
27157 expand_strncmp_align_check, expand_strn_compare,
27158 expand_block_move, rs6000_output_load_multiple)
27159 Move functions related to string/block move/compare
27160 to a separate file.
27161 * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27162 * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27163 for this function which is now used in two files.
27164 * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27165 * config.gcc: Add rs6000-string.o to extra_objs for
27166 targets powerpc*-*-* and rs6000*-*-*.
27168 2017-06-23 Michael Meissner <meissner@linux.vnet.ibm.com>
27171 * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27172 32-bit, since indexed is not valid for DImode.
27173 (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27174 3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27175 (define_peephole2 for Altivec d-form load): Add 32-bit support.
27176 (define_peephole2 for Altivec d-form store): Likewise.
27179 * multiple_target.c (create_dispatcher_calls): Only create the
27180 dispatcher call if the function is the default clone of a
27181 versioned function.
27183 2017-06-23 Segher Boessenkool <segher@kernel.crashing.org>
27185 PR middle-end/80902
27186 * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27187 a call, force the call to not be a tail call.
27189 2017-06-23 Jeff Law <law@redhat.com>
27191 * doc/contrib.texi: Add entry for Steven Pemberton's work on
27194 2017-06-23 Will Schmidt <will_schmidt@vnet.ibm.com>
27196 * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27197 update_call_from_tree(). (rs6000_gimple_fold_builtin): Add
27198 handling for early expansion of vector shifts (sl,sr,sra,rl).
27199 (builtin_function_type): Add vector shift right instructions
27200 to the unsigned argument list.
27202 2017-06-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
27204 rtl-optimizatoin/79286
27205 * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27206 * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27207 trap. PIC register plus a const unspec without offset can never trap.
27209 2017-06-23 Marc Glisse <marc.glisse@inria.fr>
27211 * tree.h (builtin_structptr_type): New type.
27212 (builtin_structptr_types): Declare new array.
27213 * tree.c (builtin_structptr_types): New array.
27214 (free_lang_data, build_common_tree_nodes): Use it.
27216 2017-06-23 Jonathan Wakely <jwakely@redhat.com>
27219 * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27222 2017-06-22 Matt Turner <mattst88@gmail.com>
27224 * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27225 Lake models to skylake case. Assume skylake for unknown
27226 models with clflushopt.
27228 2017-06-22 Jeff Law <law@redhat.com>
27230 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27231 frame sizes that do not satisfy aarch64_uimm12_shift.
27233 2017-06-22 Jan Hubicka <hubicka@ucw.cz>
27235 * profile-count.h (apply_probability,
27236 apply_scale, probability_in): Fix checks for zero.
27238 2017-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27240 * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27241 * doc/cppdiropts.texi (-I @var{dir}): Document it.
27243 2016-06-22 Richard Biener <rguenther@suse.de>
27245 * tree-vect-loop.c (vect_model_reduction_cost): Handle
27246 COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27247 REDUC_MAX_EXPR support.
27248 (vectorizable_reduction): Likewise.
27249 (vect_create_epilog_for_reduction): Likewise.
27251 2017-06-22 James Greenhalgh <james.greenhalgh@arm.com>
27253 * match.pd (A / (1 << B) -> A >> B): New.
27254 * generic-match-head.c: Include optabs-tree.h.
27255 * gimple-match-head.c: Likewise.
27256 * optabs-tree.h (target_supports_op_p): New.
27257 * optabs-tree.c (target_supports_op_p): New.
27259 2017-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27261 * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27262 $gcc_cv_ld --help output.
27263 (gcc_cv_ld_demangle): Likewise.
27264 (gcc_cv_ld_eh_frame_hdr): Likewise.
27265 (gcc_cv_ld_pie): Likewise.
27266 (gcc_cv_ld_as_needed): Likewise. Prefer native forms unless $gnu_ld.
27267 (gcc_cv_ld_buildid): Likewise.
27268 (gcc_cv_ld_sysroot): Likewise.
27269 (ld_bndplt_support): Likewise.
27270 (ld_pushpopstate_support): Likewise.
27271 * configure: Regenerate.
27272 * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27274 2017-06-21 Jakub Jelinek <jakub@redhat.com>
27277 * config/i386/sse.md (round<mode>2): Renumber match_dup and
27278 operands indexes to avoid gap between operands and match_dups.
27280 2017-06-21 Andrew Pinski <apinski@cavium.com>
27282 * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27283 Increment Arith_shift and Arith_shift_reg by 1.
27284 * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27286 * config/aarch64/aarch64.c (thunderx_tunings): Enable
27287 AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27288 (aarch64_strip_extend): Add new argument and test for it.
27289 (aarch64_cheap_mult_shift_p): New function.
27290 (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27291 add a cost if it is true.
27292 Update calls to aarch64_strip_extend.
27293 (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27295 2017-06-21 Andrew Pinski <apinski@cavium.com>
27297 * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27299 (thunderxt88): Likewise.
27300 * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27301 (thunderx_prefetch_tune): New variable.
27302 (thunderx2t99_prefetch_tune): Update for the correct values.
27303 (thunderxt88_tunings): New variable.
27304 (thunderx_tunings): Use thunderx_prefetch_tune instead of
27305 generic_prefetch_tune.
27306 (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27308 2017-06-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27310 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27311 SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27312 (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27313 (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27314 (aarch64_atomic_cas<mode>, GPI): Likewise.
27316 2017-06-21 Martin Liska <mliska@suse.cz>
27318 * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27319 statements on cold and hot labels.
27320 * predict.c (tree_estimate_probability_bb): Remove the
27321 prediction from this place.
27323 2017-06-21 Martin Liska <mliska@suse.cz>
27325 PR tree-optimization/79489
27326 * gimplify.c (maybe_add_early_return_predict_stmt): New
27328 (gimplify_return_expr): Call the function.
27329 * predict.c (tree_estimate_probability_bb): Remove handling
27331 * predict.def: Update comment about early return predictor.
27332 * gimple-predict.h (is_gimple_predict): New function.
27333 * predict.def: Change default value of early return to 66.
27334 * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27336 * passes.def: Put pass_strip_predict_hints to the beginning of
27339 2017-06-21 Pierre-Marie de Rodat <derodat@adacore.com>
27341 * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27342 FUNCTION_DECL declarations.
27343 (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27345 (dwaf2out_decl): Likewise.
27346 * godump.c (go_early_global_decl): Skip call to the real debug hook
27347 for FUNCTION_DECL declarations.
27348 * passes.c (rest_of_decl_compilation): Skip call to the
27349 early_global_decl debug hook for FUNCTION_DECL declarations, unless
27350 -fdump-go-spec is passed.
27352 2017-06-21 Marc Glisse <marc.glisse@inria.fr>
27354 * config/i386/i386.c (struct builtin_isa): New field pure_p.
27355 Reorder for compactness.
27356 (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27357 (def_builtin_pure, def_builtin_pure2): New functions.
27358 (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27360 2017-06-21 Marc Glisse <marc.glisse@inria.fr>
27362 * match.pd (nop_convert): New predicate.
27363 ((A +- CST1) +- CST2): Allow some NOP conversions.
27365 2017-06-21 Jakub Jelinek <jakub@redhat.com>
27368 * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27369 with ctors/dtors if GOVD_SHARED is set.
27371 2017-06-21 Wilco Dijkstra <wdijkstr@arm.com>
27373 * config/aarch64/aarch64.md (movti_aarch64):
27374 Emit mov rather than orr.
27375 (movtf_aarch64): Likewise.
27376 * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27377 Emit mov rather than orr.
27379 2017-06-21 Wilco Dijkstra <wdijkstr@arm.com>
27381 * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27382 Swap alternatives, make integer dup more expensive.
27384 2017-06-21 Wilco Dijkstra <wdijkstr@arm.com>
27386 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27387 Return true for non-tls symbols.
27389 2017-06-21 James Greenhalgh <james.greenhalgh@arm.com>
27391 * config/aarch64/aarch64-cores.def (cortex-a55): New.
27392 (cortex-a75): Likewise.
27393 (cortex-a75.cortex-a55): Likewise.
27394 * config/aarch64/aarch64-tune.md: Regenerate.
27395 * doc/invoke.texi (-mtune): Document new values for -mtune.
27397 2017-06-21 Tom de Vries <tom@codesourcery.com>
27399 * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27400 stack_size feature.
27401 (Effective-Target Keywords, Other attributes): Suggest using
27402 dg-add-options stack_size feature to get stack limit in stack_size
27403 effective target documentation.
27405 2017-06-21 Julian Brown <julian@codesourcery.com>
27406 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
27408 * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27409 (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27410 * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27412 * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27413 attribute type list for neon_multiply.
27414 * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27415 attribute type list for neon_multiply.
27416 * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27417 * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27418 attribute type list for neon_multiply.
27419 * config/arm/types.md (crypto_pmull): Add.
27420 * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27421 attribute type list.
27423 2017-06-20 Andreas Tobler <andreast@gcc.gnu.org>
27425 * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27428 2017-06-20 Jakub Jelinek <jakub@redhat.com>
27430 * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27431 to make sure not to dereference a NULL cost_classes_ptr pointer.
27433 2017-06-20 Carl Love <cel@us.ibm.com>
27435 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27436 ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27437 ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27438 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27439 builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27440 * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27441 VMULOSW): New enum "unspec" values.
27442 (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27443 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27444 altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27445 altivec_vmulosw): New patterns.
27446 * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27447 VMULOSW): Add definitions.
27449 2017-06-20 Julia Koval <julia.koval@intel.com>
27451 * config/i386/i386.c: Fix rounding expand for new pattern.
27452 * config/i386/subst.md: Fix pattern (parallel -> unspec).
27454 2017-06-20 James Greenhalgh <james.greenhalgh@arm.com>
27456 * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27457 * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27459 2017-06-20 James Greenhalgh <james.greenhalgh@arm.com>
27461 * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27464 2017-06-20 James Greenhalgh <james.greenhalgh@arm.com>
27466 * config/aarch64/aarch64-cores.def: Rearrange to sort by
27467 architecture, then by implementer ID.
27468 * config/aarch64/aarch64-tune.md: Regenerate.
27470 2017-06-20 Richard Biener <rguenther@suse.de>
27472 PR middle-end/81097
27473 * fold-const.c (split_tree): Fold to type before negating.
27475 2017-06-20 David Malcolm <dmalcolm@redhat.com>
27477 * diagnostic-show-locus.c
27478 (selftest::test_fixit_deletion_affecting_newline): New function.
27479 (selftest::diagnostic_show_locus_c_tests): Call it.
27481 2017-06-20 Andreas Schwab <schwab@suse.de>
27484 * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27487 2017-06-20 Prakhar Bahuguna <prakhar.bahuguna@arm.com>
27489 * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27490 __ARM_FEATURE_COPROC according to support.
27492 2017-06-20 Jakub Jelinek <jakub@redhat.com>
27494 * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27495 Rewritten to avoid overflow for > 32-bit pointers.
27498 * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27499 by removing enum keyword.
27500 (ubsan_type_descriptor): Likewise. Formatting fix.
27503 * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27504 splitter): Require TARGET_SSE2 in the condition.
27506 2017-06-20 Michael Meissner <meissner@linux.vnet.ibm.com>
27509 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27510 for doing vector set of SFmode on ISA 3.0.
27511 * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27512 (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27514 (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27515 SFmode value into a V4SF variable that was extracted from another
27516 V4SF variable without converting the element to double precision
27517 and back to single precision vector format.
27518 (vsx_insert_extract_v4sf_p9_2): Likewise.
27520 2017-06-19 Jakub Jelinek <jakub@redhat.com>
27522 * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27523 in UWHI to avoid undefined overflow.
27526 * ubsan.h (enum ubsan_encode_value_phase): New.
27527 (ubsan_encode_value): Change second argument to
27528 enum ubsan_encode_value_phase with default value of
27529 UBSAN_ENCODE_VALUE_GENERIC.
27530 * ubsan.c (ubsan_encode_value): Change second argument to
27531 enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27532 adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27533 create_tmp_var_raw instead of create_tmp_var and use a
27535 (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27536 instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27537 ubsan_encode_value callers.
27540 * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27541 use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27542 just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27544 2017-06-19 Richard Biener <rguenther@suse.de>
27546 PR middle-end/81118
27547 * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27548 estimates if we changed anything.
27550 2017-06-19 Richard Biener <rguenther@suse.de>
27552 PR tree-optimization/80887
27553 * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27554 (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27555 simplified lookups, then reset mprts_hook.
27556 (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27558 (try_to_simplify): Likewise.
27560 2017-06-19 Martin Liska <mliska@suse.cz>
27563 * gimplify.c (gimplify_switch_expr):
27564 Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27566 2017-06-19 Martin Liska <mliska@suse.cz>
27568 * doc/install.texi: Document that PGO runs in 4 stages.
27570 2017-06-19 Martin Liska <mliska@suse.cz>
27573 * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27574 to dispatcher function name.
27575 * multiple_target.c (replace_function_decl): New function.
27576 (create_dispatcher_calls): Redirect both edges and references.
27578 2017-06-19 Jan Hubicka <hubicka@ucw.cz>
27580 * profile-count.c (profile_count::dump): Dump quality.
27581 (profile_count::differs_from_p): Update for unsigned val.
27582 * profile-count.h (profile_count_quality): New enum.
27583 (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27585 2017-06-19 Richard Biener <rguenther@suse.de>
27587 * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27588 struct function as arg.
27589 (estimate_numbers_of_iterations): Export overload with loop arg.
27590 (free_numbers_of_iterations_estimates_loop): Use an overload of
27591 free_numbers_of_iterations_estimates instead.
27592 * tree-cfg.c (remove_bb): Adjust.
27593 * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27594 * tree-parloops.c (gen_parallel_loop): Likewise.
27595 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27597 (tree_unroll_loops_completely): Likewise.
27598 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27599 Use an overload instead and export.
27600 (estimated_loop_iterations): Adjust.
27601 (max_loop_iterations): Likewise.
27602 (likely_max_loop_iterations): Likewise.
27603 (estimate_numbers_of_iterations): Take struct function as arg
27605 (loop_exits_before_overflow): Adjust.
27606 (free_numbers_of_iterations_estimates_loop): Use an overload.
27607 * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27608 * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27610 2017-06-19 Richard Biener <rguenther@suse.de>
27613 * ipa-prop.c (find_constructor_constant_at_offset): Handle
27614 RANGE_EXPR conservatively.
27616 2017-06-16 Carl Love <cel@us.ibm.com>
27618 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27619 definitions for vec_float, vec_float2, vec_floato,
27620 vec_floate built-ins.
27621 * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27622 for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27624 * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27625 FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27626 UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27627 * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27628 define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27629 * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27630 vec_floato): Add builtin defines.
27631 * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27632 Update the built-in documentation file for the new built-in
27635 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27637 * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27638 (mthumb): Mark as the negative of -marm.
27640 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27642 * doc/invoke.texi (ARM Options, -mcpu): Document supported
27644 (ARM Options, -mtune): Document that this accepts the same
27645 extension options as -mcpu.
27646 (ARM Options, -mfpu): Document addition of -mfpu=auto.
27648 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27650 * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27651 permitted extensions.
27653 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27655 * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27656 (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27657 (armv8-m.main): Add option +nodsp.
27658 * config/arm/arm-cpu-cdata.h: Regenerated.
27660 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27662 * config/arm/t-fuchsia: New file.
27663 * config.gcc (arm*-*-fuchsia*): Use it.
27665 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27667 * config/arm/t-symbian: Rewrite for new option infrastructure.
27669 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27671 * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27672 (MULTILIB_REQUIRED): Likewise.
27674 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27676 * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27677 (MULTILIB_RESUE): Likewise.
27678 (MULTILIB_MATCHES): Likewise.
27679 (MULTLIB_REQUIRED): Likewise.
27681 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27683 * config/arm/t-rtems: Rewrite for new option framework.
27685 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27687 * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27688 (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27689 (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27690 (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27691 * config/arm/t-multilib: ... here.
27692 (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27693 (MULTILIB_MATCHES): Use armv7 libraries for armv7-r. Also use for
27694 armv7-a and armv8*-a when A-profile libraries have not been built.
27695 * config/arm/t-rmprofile: Rewrite.
27697 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27699 * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27700 with a backslash. Remove the backslash after substituting unescaped
27702 * doc/fragments.texi (MULTILIB_REUSE): Document it.
27704 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27706 * config.gcc: (arm*-*-*): When building a-profile libraries, force
27707 the driver to pass through the default setting of -mfloat-abi.
27708 * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27710 * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27711 (all_feat_combs): New rule.
27712 (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories. Rework
27714 * config/arm/t-aprofile: Rewrite.
27716 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27718 * config/arm/arm.h (FPUTYPE_AUTO): Define.
27719 * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27720 fpu is not specified by the user/command-line.
27721 * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27722 * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27723 * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27724 * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27725 * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27726 FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27728 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27730 * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27731 * config/arm/t-arm-elf: Rewritten.
27733 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27735 * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27736 have some floating-point instructions.
27737 (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27738 (TARGET_MAYBE_HARD_FLOAT): New macro.
27739 * config/arm/arm-builtins.c (arm_init_builtins): Use
27740 TARGET_MAYBE_HARD_FLOAT.
27741 * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27743 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27745 * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27746 (configargs.h): Include it.
27747 (arm_print_hint_for_fpu_option): New function.
27748 (arm_parse_fpu_option): New function.
27749 (candidate_extension): New class.
27750 (arm_canon_for_multilib): New function.
27751 * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27752 (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27753 (ARCH_CANONICAL_SPECS): New macro.
27754 (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27756 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27758 * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27759 are set after handling multilib fragments. Set target_cpu_default2
27762 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27764 * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27766 (arm*-*-*): Set target_cpu_default2 to a quoted string.
27767 * config/arm/parsecpu.awk (check_cpu): Validate any extension
27769 (check_arch): Likewise.
27770 * config/arm/arm.c (arm_configure_build_target): Handle
27771 TARGET_CPU_DEFAULT being a string constant. Scan any feature
27772 options in the default.
27774 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27776 * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27777 when an option is an alias of another.
27778 * config/arm/parsecpu.awk (optalias): New parser token.
27779 (gen_comm_data): Mark non-alias options as such. Emit entries
27780 for extension aliases.
27781 * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27782 (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27783 (armv6kz, armv6zk, armv6t2): Likewise.
27784 (armv7): Make vfpv3-d16 an alias.
27785 (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases. Sort in
27787 (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27788 Sort in canonical order.
27789 (armv8-a): Sort in canonical order.
27790 (armv8.1-a, armv8.2-a): Likewise.
27791 (generic-armv7-a): Make neon and neon-vfpv3 aliases. Sort in
27793 (cortex-a9): Sort in canonical order.
27794 * config/arm/arm.c (selftests.h): Include it.
27795 (arm_test_cpu_arch_data): New function.
27796 (arm_run_self_tests): New function.
27797 (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27798 (targetm): Move declaration to the end of the file.
27799 * arm-cpu-cdata.h: Regenerated.
27801 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27803 * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27804 call to target_mode_check describing the type of option passed.
27805 * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27806 (arm_target_thumb_only): Use arm_parse_arch_option_name or
27807 arm_parse_cpu_option_name to match parameters against list of
27809 * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27810 arm_arch_core_flags data structure.
27811 * config/arm/arm-cpu_cdata.h: Regenerated.
27813 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27815 * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27817 (arm_print_hint_for_cpu_option): Likewise.
27818 (arm_print_hint_for_arch_option): Likewise.
27819 (arm_parse_cpu_option_name): Likewise.
27820 (arm_parse_arch_option_name): Likewise.
27821 * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27822 of entries in the all_fpus list.
27823 * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27824 (arm_parse_cpu_option_name): Declare.
27825 (arm_parse_arch_option_name): Declare.
27826 (arm_parse_option_features): Declare.
27827 (arm_intialize_isa): Declare.
27828 * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27830 (gen_comm_data): ... here. Make definitions non-static.
27831 * config/arm/arm-cpu-data.h: Regenerated.
27832 * config/arm/arm-cpu-cdata.h: Regenerated.
27834 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27836 * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27837 (cpu_arch_extension): New structure.
27838 (cpu_arch_option, arch_option, cpu_option): New structures.
27839 * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27840 architecture types.
27841 (gen_data): Generate new format data tables.
27842 * config/arm/arm.c (cpu_tune): New structure.
27843 (cpu_option, processors): Delete.
27844 (arm_print_hint_for_core_or_arch): Delete. Replace with ...
27845 (arm_print_hint_for_cpu_option): ... this and ...
27846 (arm_print_hint_for_arch_option): ... this.
27847 (arm_parse_arch_cpu_name): Delete. Replace with ...
27848 (arm_parse_cpu_option_name): ... this and ...
27849 (arm_parse_arch_option_name): ... this.
27850 (arm_unrecognized_feature): Change type of target parameter to
27852 (arm_parse_arch_cpu_features): Delete. Replace with ...
27853 (arm_parse_option_features): ... this.
27854 (arm_configure_build_target): Rework to use new configuration data
27856 (arm_print_tune_info): Rework for new configuration data tables.
27857 * config/arm/arm-cpu-data.h: Regenerated.
27858 * config/arm/arm-cpu.h: Regenerated.
27860 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27862 * Makefile.in (OBJS): Move sbitmap.o from here ...
27863 (OBJS-libcommon): ... to here.
27865 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27867 * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
27868 (ISA_ALL_CRYPTO): New macro.
27869 (ISA_ALL_SIMD): New macro
27870 (ISA_ALL_FP): New macro.
27871 * config/arm/arm.c (fpu_bitlist): Update initializer.
27872 * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
27874 (arm9e): Add fpu. Add option for nofp
27875 (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
27876 (arm926ej-s, arm1026ej-s): Likewise.
27877 (generic-armv7-a): Add fpu. Add options for simd, vfpv3, vfpv3-d16,
27878 vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
27879 neon-fp16, neon-vfpv4, nofp and nosimd.
27880 (cortex-a5, cortex-a7): Add fpu. Add options for nosimd and nofp.
27881 (cortex-a8): Add fpu. Add option for nofp.
27882 (cortex-a9): Add fpu. Add options for nosimd and nofp.
27883 (cortex-a12, cortex-a15, cortex-a17): Add fpu. Add option for nofp.
27884 (cortex-r4f): Add fpu.
27885 (cortex-r5): Add fpu. Add options for nofp.dp and nofp.
27886 (cortex-r7): Use idiv option from architecture. Add fpu. Add option
27888 (cortex-r8): Likewise.
27889 (cortex-m4): Add fpu. Add option for nofp.
27890 (cortex-a15.cortex-a7): Add fpu. Add option for nofp.
27891 (cortex-a17.cortex-a7): Likewise.
27892 (cortex-a32): Add fpu. Add options for crypto and nofp.
27893 (cortex-a35, cortex-a53): Likewise.
27894 (cortex-a57): Add fpu. Add option for crypto.
27895 (cortex-a72, cortex-a73): Likewise.
27896 (exynos-m1): Likewise.
27897 (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
27898 (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
27899 (cortex-m33): Add fpu. Add option for nofp.
27900 * config/arm/arm-cpu-cdata.h: Regenerated
27901 * config/arm/arm-cpu-data.h: Regenerated.
27903 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27905 * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
27906 (armv5te, armv5tej): Likewise.
27907 (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
27908 (armv7): Add options fp and vfpv3-d16.
27909 (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
27910 vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
27912 (armv7ve): Likewise.
27913 (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
27914 (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
27915 (armv8-a): Add nocrypto option.
27916 (armv8.1-a, armv8.2-a): Likewise.
27917 (armv8-m.main): add options fp, fp.dp and nofp.
27919 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27921 * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
27923 (armv8-a+crc): Delete.
27924 (armv8.1-a): Add options simd, crypto and nofp.
27925 (armv8.2-a): Add options fp16, simd, crypto and nofp.
27926 (armv8.2-a+fp16): Delete.
27927 (armv8-m.main): Add option dsp.
27928 (armv8-m.main+dsp): Delete.
27929 (cortex-a8): Add fpu. Add nofp option.
27930 (cortex-a9): Add fpu. Add nofp and nosimd options.
27931 * config/arm/parsecpu.awk (gen_data): Generate option tables and
27932 link to main cpu and architecture data structures.
27933 (gen_comm_data): Only put isa attributes from the main architecture
27935 (option): New statement for architecture and CPU entries.
27936 * arm.c (struct cpu_option): New structure.
27937 (struct processors): Add entry for options.
27938 (arm_unrecognized_feature): New function.
27939 (arm_parse_arch_cpu_name): Ignore any characters after the first
27941 (arm_parse_arch_cpu_feature): New function.
27942 (arm_configure_build_target): Separate out any CPU and architecture
27943 features and parse separately. Don't error out if -mfpu=auto is
27944 used with only an architecture string.
27945 (arm_print_asm_arch_directives): New function.
27946 (arm_file_start): Call it.
27947 * config/arm/arm-cpu-cdata.h: Regenerated.
27948 * config/arm/arm-cpu-data.h: Likewise.
27949 * config/arm/arm-tables.opt: Likewise.
27951 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27953 * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
27954 assembler when it is not -mfpu=auto.
27956 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27958 * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
27959 (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
27960 (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
27961 (ASM_CPU_SPEC): Rewrite.
27962 (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
27963 (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef. Use
27964 MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS. Remove
27965 reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
27966 * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
27967 copied string is NUL-terminated. Also strip any characters prefixed
27969 (arm_rewrite_selected_arch): New function.
27970 (arm_rewrite_march): New function.
27972 2017-06-16 Richard Earnshaw <rearnsha@arm.com>
27974 * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
27975 (x_arm_cpu_string, x_arm_tune_string): Likewise.
27976 (march, mcpu, mtune): Convert to string-based options.
27977 * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
27978 (arm_parse_arch_cpu_name): New function.
27979 (arm_configure_build_target): Use arm_parse_arch_cpu_name to
27980 identify selected architecture or CPU.
27981 (arm_option_save): New function.
27982 (TARGET_OPTION_SAVE): Redefine.
27983 (arm_option_restore): Restore string options.
27984 (arm_option_print): Print string options.
27986 2017-06-16 Martin Sebor <msebor@redhat.com>
27988 PR tree-optimization/80933
27989 PR tree-optimization/80934
27990 * builtins.c (fold_builtin_3): Do not handle bcmp here.
27991 * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
27992 (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
27993 (gimple_fold_builtin): Call them.
27995 2017-06-16 Jan Hubicka <hubicka@ucw.cz>
27997 * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
27998 as unlikely; update profile.
28000 2017-06-16 Jan Hubicka <hubicka@ucw.cz>
28002 * predict.c (force_edge_cold): Handle declaring edges impossible
28005 2017-06-16 Jan Hubicka <hubicka@ucw.cz>
28007 * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28009 (try_unroll_loop_completely): Fix reporting.
28011 2017-06-16 Jan Hubicka <hubicka@ucw.cz>
28013 * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28015 2017-06-16 James Greenhalgh <james.greenhalgh@arm.com>
28018 * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28019 if given a non-constant argument for an intrinsic which requires a
28022 2017-06-16 Jan Hubicka <hubicka@ucw.cz>
28024 * profile.c (compare_freqs): New function.
28025 (branch_prob): Sort edge list.
28026 (find_spanning_tree): Assume that the list is priority sorted.
28028 2017-06-16 Richard Biener <rguenther@suse.de>
28030 PR tree-optimization/81090
28031 * passes.def (pass_record_bounds): Remove.
28032 * tree-pass.h (make_pass_record_bounds): Likewise.
28033 * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28034 make_pass_record_bounds): Likewise.
28035 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28036 not free niter estimates at the beginning but at the end.
28037 * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28039 2017-06-16 Richard Biener <rguenther@suse.de>
28041 * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28042 initializer to workaround ICE in host GCC 4.8.
28044 2017-06-16 Jan Hubicka <hubicka@ucw.cz>
28046 * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28048 (clone_inlined_nodes): Update.
28050 2017-06-16 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
28052 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28053 prefetch settings, and enable prefetching by default at -O3.
28055 2017-06-16 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
28057 * config/aarch64/aarch64.c (aarch64_override_options_internal):
28058 Set flag_prefetch_loop_arrays according to tuning data.
28060 2017-06-16 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
28062 * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28063 New tune structure.
28064 (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28065 [Unrelated to main purpose of the patch] Place the pointer field last
28066 to enable type checking errors when tune structure are wrongly merged.
28067 * config/aarch64/aarch64.c (generic_prefetch_tune,)
28068 (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28069 (thunderx2t99_prefetch_tune): New tune constants.
28070 (tune_params *_tunings): Update all tunings (no functional change).
28071 (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28072 PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28073 from tunings structures.
28075 2017-06-16 Jakub Jelinek <jakub@redhat.com>
28078 * ubsan.c (instrument_null): Add T argument, use it instead
28079 of computing it based on IS_LHS.
28080 (instrument_object_size): Likewise.
28081 (pass_ubsan::execute): Adjust instrument_null and
28082 instrument_object_size callers to pass gimple_get_lhs or
28083 gimple_assign_rhs1 result to it. Use instrument_null instead of
28084 calling get_base_address and instrument_mem_ref. Handle
28085 aggregate call arguments for object-size sanitization.
28087 2017-06-16 Yury Gribov <tetra2005@gmail.com>
28089 PR tree-optimization/81089
28090 * tree-vrp.c (is_masked_range_test): Validate operands of
28093 2017-06-15 Martin Sebor <msebor@redhat.com>
28096 * dumpfile.c (dump_register): Avoid calling memset to initialize
28097 a class with a default ctor.
28098 * gcc.c (struct compiler): Remove const qualification.
28099 * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28100 * hash-table.h: Ditto.
28101 * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28103 * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28104 * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28106 * params.h (struct param_info): Make struct members non-const.
28107 * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28108 with default initialization.
28109 * vec.h (vec_copy_construct, vec_default_construct): New helper
28111 (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28112 with vec_copy_construct.
28113 (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28114 (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28115 * doc/invoke.texi (-Wclass-memaccess): Document.
28117 2017-06-15 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
28119 * emit-rtl.h (is_leaf): Update comment about local
28120 register allocator.
28122 2017-06-15 Jozef Lawrynowicz <jozef.l@somniumtech.com>
28125 * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28126 for a variable to have a section before checking if the section has a
28128 Set section to.persistent if persistent attribute is set.
28129 Warn if .persistent attribute is used on an automatic variable.
28131 2017-06-15 Eric Botcazou <ebotcazou@adacore.com>
28133 PR rtl-optimization/80474
28134 * reorg.c (update_block): Do not ignore instructions in a delay slot.
28136 2017-06-15 Segher Boessenkool <segher@kernel.crashing.org>
28138 * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28141 2017-06-14 Maciej W. Rozycki <macro@imgtec.com>
28143 * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28144 (casesi): Emit bounds checking as RTL.
28145 (casesi_internal_mips16_<mode>): Remove bounds checking.
28147 2017-06-14 Max Filippov <jcmvbkbc@gmail.com>
28149 * config/xtensa/xtensa.c (xtensa_option_override): Append
28150 MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28151 (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28152 xtensa_doloop_hooks): Define unconditionally.
28153 (xtensa_reorg_loops): Only call reorg_loops in the presence of
28155 * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28156 (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28157 for it in xtensa_option_override.
28158 (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28159 HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28161 2017-06-14 Boris Kolpackov <boris@codesynthesis.com>
28163 * doc/cppopts.texi: Document '-' special value to -MF.
28165 2017-06-14 Wilco Dijkstra <wdijkstr@arm.com>
28167 * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28168 (cortex_a53_fconst): Likewise.
28169 (cortex_a53_fpmul): Likewise.
28170 (cortex_a53_f_load_64): Likewise.
28171 (cortex_a53_f_load_many): Likewise.
28172 (cortex_a53_advsimd_alu): Likewise.
28173 (cortex_a53_advsimd_alu_q): Likewise.
28174 (cortex_a53_advsimd_mul): Likewise.
28175 (cortex_a53_advsimd_mul_q): Likewise.
28176 (fpmac bypass): Add new bypass for fpmac-fpmac case.
28177 Add missing fmul, r2f_cvt and fconst cases.
28179 2017-06-14 Richard Biener <rguenther@suse.de>
28181 PR middle-end/81088
28182 * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28184 (fold_binary_loc): When associating do not treat pre-existing
28185 TREE_OVERFLOW on literal constants as a reason to allow
28186 TREE_OVERFLOW on associated literal constants.
28188 2017-06-14 Eric Botcazou <ebotcazou@adacore.com>
28190 * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28191 (MASK_FEATURES): New macro.
28192 * config/sparc/sparc.c (sparc_option_override): Remove the special
28193 handling of -mfpu and generalize it to all MASK_FEATURES switches.
28195 2017-06-14 Eric Botcazou <ebotcazou@adacore.com>
28197 * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28198 a division of 0 if non-call exceptions are enabled.
28200 2017-06-14 Andrew Pinski <apinski@cavium.com>
28201 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
28204 * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28205 Improve vector initialization code gen for only variable case.
28207 2017-06-14 Eric Botcazou <ebotcazou@adacore.com>
28209 * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28211 2017-06-14 Richard Biener <rguenther@suse.de>
28213 PR tree-optimization/81083
28214 * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28217 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28219 * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28220 (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28221 * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28222 * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete. Adjust former use.
28223 * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28224 * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28226 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28228 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28229 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28231 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28233 * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28235 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28237 * config/rs6000/t-rtems: Don't handle SPE.
28239 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28241 * config/rs6000/t-linux: Don't handle SPE.
28243 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28245 * config/rs6000/eabispe.h: Delete file.
28247 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28249 * config/rs6000/t-spe: Delete file.
28251 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28253 * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28254 (rs6000_legitimate_offset_address_p): Return false for anything in
28255 V2SImode or V2SFmode.
28257 2017-06-13 Segher Boessenkool <segher@kernel.crashing.org>
28259 * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28260 except V2SF and V2SI. Rearrange the vector modes, and add comments.
28261 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28263 (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28264 (rs6000_legitimate_offset_address_p): Ditto.
28265 (rs6000_emit_move): Ditto.
28266 (rs6000_init_builtins): Remove V4HI_type_node.
28268 2017-06-13 Martin Liska <mliska@suse.cz>
28271 * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28272 (gate_asan): Likewise.
28273 * asan.h (asan_no_sanitize_address_p): Remove the function.
28274 (sanitize_flags_p): New function.
28275 * builtins.def: Fix coding style.
28276 * common.opt: Use renamed enum value.
28277 * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28278 * doc/extend.texi: Document no_sanitize attribute.
28279 * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28280 to SANITIZE_UNDEFINED_NONDEFAULT.
28281 * gcc.c (sanitize_spec_function): Use the renamed enum value.
28282 * gimple-fold.c (optimize_atomic_compare_exchange_p):
28283 Use sanitize_flags_p.
28284 * gimplify.c (gimplify_function_tree): Likewise.
28285 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28286 * opts.c (parse_no_sanitize_attribute): New function.
28287 (common_handle_option): Use renamed enum value.
28288 * opts.h (parse_no_sanitize_attribute): Declare.
28289 * tree.c (sanitize_flags_p): New function.
28290 * tree.h: Declared here.
28291 * tsan.c: Use sanitize_flags_p.
28292 * ubsan.c (ubsan_expand_null_ifn): Likewise.
28293 (instrument_mem_ref): Likewise.
28294 (instrument_bool_enum_load): Likewise.
28295 (do_ubsan_in_current_function): Remove the function.
28296 (pass_ubsan::execute): Use sanitize_flags_p.
28297 * ubsan.h: Remove do_ubsan_in_current_function
28298 * tree-cfg.c (print_no_sanitize_attr_value): New function.
28299 (dump_function_to_file): Use it here.
28301 2017-06-13 Martin Jambor <mjambor@suse.cz>
28303 PR tree-optimization/80803
28304 PR tree-optimization/81063
28305 * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28306 (propagate_subaccesses_across_link): Enqueue subtree whenever
28307 necessary instead of relying on the caller.
28309 2017-06-13 Martin Jambor <mjambor@suse.cz>
28311 * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28312 that have a first_link.
28313 (sort_and_splice_var_accesses): Do not check first_link before
28315 (subtree_mark_written_and_enqueue): Likewise.
28316 (propagate_all_subaccesses): Likewise and do not stop at first
28317 parent with a first_link.
28319 2017-06-13 Martin Jambor <mjambor@suse.cz>
28321 * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28324 2017-06-13 Yury Gribov <tetra2005@gmail.com>
28326 * match.pd: New pattern.
28328 2017-06-13 Yury Gribov <tetra2005@gmail.com>
28330 * tree-vrp.c (is_masked_range_test): New function.
28331 (register_edge_assert_for): Determine ranges for
28334 2017-06-13 Yury Gribov <tetra2005@gmail.com>
28336 PR tree-optimization/67328
28337 * fold-const.c (maskable_range_p): New function.
28338 (build_range_check): Generate bittests if possible.
28340 2017-06-13 Martin Liska <mliska@suse.cz>
28342 * gimple-pretty-print.c (dump_probability): Add new argument.
28343 (dump_edge_probability): Dump both probability and count.
28344 (dump_gimple_label): Likewise.
28345 (dump_gimple_bb_header): Likewise.
28347 2017-06-13 Georg-Johann Lay <avr@gjlay.de>
28350 * config/avr/avr-devices.c: Fix indentation.
28351 * config/avr/gen-avr-mmcu-specs.c: Dito.
28353 2017-06-13 Richard Biener <rguenther@suse.de>
28355 * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28356 instead get vector type from stmt_info.
28357 (vectorizable_reduction): Adjust. Remove dead code.
28359 2017-06-13 Richard Biener <rguenther@suse.de>
28361 PR middle-end/81065
28362 * fold-const.c (extract_muldiv_1): Remove bogus distribution
28363 case of C * (x * C2 + C3).
28364 (fold_addr_of_array_ref_difference): Properly fold index difference.
28366 2017-06-12 David S. Miller <davem@davemloft.net>
28369 * config/sparc/sparc.md (return expander): Emit frame blockage if
28370 function uses alloca.
28372 2017-06-12 Richard Sandiford <richard.sandiford@linaro.org>
28374 * combine.c (make_field_assignment): Check len rather than the mode
28375 precision when calling force_to_mode.
28377 2017-06-12 Georg-Johann Lay <avr@gjlay.de>
28379 Support multilibs and devices that see flash in RAM address range.
28382 * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28383 (avr_mcu_t) <flash_pm_offset>: New field.
28384 (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28385 * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28386 (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28387 (AVR_TINY_PM_OFFSET): Remove macro.
28388 * config/avr/avr.opt (-mshort-calls): New option.
28389 * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28390 [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28391 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28392 <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28393 <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28394 instead of avr_arch->have_jmp_call.
28395 <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28396 [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28397 avr_arch->flash_pm_offset to define.
28398 * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28399 new field flash_pm_offset. Add entry for avrxmega3.
28400 (avr_texinfo): Add entry for avrxmega3.
28401 * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28402 attiny212, attiny214,
28403 attiny412, attiny414, attiny416, attiny417,
28404 attiny814, attiny816, attiny817,
28405 attiny1614, attiny1616, attiny1617,
28406 attiny3214, attiny3216, attiny3217.
28407 * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28408 avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28409 (avr_print_operand_address) [AVR_TINY]: Same.
28410 (avr_asm_init_sections) <readonly_data_section>: Only patch
28411 callback if avr_arch->flash_pm_offset = 0.
28412 (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28413 for rodata if avr_arch->flash_pm_offset != 0.
28414 (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28415 * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28416 (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28417 (m_options): Append opt_rcall.
28418 (m_dirnames): Append dir_rcall.
28419 * config/avr/t-multilib: Regenerate.
28421 * configure.ac [target=avr]: Check whether avrxmega3 default
28422 linker description file works as needed.
28423 * configure: Regenerate.
28424 * doc/avr-mmcu.texi: Regenerate.
28425 * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28426 <__AVR_ARCH__>: Document avrxmega3 and 103.
28427 <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28428 <__AVR_SHORT_CALLS__>: Document it.
28429 <__AVR_PM_BASE_ADDRESS__>: Document it.
28430 * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28431 (AVR Variable Attributes) <progmem>: Document this is
28432 not needed for avrxmega3.
28433 (AVR Named Address Spaces) <__flash>: Dito.
28435 2017-06-12 Jan Hubicka <hubicka@ucw.cz>
28437 * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28439 2017-06-12 Doug Rupp <rupp@adacore.com>
28441 * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28442 Append vxworks-stdint.h to the tm_file list.
28443 * config/vxworks-stdint.h: New file.
28445 2017-06-12 Martin Liska <mliska@suse.cz>
28447 PR tree-optimization/81041
28448 * tree-profile.c (gimple_gen_ic_func_profiler):
28449 Create an extra BB in profile-generate
28450 (gimple_gen_time_profiler): Likewise.
28452 2017-06-12 Jakub Jelinek <jakub@redhat.com>
28454 PR tree-optimization/81003
28455 * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28456 (update_range_test): Use it instead of force_gimple_operand_gsi.
28458 2017-06-12 Richard Biener <rguenther@suse.de>
28460 PR tree-optimization/81053
28461 * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28462 with backedge value not defined in loop. Simplify def stmt
28465 2017-06-11 Tom de Vries <tom@codesourcery.com>
28468 * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28470 (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28471 nvptx_cannot_force_const_mem.
28473 2017-06-10 Jan Hubicka <hubicka@ucw.cz>
28475 * opts.c (finish_options): Move test for flag_split_stack after
28476 it has been initialized.
28478 2017-06-11 Jason Merrill <jason@redhat.com>
28480 * tree.h (id_equal): New.
28481 * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28482 omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28483 instead of strcmp of IDENTIFIER_POINTER.
28485 2017-06-10 Jan Hubicka <hubicka@ucw.cz>
28487 * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28488 (mark_all_inlined_calls_cdtor): Fix formating.
28489 (inline_transform): Rescale profile before inlining.
28491 2017-06-10 Jan Hubicka <hubicka@ucw.cz>
28493 * cgraph.h (cgraph_edge::clone): Update prototype.
28494 * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28495 (cgraph_node::create_clone): Update.
28496 (cgraph_node::create_version_clone): Update.
28497 * tree-inline.c (copy_bb): Update.
28498 (expand_call_inline): Update.
28500 2017-06-10 Segher Boessenkool <segher@kernel.crashing.org>
28502 * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28503 factored out from ...
28504 (rs6000_emit_prologue): ... here.
28506 2017-06-10 Segher Boessenkool <segher@kernel.crashing.org>
28508 * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28509 factored out from ...
28510 (rs6000_emit_prologue): ... here.
28512 2017-06-10 Jan Hubicka <hubicka@ucw.cz>
28514 * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28516 (handle_missing_profiles): Fix computation of tp_first_run.
28517 (counts_to_freqs): Do not touch freqs when count is 0.
28519 2017-06-10 Jan Hubicka <hubicka@ucw.cz>
28521 * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28524 2017-06-10 Tom de Vries <tom@codesourcery.com>
28526 * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28527 attributes): Document signal effective target.
28529 2017-06-10 Tom de Vries <tom@codesourcery.com>
28531 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28532 Document effective target stack_size.
28534 2017-06-09 David Malcolm <dmalcolm@redhat.com>
28536 * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28537 to the edit_context if they can be auto-applied.
28539 2017-06-9 Ian Lance Taylor <iant@golang.org>
28541 * opts.c (finish_options): If -fsplit-stack, disable implicit
28542 -forder-blocks-and-partition.
28543 * doc/invoke.texi (Optimize Options): Document that when using
28544 -fsplit-stack -forder-blocks-and-partition is not implicitly
28547 2017-06-09 Jan Hubicka <hubicka@ucw.cz>
28549 * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28550 ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28551 ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28552 * builtins.def (abort, trap, unreachable): Declare cold.
28553 * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28554 * tree-core.h (ECF_COLD): New.
28555 * tree.c (set_call_expr_flags): Handle ECF_COLD.
28556 (build_common_builtin_nodes): Mark unreachable and abort as cold.
28558 2017-06-09 Jan Hubicka <hubicka@ucw.cz>
28560 * predict.c (unlikely_executed_stmt_p): Cleanup.
28562 2017-06-09 Richard Biener <rguenther@suse.de>
28564 * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28565 model if the ref is always written to.
28567 2017-06-09 Tamar Christina <tamar.christina@arm.com>
28569 * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28571 2017-06-09 Tamar Christina <tamar.christina@arm.com>
28573 * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28576 2017-06-09 Tom de Vries <tom@codesourcery.com>
28579 * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28580 "target cannot support label values" when encountering LABEL_REF.
28582 2017-06-09 Martin Liska <mliska@suse.cz>
28584 * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28585 (gimple_gen_ic_func_profiler): Emit direct comparison
28586 of __gcov_indirect_call_callee with NULL.
28587 (gimple_gen_time_profiler): Change probability from
28588 PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28590 2017-06-09 Jan Hubicka <hubicka@ucw.cz>
28592 * profile.c (edge_gcov_counts): Turn to pointer.
28593 (compute_branch_probabilities, compute_branch_probabilities): Update.
28594 (branch_prob): Do not clear edge_gcov_count.
28595 * profile.h (edge_gcov_counts): Turn to pointer.
28596 (edge_gcov_count): Update.
28598 2017-06-09 Jan Hubicka <hubicka@ucw.cz>
28600 * gimple.h (gimple_check_failed): Mark cold.
28602 2017-06-09 Richard Biener <rguenther@suse.de>
28604 PR tree-optimization/66623
28605 * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28606 refactor check_reduction into two parts, properly computing
28607 whether we have to check reduction validity for outer loop
28610 2017-06-09 Richard Biener <rguenther@suse.de>
28612 PR tree-optimization/79483
28613 * graphite-scop-detection.c (order): New global.
28614 (get_order): Compute bb to order mapping that satisfies code
28615 generation constraints.
28616 (cmp_pbbs): New helper.
28617 (build_scops): Start domwalk at entry block, sort generated
28620 2017-06-09 Richard Biener <rguenther@suse.de>
28622 PR middle-end/81007
28623 * ipa-polymorphic-call.c
28624 (ipa_polymorphic_call_context::restrict_to_inner_class):
28625 Skip FIELD_DECLs with error_mark_node type.
28626 * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28629 2017-06-09 Martin Liska <mliska@suse.cz>
28631 * predict.c (struct branch_predictor): New struct.
28632 (test_prediction_value_range): New test.
28633 (predict_c_tests): New function.
28634 * selftest-run-tests.c (selftest::run_tests): Run the function.
28635 * selftest.h: Declare new tests.
28637 2017-06-09 Segher Boessenkool <segher@kernel.crashing.org>
28640 * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28641 gen_add3_insn did not fail.
28642 * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28643 r0, construct that number in a temporary reg and add that reg to r0.
28644 If asked to put the result in r0 as well, fail.
28646 2017-06-08 Will Schmidt <will_schmidt@vnet.ibm.com>
28648 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28649 for early expansion of vec_eqv.
28651 2017-06-08 Jakub Jelinek <jakub@redhat.com>
28653 PR middle-end/81005
28654 * ubsan.c (instrument_null): Avoid pointless code temporary.
28655 (pass_ubsan::execute): Instrument aggregate arguments of calls.
28657 2017-06-08 Uros Bizjak <ubizjak@gmail.com>
28661 2016-12-14 Uros Bizjak <ubizjak@gmail.com>
28664 * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28667 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28669 * predict.c (unlikely_executed_edge_p): Move ahead.
28670 (probably_never_executed_edge_p): Use it.
28672 2017-06-08 Alexander Ivchenko <alexander.ivchenko@intel.com>
28674 PR middle-end/79988
28675 * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28676 gimple_call_builtin_p call.
28678 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28680 * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28681 * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28682 rtl_check_failed_type2, rtl_check_failed_code1,
28683 rtl_check_failed_code2, rtl_check_failed_code_mode,
28684 rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28685 rtvec_check_failed_bounds, rtl_check_failed_flag,
28686 _fatal_insn_not_found, _fatal_insn): Likewise.
28687 * tree.h (tree_contains_struct_check_failed,
28688 tree_check_failed, tree_not_check_failed,
28689 tree_class_check_failed, tree_range_check_failed,
28690 tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28691 tree_vec_elt_check_failed, phi_node_elt_check_failed,
28692 tree_operand_check_failed, omp_clause_check_failed,
28693 omp_clause_operand_check_failed, omp_clause_range_check_failed):
28696 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28698 * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28699 flag_branch_probabilities.
28700 * ipa-inline.c (edge_badness): Likewise.
28701 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28702 * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28703 * predict.c (maybe_hot_frequency_p): Likewise.
28704 (probably_never_executed): Likewise.
28705 * sched-ebb.c (schedule_ebbs): Likewise.
28706 * sched-rgn.c (find_single_block_region): Likewise.
28707 * tracer.c (tail_duplicate): Likewise.
28709 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28711 * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28712 longer requires x_flag_profile_use.
28714 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28716 * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28717 instead of flag_reorder_blocks_and_partition.
28718 * dbxout.c (dbxout_function_end): Likewise.
28719 * dwarf2out.c (gen_subprogram_die): Likewise.
28720 * haifa-sched.c (sched_create_recovery_edges): Likewise.
28721 * hw-doloop.c (reorg_loops): Likewise.
28722 * varasm.c (assemble_start_function,
28723 assemble_end_function): Likewise.
28724 (decide_function_section): Do not check for
28725 flag_reorder_blocks_and_partition.
28727 2017-06-08 Alexander Ivchenko <alexander.ivchenko@intel.com>
28729 * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28731 (chkp_get_hard_register_fake_addr_expr): Ditto.
28732 (chkp_build_addr_expr): Add check for hard reg case.
28733 (chkp_parse_array_and_component_ref): Ditto.
28734 (chkp_find_bounds_1): Ditto.
28735 (chkp_process_stmt): Don't generate bounds store for
28738 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28740 * predict.c (maybe_hot_bb_p): Do not check profile status.
28741 (maybe_hot_edge_p): Likewise.
28742 (probably_never_executed): Check for zero counts even if profile
28744 (unlikely_executed_edge_p): New function.
28745 (unlikely_executed_stmt_p): New function.
28746 (unlikely_executed_bb_p): New function.
28747 (set_even_probabilities): Use unlikely predicates.
28748 (combine_predictions_for_bb): Likewise.
28749 (predict_paths_for_bb): Likewise.
28750 (predict_paths_leading_to_edge): Likewise.
28751 (determine_unlikely_bbs): New function.
28752 (estimate_bb_frequencies): Use it.
28753 (compute_function_frequency): Use zero counts even if profile is
28755 * profile-count.h: Fix typo.
28757 2017-08-08 Julia Koval <julia.koval@intel.com>
28759 * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28760 _mm512_mask_cvtsepi16_storeu_epi8,
28761 _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28762 * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28763 _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28764 _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28765 _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28766 * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28767 (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28768 VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28769 * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28770 __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28771 __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28772 __builtin_ia32_pmovuswb256mem_mask,
28773 __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28774 __builtin_ia32_pmovwb512mem_mask): New builtins.
28776 2017-08-08 Julia Koval <julia.koval@intel.com>
28778 PR target/73350,80862
28779 * config/i386/subst.md (round): Fix round pattern.
28780 * config/i386/i386.c (ix86_erase_embedded_rounding):
28781 Fix erasing rounding for the fixed pattern.
28783 2017-06-08 Jan Hubicka <hubicka@ucw.cz>
28785 * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28787 2017-06-08 Martin Liska <mliska@suse.cz>
28789 PR gcov-profile/80911
28790 * gcov.c (block_info::block_info): New constructor.
28792 2017-06-07 Carl Love <cel@us.ibm.com>
28794 * config/rs6000/rs6000-c: The return type of the following
28795 built-in functions was implemented as int not long long. Fix sign
28796 of return value for the unsigned version of vec_mulo and vec_mule.
28797 vector unsigned long long vec_bperm (vector unsigned long long,
28798 vector unsigned char)
28799 vector signed long long vec_mule (vector signed int,
28801 vector unsigned long long vec_mule (vector unsigned int,
28802 vector unsigned int)
28803 vector signed long long vec_mulo (vector signed int,
28805 vector unsigned long long vec_mulo (vector unsigned int,
28806 vector unsigned int)
28807 * doc/extend.texi: Fix the documentation for the built-in
28810 2017-06-07 Carl Love <cel@us.ibm.com>
28813 * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28816 2017-06-07 Carl Love <cel@us.ibm.com>
28818 * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28819 support, Generate doublehv for signed int/float for BE case only.
28821 2017-06-07 Alexander Monakov <amonakov@ispras.ru>
28823 * doc/invoke.texi (mcx16): Rewrite.
28825 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28827 * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28828 * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28829 *mov<mode>_softfloat, and an anonymous splitter): Use
28830 nonimmediate_operand instead of rs6000_nonimmediate_operand.
28832 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28834 * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28836 * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28837 (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28838 (rs6000_debug_reg_global): Adjust.
28839 (rs6000_init_hard_regno_mode_ok): Adjust.
28840 (rs6000_dbx_register_number): Adjust.
28841 * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28842 (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28843 Remove SPE_ACC and SPEFSCR.
28844 (REG_ALLOC_ORDER): Ditto.
28845 (FRAME_POINTER_REGNUM): Change to 111.
28846 (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
28847 (REG_CLASS_NAMES): Ditto.
28848 (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
28849 (REGISTER_NAMES): Ditto.
28850 (ADDITIONAL_REG_NAMES): Ditto.
28851 (rs6000_reg_names): Ditto.
28852 * config/rs6000/rs6000.md: Renumber some register number
28855 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28857 * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
28859 * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28860 * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
28862 (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
28863 (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28864 Delete the SPE high registers.
28865 (REG_ALLOC_ORDER): Ditto.
28866 (enum reg_class): Remove SPE_HIGH_REGS.
28867 (REG_CLASS_NAMES): Ditto.
28868 (REG_CLASS_CONTENTS): Delete the SPE high registers.
28869 (REGISTER_NAMES): Ditto.
28870 (rs6000_reg_names): Ditto.
28871 * doc/tm.texi.in: Remove SPE as example.
28872 * doc/tm.texi: Regenerate.
28874 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28876 * config/rs6000/8540.md (ppc8540_brinc): Delete.
28877 * config/rs6000/e500mc.md (e500mc_brinc): Delete.
28878 * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
28879 * config/rs6000/rs6000.md (type): Remove "brinc".
28881 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28883 * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
28884 (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
28885 * config/rs6000/linuxspe.h: Delete file.
28886 * config/rs6000/rs6000.md: Don't include spe.md.
28887 * config/rs6000/spe.h: Delete file.
28888 * config/rs6000/spe.md: Delete file.
28889 * config/rs6000/t-rs6000: Remove spe.md.
28891 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28893 * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
28894 (reg_or_none500mem_operand): Delete.
28895 * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
28896 instead of reg_or_none500mem_operand.
28898 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28900 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28901 handling of SPE flags.
28902 * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
28904 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28906 * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
28908 * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
28909 (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
28910 paired_divv2sf3): Similar.
28911 * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
28912 SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
28913 * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
28915 Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
28916 Rename the paired_* instruction patterns.
28917 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
28919 * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
28920 * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
28921 (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
28922 spe_padding_size, and spe_64bit_regs_used. Replace TARGET_SPE and
28923 TARGET_SPE_ABI with 0, simplify. Replace SPE_VECTOR_MODE with
28924 PAIRED_VECTOR_MODE.
28925 (struct machine_function): Delete field spe_insn_chain_scanned_p.
28926 (spe_func_has_64bit_regs_p): Delete.
28927 (spe_expand_predicate_builtin): Delete.
28928 (spe_expand_evsel_builtin): Delete.
28929 (TARGET_DWARF_REGISTER_SPAN): Do not define.
28930 (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
28931 (invalid_e500_subreg): Delete.
28932 (rs6000_legitimize_address): Always force_reg op2 as well, for
28933 paired single memory accesses.
28934 (rs6000_member_type_forces_blk): Delete.
28935 (rs6000_spe_function_arg): Delete.
28936 (rs6000_expand_unop_builtin): Delete SPE handling.
28937 (rs6000_expand_binop_builtin): Ditto.
28938 (spe_expand_stv_builtin): Delete.
28939 (bdesc_2arg_spe): Delete.
28940 (spe_expand_builtin): Delete.
28941 (spe_expand_predicate_builtin): Delete.
28942 (spe_expand_evsel_builtin): Delete.
28943 (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
28944 (spe_init_builtins): Delete.
28945 (spe_func_has_64bit_regs_p): Delete.
28946 (savres_routine_name): Delete "info" parameter. Adjust callers.
28947 (rs6000_emit_stack_reset): Ditto.
28948 (rs6000_dwarf_register_span): Delete.
28949 * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
28950 UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
28951 SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
28953 * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
28955 * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
28956 * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
28957 * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
28958 mulv2sf3, divv2sf3): Delete expanders.
28960 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28962 config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
28964 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28966 * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
28967 * config/rs6000/rs6000.c: Ditto.
28969 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28971 * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
28972 * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
28973 comparison_operator.
28975 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28977 * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
28978 * config/rs6000/rs6000.opt: Ditto.
28979 * config/rs6000/t-rtems: Ditto.
28981 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28983 * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
28984 TARGET_E500_SINGLE by 0, simplify.
28985 * config/rs6000/rs6000.c: Ditto.
28986 (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
28987 (spe_build_register_parallel): Delete.
28988 * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
28989 TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
28990 * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
28991 TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
28992 (E500_CONVERT): Delete.
28993 * config/rs6000/spe.md: Remove many patterns and all define_constants.
28995 2017-06-07 Segher Boessenkool <segher@kernel.crashing.org>
28997 * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
28998 * config/rs6000/dfp.md: Ditto.
28999 (negdd2, *negdd2_fpr): Merge.
29000 (absdd2, *absdd2_fpr): Merge.
29001 (negtd2, *negtd2_fpr): Merge.
29002 (abstd2, *abstd2_fpr): Merge.
29003 * config/rs6000/e500.h: Delete file.
29004 * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29005 TARGET_FPRS by 1 and simplify.
29006 * config/rs6000/rs6000-c.c: Ditto.
29007 * config/rs6000/rs6000.c: Ditto. Also replace TARGET_SF_SPE and
29008 TARGET_DF_SPE by 0.
29009 * config/rs6000/rs6000.h: Ditto. Delete TARGET_SF_SPE and
29011 * config/rs6000/rs6000.md: Ditto.
29012 (floatdidf2, *floatdidf2_fpr): Merge.
29013 (move_from_CR_gt_bit): Delete.
29014 * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29015 (E500_CR_IOR_COMPARE): Delete.
29016 (All patterns that require !TARGET_FPRS): Delete.
29017 * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29019 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29021 * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29023 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29025 * graphds.c (add_edge): Intitialize edge's attached data.
29026 (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29027 pointer parameter. Call pointed function on each edge during
29028 graph traversing. Skip traversing the edge when the function
29030 (graphds_dfs, graphds_scc): Ditto.
29031 (for_each_edge): New parameter. Pass the new parameter to callback
29033 * graphds.h (skip_edge_callback): New function pointer type.
29034 (graphds_dfs, graphds_scc): New function pointer parameter.
29035 (graphds_edge_callback, for_each_edge): New parameter.
29037 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29039 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29040 out code checking if runtime alias check is possible to below ...
29041 Call the new function.
29042 * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29043 * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29045 2017-06-07 Marek Polacek <polacek@redhat.com>
29048 * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29049 TYPE_OVERFLOW_WRAPS check.
29051 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29053 * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29054 if versioning is required.
29055 * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29056 peeling with the check for versioning.
29058 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29060 * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29061 * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29062 Set true to new parameter if new ssa variable is defined.
29063 (vect_gen_vector_loop_niters): Refactor. Set range information
29064 for the new vector loop bound variable.
29065 (vect_do_peeling): Ditto.
29067 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29069 * tree-affine.c (ssa.h): Include header file.
29070 (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29071 has wrapping overflow behavior.
29073 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29075 * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29077 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29079 (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29080 (tree_to_aff_combination): ... here.
29082 2017-06-07 Bin Cheng <bin.cheng@arm.com>
29084 * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29085 reg_pressure model function.
29086 (ivopts_global_cost_for_size): Delete.
29087 (determine_set_costs, iv_ca_recount_cost): Call new model function
29088 ivopts_estimate_reg_pressure.
29090 2017-06-07 Tamar Christina <tamar.christina@arm.com>
29092 * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29093 expensive than udiv. Remove floating point cases from mod.
29095 2017-06-07 Tamar Christina <tamar.christina@arm.com>
29097 * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29098 Increase idiv cost.
29100 2017-06-07 Tamar Christina <tamar.christina@arm.com>
29102 * config/aarch64/aarch64.md
29103 (copysignsf3): Fix mask generation.
29105 2017-06-07 Jakub Jelinek <jakub@redhat.com>
29107 * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29109 (class dump_manager): Add register_dumps method.
29110 * dumpfile.c: Include langhooks.h.
29111 (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29112 (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29113 (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29114 (dump_manager::dump_register): Start with 512 entries instead of 32.
29115 (dump_manager::register_dumps): New method.
29116 * toplev.c (general_init): Instead of invoking register_dumps
29117 langhook, invoke register_dumps method on the dump manager.
29118 * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29121 2017-06-07 Richard Sandiford <richard.sandiford@linaro.org>
29123 * doc/md.texi: Clarify the restrictions on a define_insn condition.
29124 Say that # requires an associated define_split to exist, and that
29125 the define_split must be suitable for use after register allocation.
29127 2017-06-06 Jan Hubicka <hubicka@ucw.cz>
29129 * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29130 (compute_outgoing_frequencies): Also initialize zero counts.
29131 (find_many_sub_basic_blocks): Do not produce uninitialized profile
29132 around loops; preserve more of profile when nothing changes.
29134 2017-06-06 Jim Wilson <jim.wilson@linaro.org>
29136 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29138 * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29139 * config/arm/arm-cpu-cdata.h: Regenerate.
29140 * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29141 * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29142 * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29143 * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29144 * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29146 * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29148 * config/arm/t-rmprofile: Likewise.
29149 * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29151 2017-06-06 David S. Miller <davem@davemloft.net>
29154 * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29155 blockage if function uses alloca.
29157 2017-06-06 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
29159 * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29160 New "uid" fields to hold pretty-print IDs of group and ref.
29161 Memory references are now identified as <group_id>:<ref_id>
29162 instead of using [random] addresses.
29163 (dump_mem_details): Simplify, no functional change.
29164 (dump_mem_ref): Simplify and make output more concise.
29165 Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29166 (find_or_create_group): Initialize group uid.
29167 (record_ref): Initialize ref uid. Improve debug output.
29168 (prune_group_by_reuse, should_issue_prefetch_p,)
29169 (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29170 (mark_nontemporal_store, determine_loop_nest_reuse):
29171 Improve debug output.
29173 2017-06-06 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
29175 * dbgcnt.def (prefetch): New debug counter.
29176 * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29177 (schedule_prefetches): Stop issueing prefetches if debug counter
29180 2017-06-06 Tom de Vries <tom@codesourcery.com>
29182 * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29183 gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29185 2017-06-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29187 * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29188 Use aarch64_reg_or_zero predicate for operand 4.
29189 (aarch64_compare_and_swap<mode> define_insn_and_split):
29190 Use aarch64_reg_or_zero predicate for operand 3. Add 'Z' constraint.
29191 (aarch64_store_exclusive<mode>): Likewise for operand 2.
29193 2017-06-06 Thomas Preud'homme <thomas.preudhomme@arm.com>
29195 * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29196 (arm_compute_save_core_reg_mask): This.
29197 (thumb1_compute_save_reg_mask): Rename into ...
29198 (thumb1_compute_save_core_reg_mask): This.
29199 (arm_compute_save_reg0_reg12_mask): Adapt comment.
29200 (arm_compute_frame_layout): Likewise.
29202 2017-06-06 Richard Biener <rguenther@suse.de>
29204 PR tree-optimization/80974
29205 * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29206 keep or clear leaders SSA info.
29208 2017-06-06 Tom de Vries <tom@codesourcery.com>
29210 * config/nvptx/nvptx.c (split_mode_p): New function.
29211 (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29213 2017-06-06 Tom de Vries <tom@codesourcery.com>
29215 * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29217 2017-06-06 Jan Hubicka <hubicka@ucw.cz>
29220 * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29223 2017-06-06 Jan Hubicka <hubicka@ucw.cz>
29225 * shrink-wrap.c (handle_simple_exit): Update profile.
29226 (try_shrink_wrapping): Upate profile.
29228 2017-06-06 Jan Hubicka <hubicka@ucw.cz>
29230 * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29231 (tree_guess_outgoing_edge_probabilities): New.
29232 * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29233 * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29235 2017-06-06 Jan Hubicka <hubicka@ucw.cz>
29237 * ipa-split.c (split_function): Initialize return bb profile.
29239 2017-06-06 Jan Hubicka <hubicka@ucw.cz>
29241 * profile.c (compute_branch_probabilities): Also initialize
29242 EXIT_BLOCK profile.
29244 2017-06-06 Richard Biener <rguenther@suse.de>
29246 PR tree-optimization/80928
29247 * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29248 (vect_analyze_loop_operations): Properly guard analysis for
29250 (vect_transform_loop): Likewise.
29251 (vect_analyze_loop_2): Also reset SLP type on PHIs.
29252 (vect_model_induction_cost): Do not cost for pure SLP.
29253 (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29254 of induction in inner loop vectorization.
29255 * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29256 (vect_get_and_check_slp_defs): Handle vect_induction_def.
29257 (vect_build_slp_tree): Likewise. Handle PHIs as terminating the
29259 (vect_analyze_slp_cost_1): Cost induction.
29260 (vect_detect_hybrid_slp_stmts): Handle PHIs.
29261 (vect_get_slp_vect_defs): Likewise.
29262 * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29263 (vect_transform_stmt): Handle SLP reductions.
29264 * tree-vectorizer.h (vectorizable_induction): Adjust.
29266 2017-06-05 Michael Meissner <meissner@linux.vnet.ibm.com>
29268 * config/rs6000/rs6000.c (make_resolver_func): Update
29269 init_lowered_empty_function call.
29271 2017-06-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
29273 * doc/invoke.texi: Document the -fprofile-abs-path option.
29274 * common.opt (fprofile-abs-path): New option.
29275 * gcov-io.h (gcov_write_filename): Declare.
29276 * gcov-io.c (gcov_write_filename): New function.
29277 * coverage.c (coverage_begin_function): Use gcov_write_filename.
29278 * profile.c (output_location): Likewise.
29280 2017-06-05 Jan Hubicka <hubicka@ucw.cz>
29282 * shring-wrap.c: Revert accidental commit.
29284 2017-06-05 Volker Reichelt <v.reichelt@netcologne.de>
29286 * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29288 2017-06-05 Jan Hubicka <hubicka@ucw.cz>
29290 * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29292 * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29293 profile in callgraph edge.
29294 * profile-count.h (apply_probability): If THIS is 0, then result is 0
29295 (apply_scale): Likewise.
29296 * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29297 Also scale profile when inlining function with zero profile.
29298 (initialize_cfun): Update exit block profile even when it is zero.
29299 * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29300 when profile is read.
29302 2017-06-05 Michael Meissner <meissner@linux.vnet.ibm.com>
29304 * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29305 (CLONE_*): New constants to define the processors we can generate
29306 code for with the target_clone attribute.
29307 (rs6000_clone_map): New array to identify which clone processors
29308 the current program is running on.
29309 (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29310 target_clone attribute.
29311 (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29312 (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29313 (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29314 (cpu_expand_builtin): Add support for target_clone attribute.
29315 (rs6000_valid_attribute_p): Allow "default" attribute.
29316 (get_decl_name): New debug function to simplify printing the
29317 current function name in debugging statements.
29318 (rs6000_clone_priority): New functions to support the target_clone
29319 attribute, and be able to generate code to switch between ISA 2.05
29320 through ISA 3.0 (power6 through power9).
29321 (rs6000_compare_version_priority): Likewise.
29322 (rs6000_get_function_versions_dispatcher): Likewise.
29323 (make_resolver_func): Likewise.
29324 (add_condition_to_bb): Likewise.
29325 (dispatch_function_versions): Likewise.
29326 (rs6000_generate_version_dispatcher_body): Likewise.
29327 (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29328 (fusion_gpr_load_p): Fix a spacing issue.
29329 * doc/extend.texi (Common Function Attributes): Document that the
29330 PowerPC supports the target_clone attribute.
29332 2017-06-05 Thomas Preud'homme <thomas.preudhomme@arm.com>
29334 * config/arm/arm.h: explain F symbol found in description of ARM
29335 register allocation in its legend.
29337 2017-06-05 Jan Hubicka <hubicka@ucw.cz>
29339 * config/mips/frame-header-opt.c: Include profile-count.h.
29340 * config/riscv/riscv.c: Include profile-count.h
29342 2017-06-05 Jan Hubicka <hubicka@ucw.cz>
29344 * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29346 (sm_set_flag_if_changed): Add bbs field.
29347 (execute_sm_if_changed_flag_set): Pass BBS.
29348 (execute_sm): Update.
29350 2017-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29352 * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29355 2017-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29357 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29358 (peephole2): New peephole2 to emit the above.
29359 * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29361 2017-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29363 * config/aarch64/aarch64.c (define_peephole2 above
29364 *sub_<shift>_<mode>): New peephole.
29366 2017-05-23 Jan Hubicka <hubicka@ucw.cz>
29368 * config/i386/i386.c (make_resolver_func): Update.
29369 * Makefile.in: Add profile-count.h and profile-count.o
29370 * auto-profile.c (afdo_indirect_call): Update to new API.
29371 (afdo_set_bb_count): Update.
29372 (afdo_propagate_edge): Update.
29373 (afdo_propagate_circuit): Update.
29374 (afdo_calculate_branch_prob): Update.
29375 (afdo_annotate_cfg): Update.
29376 * basic-block.h: Include profile-count.h
29377 (struct edge_def): Turn count to profile_count.
29378 (struct basic_block_def): Likewie.
29379 (REG_BR_PROB_BASE): Move to profile-count.h
29380 (RDIV): Move to profile-count.h
29381 * bb-reorder.c (max_entry_count): Turn to profile_count.
29382 (find_traces): Update.
29383 (rotate_loop):Update.
29384 (connect_traces):Update.
29385 (sanitize_hot_paths):Update.
29386 * bt-load.c (migrate_btr_defs): Update.
29387 * cfg.c (RDIV): Remove.
29388 (init_flow): Use alloc_block.
29389 (alloc_block): Uninitialize count.
29390 (unchecked_make_edge): Uninitialize count.
29391 (check_bb_profile): Update.
29392 (dump_edge_info): Update.
29393 (dump_bb_info): Update.
29394 (update_bb_profile_for_threading): Update.
29395 (scale_bbs_frequencies_int): Update.
29396 (scale_bbs_frequencies_gcov_type): Update.
29397 (scale_bbs_frequencies_profile_count): New.
29398 * cfg.h (update_bb_profile_for_threading): Update.
29399 (scale_bbs_frequencies_profile_count): Declare.
29400 * cfgbuild.c (compute_outgoing_frequencies): Update.
29401 (find_many_sub_basic_blocks): Update.
29402 * cfgcleanup.c (try_forward_edges): Update.
29403 (try_crossjump_to_edge): Update.
29404 * cfgexpand.c (expand_gimple_tailcall): Update.
29405 (construct_exit_block): Update.
29406 * cfghooks.c (verify_flow_info): Update.
29407 (dump_bb_for_graph): Update.
29408 (split_edge): Update.
29409 (make_forwarder_block): Update.
29410 (duplicate_block): Update.
29411 (account_profile_record): Update.
29412 * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29413 (get_estimated_loop_iterations): Update.
29414 * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29415 (single_likely_exit): Update.
29416 * cfgloopmanip.c (scale_loop_profile): Update.
29418 (set_zero_probability): Update.
29419 (lv_adjust_loop_entry_edge): Update.
29420 * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29421 (purge_dead_edges): Update.
29422 (rtl_account_profile_record): Update.
29423 * cgraph.c (cgraph_node::create): Uninitialize count.
29424 (symbol_table::create_edge): Uninitialize count.
29425 (cgraph_update_edges_for_call_stmt_node): Update.
29426 (cgraph_edge::dump_edge_flags): Update.
29427 (cgraph_node::dump): Update.
29428 (cgraph_edge::maybe_hot_p): Update.
29429 * cgraph.h: Include profile-count.h
29430 (create_clone), create_edge, create_indirect_edge): Update.
29431 (cgraph_node): Turn count to profile_count.
29432 (cgraph_edge0: Likewise.
29433 (make_speculative, clone): Update.
29434 (create_edge): Update.
29435 (init_lowered_empty_function): Update.
29436 * cgraphclones.c (cgraph_edge::clone): Update.
29437 (duplicate_thunk_for_node): Update.
29438 (cgraph_node::create_clone): Update.
29439 * cgraphunit.c (cgraph_node::analyze): Update.
29440 (cgraph_node::expand_thunk): Update.
29441 * final.c (dump_basic_block_info): Update.
29442 * gimple-streamer-in.c (input_bb): Update.
29443 * gimple-streamer-out.c (output_bb): Update.
29444 * graphite.c (print_global_statistics): Update.
29445 (print_graphite_scop_statistics): Update.
29446 * hsa-brig.c: Include basic-block.h.
29447 * hsa-dump.c: Include basic-block.h.
29448 * hsa-gen.c (T sum_slice): Update.
29449 (convert_switch_statements):Update.
29450 * hsa-regalloc.c: Include basic-block.h.
29451 * ipa-chkp.c (chkp_produce_thunks): Update.
29452 * ipa-cp.c (struct caller_statistics): Update.
29453 (init_caller_stats): Update.
29454 (gather_caller_stats): Update.
29455 (ipcp_cloning_candidate_p): Update.
29456 (good_cloning_opportunity_p): Update.
29457 (get_info_about_necessary_edges): Update.
29458 (dump_profile_updates): Update.
29459 (update_profiling_info): Update.
29460 (update_specialized_profile): Update.
29461 (perhaps_add_new_callers): Update.
29462 (decide_about_value): Update.
29463 (ipa_cp_c_finalize): Update.
29464 * ipa-devirt.c (struct odr_type_warn_count): Update.
29465 (struct decl_warn_count): Update.
29466 (struct final_warning_record): Update.
29467 (possible_polymorphic_call_targets): Update.
29468 (ipa_devirt): Update.
29469 * ipa-fnsummary.c (redirect_to_unreachable): Update.
29470 * ipa-icf.c (sem_function::merge): Update.
29471 * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29472 * ipa-inline.c (compute_uninlined_call_time): Update.
29473 (compute_inlined_call_time): Update.
29474 (want_inline_small_function_p): Update.
29475 (want_inline_self_recursive_call_p): Update.
29476 (edge_badness): Update.
29477 (lookup_recursive_calls): Update.
29478 (recursive_inlining): Update.
29479 (inline_small_functions): Update.
29480 (dump_overall_stats): Update.
29481 (dump_inline_stats): Update.
29482 * ipa-profile.c (ipa_profile_generate_summary): Update.
29483 (ipa_propagate_frequency): Update.
29484 (ipa_profile): Update.
29485 * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29486 * ipa-utils.c (ipa_merge_profiles): Update.
29487 * loop-doloop.c (doloop_modify): Update.
29488 * loop-unroll.c (report_unroll): Update.
29489 (unroll_loop_runtime_iterations): Update.
29490 * lto-cgraph.c (lto_output_edge): Update.
29491 (lto_output_node): Update.
29492 (input_node): Update.
29493 (input_edge): Update.
29494 (merge_profile_summaries): Update.
29495 * lto-streamer-in.c (input_cfg): Update.
29496 * lto-streamer-out.c (output_cfg): Update.
29497 * mcf.c (create_fixup_graph): Update.
29498 (adjust_cfg_counts): Update.
29499 (sum_edge_counts): Update.
29500 * modulo-sched.c (sms_schedule): Update.
29501 * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29502 * predict.c (maybe_hot_count_p): Update.
29503 (probably_never_executed): Update.
29504 (dump_prediction): Update.
29505 (combine_predictions_for_bb): Update.
29506 (propagate_freq): Update.
29507 (handle_missing_profiles): Update.
29508 (counts_to_freqs): Update.
29509 (rebuild_frequencies): Update.
29510 (force_edge_cold): Update.
29511 * predict.h: Include profile-count.h
29512 (maybe_hot_count_p, counts_to_freqs): UPdate.
29513 * print-rtl-function.c: Do not include cfg.h
29514 * print-rtl.c: Include basic-block.h
29515 * profile-count.c: New file.
29516 * profile-count.h: New file.
29517 * profile.c (is_edge_inconsistent): Update.
29518 (correct_negative_edge_counts): Update.
29519 (is_inconsistent): Update.
29520 (set_bb_counts): Update.
29521 (read_profile_edge_counts): Update.
29522 (compute_frequency_overlap): Update.
29523 (compute_branch_probabilities): Update; Initialize and deinitialize
29525 (branch_prob): Update.
29526 * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29527 (edge_gcov_count): New.
29528 (bb_gcov_count): New.
29529 * shrink-wrap.c (try_shrink_wrapping): Update.
29530 * tracer.c (better_p): Update.
29531 * trans-mem.c (expand_transaction): Update.
29532 (ipa_tm_insert_irr_call): Update.
29533 (ipa_tm_insert_gettmclone_call): Update.
29534 * tree-call-cdce.c: Update.
29535 * tree-cfg.c (gimple_duplicate_sese_region): Update.
29536 (gimple_duplicate_sese_tail): Update.
29537 (gimple_account_profile_record): Update.
29538 (execute_fixup_cfg): Update.
29539 * tree-inline.c (copy_bb): Update.
29540 (copy_edges_for_bb): Update.
29541 (initialize_cfun): Update.
29542 (freqs_to_counts): Update.
29543 (copy_cfg_body): Update.
29544 (expand_call_inline): Update.
29545 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29546 * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29547 (try_unroll_loop_completely): Update.
29548 (try_peel_loop): Update.
29549 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29550 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29551 * tree-ssa-loop-split.c (connect_loops): Update.
29552 * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29553 * tree-ssa-reassoc.c (branch_fixup): Update.
29554 * tree-ssa-tail-merge.c (replace_block_by): Update.
29555 * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29556 (compute_path_counts): Update.
29557 (update_profile): Update.
29558 (recompute_probabilities): Update.
29559 (update_joiner_offpath_counts): Update.
29560 (estimated_freqs_path): Update.
29561 (freqs_to_counts_path): Update.
29562 (clear_counts_path): Update.
29563 (ssa_fix_duplicate_block_edges): Update.
29564 (duplicate_thread_path): Update.
29565 * tree-switch-conversion.c (case_bit_test_cmp): Update.
29566 (struct switch_conv_info): Update.
29567 * tree-tailcall.c (decrease_profile): Update.
29568 * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29569 * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29570 * value-prof.c (check_counter): Update.
29571 (gimple_divmod_fixed_value): Update.
29572 (gimple_mod_pow2): Update.
29573 (gimple_mod_subtract): Update.
29574 (gimple_ic_transform): Update.
29575 (gimple_stringop_fixed_value): Update.
29576 * value-prof.h (gimple_ic): Update.
29578 2017-06-02 Carl Love <cel@us.ibm.com>
29580 * config/rs6000/rs6000-c: Add support for built-in functions
29581 vector double vec_doublee (vector signed int);
29582 vector double vec_doublee (vector unsigned int);
29583 vector double vec_doublee (vector float);
29584 vector double vec_doubleh (vector signed int);
29585 vector double vec_doubleh (vector unsigned int);
29586 vector double vec_doubleh (vector float);
29587 vector double vec_doublel (vector signed int);
29588 vector double vec_doublel (vector unsigned int);
29589 vector double vec_doublel (vector float);
29590 vector double vec_doubleo (vector signed int);
29591 vector double vec_doubleo (vector unsigned int);
29592 vector double vec_doubleo (vector float);.
29593 * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29594 DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29596 * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29597 unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29598 unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29600 * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29601 vec_doublel, vec_doubleh.
29602 * doc/extend.texi: Update the built-in documentation file for the
29603 new built-in functions.
29605 2017-06-02 David Malcolm <dmalcolm@redhat.com>
29608 * ipa-inline-analysis.c (free_growth_caches): Set
29609 edge_removal_hook_holder to NULL after removing it.
29611 2017-06-02 Sudakshina Das <sudi.das@arm.com>
29613 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29614 comparision with zero.
29616 2017-06-02 Will Schmidt <will_schmidt@vnet.ibm.com>
29617 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29618 for early expansion of vec_min and vec_max builtins.
29619 (builtin_function_type): Add min/max unsigned variants to those
29620 identified as having unsigned arguments.
29622 2017-06-02 Olivier Hainque <hainque@adacore.com>
29624 * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29626 2017-06-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29628 * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29629 Use VALL_F16 iterator rather than VALL.
29631 2017-06-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29633 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29634 Emit CBNZ inside loop when doing a strong exchange and comparing
29635 against zero. Generate the CC flags after the loop.
29637 2017-06-02 David Edelsohn <dje.gcc@gmail.com>
29639 * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29640 (dl_section_ref): New.
29641 (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29642 On AIX, append an expression to subtract the size of the
29643 section length to dl_section_ref.
29645 2017-06-02 Will Schmidt <will_schmidt@vnet.ibm.com>
29647 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29648 for early expansion of vector absolute builtins.
29650 2017-06-02 Richard Biener <rguenther@suse.de>
29652 * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29653 what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29655 2017-06-02 Richard Biener <rguenther@suse.de>
29657 PR tree-optimization/80948
29658 * tree-tailcall.c (find_tail_calls): Track stmts to move in
29659 stmt order as well.
29661 2017-06-02 Richard Biener <rguenther@suse.de>
29663 * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29665 * tree-vect-stmts.c (process_use): Do not mark backedge defs
29666 for inductions as relevant.
29668 2017-06-02 Richard Biener <rguenther@suse.de>
29670 * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29671 (vectorizable_induction): ... this. Remove dead code.
29673 2017-06-02 Eric Botcazou <ebotcazou@adacore.com>
29675 * builtins. (expand_builtin_alloca): Remove second parameter and
29676 infer its value from the first parameter instead.
29677 (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29679 2017-06-02 Jakub Jelinek <jakub@redhat.com>
29681 PR rtl-optimization/80903
29682 * loop-doloop.c (add_test): Unshare sequence.
29684 2017-06-02 Bernd Edlinger <bernd.edlinger@hotmail.de>
29686 * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29688 2017-06-01 Bernd Edlinger <bernd.edlinger@hotmail.de>
29690 * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29692 (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29693 xlogue_layout::get_instance, logue_layout::xlogue_layout,
29694 sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29695 (xlogue_layout::get_stub_rtx): Make static.
29696 (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29697 (xlogue_layout::compute_stub_managed_regs): Rename to...
29698 (xlogue_layout::count_stub_managed_regs): ...this.
29699 (xlogue_layout::is_stub_managed_reg): New function.
29700 (xlogue_layout::m_stub_names): Rename to...
29701 (xlogue_layout::s_stub_names): ...this, make static.
29702 (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29703 xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29704 xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29705 xlogue_layout::s_stub_names): Instantiate statics.
29706 (stub_managed_regs): Remove.
29707 (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29708 (disable_call_ms2sysv_xlogues): Rename to...
29709 (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29710 (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29712 (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29713 change after reload_completed.
29714 (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29716 (ix86_expand_prologue): Likewise.
29717 (ix86_expand_epilogue): Likewise.
29718 (ix86_expand_split_stack_prologue): Likewise.
29719 (ix86_compute_frame_layout): Remove frame parameter ...
29720 (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29721 (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29723 (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29724 (ix86_frame): Move from here ...
29725 * config/i386/i386.h (ix86_frame): ... to here.
29726 (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29727 complete ix86_frame data structure instead. Remove some_ld_name.
29729 2017-06-01 Pierre-Marie de Rodat <derodat@adacore.com>
29731 * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29732 symbols that hold a DECL_VALUE_EXPR.
29734 2017-06-01 Martin Jambor <mjambor@suse.cz>
29736 PR tree-optimization/80898
29737 * tree-sra.c (process_subtree_disqualification): Removed.
29738 (disqualify_candidate): Do not acll
29739 process_subtree_disqualification.
29740 (subtree_mark_written_and_enqueue): New function.
29741 (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29742 RHS has been disqualified and re-queue LHS if necessary. Apart
29743 from that, ignore disqualified RHS.
29745 2017-06-01 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
29747 * config/s390/s390.c (s390_emit_epilogue): Disable early return
29748 address fetch for z10 or later.
29750 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29752 * config/arc/arc.md (tst_movb): Add guard when splitting.
29754 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29756 * config/arc/arc.c (arc_can_eliminate): Test against
29757 arc_frame_pointer_needed.
29759 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29761 * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29762 to prevent store reordering.
29763 * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29764 (type): Add block type.
29765 (stack_tie): Define special instruction to be used in
29768 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29770 * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29771 constraint. It is not valid for the pattern.
29772 (noncommutative_binary_comparison): Likewise.
29774 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29776 * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29779 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29781 * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29782 be used by the reg-alloc.
29784 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29786 * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29787 reg-alloc when having mul64 or mul32x16 instructions.
29788 (mulsidi3): Likewise.
29789 (umulsidi3): Likewise.
29790 (mulsi32x16): New pattern.
29791 (mulsi64): Likewise.
29792 (mulsidi64): Likewise.
29793 (umulsidi64): Likewise.
29794 (MUL32x16_REG): Define.
29795 (mul64_600): Use MUL32x16_REG.
29796 (mac64_600): Likewise.
29797 (umul64_600): Likewise.
29798 (umac64_600): Likewise.
29800 2017-06-01 Claudiu Zissulescu <claziss@synopsys.com>
29802 * config/arc/arc.md (mulsi3_700): Make it commutative.
29804 2017-06-01 Jose E. Marchesi <jose.marchesi@oracle.com>
29806 * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29808 (*sign_extendsidi2_insn): Likewise for movstosw.
29810 2017-06-01 Pierre-Marie de Rodat <derodat@adacore.com>
29812 * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29813 the type of the input discriminant value. Convert the
29814 discriminant value of signedness vary.
29816 2017-06-01 Volker Reichelt <v.reichelt@netcologne.de>
29818 * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29819 Add to -Wall section.
29821 2017-06-01 Richard Biener <rguenther@suse.de>
29823 PR middle-end/66313
29824 * fold-const.c (fold_plusminus_mult_expr): If the factored
29825 factor may be zero use a wrapping type for the inner operation.
29826 * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29827 and handle moved defs.
29828 (process_assignment): Properly guard the unary op case. Return a
29829 tri-state indicating that moving the stmt before the call may allow
29830 to continue. Pass through to_move.
29831 (find_tail_calls): Handle moving unrelated defs before
29834 2017-05-31 Segher Boessenkool <segher@kernel.crashing.org>
29837 * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29838 splitter result in the canonical way.
29840 2017-05-31 Uros Bizjak <ubizjak@gmail.com>
29842 * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29843 also for 32bit target. Update insn attributes.
29844 (zero-extendsidi2 splitter): Allow all registers for operand 1.
29846 2017-05-31 Sebastian Peryt <sebastian.peryt@intel.com>
29848 * config/i386/avx512fintrin.h (_mm_mask_max_sd)
29849 (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
29850 (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
29851 (_mm_maskz_min_ss): New intrinsics.
29853 2017-05-31 Martin Liska <mliska@suse.cz>
29855 * tree-vect-loop.c (vect_create_epilog_for_reduction):
29856 Change comment style to one we normally use.
29857 (vectorizable_reduction): Likewise.
29858 (vectorizable_induction): Likewise.
29859 * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
29860 (vectorizable_call): Likewise.
29861 (vectorizable_simd_clone_call): Likewise.
29862 (vectorizable_conversion): Likewise.
29863 (vectorizable_assignment): Likewise.
29864 (vectorizable_shift): Likewise.
29865 (vectorizable_operation): Likewise.
29866 (vectorizable_store): Likewise.
29867 (vectorizable_load): Likewise.
29868 * tree-vectorizer.h: Likewise.
29870 2017-05-31 Alexander Monakov <amonakov@ispras.ru>
29872 * passes.c (emergency_dump_function): New.
29873 * tree-pass.h (emergency_dump_function): Declare.
29874 * plugin.c (plugins_internal_error_function): Remove.
29875 * plugin.h (plugins_internal_error_function): Remove declaration.
29876 * toplev.c (internal_error_function): New static function. Use it...
29877 (general_init): ...here.
29879 2017-05-31 Graham Markall <graham.markall@embecosm.com>
29881 * config/arc/arc.c (arc_print_operand): Handle constant operands.
29882 (arc_rtx_costs): Add costs for new patterns.
29883 * config/arc/arc.md: Additional *add_n and *sub_n patterns.
29884 * config/arc/predicates.md: Add _1_2_3_operand predicate.
29886 2017-05-31 Richard Sandiford <richard.sandiford@linaro.org>
29888 * tree-ssa-strlen.c (get_next_strinfo): New function.
29889 (get_stridx_plus_constant): Use it.
29890 (zero_length_string): Likewise.
29891 (adjust_related_strinfos): Likewise.
29892 (adjust_last_stmt): Likewise.
29894 2017-05-31 Richard Biener <rguenther@suse.de>
29897 * config/i386/i386.c (ix86_expand_builtin): Remove assert
29898 for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
29900 2017-05-31 Richard Sandiford <richard.sandiford@linaro.org>
29902 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
29903 loop_vinfo argument and use of dependence distance vectors.
29904 Check instead whether the two references differ only in their
29905 initial value and assume that they have the same alignment if the
29906 difference is a multiple of the vector alignment.
29907 (vect_analyze_data_refs_alignment): Update call accordingly.
29909 2017-05-31 Martin Liska <mliska@suse.cz>
29912 * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
29914 2017-05-31 Bin Cheng <bin.cheng@arm.com>
29916 * tree-vect-loop-manip.c (create_intersect_range_checks_index)
29917 (create_intersect_range_checks): Move from ...
29918 * tree-data-ref.c (create_intersect_range_checks_index)
29919 (create_intersect_range_checks): ... to here.
29920 (create_runtime_alias_checks): New function factored from ...
29921 * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
29922 here. Call above function.
29923 * tree-data-ref.h (create_runtime_alias_checks): New function.
29925 2017-05-31 Bin Cheng <bin.cheng@arm.com>
29927 * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
29928 segment length for dr_b and compute it in wide_int.
29930 2017-05-31 Richard Biener <rguenther@suse.de>
29932 PR tree-optimization/80906
29933 * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
29934 and pass through iv_map.
29935 (copy_bb_and_scalar_dependences): Adjust.
29936 (translate_pending_phi_nodes): Likewise.
29937 (copy_loop_close_phi_args): Handle code-generating IVs instead
29940 2017-05-30 David Malcolm <dmalcolm@redhat.com>
29942 * diagnostic-color.c (color_dict): Add "type-diff".
29943 (parse_gcc_colors): Update comment.
29944 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
29945 -fdiagnostics-show-template-tree and -fno-elide-type.
29946 (GCC_COLORS): Add type-diff to example.
29948 (-fdiagnostics-show-template-tree): New.
29949 (-fno-elide-type): New.
29950 * pretty-print.c (pp_format): Pass quote and formatters[argno] to
29951 the pp_format_decoder callback. Call any m_format_postprocessor's
29953 (pretty_printer::pretty_printer): Initialize
29954 m_format_postprocessor.
29955 (pretty_printer::~pretty_printer): Delete any
29956 m_format_postprocessor.
29957 * pretty-print.h (printer_fn): Add bool and const char ** parameters.
29958 (class format_postprocessor): New class.
29959 (struct pretty_printer::format_decoder): Document the new parameters.
29960 (struct pretty_printer::m_format_postprocessor): New field.
29961 * tree-diagnostic.c (default_tree_printer): Update for new
29962 bool and const char ** params.
29963 * tree-diagnostic.h (default_tree_printer): Likewise.
29965 2017-05-30 Segher Boessenkool <segher@kernel.crashing.org>
29967 * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
29968 (lwa_operand): Delete rs6000_gen_cell_microcode test.
29969 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29970 rs6000_gen_cell_microcode code.
29971 (rs6000_final_prescan_insn): Delete.
29972 (rs6000_opt_vars): Delete the "gen-cell-microcode" and
29973 "warn-cell-microcode" entries.
29974 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
29975 * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
29976 throughout. Change cc_reg_not_micro_cr0_operand to
29977 cc_reg_not_cr0_operand throughout.
29978 (*extendhi<mode>2_noload): Delete.
29979 * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
29980 (mwarn-cell-microcode): Delete.
29981 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
29982 -mgen-cell-microcode and -mwarn-cell-microcode.
29984 2017-05-30 Uros Bizjak <ubizjak@gmail.com>
29987 * config/i386/constraints.md (Yd): New constraint.
29989 * config/i386/i386.md (*movti_internal): Add (?r, Ye)
29990 and (?Yd, r) alternatives. Update insn attributes.
29991 * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
29992 and (?*Yd, r) alternatives. Update insn attributes.
29993 (double-mode inter-unit splitters): Add new GR<->XMM splitters.
29995 2017-05-30 Pierre-Marie de Rodat <derodat@adacore.com>
29997 * gimplify.c (gimplify_modify_expr): Don't create a
29998 DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30001 2017-05-30 Wilco Dijkstra <wdijkstr@arm.com>
30003 * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30005 2017-05-30 Richard Biener <rguenther@suse.de>
30007 * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30008 and reduc_def fields.
30009 (STMT_VINFO_REDUC_TYPE): New define.
30010 (STMT_VINFO_REDUC_DEF): Likewise.
30011 (vect_force_simple_reduction): Adjust prototype.
30012 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30013 (vect_is_simple_reduction): Remove check_reduction argument.
30014 (vect_force_simple_reduction): Adjust and set
30015 STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30016 (vectorizable_reduction): Do not re-do reduction analysis
30017 but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30018 * tree-parloops.c (gather_scalar_reductions): Adjust.
30020 2017-05-30 Richard Biener <rguenther@suse.de>
30022 PR middle-end/80901
30023 * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30026 2017-05-24 Robin Dapp <rdapp@linux.vnet.ibm.com>
30028 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30029 Introduce unknown_misalignment parameter and remove vf.
30030 (vect_peeling_hash_get_lowest_cost):
30031 Pass unknown_misalignment parameter.
30032 (vect_enhance_data_refs_alignment):
30033 Fix unsupportable data ref treatment.
30035 2017-05-30 Robin Dapp <rdapp@linux.vnet.ibm.com>
30037 * tree-vect-data-refs.c (vect_get_data_access_cost):
30038 Workaround for SLP handling.
30039 (vect_enhance_data_refs_alignment):
30040 Compute costs for doing no peeling at all, compare to the best
30041 peeling costs so far and avoid peeling if cheaper.
30043 2017-05-30 Robin Dapp <rdapp@linux.vnet.ibm.com>
30045 * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30046 Return peeling info and set costs to zero for unlimited cost
30048 (vect_enhance_data_refs_alignment): Also inspect all datarefs
30049 with unknown misalignment. Compute and costs for unknown
30050 misalignment, compare them to the costs for known misalignment
30051 and choose the cheapest for peeling.
30053 2017-05-30 Robin Dapp <rdapp@linux.vnet.ibm.com>
30055 * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30056 (vect_get_peeling_costs_all_drs): Create function.
30057 (vect_peeling_hash_get_lowest_cost):
30058 Use vect_get_peeling_costs_all_drs.
30059 (vect_peeling_supportable): Create function.
30060 (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30062 2017-05-30 Robin Dapp <rdapp@linux.vnet.ibm.com>
30064 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30065 DR_HAS_NEGATIVE_STEP.
30066 (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30067 (vect_enhance_data_refs_alignment): Use.
30068 (vect_duplicate_ssa_name_ptr_info): Use.
30069 * tree-vectorizer.h (dr_misalignment): Use.
30070 (known_alignment_for_access_p): Use.
30072 2017-05-30 Jozef Lawrynowicz <jozef.l@somniumtech.com>
30075 * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30077 (has_section_name): New function.
30079 2017-05-30 Martin Liska <mliska@suse.cz>
30082 * auto-profile.c (get_function_decl_from_block): Fix
30085 2017-05-30 Richard Biener <rguenther@suse.de>
30087 PR middle-end/80876
30088 * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30090 2017-05-30 Martin Liska <mliska@suse.cz>
30092 * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30093 * dumpfile.h (struct dump_file_info): Remove ctors.
30095 2017-05-30 Martin Liska <mliska@suse.cz>
30097 * predict.def: Fix GNU coding style.
30099 2017-05-29 Max Filippov <jcmvbkbc@gmail.com>
30101 * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30102 Mark 'to' argument with ATTRIBUTE_UNUSED.
30104 2017-05-29 Max Filippov <jcmvbkbc@gmail.com>
30106 * config/xtensa/xtensa.c (xtensa_emit_call): Use
30107 HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30108 (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30111 2017-05-29 Eric Botcazou <ebotcazou@adacore.com>
30113 * doc/install.texi (Options specification): Restore entry of
30114 --enable-sjlj-exceptions.
30116 2017-05-27 Michael Eager <eager@eagercon.com>
30119 2016-01-21 Ajit Agarwal <ajitkum@xilinx.com>
30121 See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30123 * config/microblaze/microblaze.h
30124 (FIXED_REGISTERS): Update in macro.
30125 (CALL_USED_REGISTERS): Update in macro.
30127 2017-05-27 François-Xavier Coudett <fxcoudert@gcc.gnu.org>
30129 * doc/install.texi: Add links to macOS binary distributions.
30131 2017-05-27 Jakub Jelinek <jakub@redhat.com>
30135 2017-05-25 Marc Glisse <marc.glisse@inria.fr>
30137 * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30139 2017-05-26 Martin Liska <mliska@suse.cz>
30141 * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30143 2017-05-26 Martin Liska <mliska@suse.cz>
30145 * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30146 always leading ';; '.
30147 (dump_bb_info): Likewise.
30148 (brief_dump_cfg): Likewise.
30149 * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30150 * dumpfile.c: Remove usage of TDF_VERBOSE.
30151 * dumpfile.h (enum dump_kind): Likewise.
30152 (dump_gimple_bb_header): Do not use TDF_COMMENT.
30153 * print-tree.c (debug_verbose): Remove.
30154 * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30155 (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30156 * tree-diagnostic.c (default_tree_printer): Replace
30157 TDF_DIAGNOSTIC with TDF_SLIM.
30159 2017-05-26 Bin Cheng <bin.cheng@arm.com>
30161 * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30162 in parameter loop, rather than loop_vinfo.
30163 (create_intersect_range_checks): Ditto.
30164 (vect_create_cond_for_alias_checks): Update call to above functions.
30166 2017-05-26 Bin Cheng <bin.cheng@arm.com>
30168 PR tree-optimization/80815
30169 * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30170 for merging runtime alias checks. Handle negative DR_STEPs.
30172 2017-05-26 Bin Cheng <bin.cheng@arm.com>
30174 * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30176 * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30177 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30178 out code pruning runtime alias checks.
30179 * tree-data-ref.c (prune_runtime_alias_test_list): New function
30180 factored out from above.
30181 * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30183 * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30185 (prune_runtime_alias_test_list): New decalaration.
30187 2017-05-26 Bin Cheng <bin.cheng@arm.com>
30189 * tree-vect-data-refs.c (compare_tree): Rename and move ...
30190 * tree-data-ref.c (data_ref_compare_tree): ... to here.
30191 * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30192 * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30193 (operator==, comp_dr_with_seg_len_pair): Ditto.
30194 (vect_prune_runtime_alias_test_list): Ditto.
30196 2017-05-26 Martin Liska <mliska@suse.cz>
30199 * params.def: Bound partial-inlining-entry-probability param.
30201 2017-05-26 Marek Polacek <polacek@redhat.com>
30204 * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30207 2017-05-26 Richard Biener <rguenther@suse.de>
30209 PR tree-optimization/80842
30210 * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30213 2017-05-26 Richard Biener <rguenther@suse.de>
30215 PR tree-optimization/80844
30216 * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30218 2017-05-25 Sebastian Peryt <sebastian.peryt@intel.com>
30220 * doc/md.texi (Machine Constraints): Update x86 family
30221 machine constraints section to match 'config/i386/constraints.md'.
30223 2017-05-25 Volker Reichelt <v.reichelt@netcologne.de>
30225 * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30227 2017-05-25 Nathan Sidwell <nathan@acm.org>
30229 * doc/invoke.texi (--enable-languages): Update documentation.
30231 2017-05-25 Martin Liska <mliska@suse.cz>
30233 * dumpfile.c: Add TDF_FOLDING.
30234 * dumpfile.h (enum dump_kind): Likewise.
30235 * genmatch.c (dt_simplify::gen_1): Use it.
30237 2017-05-25 Marc Glisse <marc.glisse@inria.fr>
30239 * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30241 2017-05-25 Marc Glisse <marc.glisse@inria.fr>
30243 * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30244 * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30246 2017-05-25 Marc Glisse <marc.glisse@inria.fr>
30248 * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30249 * match.pd (X == C): Rewrite it here.
30250 (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30251 with_certain_nonzero_bits2): New predicates.
30252 * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30254 2017-05-24 Nathan Sidwell <nathan@acm.org>
30256 * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30259 * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30262 2017-05-24 Segher Boessenkool <segher@kernel.crashing.org>
30264 * config/powerpcspe: New port. Files are copied from the rs6000
30265 port, with "rs6000" in filenames replaced by "powerpcspe".
30267 2017-05-24 Wilco Dijkstra <wdijkstr@arm.com>
30269 PR rtl-optimization/80754
30270 * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30272 2017-05-24 Sheldon Lobo <smlobo@sheldon.us.oracle.com>
30274 * config/sparc/sparc.md (length): Return the correct value for -mflat
30275 sibcalls to match output_sibcall.
30277 2017-05-24 Segher Boessenkool <segher@kernel.crashing.org>
30281 * config/rs6000/rs6000.c (struct machine_function): Add new field
30283 (rs6000_get_separate_components): Init that field, use it.
30284 (rs6000_components_for_bb): Use the field.
30286 2017-05-24 Thomas Preud'homme <thomas.preudhomme@arm.com>
30288 * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30290 2017-05-24 Peter Bergner <bergner@vnet.ibm.com>
30292 PR middle-end/80823
30293 * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30295 2017-05-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
30298 * config/s390/s390.c (s390_check_qrst_address): Check incoming
30299 address against address_operand predicate.
30300 * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30302 2017-05-24 Eric Botcazou <ebotcazou@adacore.com>
30304 * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30305 parameters passed indirectly.
30307 2017-05-23 Uros Bizjak <ubizjak@gmail.com>
30309 * config/i386/i386.md (*movdi_internal): Remove SSE4
30310 alternative 18 (?r, *v). Update insn attributes.
30311 (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30312 Update insn attributes.
30313 (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30314 Update insn attributes.
30315 * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30316 alternative 1 (r, v). Remove isa attribute.
30317 * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30318 Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30319 and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30321 2017-05-23 Tom de Vries <tom@codesourcery.com>
30323 * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30326 2017-05-23 Jan Hubicka <hubicka@ucw.cz>
30328 * cgraphunit.c (symbol_table::process_new_functions): Update.
30329 * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30330 (inline_generate_summary): Rename to ...
30331 (ipa_fn_summary_generate): ... this one.
30332 (inline_read_summary): Rename to ...
30333 (ipa_fn_summary_read): ... this one.
30334 (inline_write_summary): Rename to ...
30335 (ipa_fn_summary_write): ... this one.
30336 (inline_free_summary): Rename to ...
30337 (ipa_free_fn_summary): ... this one.
30338 (pass_data_local_fn_summary, pass_local_fn_summary,
30339 make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30340 pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30341 pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30342 make_pass_ipa_fn_summary): New.
30343 * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30344 inline_write_summary, inline_free_summary): Remove.
30345 (ipa_free_fn_summary) : New.
30346 * ipa-inline.c (ipa_inline): Update.
30347 (pass_ipa_inline): Do not generate summaries.
30348 * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30350 * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30351 and add pass_ipa_fn_summary.
30352 * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30354 (make_pass_inline_parameters): Remove.
30356 2017-05-23 Thomas Schwinge <thomas@codesourcery.com>
30358 * omp-low.c (struct omp_context): Remove "default_kind" member.
30361 * omp-offload.c (execute_oacc_device_lower): Remove the
30362 parallelism dimensions function attributes for unparallelized
30363 OpenACC kernels constructs.
30365 2017-05-23 Martin Liska <mliska@suse.cz>
30367 * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30369 (cgraph_edge::make_speculative): Likewise.
30370 (cgraph_edge::resolve_speculation): Likewise.
30371 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30372 (cgraph_node::dump): Likewise.
30373 * cgraph.h: Likewise.
30374 * cgraphunit.c (analyze_functions): Likewise.
30375 (symbol_table::compile): Likewise.
30376 * ipa-cp.c (print_all_lattices): Likewise.
30377 (determine_versionability): Likewise.
30378 (initialize_node_lattices): Likewise.
30379 (ipcp_verify_propagated_values): Likewise.
30380 (estimate_local_effects): Likewise.
30381 (update_profiling_info): Likewise.
30382 (create_specialized_node): Likewise.
30383 (perhaps_add_new_callers): Likewise.
30384 (decide_about_value): Likewise.
30385 (decide_whether_version_node): Likewise.
30386 (identify_dead_nodes): Likewise.
30387 (ipcp_store_bits_results): Likewise.
30388 * ipa-devirt.c (dump_targets): Likewise.
30389 (ipa_devirt): Likewise.
30390 * ipa-icf.c (sem_item::dump): Likewise.
30391 (sem_function::equals): Likewise.
30392 (sem_variable::equals): Likewise.
30393 (sem_item_optimizer::read_section): Likewise.
30394 (sem_item_optimizer::execute): Likewise.
30395 (congruence_class::dump): Likewise.
30396 * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30397 (dump_inline_summary): Likewise.
30398 (estimate_node_size_and_time): Likewise.
30399 (inline_analyze_function): Likewise.
30400 * ipa-inline-transform.c (inline_call): Likewise.
30401 * ipa-inline.c (report_inline_failed_reason): Likewise.
30402 (want_early_inline_function_p): Likewise.
30403 (edge_badness): Likewise.
30404 (update_edge_key): Likewise.
30405 (inline_small_functions): Likewise.
30406 * ipa-profile.c (ipa_profile): Likewise.
30407 * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30408 (ipa_make_edge_direct_to_target): Likewise.
30409 (remove_described_reference): Likewise.
30410 (ipa_impossible_devirt_target): Likewise.
30411 (propagate_controlled_uses): Likewise.
30412 (ipa_print_node_params): Likewise.
30413 (ipcp_transform_function): Likewise.
30414 * ipa-pure-const.c (pure_const_read_summary): Likewise.
30415 (propagate_pure_const): Likewise.
30416 * ipa-reference.c (generate_summary): Likewise.
30417 (read_write_all_from_decl): Likewise.
30418 (propagate): Likewise.
30419 (ipa_reference_read_optimization_summary): Likewise.
30420 * ipa-utils.c (ipa_merge_profiles): Likewise.
30421 * ipa.c (walk_polymorphic_call_targets): Likewise.
30422 (symbol_table::remove_unreachable_nodes): Likewise.
30423 (ipa_single_use): Likewise.
30424 * passes.c (execute_todo): Likewise.
30425 * predict.c (drop_profile): Likewise.
30426 * symtab.c (symtab_node::get_dump_name): New function.
30427 (symtab_node::dump_name): Likewise.
30428 (symtab_node::dump_asm_name): Likewise.
30429 (symtab_node::dump_references): Likewise.
30430 (symtab_node::dump_referring): Likewise.
30431 (symtab_node::dump_base): Likewise.
30432 (symtab_node::debug_symtab): Likewise.
30433 * tree-sra.c (convert_callers_for_node): Likewise.
30434 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30435 * value-prof.c (init_node_map): Likewise.
30437 2017-05-23 Martin Liska <mliska@suse.cz>
30439 * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30440 and symtab_node::debug_symtab to symbol_table::debug.
30441 * cgraphunit.c (analyze_functions): Use the renamed function.
30442 (symbol_table::compile): Likewise.
30443 * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30444 * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30445 * passes.c (execute_todo): Likewise.
30446 * symtab.c (symbol_table::dump): New function.
30447 * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30449 2017-05-23 Jan Hubicka <hubicka@ucw.cz>
30451 * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30452 that nonconst implies exec.
30454 2017-05-23 Jan Hubicka <hubicka@ucw.cz>
30456 * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30457 inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30458 (inline_edge_summary_vec): Turn into ...
30459 (ipa_call_summaries): ... this one.
30460 (redirect_to_unreachable, edge_set_predicate,
30461 evaluate_properties_for_edge, inline_summary_alloc,
30462 reset_ipa_call_summary, reset_inline_summary,
30463 inline_summary_t::duplicate): Update.
30464 (inline_edge_duplication_hook): Turn to ...
30465 (ipa_call_summary_t::duplicate): ... this one.
30466 (inline_edge_removal_hook): Turn to ...
30467 (ipa_call_summary_t::remove): ... this one.
30468 (dump_inline_edge_summary): Turn to ...
30469 (dump_ipa_call_summary): ... this one.
30470 (estimate_function_body_sizes): Update.
30471 (inline_update_callee_summaries): Update.
30472 (remap_edge_change_prob): Update.
30473 (remap_edge_summaries): Update.
30474 (inline_merge_summary): Update.
30475 (do_estimate_edge_time): Update.
30476 (inline_generate_summary): Update.
30477 (inline_read_section): Update.
30478 (inline_read_summary): Update.
30479 (inline_free_summary): Update.
30480 * ipa-inline.c (can_inline_edge_p): Update.
30481 (compute_inlined_call_time): Update.
30482 (want_inline_small_function_p): Update.
30483 (edge_badness): Update.
30484 (early_inliner): Update.
30485 * ipa-inline.h (inline_edge_summary): Turn to ...
30486 (ipa_call_summary): ... this one.
30487 (ipa_call_summary_t): New class.
30488 (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30489 (ipa_call_summaries): New.
30490 (inline_edge_summary): Remove.
30491 (estimate_edge_growth): Update.
30492 * ipa-profile.c (ipa_propagate_frequency_1): Update.
30493 * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30494 * ipa-split.c (execute_split_functions): Update.
30495 * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30497 2017-05-23 Tom de Vries <tom@codesourcery.com>
30499 * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30500 attributes): Document rdrand effective target.
30502 2017-05-23 Tom de Vries <tom@codesourcery.com>
30504 * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30505 attributes): Sort alphabetically.
30507 2017-05-23 Georg-Johann Lay <avr@gjlay.de>
30509 * config/avr/genmultilib.awk: Use gsub instead of gensub.
30511 2017-05-22 Michael Meissner <meissner@linux.vnet.ibm.com>
30514 * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30515 V2DF/V2DI splat into two separate patterns, one that handles
30516 registers, and the other that only handles memory. Drop support
30517 for splatting from a GPR on ISA 2.07 and then splitting the
30518 splat into direct move and splat.
30519 (vsx_splat_<mode>_reg): Likewise.
30520 (vsx_splat_<mode>_mem): Likewise.
30522 2017-05-22 Segher Boessenkool <segher@kernel.crashing.org>
30524 * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30526 2017-05-22 Jakub Jelinek <jakub@redhat.com>
30528 PR middle-end/80809
30529 * omp-low.c (finish_taskreg_remap): New function.
30530 (finish_taskreg_scan): If unit size of ctx->record_type
30531 is non-constant, unshare the size expression and replace
30532 decls in it with possible outer var refs.
30534 PR middle-end/80809
30535 * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30536 GOVD_SHARED rather than GOVD_PRIVATE with it.
30537 (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30538 GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30540 PR middle-end/80853
30541 * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30542 as last argument to build_outer_var_ref for pointer bases of array
30543 section reductions.
30545 2017-05-19 Martin Sebor <msebor@redhat.com>
30547 * print-tree.c (print_node): Print DECL_READ_P flag.
30549 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
30551 * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30552 * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30553 * cgraph.c: Likewise.
30554 * cgraphunit.c: Likewise.
30555 * gengtype.c: Likewise.
30556 * ipa-cp.c: Likewise.
30557 * ipa-devirt.c: Likewise.
30558 * ipa-icf.c: Likewise.
30559 * ipa-predicate.c: Likewise.
30560 * ipa-profile.c: Likewise.
30561 * ipa-prop.c: Likewise.
30562 * ipa-split.c: Likewise.
30564 * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30565 edge_predicate_pool, dump_inline_hints,
30566 inline_summary::account_size_time, redirect_to_unreachable,
30567 edge_set_predicate, set_hint_predicate,
30568 evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30569 inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30570 inline_summary_t::remove, remap_hint_predicate_after_duplication,
30571 inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30572 ipa_call_summary_t::remove, initialize_growth_caches,
30573 free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30574 debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30575 mark_modified, unmodified_parm_1, unmodified_parm,
30576 unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30577 set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30578 compute_bb_predicates, will_be_nonconstant_expr_predicate,
30579 will_be_nonconstant_predicate, record_modified_bb_info,
30580 get_minimal_bb, record_modified, param_change_prob,
30581 phi_result_unknown_predicate, predicate_for_phi_result,
30582 array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30583 estimate_function_body_sizes, compute_inline_parameters,
30584 compute_inline_parameters_for_curren, pass_data_inline_parameters,
30585 estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30586 inline_update_callee_summaries, remap_edge_change_prob,
30587 remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30588 inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30589 inline_analyze_function, inline_summary_t::insert,
30590 inline_generate_summary, read_ipa_call_summary, inline_read_section,
30591 inline_read_summary, write_ipa_call_summary, inline_write_summary,
30592 inline_free_summary): Move to ipa-fnsummary.h
30593 (predicate_t): Remove.
30594 * ipa-fnsummary.c: New file.
30595 * ipa-inline.h: Do not include sreal.h and ipa-predicate.h
30596 (enum inline_hints_vals, inline_hints, agg_position_info,
30597 INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30598 inline_summaries, ipa_call_summary, ipa_call_summary_t,
30599 ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30600 dump_inline_summary, dump_inline_hints, inline_generate_summary,
30601 inline_read_summary, inline_write_summary, inline_free_summary,
30602 inline_analyze_function, initialize_inline_failed,
30603 inline_merge_summary, inline_update_overall_summary,
30604 compute_inline_parameters): Move to ipa-fnsummary.h
30605 * ipa-fnsummary.h: New file.
30606 * ipa-inline-transform.h: Include ipa-inline.h.
30607 * ipa-inline.c: LIkewise.
30609 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
30611 * ipa-inline.c (edge_badness): Use inlined_time instead of
30612 inline_summaries->get.
30614 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
30616 * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30618 2017-05-22 Nathan Sidwell <nathan@acm.org>
30620 * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30621 (fdump-lang): Document 'raw' option.
30622 * dumpfile.h (TDI_tu): Delete.
30623 * dumpfile.c (dump_files): Remove translation-unit.
30624 (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30626 2017-05-22 Georg-Johann Lay <avr@gjlay.de>
30628 * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30629 command option from $(AWK) call.
30630 * config/avr/genmultilib.awk: Simplify and rewrite so that it
30631 generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30632 [FORMAT]: Remove handling of variable.
30633 * config/avr/t-multilib: Regenerate.
30635 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
30637 * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30639 (dump_inline_summary): Do not print self_time.
30640 (estimate_function_body_sizes): Do not set self_time.
30641 (compute_inline_parameters): Likewise.
30642 (inline_read_section, inline_write_summary): Do not stream self_time.
30643 * ipa-inline.h (inline_summary): Drop self_time.
30645 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
30647 * ipa-inline-analysis.c (account_size_time): Rename to ...
30648 (inline_summary::account_size_time): ... this one.
30649 (reset_ipa_call_summary): Turn to ...
30650 (ipa_call_summary::reset): ... this one.
30651 (reset_inline_summary): Turn to ...
30652 (inline_summary::reset): ... this one.
30653 (inline_summary_t::remove): Update.
30654 (inline_summary_t::duplicate): Update.
30655 (ipa_call_summary_t::remove): Update.
30656 (dump_inline_summary): Update.
30657 (estimate_function_body_sizes): Update.
30658 (compute_inline_parameters): Update.
30659 (estimate_node_size_and_time): Update.
30660 (inline_merge_summary): Update.
30661 (inline_update_overall_summary): Update.
30662 (inline_read_section): Update.
30663 (inline_write_summary): Update.
30664 * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30665 add account_size_time and reset member functions.
30666 (ipa_call_summary): Add reset function.
30667 * ipa-predicate.h (predicate::operator &): Constify.
30669 2017-05-22 Richard Biener <rguenther@suse.de>
30671 * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30673 2017-05-19 Jason Merrill <jason@redhat.com>
30675 * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30677 2017-05-19 Marek Polacek <polacek@redhat.com>
30680 * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30681 TYPE_OVERFLOW_WRAPS checks.
30683 2017-05-19 Thomas Schwinge <thomas@codesourcery.com>
30685 * tree-core.h (enum omp_clause_default_kind): Add
30686 "OMP_CLAUSE_DEFAULT_PRESENT".
30687 * tree-pretty-print.c (dump_omp_clause): Handle it.
30688 * gimplify.c (enum gimplify_omp_var_data): Add
30689 "GOVD_MAP_FORCE_PRESENT".
30690 (gimplify_adjust_omp_clauses_1): Map it to
30691 "GOMP_MAP_FORCE_PRESENT".
30692 (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30694 * gimplify.c (oacc_default_clause): Clarify.
30696 2017-05-19 Nathan Sidwell <nathan@acm.org>
30698 LANG_HOOK_REGISTER_DUMPS
30699 * toplev.c (general_init): Call register dump lang hook.
30700 * doc/invoke.texi: Document -fdump-lang option family.
30701 * dumpfile.c (dump_files): Remove class dump here.
30702 (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30703 * dumpfile.h (tree_dump_index): Remove TDI_class.
30704 * langhooks-def.h (lhd_register_dumps): Declare.
30705 (LANG_HOOKS_REGISTER_DUMPS): Define.
30706 (LANG_HOOKS_INITIALIZER): Add it.
30707 * langhooks.c (lhd_register_dumps): Define.
30708 * langhooks.h (struct lang_hooks): Add register_dumps.
30710 2017-05-19 Nathan Sidwell <nathan@acm.org>
30712 * context.h (context::set_passes): New.
30713 * context.c (context::context): Do not create pass manager.
30714 * toplev.c (general_init): Create pass manager here.
30716 2017-05-19 Segher Boessenkool <segher@kernel.crashing.org>
30718 * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30719 use this splitter if two add or or instructions would also work for
30720 the constant we want to generate.
30722 2017-05-19 Richard Biener <rguenther@suse.de>
30725 * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30726 predicate evaluation.
30728 2017-05-19 Jan Hubicka <hubicka@ucw.cz>
30730 * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30732 * ipa-inline.c (want_inline_small_function_p): Do not cast to
30735 2017-05-19 Jan Hubicka <hubicka@ucw.cz>
30737 * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30738 inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30739 (inline_edge_summary_vec): Turn into ...
30740 (ipa_call_summaries): ... this one.
30741 (redirect_to_unreachable, edge_set_predicate,
30742 evaluate_properties_for_edge, inline_summary_alloc,
30743 reset_ipa_call_summary, reset_inline_summary,
30744 inline_summary_t::duplicate): Update.
30745 (inline_edge_duplication_hook): Turn to ...
30746 (ipa_call_summary_t::duplicate): ... this one.
30747 (inline_edge_removal_hook): Turn to ...
30748 (ipa_call_summary_t::remove): ... this one.
30749 (dump_inline_edge_summary): Turn to ...
30750 (dump_ipa_call_summary): ... this one.
30751 (estimate_function_body_sizes): Update.
30752 (inline_update_callee_summaries): Update.
30753 (remap_edge_change_prob): Update.
30754 (remap_edge_summaries): Update.
30755 (inline_merge_summary): Update.
30756 (do_estimate_edge_time): Update.
30757 (inline_generate_summary): Update.
30758 (inline_read_section): Update.
30759 (inline_read_summary): Update.
30760 (inline_free_summary): Update.
30761 * ipa-inline.c (can_inline_edge_p): Update.
30762 (compute_inlined_call_time): Update.
30763 (want_inline_small_function_p): Update.
30764 (edge_badness): Update.
30765 (early_inliner): Update.
30766 * ipa-inline.h (inline_edge_summary): Turn to ...
30767 (ipa_call_summary): ... this one.
30768 (ipa_call_summary_t): New class.
30769 (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30770 (ipa_call_summaries): New.
30771 (inline_edge_summary): Remove.
30772 (estimate_edge_growth): Update.
30773 * ipa-profile.c (ipa_propagate_frequency_1): Update.
30774 * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30775 * ipa-split.c (execute_split_functions): Update.
30776 * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30778 2017-05-19 Richard Biener <rguenther@suse.de>
30780 PR middle-end/80764
30781 * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30783 2017-05-18 Segher Boessenkool <segher@kernel.crashing.org>
30785 * config/rs6000/rs6000.c (struct machine_function): Add field
30786 fpr_is_wrapped_separately.
30787 (rs6000_get_separate_components): Use 64 components. Handle the
30788 new FPR components.
30789 (rs6000_components_for_bb): Handle the FPR components.
30790 (rs6000_emit_prologue_components): Handle the FPR components.
30791 (rs6000_emit_epilogue_components): Handle the FPR components.
30792 (rs6000_set_handled_components): Handle the FPR components.
30793 (rs6000_emit_prologue): Don't output prologue code for those FPRs
30794 that are already separately shrink-wrapped.
30795 (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30796 that are already separately shrink-wrapped.
30798 2017-05-18 Michael Meissner <meissner@linux.vnet.ibm.com>
30801 * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30804 * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30805 (define_peephole2 for Altivec d-form load): Add peepholes to catch
30806 cases where the register allocator uses a move and an offsettable
30807 memory operation to/from a FPR register on ISA 2.06/2.07.
30808 (define_peephole2 for Altivec d-form store): Likewise.
30810 2017-05-18 Uros Bizjak <ubizjak@gmail.com>
30813 * config/i386/mmx.md (*mov<mode>_internal): Enable
30814 alternatives 11, 12, 13 and 14 also for 32bit targets.
30815 Remove alternatives 15, 16, 17 and 18.
30816 * config/i386/sse.md (vec_concatv2di): Change
30817 alternative (!x, *y) to (x, ?!*Yn).
30819 2017-05-18 Paolo Carlini <paolo.carlini@oracle.com>
30821 * dumpfile.h (enum dump_kind): Remove stray comma.
30823 2017-05-18 Jan Hubicka <hubicka@ucw.cz>
30825 * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30826 * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30827 predicate::num_conditions
30828 (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30829 (CHANGED): turn into predicate::changed.
30830 (agg_position_info): Move to ipa-predicate.h
30831 (add_condition, predicate::add_clause, predicate::operator &=,
30832 predicate::or_with, predicate::evaluate, predicate::probability,
30833 dump_condition, dump_clause, predicate::dump,
30834 predicate::remap_after_duplication, predicate::remap_after_inlining,
30835 predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30836 (evaluate_conditions_for_known_args): Update.
30837 (set_cond_stmt_execution_predicate): Update.
30838 * ipa-inline.h: Include ipa-predicate.h
30839 (condition, inline_param_summary, conditions, agg_position_info,
30840 predicate): Move to ipa-predicate.h
30841 * ipa-predicate.c: New file.
30842 * ipa-predicate.h: New file.
30844 2017-05-18 Wilco Dijkstra <wdijkstr@arm.com>
30846 * final.c (leaf_function_p): Check we are not in a sequence.
30848 2017-05-18 Martin Liska <mliska@suse.cz>
30850 * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
30851 * dumpfile.c (dump_register): Use new enum dump_kind.
30852 (get_dump_file_name): Likewise.
30853 (dump_enable_all): Likewise.
30854 (dump_switch_p_1): Likewise.
30855 (enable_rtl_dump_file): Remove usage of TDF_RTL.
30856 * dumpfile.h (enum dump_kind): New enum type.
30857 (struct dump_file_info): Create constructor and
30858 format fields and comments.
30859 * passes.c (pass_manager::register_one_dump_file):
30861 * statistics.c (statistics_early_init): Likewise.
30862 * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
30863 TDF_TREE with TDF_SLIM.
30864 (gather_memory_references_ref): Likewise.
30866 2017-05-18 Martin Liska <mliska@suse.cz>
30868 * vec.h (struct vnull): Use it.
30870 2017-05-18 Jan Hubicka <hubicka@ucw.cz>
30872 * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
30873 (true_predicate, false_predicate, true_predicate_p,
30874 false_predicate_p): Remove.
30875 (single_cond_predicate, not_inlined_predicate): Turn to member function
30877 (add_condition): Update.
30878 (add_clause): Turn to...
30879 (predicate::add_clause): ... this one; update; allow passing NULL
30881 (and_predicates): Turn to ...
30882 (predicate::operator &=): ... this one.
30883 (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
30884 (or_predicates): Turn to ...
30885 (predicate::or_with): ... this one.
30886 (evaluate_predicate): Turn to ...
30887 (predicate::evaluate): ... this one.
30888 (predicate_probability): Turn to ...
30889 (predicate::probability): ... this one.
30890 (dump_condition): Update.
30891 (dump_predicate): Turn to ...
30892 (predicate::dump): ... this one.
30893 (account_size_time): Update.
30894 (edge_set_predicate): Update.
30895 (set_hint_predicate): UPdate.
30896 (evaluate_conditions_for_known_args): Update.
30897 (evaluate_properties_for_edge): Update.
30898 (remap_predicate_after_duplication): Turn to...
30899 (predicate::remap_after_duplication): ... this one.
30900 (remap_hint_predicate_after_duplication): Update.
30901 (inline_summary_t::duplicate): UPdate.
30902 (dump_inline_edge_summary): Update.
30903 (dump_inline_summary): Update.
30904 (set_cond_stmt_execution_predicate): Update.
30905 (set_switch_stmt_execution_predicate): Update.
30906 (compute_bb_predicates): Update.
30907 (will_be_nonconstant_expr_predicate): Update.
30908 (will_be_nonconstant_predicate): Update.
30909 (phi_result_unknown_predicate): Update.
30910 (predicate_for_phi_result): Update.
30911 (array_index_predicate): Update.
30912 (estimate_function_body_sizes): Update.
30913 (estimate_node_size_and_time): Update.
30914 (estimate_ipcp_clone_size_and_time): Update.
30915 (remap_predicate): Rename to ...
30916 (predicate::remap_after_inlining): ... this one.
30917 (remap_hint_predicate): Update.
30918 (inline_merge_summary): Update.
30919 (inline_update_overall_summary): Update.
30920 (estimate_size_after_inlining): Update.
30921 (read_predicate): Rename to ...
30922 (predicate::stream_in): ... this one.
30923 (read_inline_edge_summary): Update.
30924 (write_predicate): Rename to ...
30925 (predicate::stream_out): ... this one.
30926 (write_inline_edge_summary): Update.
30927 * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
30928 (clause_t): Turn to uint32_t
30929 (predicate): Turn to class; implement constructor and operators
30931 (size_time_entry): Update.
30932 (inline_summary): Update.
30933 (inline_edge_summary): Update.
30935 2017-05-18 Marc Glisse <marc.glisse@inria.fr>
30937 * fold-const.c (fold_binary_loc): Move transformation...
30938 * match.pd (C - X CMP X): ... here.
30940 2017-05-18 Sheldon Lobo <sheldon.lobo@oracle.com>
30942 * config/sparc/sparc.c (sparc_option_override): Set function
30943 alignment for -mcpu=niagara7 to 64 to match the I$ line.
30944 * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
30946 * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
30948 * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
30950 2017-05-18 Marek Polacek <polacek@redhat.com>
30953 * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
30954 (pass_ubsan::execute): Call gimple_assign_single_p instead of
30955 gimple_assign_load_p.
30957 2017-05-17 Segher Boessenkool <segher@kernel.crashing.org>
30959 PR middle-end/80692
30960 * real.c (do_compare): Give decimal_do_compare preference over
30961 comparing just the signs.
30963 2017-05-17 Uros Bizjak <ubizjak@gmail.com>
30965 * doc/md.texi (Canonicalization of Instructions): Describe the
30966 canonical form of instructions that inherently set a condition
30969 2017-05-17 Peter Bergner <bergner@vnet.ibm.com>
30971 PR middle-end/80775
30972 * tree-cfg.c: Move deletion of unreachable case statements to after
30973 the merging of consecutive case labels.
30975 2017-05-17 Thomas Preud'homme <thomas.preudhomme@arm.com>
30977 * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
30978 readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
30979 restoring of callee-saved registers.
30981 2017-05-17 Eric Botcazou <ebotcazou@adacore.com>
30983 * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
30984 * config/visium/visium.c (single_set_and_flags): Likewise.
30985 * config/visium/visium.md (Substitutions): Likewise.
30987 2017-05-17 Martin Liska <mliska@suse.cz>
30989 * cfg.c: Introduce dump_flags_t type and
30990 use it instead of int type.
30992 * cfghooks.c: Likewise.
30993 * cfghooks.h (struct cfg_hooks): Likewise.
30994 * cfgrtl.c: Likewise.
30995 * cfgrtl.h: Likewise.
30996 * cgraph.c (cgraph_node::get_body): Likewise.
30997 * coretypes.h: Likewise.
30998 * domwalk.c: Likewise.
30999 * domwalk.h: Likewise.
31000 * dumpfile.c (struct dump_option_value_info): Likewise.
31001 (dump_enable_all): Likewise.
31002 (dump_switch_p_1): Likewise.
31003 (opt_info_switch_p): Likewise.
31004 * dumpfile.h (enum tree_dump_index): Likewise.
31005 (struct dump_file_info): Likewise.
31006 * genemit.c: Likewise.
31007 * generic-match-head.c: Likewise.
31008 * gengtype.c (open_base_files): Likewise.
31009 * gimple-pretty-print.c: Likewise.
31010 * gimple-pretty-print.h: Likewise.
31011 * graph.c (print_graph_cfg): Likewise.
31012 * graphite-scop-detection.c (dot_all_sese): Likewise.
31013 * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31014 * loop-unroll.c (report_unroll): Likewise.
31015 * passes.c (pass_manager::register_one_dump_file): Likewise.
31016 * print-tree.c: Likewise.
31017 * statistics.c: Likewise.
31018 * tree-cfg.c: Likewise.
31019 * tree-cfg.h: Likewise.
31020 * tree-dfa.c: Likewise.
31021 * tree-dfa.h: Likewise.
31022 * tree-dump.c (dump_function): Likewise.
31023 * tree-dump.h (struct dump_info): Likewise.
31024 * tree-pretty-print.c: Likewise.
31025 * tree-pretty-print.h: Likewise.
31026 * tree-ssa-live.c: Likewise.
31027 * tree-ssa-live.h: Likewise.
31028 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31029 * tree-vect-loop.c: Likewise.
31030 * tree-vect-slp.c: Likewise.
31032 2017-05-16 James Greenhalgh <james.greenhalgh@arm.com>
31033 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31035 PR tree-optimization/80457
31036 * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31037 of all arguments to a statement as scalar_to_vec operations.
31038 (vectorizable_call): Adjust call to vect_model_simple_cost for
31040 (vectorizable_conversion): Likewise.
31041 (vectorizable_assignment): Likewise.
31042 (vectorizable_shift): Likewise.
31043 (vectorizable_operation): Likewise.
31044 (vectorizable_comparison): Likewise.
31045 (vect_is_simple_cond): Record the def types for operands.
31046 (vectorizable_condition): Likewise, call vect_model_simple_cost.
31047 * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31048 for statement argument count.
31050 2017-05-16 Carl Love <cel@us.ibm.com>
31052 * config/rs6000/rs6000-c: Add support for built-in functions
31053 vector unsigned long long vec_bperm (vector unsigned long long,
31054 vector unsigned char)
31055 vector signed long long vec_mule (vector signed int,
31057 vector unsigned long long vec_mule (vector unsigned int,
31058 vector unsigned int)
31059 vector signed long long vec_mulo (vector signed int,
31061 vector unsigned long long vec_mulo (vector unsigned int,
31062 vector unsigned int)
31063 vector signed char vec_sldw (vector signed char,
31064 vector signed char,
31066 vector unsigned char vec_sldw (vector unsigned char,
31067 vector unsigned char,
31069 vector signed short vec_sldw (vector signed short,
31070 vector signed short,
31072 vector unsigned short vec_sldw (vector unsigned short,
31073 vector unsigned short,
31075 vector signed int vec_sldw (vector signed int,
31078 vector unsigned int vec_sldw (vector unsigned int,
31079 vector unsigned int,
31081 vector signed long long vec_sldw (vector signed long long,
31082 vector signed long long,
31084 vector unsigned long long vec_sldw (vector unsigned long long,
31085 vector unsigned long long,
31087 * config/rs6000/rs6000-c: Add support for built-in functions
31088 * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31089 * config/rs6000/altivec.h: Add defintion for vec_sldw.
31090 * doc/extend.texi: Update the built-in documentation for the
31091 new built-in functions.
31093 2017-05-16 Marek Polacek <polacek@redhat.com>
31097 * tree.c (save_expr): Don't fold the expression.
31099 2017-05-16 Uros Bizjak <ubizjak@gmail.com>
31101 * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31102 to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31103 and (?*y,m). Update insn attributes.
31105 2017-05-16 Martin Liska <mliska@suse.cz>
31107 * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31108 flags argument of print_gimple_stmt, print_gimple_expr,
31109 print_generic_stmt and print_generic_expr.
31110 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31111 * coretypes.h: Likewise.
31112 * except.c (dump_eh_tree): Likewise.
31113 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31114 * gimple-pretty-print.h: Likewise.
31115 * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31116 (backprop::push_to_worklist): Likewise.
31117 (backprop::pop_from_worklist): Likewise.
31118 (backprop::process_use): Likewise.
31119 (backprop::intersect_uses): Likewise.
31120 (note_replacement): Likewise.
31121 * gimple-ssa-store-merging.c
31122 (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31123 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31124 (pass_store_merging::execute): Likewise.
31125 * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31126 (ssa_base_cand_dump_callback): Likewise.
31127 (dump_incr_vec): Likewise.
31128 (replace_refs): Likewise.
31129 (replace_mult_candidate): Likewise.
31130 (create_add_on_incoming_edge): Likewise.
31131 (create_phi_basis): Likewise.
31132 (insert_initializers): Likewise.
31133 (all_phi_incrs_profitable): Likewise.
31134 (introduce_cast_before_cand): Likewise.
31135 (replace_one_candidate): Likewise.
31136 * gimplify.c (gimplify_expr): Likewise.
31137 * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31138 (set_rename): Likewise.
31139 (rename_uses): Likewise.
31140 (copy_loop_phi_nodes): Likewise.
31141 (add_close_phis_to_merge_points): Likewise.
31142 (copy_loop_close_phi_args): Likewise.
31143 (copy_cond_phi_args): Likewise.
31144 (graphite_copy_stmts_from_block): Likewise.
31145 (translate_pending_phi_nodes): Likewise.
31146 * graphite-poly.c (print_pdr): Likewise.
31147 (dump_gbb_cases): Likewise.
31148 (dump_gbb_conditions): Likewise.
31149 (print_scop_params): Likewise.
31150 * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31151 (build_cross_bb_scalars_use): Likewise.
31152 (gather_bbs::before_dom_children): Likewise.
31153 * hsa-dump.c (dump_hsa_immed): Likewise.
31154 * ipa-cp.c (print_ipcp_constant_value): Likewise.
31155 (get_replacement_map): Likewise.
31156 * ipa-inline-analysis.c (dump_condition): Likewise.
31157 (estimate_function_body_sizes): Likewise.
31158 * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31159 (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31160 * ipa-prop.c (ipa_dump_param): Likewise.
31161 (ipa_print_node_jump_functions_for_edge): Likewise.
31162 (ipa_modify_call_arguments): Likewise.
31163 (ipa_modify_expr): Likewise.
31164 (ipa_dump_param_adjustments): Likewise.
31165 (ipa_dump_agg_replacement_values): Likewise.
31166 (ipcp_modif_dom_walker::before_dom_children): Likewise.
31167 * ipa-pure-const.c (check_stmt): Likewise.
31168 (pass_nothrow::execute): Likewise.
31169 * ipa-split.c (execute_split_functions): Likewise.
31170 * omp-offload.c (dump_oacc_loop_part): Likewise.
31171 (dump_oacc_loop): Likewise.
31172 * trans-mem.c (tm_log_emit): Likewise.
31173 (tm_memopt_accumulate_memops): Likewise.
31174 (dump_tm_memopt_set): Likewise.
31175 (dump_tm_memopt_transform): Likewise.
31176 * tree-cfg.c (gimple_verify_flow_info): Likewise.
31177 (print_loop): Likewise.
31178 * tree-chkp-opt.c (chkp_print_addr): Likewise.
31179 (chkp_gather_checks_info): Likewise.
31180 (chkp_get_check_result): Likewise.
31181 (chkp_remove_check_if_pass): Likewise.
31182 (chkp_use_outer_bounds_if_possible): Likewise.
31183 (chkp_reduce_bounds_lifetime): Likewise.
31184 * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31185 (chkp_mark_completed_bounds): Likewise.
31186 (chkp_register_incomplete_bounds): Likewise.
31187 (chkp_mark_invalid_bounds): Likewise.
31188 (chkp_maybe_copy_and_register_bounds): Likewise.
31189 (chkp_build_returned_bound): Likewise.
31190 (chkp_get_bound_for_parm): Likewise.
31191 (chkp_build_bndldx): Likewise.
31192 (chkp_get_bounds_by_definition): Likewise.
31193 (chkp_generate_extern_var_bounds): Likewise.
31194 (chkp_get_bounds_for_decl_addr): Likewise.
31195 * tree-chrec.c (chrec_apply): Likewise.
31196 * tree-data-ref.c (dump_data_reference): Likewise.
31197 (dump_subscript): Likewise.
31198 (dump_data_dependence_relation): Likewise.
31199 (analyze_overlapping_iterations): Likewise.
31200 * tree-inline.c (expand_call_inline): Likewise.
31201 (tree_function_versioning): Likewise.
31202 * tree-into-ssa.c (dump_defs_stack): Likewise.
31203 (dump_currdefs): Likewise.
31204 (dump_names_replaced_by): Likewise.
31205 (dump_update_ssa): Likewise.
31206 (update_ssa): Likewise.
31207 * tree-object-size.c (pass_object_sizes::execute): Likewise.
31208 * tree-parloops.c (build_new_reduction): Likewise.
31209 (try_create_reduction_list): Likewise.
31210 (ref_conflicts_with_region): Likewise.
31211 (oacc_entry_exit_ok_1): Likewise.
31212 (oacc_entry_exit_single_gang): Likewise.
31213 * tree-pretty-print.h: Likewise.
31214 * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31215 (get_scalar_evolution): Likewise.
31216 (add_to_evolution): Likewise.
31217 (get_loop_exit_condition): Likewise.
31218 (analyze_evolution_in_loop): Likewise.
31219 (analyze_initial_condition): Likewise.
31220 (analyze_scalar_evolution): Likewise.
31221 (instantiate_scev): Likewise.
31222 (number_of_latch_executions): Likewise.
31223 (gather_chrec_stats): Likewise.
31224 (final_value_replacement_loop): Likewise.
31225 (scev_const_prop): Likewise.
31226 * tree-sra.c (dump_access): Likewise.
31227 (disqualify_candidate): Likewise.
31228 (create_access): Likewise.
31229 (reject): Likewise.
31230 (maybe_add_sra_candidate): Likewise.
31231 (create_access_replacement): Likewise.
31232 (analyze_access_subtree): Likewise.
31233 (analyze_all_variable_accesses): Likewise.
31234 (sra_modify_assign): Likewise.
31235 (initialize_constant_pool_replacements): Likewise.
31236 (find_param_candidates): Likewise.
31237 (decide_one_param_reduction): Likewise.
31238 (replace_removed_params_ssa_names): Likewise.
31239 * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31240 * tree-ssa-copy.c (dump_copy_of): Likewise.
31241 (copy_prop_visit_cond_stmt): Likewise.
31242 * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31243 * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31244 (record_equivalences_from_stmt): Likewise.
31245 * tree-ssa-dse.c (compute_trims): Likewise.
31246 (delete_dead_call): Likewise.
31247 (delete_dead_assignment): Likewise.
31248 * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31249 (forward_propagate_into_cond): Likewise.
31250 (pass_forwprop::execute): Likewise.
31251 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31252 * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31254 (move_computations_worker): Likewise.
31255 (execute_sm): Likewise.
31256 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31257 (remove_exits_and_undefined_stmts): Likewise.
31258 (remove_redundant_iv_tests): Likewise.
31259 * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31260 (adjust_iv_update_pos): Likewise.
31261 * tree-ssa-math-opts.c (bswap_replace): Likewise.
31262 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31263 (value_replacement): Likewise.
31264 * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31265 * tree-ssa-pre.c (print_pre_expr): Likewise.
31266 (get_representative_for): Likewise.
31267 (create_expression_by_pieces): Likewise.
31268 (insert_into_preds_of_block): Likewise.
31269 (eliminate_insert): Likewise.
31270 (eliminate_dom_walker::before_dom_children): Likewise.
31271 (eliminate): Likewise.
31272 (remove_dead_inserted_code): Likewise.
31273 * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31274 * tree-ssa-reassoc.c (get_rank): Likewise.
31275 (eliminate_duplicate_pair): Likewise.
31276 (eliminate_plus_minus_pair): Likewise.
31277 (eliminate_not_pairs): Likewise.
31278 (undistribute_ops_list): Likewise.
31279 (eliminate_redundant_comparison): Likewise.
31280 (update_range_test): Likewise.
31281 (optimize_range_tests_var_bound): Likewise.
31282 (optimize_vec_cond_expr): Likewise.
31283 (rewrite_expr_tree): Likewise.
31284 (rewrite_expr_tree_parallel): Likewise.
31285 (linearize_expr): Likewise.
31286 (break_up_subtract): Likewise.
31287 (linearize_expr_tree): Likewise.
31288 (attempt_builtin_powi): Likewise.
31289 (attempt_builtin_copysign): Likewise.
31290 (transform_stmt_to_copy): Likewise.
31291 (transform_stmt_to_multiply): Likewise.
31292 (dump_ops_vector): Likewise.
31293 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31294 (print_scc): Likewise.
31295 (set_ssa_val_to): Likewise.
31296 (visit_reference_op_store): Likewise.
31297 (visit_use): Likewise.
31298 (sccvn_dom_walker::before_dom_children): Likewise.
31299 (run_scc_vn): Likewise.
31300 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31302 (expr_hash_elt::print): Likewise.
31303 (const_and_copies::pop_to_marker): Likewise.
31304 (const_and_copies::record_const_or_copy_raw): Likewise.
31305 * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31306 * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31307 (dump_predicates): Likewise.
31308 (find_uninit_use): Likewise.
31309 (warn_uninitialized_phi): Likewise.
31310 (pass_late_warn_uninitialized::execute): Likewise.
31311 * tree-ssa.c (verify_vssa): Likewise.
31312 (verify_ssa): Likewise.
31313 (maybe_optimize_var): Likewise.
31314 * tree-vrp.c (dump_value_range): Likewise.
31315 (dump_all_value_ranges): Likewise.
31316 (dump_asserts_for): Likewise.
31317 (register_edge_assert_for_2): Likewise.
31318 (vrp_visit_cond_stmt): Likewise.
31319 (vrp_visit_switch_stmt): Likewise.
31320 (vrp_visit_stmt): Likewise.
31321 (vrp_visit_phi_node): Likewise.
31322 (simplify_cond_using_ranges_1): Likewise.
31323 (fold_predicate_in): Likewise.
31324 (evrp_dom_walker::before_dom_children): Likewise.
31325 (evrp_dom_walker::push_value_range): Likewise.
31326 (evrp_dom_walker::pop_value_range): Likewise.
31327 (execute_early_vrp): Likewise.
31329 2017-05-16 Richard Biener <rguenther@suse.de>
31331 * dwarf2out.c (loc_list_from_tree_1): Do not create
31332 DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31334 2017-05-16 Richard Biener <rguenther@suse.de>
31336 * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31338 (note_variable_value_in_expr): If we resolved the decl ref
31339 do not push to the stack.
31341 2017-05-16 Matthew Wahab <matthew.wahab@arm.com>
31343 * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31344 operations in fast-math mode.
31345 (vaddq_f16): Likewise.
31346 (vmul_f16): Likewise.
31347 (vmulq_f16): Likewise.
31348 (vsub_f16): Likewise.
31349 (vsubq_f16): Likewise.
31350 * config/arm/neon.md (add<mode>3): New.
31352 (fma:<VH:mode>3): New. Also remove outdated comment.
31355 2017-05-16 Martin Liska <mliska@suse.cz>
31359 * ipa-devirt.c (warn_types_mismatch): Fix typo.
31360 (odr_types_equivalent_p): Likewise.
31362 2017-05-15 Sylvestre Ledru <sylvestre@debian.org>
31364 * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31366 2017-05-15 Uros Bizjak <ubizjak@gmail.com>
31369 * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31370 non-interunit SSE move alternatives with '?'.
31371 (zero-extendsidi peephole2): New peephole to skip intermediate
31372 general register in SSE zero-extend sequence.
31374 2017-05-15 Jeff Law <law@redhat.com>
31376 * reorg.c (relax_delay_slots): Create a new variable to hold
31377 the temporary target rather than clobbering TARGET_LABEL.
31379 * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31380 missing argument to extract_bit_field call.
31381 * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31383 2017-05-15 Martin Liska <mliska@suse.cz>
31386 * gcc.c (process_command): Do not allow empty argument of -o option.
31388 2017-05-15 Renlin Li <renlin.li@arm.com>
31390 * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31391 * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31392 * config/aarch64/constraints.md (Usf): Add long call check.
31393 * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31394 (call_value): Likewise.
31395 (sibcall): Likewise.
31396 (sibcall_value): Likewise.
31398 (call_value_insn): New.
31399 (sibcall_insn): Update rtx pattern.
31400 (sibcall_value_insn): Likewise.
31401 (call_internal): Remove.
31402 (call_value_internal): Likewise.
31403 (sibcall_internal): Likewise.
31404 (sibcall_value_internal): Likewise.
31405 (call_reg): Likewise.
31406 (call_symbol): Likewise.
31407 (call_value_reg): Likewise.
31408 (call_value_symbol): Likewise.
31410 2017-05-14 Krister Walfridsson <krister.walfridsson@gmail.com>
31413 * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31415 2017-05-14 Uros Bizjak <ubizjak@gmail.com>
31417 * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31418 compatible with CCGOCmode and with CCZmode.
31420 2017-05-14 Martin Sebor <msebor@redhat.com>
31422 PR middle-end/77671
31423 * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31424 (gimple_fold_builtin_snprintf): Same.
31425 * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31426 (gimple_fold_builtin_snprintf): Same.
31427 * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31428 of character types.
31429 (is_call_safe): New function.
31430 (try_substitute_return_value): Call it.
31431 (try_simplify_call): New function.
31432 (pass_sprintf_length::handle_gimple_call): Call it.
31434 2017-05-14 Martin Sebor <msebor@redhat.com>
31436 PR middle-end/80669
31437 * builtins.c (expand_builtin_stpncpy): Simplify.
31439 2017-05-14 Daniel Santos <daniel.santos@pobox.com>
31441 * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31442 * config/i386/i386.h
31443 (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31444 (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31445 (struct machine_function): Add new members call_ms2sysv,
31446 call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31447 (struct machine_frame_state): New fields sp_realigned and
31448 sp_realigned_offset.
31449 * config/i386/i386.c
31450 (enum xlogue_stub): New enum.
31451 (enum xlogue_stub_sets): New enum.
31452 (class xlogue_layout): New class.
31453 (struct ix86_frame): New fields stack_realign_allocate_offset,
31454 stack_realign_offset and outlined_save_offset. Modify comments to
31455 detail stack layout when using out-of-line stubs.
31456 (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31457 (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31458 -mcall-ms2sysv-xlogues.
31459 (stub_managed_regs): New static variable.
31460 (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31461 registers managed by out-of-line stub.
31462 (disable_call_ms2sysv_xlogues): New function.
31463 (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31464 m->call_ms2sysv when appropriate and compute frame layout for
31466 (sp_valid_at, fp_valid_at): New inline functions.
31467 (choose_basereg): New function.
31468 (choose_baseaddr): Add align parameter, use choose_basereg and modify
31470 (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31471 Use align parameter of choose_baseaddr to generated aligned SSE movs
31473 (pro_epilogue_adjust_stack): Modify to track
31474 machine_frame_state::sp_realigned.
31475 (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31476 (ix86_nsaved_sseregs): Likewise.
31477 (ix86_emit_save_regs): Likewise.
31478 (ix86_emit_save_regs_using_mov): Likewise.
31479 (ix86_emit_save_sse_regs_using_mov): Likewise.
31480 (get_scratch_register_on_entry): Likewise.
31481 (gen_frame_set): New function.
31482 (gen_frame_load): Likewise.
31483 (gen_frame_store): Likewise.
31484 (emit_outlined_ms2sysv_save): Likewise.
31485 (emit_outlined_ms2sysv_restore): Likewise.
31486 (ix86_expand_prologue): Modify stack re-alignment code and call
31487 emit_outlined_ms2sysv_save when appropriate.
31488 (ix86_emit_leave): Clear machine_frame_state::sp_realigned. Add
31489 parameter rtx_insn *insn, which allows the function to be used to only
31490 generate the notes.
31491 (ix86_expand_epilogue): Modify validity checks of frame and stack
31492 pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31493 (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31494 * config/i386/predicates.md
31495 (save_multiple): New predicate.
31496 (restore_multiple): Likewise.
31497 * config/i386/sse.md
31498 (save_multiple<mode>): New pattern.
31499 (save_multiple_realign<mode>): Likewise.
31500 (restore_multiple<mode>): Likewise.
31501 (restore_multiple_and_return<mode>): Likewise.
31502 (restore_multiple_leave_return<mode>): Likewise.
31503 * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31505 2017-05-14 Julia Koval <julia.koval@intel.com>
31507 * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31508 * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31509 (__builtin_ia32_xsetbv): New builtins.
31510 * config/i386/i386.c (ix86_expand_special_args_builtin):
31512 (ix86_expand_builtin): Special expand for new intrinsics.
31513 * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31514 (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31515 * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31517 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31519 * cfganal.c (inverted_post_order_compute): Change argument type
31521 * cfganal.h (inverted_post_order_compute): Adjust prototype.
31522 * df-core.c (rest_of_handle_df_initialize): Adjust.
31523 (rest_of_handle_df_finish): Likewise.
31524 (df_analyze_1): Likewise.
31525 (df_analyze): Likewise.
31526 (loop_inverted_post_order_compute): Change argument to be a vec *.
31527 (df_analyze_loop): Adjust.
31528 (df_get_n_blocks): Likewise.
31529 (df_get_postorder): Likewise.
31530 * df.h (struct df_d): Change field to be a vec.
31531 * lcm.c (compute_laterin): Adjust.
31532 (compute_available): Likewise.
31533 * lra-lives.c (lra_create_live_ranges_1): Likewise.
31534 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31535 * tree-ssa-pre.c (compute_antic): Likewise.
31537 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31539 * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31540 (depth_first_search::depth_first_search): Change structure init
31541 function to this constructor.
31542 (depth_first_search::add_bb): Rename function to this member.
31543 (depth_first_search::execute): Likewise.
31544 (flow_dfs_compute_reverse_finish): Adjust.
31546 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31548 * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31549 (longest_simple_path): Likewise.
31550 * shrink-wrap.c (spread_components): Likewise.
31551 (disqualify_problematic_components): Likewise.
31552 (emit_common_heads_for_components): Likewise.
31553 (emit_common_tails_for_components): Likewise.
31554 (insert_prologue_epilogue_for_components): Likewise.
31556 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31558 * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31561 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31563 * df-core.c (df_set_blocks): Start using auto_bitmap.
31564 (df_compact_blocks): Likewise.
31565 * df-problems.c (df_rd_confluence_n): Likewise.
31566 * df-scan.c (df_insn_rescan_all): Likewise.
31567 (df_process_deferred_rescans): Likewise.
31568 (df_update_entry_block_defs): Likewise.
31569 (df_update_exit_block_uses): Likewise.
31570 (df_entry_block_bitmap_verify): Likewise.
31571 (df_exit_block_bitmap_verify): Likewise.
31572 (df_scan_verify): Likewise.
31573 * lra-constraints.c (lra_constraints): Likewise.
31574 (undo_optional_reloads): Likewise.
31575 (lra_undo_inheritance): Likewise.
31576 * lra-remat.c (calculate_gen_cands): Likewise.
31577 (do_remat): Likewise.
31578 * lra-spills.c (assign_spill_hard_regs): Likewise.
31579 (spill_pseudos): Likewise.
31580 * tree-ssa-pre.c (bitmap_set_and): Likewise.
31581 (bitmap_set_subtract_values): Likewise.
31583 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31585 * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31586 management with auto_bitmap.
31587 (fix_inter_tick): Likewise.
31588 (fix_recovery_deps): Likewise.
31589 * ira.c (add_store_equivs): Likewise.
31590 (find_moveable_pseudos): Likewise.
31591 (split_live_ranges_for_shrink_wrap): Likewise.
31592 * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31593 (rtx_reuse_manager::seen_def_p): Likewise.
31594 (rtx_reuse_manager::set_seen_def): Likewise.
31595 * print-rtl.h (class rtx_reuse_manager): Likewise.
31597 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31599 * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31601 (migrate_btr_def): Likewise.
31602 * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31603 * df-core.c (loop_post_order_compute): Likewise.
31604 (loop_inverted_post_order_compute): Likewise.
31605 * hsa-common.h: Likewise.
31606 * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31607 * init-regs.c (initialize_uninitialized_regs): Likewise.
31608 * ipa-inline.c (resolve_noninline_speculation): Likewise.
31609 (inline_small_functions): Likewise.
31610 * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31611 * ira.c (combine_and_move_insns): Likewise.
31612 (build_insn_chain): Likewise.
31613 * loop-invariant.c (find_invariants): Likewise.
31614 * lower-subreg.c (propagate_pseudo_copies): Likewise.
31615 * predict.c (tree_predict_by_opcode): Likewise.
31616 (predict_paths_leading_to): Likewise.
31617 (predict_paths_leading_to_edge): Likewise.
31618 (estimate_loops_at_level): Likewise.
31619 (estimate_loops): Likewise.
31620 * shrink-wrap.c (try_shrink_wrapping): Likewise.
31621 (spread_components): Likewise.
31622 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31623 * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31624 * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31625 * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31626 * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31627 * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31628 * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31629 * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31630 (mark_threaded_blocks): Likewise.
31631 (thread_through_all_blocks): Likewise.
31632 * tree-ssa.c (verify_ssa): Likewise.
31633 (execute_update_addresses_taken): Likewise.
31634 * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31636 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31638 * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31640 (post_order_compute): Likewise.
31641 (inverted_post_order_compute): Likewise.
31642 (pre_and_rev_post_order_compute_fn): Likewise.
31644 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31646 * genrecog.c (int_set::int_set): Explicitly construct our
31647 auto_vec base class.
31648 * vec.h (auto_vec::auto_vec): New constructor.
31650 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31652 * bitmap.h (class auto_bitmap): New constructor taking
31653 bitmap_obstack * argument.
31655 2017-05-13 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
31657 * bitmap.h (class auto_bitmap): Change type of m_bits to
31658 bitmap_head, and adjust ctor / dtor and member operators.
31660 2017-05-13 Uros Bizjak <ubizjak@gmail.com>
31662 * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31663 when returned register mode doesn't match original mode.
31665 2017-05-12 Jeff Law <law@redhat.com>
31666 Jakub Jelinek <jakub@redhat.com>
31668 * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31669 we look for cc setter after the compare-elim changes.
31670 * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31671 within the vector to match what compare-elim now expects.
31672 (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31673 (xorsi3_flags, one_cmplsi2_flags): Likewise.
31675 * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31676 after the compare-elim changes.
31677 * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31678 the vector to match what compare-elim now expects.
31679 (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31680 (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31681 (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31682 (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31683 (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31685 * config/visium/visium.c (single_set_and_flags): Fix where
31686 we look for cc setter after the compare-elim changes.
31687 * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31688 with the vector to match what compare-elim now expects.
31689 (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31690 (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31691 (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31692 (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31693 (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31694 (neg<mode>2_insn_set_overflow): Likewise.
31696 2017-05-12 Jim Wilson <jim.wilson@linaro.org>
31698 PR middle-end/79794
31699 * expmed.c (extract_bit_field_1): Add alt_rtl argument. Before
31700 maybe_expand_insn call, set ops[0].target. If still set after call,
31701 set alt_rtl. Add extra arg to recursive calls.
31702 (extract_bit_field): Add alt_rtl argument. Pass to
31704 * expmed.h (extract_bit_field): Fix prototype.
31705 * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31706 (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31707 to extract_bit_field_calls.
31708 (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31709 Pass alt_rtl to extract_bit_field calls.
31710 * calls.c (store_unaligned_arguments_into_psuedos)
31711 load_register_parameters): Pass extra NULL to extract_bit_field calls.
31712 * optabs.c (maybe_legitimize_operand): Clear op->target when call
31714 * optabs.h (struct expand_operand): Add target bitfield.
31716 2017-05-12 Uros Bizjak <ubizjak@gmail.com>
31718 * compare-elim.c (try_eliminate_compare): Canonicalize
31719 operation with embedded compare to
31720 [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31721 (set (reg) (operation)].
31723 * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31725 2017-05-12 Uros Bizjak <ubizjak@gmail.com>
31728 * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31729 cost of adding a carry flag for ADC instruction.
31730 [case MINUS]: Ignore the cost of subtracting a carry flag
31731 for SBB instruction.
31733 2017-05-12 Steven Munroe <munroesj@gcc.gnu.org>
31735 * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31737 * config/rs6000/bmiintrin.h: New file.
31738 * config/rs6000/bmi2intrin.h: New file.
31739 * config/rs6000/x86intrin.h: New file.
31741 2017-05-12 Jeff Law <law@redhat.com>
31743 * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31746 2017-05-12 Peter Bergner <bergner@vnet.ibm.com>
31748 PR middle-end/80707
31749 * tree-cfg.c: Remove cfg edges of unreachable case statements.
31751 2017-05-12 Will Schmidt <will_schmidt@vnet.ibm.com>
31753 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31754 early expansion of vector divide builtins.
31755 (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31756 builtins identified as having unsigned arguments.
31758 2017-05-12 Will Schmidt <will_schmidt@vnet.ibm.com>
31760 * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31761 (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31762 expansion of vector logical operations (and, andc, or, xor,
31765 2017-05-12 Will Schmidt <will_schmidt@vnet.ibm.com>
31767 * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31768 * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31770 2017-05-12 Will Schmidt <will_schmidt@vnet.ibm.com>
31772 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31773 early GIMPLE expansion of vector multiplies.
31775 2017-05-12 Prakhar Bahuguna <prakhar.bahuguna@arm.com>
31777 * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31778 TARGET_HAVE_MOVT conditional.
31779 (movt splitter): Likewise.
31781 2017-05-12 Richard Biener <rguenther@suse.de>
31783 * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31784 * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31785 Fold all stmts not inplace.
31787 2017-05-12 Richard Biener <rguenther@suse.de>
31789 PR tree-optimization/80713
31790 * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31791 inserted_exprs bit for not removed stmts.
31793 2017-05-12 Thomas Schwinge <thomas@codesourcery.com>
31795 PR middle-end/69921
31796 * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31797 parallelized" attribute for parallelized OpenACC kernels.
31798 * omp-offload.c (execute_oacc_device_lower): Use it.
31800 * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31801 Set "oacc kernels" attribute.
31802 * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31803 parameter. Adjust all users.
31804 (oacc_fn_attrib_kernels_p): Remove function.
31805 * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31806 kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31807 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31808 * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31809 assert "oacc kernels" attribute is set.
31811 2017-05-11 Carl Love <cel@us.ibm.com>
31813 * config/rs6000/rs6000-c: Add support for built-in functions
31814 vector unsigned char vec_popcnt (vector signed char)
31815 vector unsigned char vec_popcnt (vector unsigned char)
31816 vector unsigned short vec_popcnt (vector signed short)
31817 vector unsigned short vec_popcnt (vector unsigned short)
31818 vector unsigned int vec_popcnt (vector signed int)
31819 vector unsigned int vec_popcnt (vector unsigned int)
31820 vector unsigned long long vec_popcnt (vector signed long long)
31821 vector unsigned long long vec_popcnt (vector unsigned long long)
31822 vector signed long long vec_slo (vector signed long long,
31823 vector signed char)
31824 vector signed long long vec_slo (vector signed long long,
31825 vector unsigned char)
31826 vector unsigned long long vec_slo (vector unsigned long long,
31827 vector signed char)
31828 vector unsigned long long vec_slo (vector unsigned long long,
31829 vector unsigned char)
31830 * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31831 VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31832 * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31833 vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31834 * doc/extend.texi: Update the built-in documentation file for the
31835 new built-in functions.
31837 2017-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
31839 * attribs.h (sorted_attr_string): Move machine independent
31840 functions for target clone support from the i386 port to common
31841 code. Rename ix86_function_versions to common_function_versions.
31842 Rename make_name to make_unique_name.
31843 (common_function_versions): Likewise.
31844 (make_unique_name): Likewise.
31845 (make_dispatcher_decl): Likewise.
31846 (is_function_default_version): Likewise.
31847 * attribs.c (attr_strcmp): Likewise.
31848 (sorted_attr_string): Likewise.
31849 (common_function_versions): Likewise.
31850 (make_unique_name): Likewise.
31851 (make_dispatcher_decl): Likewise.
31852 (is_function_default_version): Likewise.
31853 * config/i386/i386.c (attr_strcmp): Likewise.
31854 (sorted_attr_string): Likewise.
31855 (ix86_function_versions): Likewise.
31856 (make_name): Likewise.
31857 (make_dispatcher_decl): Likewise.
31858 (is_function_default_version): Likewise.
31859 (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
31861 2017-05-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31864 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
31865 Account for direct move costs for vec_construct of integer
31868 2017-05-11 Uros Bizjak <ubizjak@gmail.com>
31871 * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
31872 (UNSPEC_STX_ATOMIC): Ditto.
31873 (loaddi_via_sse): New insn.
31874 (storedi_via_sse): Ditto.
31875 (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
31876 Update corresponding peephole2 patterns.
31877 (atomic_storedi_fpu): Ditto.
31879 2017-05-11 Julia Koval <julia.koval@intel.com>
31881 * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
31882 (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
31884 * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
31885 (__builtin_ia32_rsqrt14ss_mask): New builtins.
31886 * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
31888 2017-05-11 Nathan Sidwell <nathan@acm.org>
31890 * graphite-poly.c: Include dumpfile.h.
31892 * dumpfle.h (dump_function): Declare here ...
31893 * tree-dump.h (dump_function): ... not here.
31894 * dumpfile.c: #include tree-cfg.h.
31895 (dump_function): Move here from ...
31896 * tree-dump.c (dump_function): ... here.
31897 * gimplify.c: #include splay-tree.h, not tree-dump.h.
31898 * graphite-poly.c: Don't include tree-dump.h.
31899 * cgraphclones.c: Include dumpfile.h not tree-dump.h.
31900 * print-tree.c: Likewise.
31901 * stor-layout.c: Likewise.
31902 * tree-nested.c: Likewise.
31904 * dumpfile.c (dump_start): Use TDF_FLAGS.
31905 (dump_enable_all): Fix TDF_KIND check thinko.
31907 2017-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
31909 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
31910 array entries to represent two legal parameterizations of the
31911 overloaded __builtin_cmpb function, as represented by the
31912 P6_OV_BUILTIN_CMPB constant.
31913 (altivec_resolve_overloaded_builtin): Add special case handling
31914 for the __builtin_cmpb function, as represented by the
31915 P6_OV_BUILTIN_CMPB constant.
31916 * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
31917 (BU_P6_64BIT_2): New macro.
31918 (BU_P6_OVERLOAD_2): New macro
31919 (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
31920 (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
31921 (CMPB): Add overload support to represent both 32-bit and 64-bit
31922 compare-bytes function.
31923 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
31924 support for TARGET_CMPB.
31925 * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
31926 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
31927 documentation of the __builtin_cmpb overloaded built-in function.
31929 2017-05-11 Richard Biener <rguenther@suse.de>
31931 PR tree-optimization/80705
31932 * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
31933 bases are not vectorizable.
31935 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31937 * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
31938 when counting register pressure.
31940 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31942 * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
31943 (iv_ca_more_deps): Renamed to ...
31944 (iv_ca_compare_deps): ... this.
31945 (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
31947 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31949 * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
31951 (determine_group_iv_costs): ... here.
31952 (find_inv_vars_cb): Record inv var if it's not recorded before.
31954 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31956 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
31957 (get_shiftadd_cost): Ditto.
31959 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31961 * tree-ssa-address.c: Include header file.
31962 (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
31964 (add_to_parts): Refactor.
31965 (addr_to_parts): New parameter. Update use of move_hint_to_base.
31966 (create_mem_ref): Update use of addr_to_parts. Re-associate addr
31969 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31971 PR tree-optimization/53090
31972 * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
31974 (extract_cond_operands): Detect condition with IV on both sides
31975 and return COMP_IV_EXPR_2.
31976 (find_interesting_uses_cond): Add iv_use for both IVs in condition.
31977 (rewrite_use_compare): Simplify by removing call to function
31978 extract_cond_operands.
31980 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31982 * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
31983 (extract_cond_operands): Detect condition comparing against non-
31984 invariant bound and return appropriate enum value.
31985 (find_interesting_uses_cond): Update use of extract_cond_operands.
31986 Handle its return value accordingly.
31987 (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
31989 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31991 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
31992 nonlinear iv_use computation in loop invariant sensitive way.
31994 2017-05-11 Bin Cheng <bin.cheng@arm.com>
31996 * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
31997 (find_iv_candidates): Call relate_compare_use_with_all_cands.
31999 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32001 * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32002 (dump_cand): Support iv_cand.inv_exprs.
32003 (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32005 (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32008 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32010 * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32012 * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32013 as local function. Include necessary header files.
32014 * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32016 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32018 * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32020 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32022 * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32023 operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32024 RSHIFT_EXPR and BIT_NOT_EXPR.
32026 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32028 * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32029 (adjust_setup_cost): New parameter supporting round up adjustment.
32030 (struct address_cost_data): Delete.
32031 (force_expr_to_var_cost): Don't bound cost with spill_cost.
32032 (split_address_cost, ptr_difference_cost): Delete.
32033 (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32034 (struct ainc_cost_data): New struct.
32035 (get_address_cost_ainc): New function.
32036 (get_address_cost, get_computation_cost): Reimplement.
32037 (determine_group_iv_cost_address): Record inv_expr for all uses of
32039 (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32040 (iv_ca_has_deps): Reimplemented to ...
32041 (iv_ca_more_deps): ... this. Check if NEW_CP introduces more deps
32043 (iv_ca_extend): Call iv_ca_more_deps.
32045 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32047 * tree-ssa-address.c (struct mem_address): Move to header file.
32048 (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32049 * tree-ssa-address.h (struct mem_address): Move from C file.
32050 (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32052 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32054 * tree-affine.h (aff_combination_type): New interface.
32055 (aff_combination_zero_p): Remove static.
32056 (aff_combination_const_p): New interface.
32057 (aff_combination_singleton_var_p): New interfaces.
32059 2017-05-11 Richard Biener <rguenther@suse.de>
32061 * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32062 Skip unreachable blocks and destinations.
32063 (eliminate): Move stmt removal and fixup ...
32064 (fini_eliminate): ... here. Skip inserted exprs.
32065 (pass_pre::execute): Move fini_pre after fini_eliminate.
32066 * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32067 (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32068 PRE to get rid of dead code that has invalid SSA form and
32069 split critical edges again.
32071 2017-05-11 Bin Cheng <bin.cheng@arm.com>
32073 * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32075 2017-05-11 Richard Biener <rguenther@suse.de>
32077 * passes.c (execute_function_todo): Verify loops if they are
32078 said to be up-to-date.
32079 * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32080 * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32082 2017-05-10 John David Anglin <danglin@gcc.gnu.org>
32085 * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32086 handle calling assemble_external ourself.
32089 * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32090 modes with zero size. Enhance comment.
32092 2017-05-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
32094 * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32095 built-ins for vec_xl and vec_xst with short and char pointer
32098 2017-05-10 Sebastian Peryt <sebastian.peryt@intel.com>
32100 * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32101 (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32102 (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32103 (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32104 (_mm_maskz_min_round_ss): New intrinsics.
32105 * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32106 (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32107 * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32108 (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32109 (__builtin_ia32_minss_mask_round): New builtins.
32110 * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32111 (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32112 * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32114 (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32115 (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32117 (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32120 2017-05-10 Sebastian Peryt <sebastian.peryt@intel.com>
32122 * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32123 (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32124 (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32125 (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32126 (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32127 (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32128 (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32129 * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32130 (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32131 * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32132 (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32133 (__builtin_ia32_mulss_mask_round): New builtins.
32134 * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32135 (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32136 * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32138 (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32139 (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32141 (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32144 2017-05-10 Julia Koval <julia.koval@intel.com>
32146 * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32147 (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32148 (_mm256_setr_m128i): New intrinsics.
32150 2017-05-10 Julia Koval <julia.koval@intel.com>
32152 * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32153 (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32154 (_mm_maskz_rcp14_ss): New intrinsics.
32155 * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32156 (__builtin_ia32_rcp14ss_mask): New builtins.
32157 * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32159 2017-05-10 Peter Bergner <bergner@vnet.ibm.com>
32161 PR tree-optimization/51513
32162 * tree-cfg.c (gimple_seq_unreachable_p): New function.
32163 (assert_unreachable_fallthru_edge_p): Use it.
32164 (group_case_labels_stmt): Likewise.
32165 * tree-cfg.h: Prototype it.
32166 * stmt.c: Include cfghooks.h and tree-cfg.h.
32167 (emit_case_dispatch_table) <gap_label>: New local variable.
32168 Use it to fill dispatch table gaps.
32169 Test for default_label before updating probabilities.
32170 (expand_case) <default_label>: Remove unneeded initialization.
32171 Test for unreachable default case statement and remove its edge.
32172 Set default_label accordingly.
32173 * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32175 2017-05-10 Carl Love <cel@us.ibm.com>
32177 * config/rs6000/rs6000-c: Add support for built-in functions
32178 vector signed char vec_neg (vector signed char)
32179 vector signed short int vec_neg (vector short int)
32180 vector signed int vec_neg (vector signed int)
32181 vector signed long long vec_neg (vector signed long long)
32182 vector float vec_neg (vector float)
32183 vector double vec_neg (vector double)
32184 * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32186 * config/rs6000/altivec.h: Add define for vec_neg
32187 * doc/extend.texi: Update the built-in documentation for the
32188 new built-in functions.
32190 2017-05-10 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
32192 PR tree-optimization/77644
32193 * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32195 2017-05-10 Nathan Sidwell <nathan@acm.org>
32197 * dumpfile.h (TDI_lang_all): New.
32198 (TDF_KIND): New. Renumber others
32199 (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32201 * dumpfile.c (dump_files): Mark language dumps as TDF_LANG. add
32203 (get_dump_file_name): Adjust suffix generation.
32204 (dump_enable_all): Use TDF_KIND.
32205 * doc/invoke.texi (-fdump-lang-all): Document.
32207 * dumpfile.h: Tabify.
32209 2017-05-10 Wilco Dijkstra <wdijkstr@arm.com>
32212 * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32213 Move member access before delete.
32215 2017-05-10 Alexandre Oliva <aoliva@redhat.com>
32217 * tree-inline.c (expand_call_inline): Split block at stmt
32220 2017-05-09 Michael Meissner <meissner@linux.vnet.ibm.com>
32223 * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32224 are now unused after splitting mov{sf,sd}_hardfloat.
32225 (f32_lr2): Likewise.
32226 (f32_lm): Likewise.
32227 (f32_lm2): Likewise.
32228 (f32_li): Likewise.
32229 (f32_li2): Likewise.
32230 (f32_lv): Likewise.
32231 (f32_sr): Likewise.
32232 (f32_sr2): Likewise.
32233 (f32_sm): Likewise.
32234 (f32_sm2): Likewise.
32235 (f32_si): Likewise.
32236 (f32_si2): Likewise.
32237 (f32_sv): Likewise.
32238 (f32_dm): Likewise.
32239 (f32_vsx): Likewise.
32240 (f32_av): Likewise.
32241 (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32242 For movsf, order stores so the VSX stores occur before the GPR
32243 store which encourages the register allocator to use a traditional
32244 FPR instead of a GPR. For movsd, order the stores so that the GPR
32245 store comes before the VSX stores to allow the power6 to work.
32246 This is due to the power6 not having a 32-bit integer store
32247 instruction from a FPR.
32248 (movsf_hardfloat): Likewise.
32249 (movsd_hardfloat): Likewise.
32251 2017-05-09 Martin Sebor <msebor@redhat.com>
32253 PR translation/80280
32254 * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32257 PR translation/80280
32258 * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32259 data member added in r247778.
32260 (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32262 2017-05-09 Nathan Sidwell <nathan@acm.org>
32264 * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32266 * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32269 2017-05-09 Marek Polacek <polacek@redhat.com>
32271 * doc/invoke.texi: Fix typo.
32273 2017-05-09 Richard Biener <rguenther@suse.de>
32275 * tree-vrp.c (vrp_val_is_max): Adjust comment.
32276 (vrp_val_is_min): Likewise.
32277 (set_value_range_to_value): Likewise.
32278 (set_value_range_to_nonnegative): Likewise.
32279 (gimple_assign_nonzero_p): Likewise.
32280 (gimple_stmt_nonzero_p): Likewise.
32281 (vrp_int_const_binop): Likewise. Remove unreachable case.
32282 (adjust_range_with_scev): Adjust comments.
32283 (compare_range_with_value): Likewise.
32284 (extract_range_from_phi_node): Likewise.
32285 (test_for_singularity): Likewise.
32287 2017-05-09 Richard Biener <rguenther@suse.de>
32289 * tree-vrp.c (get_single_symbol): Add assert that we don't
32290 get overflowed constants as invariant part.
32291 (compare_values_warnv): Add comment before the TREE_NO_WARNING
32292 checks. Use wi::cmp instead of recursing for integer constants.
32293 (compare_values): Just ignore whether we assumed undefined
32294 overflow instead of failing the compare.
32295 (extract_range_for_var_from_comparison_expr): Add comment before the
32296 TREE_NO_WARNING sets.
32297 (test_for_singularity): Likewise.
32298 (extract_range_from_comparison): Do not disable optimization
32299 when we assumed undefined overflow.
32300 (extract_range_basic): Remove init of unused var.
32302 2017-05-09 Richard Biener <rguenther@suse.de>
32304 * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32305 (extract_range_from_multiplicative_op_1): Adjust.
32306 (extract_range_from_binary_expr_1): Use int_const_binop.
32308 2017-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
32311 * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32312 rs6000_store_data_bypass_p in seven define_bypass directives and
32313 in several comments.
32314 * config/rs6000/rs6000-protos.h: Add prototype for
32315 rs6000_store_data_bypass_p function.
32316 * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32317 function implements slightly different (rs6000-specific) semantics
32318 than store_data_bypass_p, returning false rather than aborting
32319 with assertion error when arguments do not satisfy the
32320 requirements of store data bypass.
32321 (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32322 rs6000_store_data_bypass_p.
32324 2017-05-08 Max Filippov <jcmvbkbc@gmail.com>
32326 * config/xtensa/xtensa-protos.h
32327 (xtensa_initial_elimination_offset): New declaration.
32328 * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32329 New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32330 macro definition, add case for FRAME_POINTER_REGNUM when
32331 FRAME_GROWS_DOWNWARD.
32332 * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32333 (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32334 xtensa_initial_elimination_offset.
32336 2017-05-08 Nathan Sidwell <nathan@acm.org>
32338 * doc/invoke.texi: Alphabetize -fdump options.
32340 2017-05-08 Martin Sebor <msebor@redhat.com>
32342 PR translation/80280
32343 * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32345 2017-05-08 Bernd Edlinger <bernd.edlinger@hotmail.de>
32347 * target.def (compute_frame_layout): New optional target hook.
32348 * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32349 * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32350 * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32352 * reload1.c (verify_initial_elim_offsets): Likewise.
32353 * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32354 (use_simple_return_p): Call arm_compute_frame_layout if needed.
32355 (arm_get_frame_offsets): Split up into this ...
32356 (arm_compute_frame_layout): ... and this function.
32358 2017-05-08 Richard Sandiford <richard.sandiford@arm.com>
32360 * config/aarch64/constraints.md (Usa): New constraint.
32361 * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32363 2017-05-08 Thomas Preud'homme <thomas.preudhomme@arm.com>
32365 * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32366 with_multilib_list after it has been checked.
32368 2017-05-08 Richard Biener <rguenther@suse.de>
32370 * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32371 (bitmap_set_subtract_values): Likewise.
32373 2017-05-08 Richard Biener <rguenther@suse.de>
32375 * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32376 (gimple_assign_nonzero): ... this and remove strict_overflow_p
32378 (gimple_stmt_nonzero_warnv_p): Rename to ...
32379 (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32381 (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32382 (extract_range_basic): Adjust, do not disable propagation on
32383 strict overflow sensitive simplification.
32384 (vrp_visit_cond_stmt): Likewise.
32386 2017-05-05 Jan Hubicka <hubicka@ucw.cz>
32388 * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32389 body size unconditionally.
32391 2017-05-07 Jeff Law <law@redhat.com>
32394 2017-05-06 Jeff Law <law@redhat.com>
32395 PR tree-optimization/78496
32396 * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32399 PR tree-optimization/78496
32400 * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32401 (simplify_stmt_using_ranges): Call it.
32402 (vrp_dom_walker::before_dom_children): Extract equivalences
32403 from an ASSERT_EXPR with an equality comparison against a
32406 2017-05-06 Jeff Law <law@redhat.com>
32408 PR tree-optimization/78496
32409 * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32412 PR tree-optimization/78496
32413 * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32414 (simplify_stmt_using_ranges): Call it.
32415 (vrp_dom_walker::before_dom_children): Extract equivalences
32416 from an ASSERT_EXPR with an equality comparison against a
32419 2017-05-06 Richard Sandiford <richard.sandiford@linaro.org>
32421 * lra-constraints.c (lra_copy_reg_equiv): New function.
32422 (split_reg): Use it to copy equivalence information from the
32423 original register to the spill register.
32425 2017-05-06 Richard Sandiford <richard.sandiford@linaro.org>
32427 PR rtl-optimization/75964
32428 * simplify-rtx.c (simplify_const_relational_operation): Remove
32429 invalid handling of comparisons of integer ABS.
32431 2017-05-06 Uros Bizjak <ubizjak@gmail.com>
32433 * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32434 initialize to zero.
32435 (init_regs): Remove declaration.
32436 (function_arg_advance_32): Initialize error_p as boolean variable.
32438 2017-05-05 Nathan Sidwell <nathan@acm.org>
32440 * store-motion.c (remove_reachable_equiv_notes): Reformat long
32441 lines. Use for (;;).
32443 2017-05-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
32445 * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32446 (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32447 (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32448 (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32449 VF=2 that require versioning.
32451 2017-05-05 David Malcolm <dmalcolm@redhat.com>
32453 * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32456 2017-05-05 David Malcolm <dmalcolm@redhat.com>
32458 * diagnostic.h (diagnostic_override_option_index): Convert from
32459 macro to inline function.
32461 2017-05-05 David Malcolm <dmalcolm@redhat.com>
32463 * diagnostic.c (last_module_changed_p): New function.
32464 (set_last_module): New function.
32465 (diagnostic_report_current_module): Convert macro usage to
32466 the above functions.
32467 * diagnostic.h (diagnostic_context::last_module): Strengthen
32468 from const line_map * to const line_map_ordinary *.
32469 (diagnostic_last_module_changed): Delete macro.
32470 (diagnostic_set_last_module): Delete macro.
32472 2017-05-05 David Malcolm <dmalcolm@redhat.com>
32474 * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32475 with diagnostic_report_diagnostic.
32476 (diagnostic_n_impl_richloc): Likewise.
32477 * diagnostic.h (report_diagnostic): Delete macro.
32478 * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32479 with diagnostic_report_diagnostic.
32480 * substring-locations.c (format_warning_va): Likewise.
32482 2017-05-05 David Malcolm <dmalcolm@redhat.com>
32484 * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32485 save/restor of format_spec. Move option-printing code to...
32486 (print_option_information): ...this new function, and
32487 reimplement by simply printing to the pretty_printer,
32488 rather than appending to the format string.
32490 2017-05-05 David Malcolm <dmalcolm@redhat.com>
32492 * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32493 handling logic into...
32494 (update_effective_level_from_pragmas): ...this new function.
32496 2017-05-04 Andrew Waterman <andrew@sifive.com>
32498 * config/riscv/riscv.opt (mstrict-align): New option.
32499 * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it. Update comment.
32500 (SLOW_UNALIGNED_ACCESS): Define.
32501 (riscv_slow_unaligned_access): Declare.
32502 * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32504 (riscv_slow_unaligned_access): New variable.
32505 (rocket_tune_info): Set slow_unaligned_access to true.
32506 (optimize_size_tune_info): Set slow_unaligned_access to false.
32507 (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32508 (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32509 (riscv_option_override): Set riscv_slow_unaligned_access.
32510 * doc/invoke.texi: Add -mstrict-align to RISC-V.
32512 2017-05-04 Kito Cheng <kito.cheng@gmail.com>
32514 * config/riscv/riscv.md: Unify indentation.
32516 2017-05-05 Michael Meissner <meissner@linux.vnet.ibm.com>
32521 * config/rs6000/rs6000.md (u code attribute): Add FIX and
32523 (extendsi<mode>2): Add support for doing sign extension via
32524 VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32525 don't have ISA 3.0 instructions.
32526 (extendsi<mode>2 splitter): Likewise.
32527 (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32528 generate the normal insns since SImode can now go in vector
32529 registers. Disallow the special UNSPECs needed for previous
32530 machines to hide SImode being used. Add new insns
32531 fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32532 (fix_trunc<mode>si2_stfiwx): Likewise.
32533 (fix_trunc<mode>si2_internal): Likewise.
32534 (fixuns_trunc<mode>si2): Likewise.
32535 (fixuns_trunc<mode>si2_stfiwx): Likewise.
32536 (fctiw<u>z_<mode>_smallint): Likewise.
32537 (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32538 of floating point to 32-bit integer from doing a direct move to
32539 the GPR registers to do a store.
32540 (fctiwz_<mode>): Break long line.
32542 2017-05-05 Bin Cheng <bin.cheng@arm.com>
32544 * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32545 * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32546 (addr_list, addr_offset_valid_p): New.
32547 (split_address_groups): Check offset validity with above function.
32548 (gt-tree-ssa-loop-ivopts.h): Include header file.
32550 2017-05-05 Nathan Sidwell <nathan@acm.org>
32552 * config.gcc (arm*-*-*): Add missing 'fi'.
32554 2017-05-05 Steve Ellcey <sellcey@cavium.com>
32556 * doc/invoke.texi (-fopt-info): Explicitly say order of options
32557 included in -fopt-info does not matter.
32558 * doc/optinfo.texi (-fopt-info): Fix description of default
32559 behavour. Explicitly say order of options included in -fopt-info
32562 2017-05-05 Thomas Preud'homme <thomas.preudhomme@arm.com>
32564 * config.gcc: Allow combinations of aprofile and rmprofile values for
32565 --with-multilib-list.
32566 * config/arm/t-multilib: New file.
32567 * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32568 variables. Remove setting of ISA and floating-point ABI in
32569 MULTILIB_OPTIONS and MULTILIB_DIRNAMES. Set architecture and FPU in
32570 MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32571 and MULTILIB_DIRNAMES respectively. Add comment to introduce all
32572 matches. Add architecture matches for marvel-pj4 and generic-armv7-a
32574 * config/arm/t-rmprofile: Likewise except for the matches changes.
32575 * doc/install.texi (--with-multilib-list): Document the combination of
32576 aprofile and rmprofile values and warn about pitfalls in doing that.
32578 2017-05-05 Wilco Dijkstra <wdijkstr@arm.com>
32580 * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32581 (movdi_aarch64): Likewise.
32583 2017-05-05 Jakub Jelinek <jakub@redhat.com>
32585 PR tree-optimization/80632
32586 * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32588 (build_arrays): Initialize it for virtual phis.
32589 (fix_phi_nodes): Use it for virtual phis.
32591 PR tree-optimization/80558
32592 * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32593 [x, y] op z into [x op, y op z] for op & or | if conditions
32596 2017-05-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
32597 Prakhar Bahuguna <prakhar.bahuguna@arm.com>
32600 * config/arm/arm.md (use_literal_pool): Remove.
32601 (64-bit immediate split): No longer takes cost into consideration
32602 if arm_disable_literal_pool is enabled.
32603 * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32604 used when arm_disable_literal_pool is enabled.
32605 (arm_max_const_double_inline_cost): Remove use of
32606 arm_disable_literal_pool.
32607 (push_minipool_fix): Add assert.
32608 (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32609 * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32610 (no_literal_pool_sf_immediate): New.
32612 2017-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
32614 PR tree-optimization/80613
32615 * tree-ssa-dce.c (propagate_necessity): Remove cases for
32616 BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32618 2017-05-05 Richard Biener <rguenther@suse.de>
32620 * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32622 2017-05-05 Georg-Johann Lay <avr@gjlay.de>
32624 * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32625 of this flag from insn conditions due to removal from r247495.
32627 2017-05-05 Wilco Dijkstra <wdijkstr@arm.com>
32629 * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32631 (arm_early_store_addr_dep_ptr): Likewise.
32632 * config/arm/aarch-common-protos.h
32633 (arm_early_load_addr_dep_ptr): Add prototype.
32634 (arm_early_store_addr_dep_ptr): Likewise.
32635 * config/arm/cortex-a53.md: Add new bypasses.
32637 2017-05-05 Jakub Jelinek <jakub@redhat.com>
32639 * tree.c (next_type_uid): Change type to unsigned.
32640 (type_hash_canon): Decrement back next_type_uid if
32641 freeing a type node with the highest TYPE_UID. For INTEGER_TYPEs
32642 also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32645 2017-05-04 Martin Sebor <msebor@redhat.com>
32647 * builtins.c: Fix a trivial typo in a comment.
32649 PR middle-end/79234
32650 * builtins.c (check_sizes): Adjust to handle reading past the end.
32651 Avoid printing excessive upper bound of ranges. Use %E to print
32652 tree nodes instead of converting them to %wu.
32653 (expand_builtin_memchr): New function.
32654 (compute_dest_size): Rename...
32655 (compute_objsize): ...to this.
32656 (expand_builtin_memcpy): Adjust.
32657 (expand_builtin_mempcpy): Adjust.
32658 (expand_builtin_strcat): Adjust.
32659 (expand_builtin_strcpy): Adjust.
32660 (check_strncat_sizes): Adjust.
32661 (expand_builtin_strncat): Adjust.
32662 (expand_builtin_strncpy): Adjust and simplify.
32663 (expand_builtin_memset): Adjust.
32664 (expand_builtin_bzero): Adjust.
32665 (expand_builtin_memcmp): Adjust.
32666 (expand_builtin): Handle memcmp.
32667 (maybe_emit_chk_warning): Check strncat just once.
32669 2017-05-04 Martin Sebor <msebor@redhat.com>
32671 PR preprocessor/79214
32672 PR middle-end/79222
32673 PR middle-end/79223
32674 * builtins.c (check_sizes): Add inlining context and issue
32675 warnings even when -Wno-system-headers is set.
32676 (check_strncat_sizes): Same.
32677 (expand_builtin_strncat): Same.
32678 (expand_builtin_memmove): New function.
32679 (expand_builtin_stpncpy): Same.
32680 (expand_builtin): Handle memmove and stpncpy.
32682 2017-05-04 Bin Cheng <bin.cheng@arm.com>
32684 * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32685 which is not used any more.
32687 2017-05-04 Wilco Dijkstra <wdijkstr@arm.com>
32689 * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32691 2017-05-04 Wilco Dijkstra <wdijkstr@arm.com>
32693 * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32694 (cortexa53_tunings): Likewise.
32695 (cortexa57_tunings): Likewise.
32696 (cortexa72_tunings): Likewise.
32697 (cortexa73_tunings): Likewise.
32699 2017-05-04 Wilco Dijkstra <wdijkstr@arm.com>
32701 * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32702 Set loop alignment to 8.
32704 2017-05-04 Martin Sebor <msebor@redhat.com>
32706 PR translation/80280
32707 * builtins.c (expand_builtin_object_size): Add missing quoting to
32708 %D and like directives.
32709 * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32710 (hsa_type_for_tree_type): Same.
32711 (verify_function_arguments): Same.
32712 * symtab.c (symbol_table::change_decl_assembler_name): Same.
32713 * varasm.c (get_section): Same.
32716 2017-05-04 Martin Sebor <msebor@redhat.com>
32718 PR translation/80280
32719 * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32721 2017-05-04 Wilco Dijkstra <wdijkstr@arm.com>
32723 * config/aarch64/aarch64.c (generic_addrcost_table):
32724 Change HI/TI mode setting.
32726 2017-05-04 Martin Jambor <mjambor@suse.cz>
32728 PR tree-optimization/80622
32729 * tree-sra.c (comes_initialized_p): New function.
32730 (build_accesses_from_assign): Only set write lazily when
32731 comes_initialized_p is false.
32732 (analyze_access_subtree): Use comes_initialized_p.
32733 (propagate_subaccesses_across_link): Assert !comes_initialized_p
32734 instead of testing for PARM_DECL.
32736 2017-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32738 * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32739 constraint on operand 0 to allow more general addressing modes.
32740 Adjust output template.
32741 * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32743 * config/aarch64/aarch64-protos.h
32744 (aarch64_address_valid_for_prefetch_p): Declare prototype.
32745 * config/aarch64/constraints.md (Dp): New address constraint.
32746 * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32749 2017-05-04 Jan Hubicka <hubicka@ucw.cz>
32751 * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32752 update use of estimate_ipcp_clone_size_and_time.
32753 (estimate_local_effects): Update use of
32754 estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32755 * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32756 * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32757 Return nonspecialized time.
32759 2017-05-04 Richard Biener <rguenther@suse.de>
32761 * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32762 for the last VUSE which def dominates the PHI. Directly call
32764 (get_continuation_for_phi_1): Remove.
32766 2017-05-04 Richard Sandiford <richard.sandiford@linaro.org>
32768 * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32769 to explain the use of truncating division. Cap the number of
32770 iterations to the maximum given by nb_iterations_upper_bound,
32773 2017-05-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
32775 * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32776 * configure: Regenerate.
32777 * config.in: Regenerate.
32778 * config/i386/driver-mingw32.c: new file.
32779 * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32780 * config.host: Link driver-mingw32.o on MinGW host.
32781 * doc/install.texi: Document new --enable-mingw-wildcard configure
32784 2017-05-04 Marek Polacek <polacek@redhat.com>
32786 PR tree-optimization/80612
32787 * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32789 2017-05-04 Prakhar Bahuguna <prakhar.bahuguna@arm.com>
32790 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
32792 * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32793 (movt splitter): Likewise.
32794 * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32795 to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32796 (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32797 block for Thumb-1 with MOVT.
32798 (thumb2_legitimate_address_p): Move code block ...
32799 (can_avoid_literal_pool_for_label_p): ... into this new function.
32800 (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32802 (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32803 * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32804 "M-profile targets with the MOVT instruction".
32806 2017-05-04 Prakhar Bahuguna <prakhar.bahuguna@arm.com>
32808 * config/arm/arm-builtins.c (arm_init_builtins): Rename
32809 __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32810 __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32812 2017-05-04 Martin Liska <mliska@suse.cz>
32814 * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32815 variable cond_code.
32817 2017-05-04 Richard Biener <rguenther@suse.de>
32819 * tree.c (array_at_struct_end_p): Handle arrays at struct
32820 end with flexarrays more conservatively. Refactor and treat
32821 arrays of arrays or aggregates more strict. Fix
32822 VIEW_CONVERT_EXPR handling. Remove allow_compref argument.
32823 * tree.c (array_at_struct_end_p): Adjust prototype.
32824 * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32825 * gimple-fold.c (get_range_strlen): Likewise.
32826 * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32828 2017-05-04 Richard Biener <rguenther@suse.de>
32830 PR tree-optimization/31130
32831 * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32833 (supports_overflow_infinity): Likewise.
32834 (is_negative_overflow_infinity): Likewise.
32835 (is_positive_overflow_infinity): Likewise.
32836 (is_overflow_infinity): Likewise.
32837 (stmt_overflow_infinity): Likewise.
32838 (overflow_infinity_range_p): Likewise.
32839 (usable_range_p): Remove as always returning true.
32840 (make_overflow_infinity): Remove.
32841 (negative_overflow_infinity): Likewise.
32842 (positive_overflow_infinity): Likewise.
32843 (avoid_overflow_infinity): Likewise.
32844 (set_value_range): Adjust accordingly.
32845 (set_value_range_to_nonnegative): Likewise, remove now unused
32846 overflow_infinity arg.
32847 (vrp_operand_equal_p): Adjust.
32848 (update_value_range): Likewise.
32849 (range_int_cst_singleton_p): Likewise.
32850 (operand_less_p): Likewise.
32851 (compare_values_warnv): Likewise.
32852 (extract_range_for_var_from_comparison_expr): Likewise.
32853 (vrp_int_const_binop): Likewise.
32854 (zero_nonzero_bits_from_vr): Likewise.
32855 (extract_range_from_multiplicative_op_1): Likewise.
32856 (extract_range_from_binary_expr_1): Likewise.
32857 (extract_range_from_unary_expr): Likewise.
32858 (extract_range_from_comparison): Likewise.
32859 (extract_range_basic): Likewise.
32860 (adjust_range_with_scev): Likewise.
32861 (compare_ranges): Likewise.
32862 (compare_range_with_value): Likewise.
32863 (dump_value_range): Likewise.
32864 (test_for_singularity): Likewise, remove strict_overflow_p parameter
32866 (simplify_cond_using_ranges): Adjust.
32868 2017-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
32870 * brig-builtins.def: Added a builtin for class_f64.
32871 * builtin-types.def: Added a builtin type needed by class_f64.
32873 2017-05-03 Jason Merrill <jason@redhat.com>
32875 * timevar.def: Add TV_CONSTEXPR.
32877 2017-05-03 David Malcolm <dmalcolm@redhat.com>
32879 * common.opt (fdiagnostics-parseable-fixits): Fix typo.
32881 2017-05-03 Martin Jambor <mjambor@suse.cz>
32883 * ipa-prop.c (ipa_update_after_lto_read): Removed.
32884 * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
32885 * ipa-cp.c (ipcp_propagate_stage): Do not call
32886 ipa_update_after_lto_read.
32887 * ipa-inline.c (ipa_inline): Likewise.
32889 2017-05-03 Martin Jambor <mjambor@suse.cz>
32891 * ipa-prop.h (ipa_edge_args): Make a class. Mark with for_user GTY
32892 tag. Added a default constructor and a destructor.
32893 (ipa_edge_args_sum_t): New class;
32894 (ipa_edge_args_sum): Declare.
32895 (ipa_edge_args_vector): Remove declaration.
32896 (IPA_EDGE_REF): Use ipa_edge_args_sum.
32897 (ipa_free_edge_args_substructures): Remove declaration.
32898 (ipa_check_create_edge_args): Use ipa_edge_args_sum.
32899 (ipa_edge_args_info_available_for_edge_p): Likewise.
32900 * ipa-prop.c (ipa_edge_args_vector): Removed.
32901 (edge_removal_hook_holder): Likewise.
32902 (edge_duplication_hook_holder): Likewise.
32903 (ipa_edge_args_sum): New variable.
32904 (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
32905 ipa_edge_args_vector.
32906 (ipa_free_edge_args_substructures): Likewise.
32907 (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
32908 ipa_edge_args_vector.
32909 (ipa_edge_removal_hook): Turned into method
32910 ipa_edge_args_sum_t::remove.
32911 (ipa_edge_duplication_hook): Turned into method
32912 ipa_edge_args_sum_t::duplicate.
32913 (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
32914 registering edge hooks.
32915 (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
32916 * ipa-inline-analysis.c (estimate_function_body_sizes): Test
32917 ipa_edge_args_sum instead of ipa_edge_args_vector.
32918 * ipa-profile.c (ipa_profile): Likewise.
32920 2017-05-03 Martin Jambor <mjambor@suse.cz>
32922 * symbol-summary.h (function_summary): New method exists.
32923 (function_summary::symtab_removal): Deallocate through release.
32924 (call_summary): New class.
32925 (gt_ggc_mx): New overload.
32926 (gt_pch_nx): Likewise.
32927 (gt_pch_nx): Likewise.
32929 2017-05-03 Jeff Law <law@redhat.com>
32931 PR tree-optimization/78496
32932 * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
32933 from simplify_cond_using_ranges. Split off code to walk
32934 backwards through casts into ...
32935 (simplify_cond_using_ranges_2): New function.
32936 (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
32937 (execute_vrp): After identifying jump threads, call
32938 simplify_cond_using_ranges_2.
32940 2017-05-03 Jan Hubicka <hubicka@ucw.cz>
32943 * ipa-inline.h (inline_summary): Add ctor.
32944 (create_ggc): Do not use ggc_cleared_alloc.
32946 2017-05-03 Jeff Downs <heydowns@somuchpressure.net>
32947 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
32949 * gcc.c (handle_braces): Support escaping in switch matching
32951 * doc/invoke.texi (Spec Files): Document it.
32952 Remove superfluous @code markup in items.
32954 2017-05-03 David Malcolm <dmalcolm@redhat.com>
32956 * diagnostic-show-locus.c (struct column_range): New struct.
32957 (get_affected_columns): New function.
32958 (get_printed_columns): New function.
32959 (struct correction): New struct.
32960 (correction::ensure_capacity): New function.
32961 (correction::ensure_terminated): New function.
32962 (struct line_corrections): New struct.
32963 (line_corrections::~line_corrections): New dtor.
32964 (line_corrections::add_hint): New function.
32965 (layout::print_trailing_fixits): Reimplement in terms of the new
32967 (selftest::test_overlapped_fixit_printing): New function.
32968 (selftest::diagnostic_show_locus_c_tests): Call it.
32970 2017-05-03 Nathan Sidwell <nathan@acm.org>
32972 Canonicalize canonical type hashing
32973 * tree.h (type_hash_canon_hash): Declare.
32974 * tree.c (type_hash_list, attribute_hash_list): Move into
32975 type_hash_canon_hash.
32976 (build_type_attribute_qual_variant): Break out hash code calc into
32977 type_hash_canon_hash.
32978 (type_hash_canon_hash): New. Generic type hash computation.
32979 (build_range_type_1, build_array_type_1, build_function_type,
32980 build_method_type_directly, build_offset_type, build_complex_type,
32981 make_vector_type): Call it.
32983 2017-05-03 Richard Biener <rguenther@suse.de>
32985 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
32986 When all DRs have unknown misaligned do not always peel
32987 when there is a store but apply the same costing model as if
32988 there were only loads.
32990 2017-05-03 Richard Biener <rguenther@suse.de>
32993 PR tree-optimization/80492
32994 * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
32995 compare_base_decls returning dont-know properly.
32997 2017-05-03 Thomas Preud'homme <thomas.preudhomme@arm.com>
32999 * config/arm/iterators.md (CCSI): New mode iterator.
33000 (arch): New mode attribute.
33001 * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33002 (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33003 (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This. Use CCSI
33004 code iterator for success result mode.
33005 * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33006 the corresponding new insn generators.
33008 2017-05-03 Bin Cheng <bin.cheng@arm.com>
33011 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33012 * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33014 2017-05-03 Richard Sandiford <richard.sandiford@linaro.org>
33016 * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33017 (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33018 (DDR_A): Wrap DDR argument in brackets.
33019 (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33020 (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33021 (DDR_REVERSED_P): Likewise.
33023 2017-05-03 Jakub Jelinek <jakub@redhat.com>
33025 PR tree-optimization/79472
33026 * tree-switch-conversion.c (struct switch_conv_info): Add
33027 contiguous_range and default_case_nonstandard fields.
33028 (collect_switch_conv_info): Compute contiguous_range and
33029 default_case_nonstandard fields, don't clear final_bb if
33030 contiguous_range and only the default case doesn't have the required
33032 (check_all_empty_except_final): Set default_case_nonstandard instead
33033 of failing if contiguous_range and the default case doesn't have empty
33035 (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33036 and only the default case doesn't have the required constants. Skip
33038 (gather_default_values): Skip virtual phis. Allow non-NULL CASE_LOW
33039 if default_case_nonstandard.
33040 (build_constructors): Build constant 1 just once. Assert that default
33041 values aren't inserted in between cases if contiguous_range. Skip
33043 (build_arrays): Skip virtual phis.
33044 (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33045 (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33046 Handle virtual phis.
33047 (gen_inbound_check): Handle default_case_nonstandard case.
33048 (process_switch): Adjust check_final_bb caller. Call
33049 gather_default_values with the first non-default case instead of
33050 default case if default_case_nonstandard.
33052 2017-05-02 Nathan Sidwell <nathan@acm.org>
33054 * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33055 check. Fix formatting.
33057 2017-05-02 Jan Hubicka <hubicka@ucw.cz>
33059 * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33060 errors when comparing specialized and unspecialized times.
33062 2017-05-02 David Malcolm <dmalcolm@redhat.com>
33064 * diagnostic-show-locus.c
33065 (layout::should_print_annotation_line_p): Make private.
33066 (layout::print_annotation_line): Make private.
33067 (layout::annotation_line_showed_range_p): Make private.
33068 (layout::show_ruler): Make private.
33069 (layout::print_source_line): Make private. Pass in line and
33070 line_width, rather than calling location_get_source_line. Drop
33072 (layout::print_leading_fixits): New method.
33073 (layout::print_any_fixits): Rename to...
33074 (layout::print_trailing_fixits): ...this, and make private.
33075 Don't print newline fixits.
33076 (diagnostic_show_locus): Move logic for printing one row into...
33077 (layout::print_line): ...this new function. Move the
33078 location_get_source_line call and error-handling from
33079 print_source_line to here. Call print_leading_fixits, and rename
33080 print_any_fixits to print_trailing_fixits.
33081 (selftest::test_fixit_insert_containing_newline): Update now that
33082 newlines are partially supported.
33083 (selftest::test_fixit_insert_containing_newline_2): New test.
33084 (selftest::test_fixit_replace_containing_newline): Update comments.
33085 (selftest::diagnostic_show_locus_c_tests): Call the new test.
33086 * edit-context.c (class added_line): New class.
33087 (class edited_line): Describe newline handling in comment.
33088 (edited_line::actually_edited_p): New method.
33089 (edited_line::print_content): Delete redundant decl.
33090 (edited_line::m_predecessors): New field.
33091 (edited_file::print_content): Call edited_line::print_content.
33092 (edited_file::print_diff): Update to support newlines.
33093 (edited_file::print_diff_hunk): Likewise.
33094 (edited_file::print_run_of_changed_lines): New function.
33095 (edited_file::print_diff_line): Convert to...
33096 (print_diff_line): ...this.
33097 (edited_file::get_effective_line_count): New function.
33098 (edited_line::edited_line): Initialize new field m_predecessors.
33099 (edited_line::~edited_line): Clean up m_predecessors.
33100 (edited_line::apply_fixit): Handle newlines.
33101 (edited_line::get_effective_line_count): New function.
33102 (edited_line::print_content): New function.
33103 (edited_line::print_diff_lines): New function.
33104 (selftest::test_applying_fixits_insert_containing_newline): New
33106 (selftest::test_applying_fixits_replace_containing_newline): New
33108 (selftest::insert_line): New function.
33109 (selftest::test_applying_fixits_multiple_lines): Add example of
33111 (selftest::edit_context_c_tests): Call the new tests.
33113 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33115 * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33116 parameter cand. Update dump information.
33117 (get_computation_cost): Update uses.
33119 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33121 * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33122 (get_computation_aff): Reorder parameters. Use get_computation_aff_1.
33123 (get_computation_at, rewrite_use_address): Update use of
33124 get_computation_aff.
33126 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33128 * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33129 (get_computation): Delete.
33130 (get_computation_cost): Implement like get_computation_cost_at.
33131 Use get_computation_at.
33132 (get_computation_cost_at): Delete.
33133 (rewrite_use_nonlinear_expr): Use get_computation_at.
33134 (rewrite_use_compare, remove_unused_ivs): Ditto.
33136 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33138 * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33140 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33142 * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33143 (ivopts_global_cost_for_size): Rename parameter and update uses.
33144 (iv_ca_recount_cost): Update uses.
33145 (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33146 candidates seperately in n_invs and n_cands.
33147 (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33149 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33151 * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33152 (find_inv_vars_cb): New.
33153 (find_depends): Renamed to ...
33154 (find_inv_vars): ... this.
33155 (add_candidate_1, force_var_cost): Call find_inv_vars.
33156 (split_address_cost, determine_group_iv_cost_cond): Ditto.
33158 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33160 * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33161 inv_vars. Add inv_exprs.
33162 (struct iv_cand): Rename depends_on to inv_vars.
33163 (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33164 max_inv_var_id/n_inv_var_uses. Move max_inv_expr_id around.
33165 Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33166 (dump_cand): Dump inv_vars.
33167 (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33168 (record_invariant, find_depends, add_candidate_1): Ditto.
33169 (set_group_iv_cost, force_var_cost): Ditto.
33170 (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33171 (get_computation_cost_at, get_computation_cost): Ditto.
33172 (determine_group_iv_cost_generic): Ditto.
33173 (determine_group_iv_cost_address): Ditto.
33174 (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33175 (determine_group_iv_costs): Ditto.
33176 (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33177 (iv_ca_set_remove_invariants): Renamed to ...
33178 (iv_ca_set_remove_invs): ... this. Support inv_vars and inv_exprs.
33179 (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33180 (iv_ca_set_add_invariants): Renamed to ...
33181 (iv_ca_set_add_invs): ... this. Support inv_vars and inv_exprs.
33182 (iv_ca_set_cp): Use iv_ca_set_add_invs.
33183 (iv_ca_has_deps): Support inv_vars and inv_exprs.
33184 (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33185 (create_new_ivs): Remove useless dump.
33187 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33189 * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33191 (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33192 (iv_ca_set_no_cp, create_new_iv): Ditto.
33194 2017-05-02 Bin Cheng <bin.cheng@arm.com>
33196 * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33198 2017-05-02 Xi Ruoyao <ryxi@stu.xidian.edu.cn>
33200 * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33201 function tree_check2.
33203 2017-05-02 Martin Liska <mliska@suse.cz>
33205 * doc/gcov.texi: Add missing preposition.
33206 * gcov.c (function_info::function_info): Properly fill up
33207 all member variables.
33209 2017-05-02 Tamar Christina <tamar.christina@arm.com>
33211 * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33213 2017-05-02 Tamar Christina <tamar.christina@arm.com>
33215 * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33217 2017-05-02 Martin Liska <mliska@suse.cz>
33220 * lto-streamer-in.c (lto_read_tree_1): Remove
33221 LTO_STREAMER_DEBUG.
33222 * lto-streamer.c (struct tree_hash_entry): Likewise.
33223 (struct tree_entry_hasher): Likewise.
33224 (tree_entry_hasher::hash): Likewise.
33225 (tree_entry_hasher::equal): Likewise.
33226 (lto_streamer_init): Likewise.
33227 (lto_orig_address_map): Likewise.
33228 (lto_orig_address_get): Likewise.
33229 (lto_orig_address_remove): Likewise.
33230 * lto-streamer.h: Likewise.
33231 * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33232 * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33234 2017-05-02 Sebastian Peryt <sebastian.peryt@intel.com>
33236 * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33237 (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33238 (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33239 (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33240 (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33241 (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33242 (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33243 (mm_maskz_sub_ss): New intrinsics.
33244 * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33245 (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33246 * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33247 (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33248 (__builtin_ia32_subss_mask_round): New builtins.
33249 * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33250 (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33251 * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33253 (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33254 (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33256 (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33259 2017-05-02 Martin Jambor <mjambor@suse.cz>
33261 PR tree-optimization/78687
33262 * tree-sra.c (access): New field parent.
33263 (process_subtree_disqualification): New function.
33264 (disqualify_candidate): Call it.
33265 (build_accesses_from_assign): Reset write flag if creating an
33267 (build_access_subtree): Fill in parent field and also prpagate
33268 down grp_write flag.
33269 (create_artificial_child_access): New parameter set_grp_write, set
33270 grp_write to its value.
33271 (propagate_subaccesses_across_link): Also propagate grp_write flag
33273 (propagate_all_subaccesses): Push the closest parent back to work
33274 queue if add_access_to_work_queue returned true.
33276 2017-05-02 Richard Biener <rguenther@suse.de>
33278 * common.opt (fstrict-overflow): Alias negative to fwrapv.
33279 * doc/invoke.texi (fstrict-overflow): Remove all traces of
33280 -fstrict-overflow documentation.
33281 * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33282 (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33283 flag_strict_overflow.
33284 * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33285 * lto-opts.c (lto_write_options): Do not stream it.
33286 * lto-wrapper.c (merge_and_complain): Do not handle it.
33287 * opts.c (default_options_table): Do not set -fstrict-overflow.
33288 (finish_options): Likewise do not clear it when sanitizing.
33289 * simplify-rtx.c (simplify_const_relational_operation): Do not
33290 test flag_strict_overflow.
33292 2017-05-02 Uros Bizjak <ubizjak@gmail.com>
33294 * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33295 using enabled attribute.
33296 (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33297 (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33298 (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33299 (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33300 (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33301 (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33302 (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33303 (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33304 (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33305 (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33307 2017-05-02 Uros Bizjak <ubizjak@gmail.com>
33309 * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33311 2017-05-02 Richard Biener <rguenther@suse.de>
33313 PR tree-optimization/80591
33315 2017-04-10 Richard Biener <rguenther@suse.de>
33317 * tree-ssa-structalias.c (find_func_aliases): Properly handle
33320 2017-05-02 Richard Biener <rguenther@suse.de>
33322 PR tree-optimization/80549
33323 * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33324 (cleanup_tree_cfg_noloop): Create forwarders to known loop
33325 headers if they do not have a preheader.
33327 2017-05-02 Martin Liska <mliska@suse.cz>
33330 * common.opt: Fix typo.
33331 * doc/invoke.texi: Likewise.
33333 2017-05-01 Jan Beulich <jbeulich@suse.com>
33335 * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33336 swapping, add (x,x,m,x,n) alternative.
33338 2017-05-01 Nathan Sidwell <nathan@acm.org>
33340 * calls.c (combine_pending_stack_adjustment_and_call): Remove
33341 unnecessary unadjusted_alignment check.
33343 2017-05-01 Xi Ruoyao <ryxi@stu.xidian.edu.cn>
33346 * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33348 * gimplify.c (gimplify_cilk_detach): New function.
33349 (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33350 * tree-core.h: Document EXPR_CILK_SPAWN.
33351 * tree.h (EXPR_CILK_SPAWN): Define.
33353 2017-05-01 David Malcolm <dmalcolm@redhat.com>
33355 * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33356 to use new fixit_hint representation, using the "replace" logic.
33357 (get_line_span_for_fixit_hint): Likewise.
33358 (layout::print_any_fixits): Likewise.
33359 (selftest::test_one_liner_many_fixits): Rename to...
33360 (selftest::test_one_liner_many_fixits_1): ...this, and update
33361 comment and expected output to reflect that the multiple fix-it
33362 hints are now consolidated into one insertion.
33363 (selftest::test_one_liner_many_fixits_2): New test.
33364 (selftest::test_diagnostic_show_locus_one_liner): Update for
33366 (selftest::test_fixit_consolidation): Update for fix-it API
33368 * diagnostic.c (print_parseable_fixits): Likewise.
33369 * edit-context.c (edited_line::m_line_events): Convert from
33370 auto_vec <line_event *> to auto_vec <line_event>.
33371 (class line_event): Convert from abstract base class to a concrete
33372 class, taking over the role of replace_event.
33373 (class insert_event): Delete.
33374 (class replace_event): Rename to class line_event. Convert to
33376 (edit_context::add_fixits): Reimplement.
33377 (edit_context::apply_insert): Delete.
33378 (edit_context::apply_replace): Rename to...
33379 (edit_context::apply_fixit): ...this. Convert to half-open range.
33380 (edited_file::apply_insert): Delete.
33381 (edited_file::apply_replace): Rename to...
33382 (edited_file::apply_fixit): ...this.
33383 (edited_line::~edited_line): Drop deletion of events.
33384 (edited_line::apply_insert): Delete.
33385 (edited_line::apply_replace): Rename to...
33386 (edited_line::apply_fixit): ...this. Convert to half-open range.
33387 Update for change to type of m_line_events.
33388 * edit-context.h (edit_context::apply_insert): Delete.
33389 (edit_context::apply_replace): Rename to...
33390 (edit_context::apply_fixit): ...this.
33392 2017-05-01 Martin Sebor <msebor@redhat.com>
33394 * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33397 2017-05-01 Uros Bizjak <ubizjak@gmail.com>
33400 * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33401 __get_cpuid_max returns 0.
33402 (__get_cpuid_count): Ditto.
33404 2017-05-01 Eric Botcazou <ebotcazou@adacore.com>
33406 * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33407 replacement expression is another instance of one of its arguments.
33409 2017-05-01 Jakub Jelinek <jakub@redhat.com>
33412 * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33413 check for stack push/pop autoinc.
33414 * config/i386/i386.c (ix86_agi_dependent): Return false
33415 if the only reason why modified_in_p returned true is that
33416 addr is SP based and set_insn is a push or pop.
33418 2017-04-29 Jan Hubicka <hubicka@ucw.cz>
33420 * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33423 2017-04-29 Jan Hubicka <hubicka@ucw.cz>
33426 * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33427 (account_size_time): Use two predicates - exec_pred and
33429 (evaluate_conditions_for_known_args): Compute both clause and
33431 (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33432 (inline_summary_t::duplicate): Update.
33433 (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33435 (compute_inline_parameters): Likewise.
33436 (estimate_edge_size_and_time): Update caluclation of time.
33437 (estimate_node_size_and_time): Compute both time and nonspecialized
33439 (estimate_ipcp_clone_size_and_time): Update.
33440 (inline_merge_summary): Update.
33441 (do_estimate_edge_time): Update.
33442 (do_estimate_edge_size): Update.
33443 (do_estimate_edge_hints): Update.
33444 (inline_read_section, inline_write_summary): Stream both new predicates.
33445 * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33447 (compute_inlined_call_time): Cleanup.
33448 (big_speedup_p): Update.
33449 (edge_badness): Update.
33450 * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33451 (size_time_entry): Replace predicate by exec_predicate and
33452 nonconst_predicate.
33453 (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33454 (estimate_edge_time): Return also nonspec_time.
33455 (reset_edge_growth_cache): Update.
33457 2017-04-29 Jakub Jelinek <jakub@redhat.com>
33459 PR rtl-optimization/80491
33460 * ifcvt.c (noce_process_if_block): When looking for x setter
33461 with missing else_bb, don't check only the insn right before
33462 cond_earliest, but look for the last insn that x is modified in
33463 within the same bb.
33465 PR rtl-optimization/80491
33466 * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33468 2017-04-29 Marc Glisse <marc.glisse@inria.fr>
33470 PR tree-optimization/80487
33471 * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33473 2017-04-29 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33475 PR tree-optimization/79697
33476 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33477 is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33478 (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33480 * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33481 (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33483 2017-04-28 Martin Sebor <msebor@redhat.com>
33485 PR tree-optimization/80523
33486 * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33487 (init_target_to_host_charmap, target_to_host, target_strtol10): New
33489 (maybe_warn, format_directive, parse_directive): Use new functions.
33490 (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33492 2017-04-28 Marc Glisse <marc.glisse@inria.fr>
33494 * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33496 2017-04-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
33498 * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33499 target_header_dir): Set correctly.
33500 * configure: Regenerated.
33501 * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33502 (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33503 instead of SYSTEM_HEADER_DIR.
33505 2017-04-28 Jan Hubicka <hubicka@ucw.cz>
33507 * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33508 (estimate_local_effects): Likewise.
33509 * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33510 edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33511 * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33512 estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33513 do_estimate_edge_time, estimate_edge_time): Likewise.
33514 * ipa-inline-analysis.c (estimate_node_size_and_time,
33515 estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33516 (estimate_time_after_inlining): Remove.
33518 2017-04-28 Martin Liska <mliska@suse.cz>
33520 * doc/gcov.texi: Enhance documentation of gcov.
33522 2017-04-28 Martin Liska <mliska@suse.cz>
33524 * doc/gcov.texi: Sort options in alphabetic order.
33525 * doc/gcov-dump.texi: Likewise.
33526 * doc/gcov-tool.texi: Likewise.
33527 * gcov.c (print_usage): Likewise.
33528 * gcov-dump.c (print_usage): Likewise.
33529 * gcov-tool.c (print_merge_usage_message): Likewise.
33530 (print_rewrite_usage_message): Likewise.
33531 (print_overlap_usage_message): Likewise.
33533 2017-04-28 Martin Liska <mliska@suse.cz>
33535 PR gcov-profile/53915
33536 * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33538 2017-04-28 Martin Liska <mliska@suse.cz>
33540 PR gcov-profile/79891
33541 * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33542 is marked by compiler as living on a line.
33543 (get_cycles_count): Remove usage of the union.
33544 (output_intermediate_file): Likewise.
33545 (find_source): Fix GNU coding style.
33546 (accumulate_line_counts): Remove old non-all block mode.
33547 (output_lines): Remove usage of the union.
33548 * profile.c (output_location): Include all BBs, even if
33549 belonging to a same line (and file) as a previous BB.
33551 2017-04-28 Martin Liska <mliska@suse.cz>
33553 * gcov.c (process_args): Handle new argument 'w'.
33554 (read_graph_file): Assign ID to BBs.
33555 (output_branch_count): Display BB # if verbose flag is set.
33556 (output_lines): Likewise for arcs.
33557 (print_usage): Add '--verbose' option help.
33558 * doc/gcov.texi: Document --verbose (-w) option.
33560 2017-04-28 Martin Liska <mliska@suse.cz>
33562 * gcov.c (struct block_location_info): New struct.
33563 (process_file): Fill up the new structure.
33564 (read_graph_file): Replace usage of encoding by the newly added
33566 (add_line_counts): Likewise.
33567 (accumulate_line_counts): Remove usage of the union.
33568 (function_info::function_info): New function.
33569 (function_info::~function_info): Likewise.
33570 (process_file): Call delete instead of release_function.
33571 (release_function): Release the function.
33572 (release_structures): Call delete instead of release_function.
33573 (solve_flow_graph): Replace usage of num_blocks.
33574 (find_exception_blocks): Likewise.
33575 (output_lines): Fix GNU coding style.
33577 2017-04-28 Martin Liska <mliska@suse.cz>
33580 * coverage.c (coverage_remove_note_file): New function.
33581 * coverage.h: Declare the function.
33582 * toplev.c (finalize): Clean if an error has been seen.
33584 2017-04-28 Martin Liska <mliska@suse.cz>
33586 PR gcov-profile/80031
33587 * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33588 * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33589 * gcov.c (read_graph_file): Read just number of blocks.
33590 * profile.c (branch_prob): Do not stream 0 flags per a basic
33593 2017-04-28 Martin Liska <mliska@suse.cz>
33595 * gcov-dump.c (tag_*): Add new argument to declarations.
33596 (dump_gcov_file): Likewise.
33597 (tag_blocks): Add and use new argument depth.
33598 (tag_arcs): Likewise.
33599 (tag_lines): Likewise.
33600 (tag_counters): Likewise.
33601 (tag_summary): Likewise.
33602 (dump_working_sets): Use depth to do a proper indentation.
33604 2017-04-28 Jakub Jelinek <jakub@redhat.com>
33607 * cgraph.h (symtab_node::debug_symtab): No longer inline.
33608 * symtab.c (symtab_node::debug_symtab): Move definition here.
33610 2017-04-28 Richard Biener <rguenther@suse.de>
33612 * lto-streamer.h (LTO_major_version): Bump to 7.
33614 2017-04-28 Richard Biener <rguenther@suse.de>
33616 * tree-vrp.c (assert_info): New struct.
33617 (add_assert_info): New helper.
33618 (register_edge_assert_for_2): Refactor to add asserts to a vector
33620 (register_edge_assert_for_1): Likewise.
33621 (register_edge_assert_for): Likewise.
33622 (finish_register_edge_assert_for): New helper actually registering
33623 asserts where live on edge.
33624 (find_conditional_asserts): Adjust.
33625 (find_switch_asserts): Likewise.
33626 (evrp_dom_walker::try_find_new_range): Generalize.
33627 (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33629 2017-04-27 Marek Polacek <polacek@redhat.com>
33632 * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33633 arg10 and arg11 to itype.
33635 2017-04-27 Jonathan Wakely <jwakely@redhat.com>
33637 * doc/extend.texi (Object Size Checking): Improve grammar.
33639 2017-04-27 Richard Earnshaw <rearnsha@arm.com>
33642 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33643 that the logic for permitting reciprocal estimates matches that
33646 2017-04-27 Jakub Jelinek <jakub@redhat.com>
33649 * tree.c (type_cache_hasher::equal): Only compare
33650 TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33651 (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33652 non-aggregate element types.
33653 * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33654 about the flag on ARRAY_TYPEs in the comment, formatting fix.
33656 2017-04-27 Richard Biener <rguenther@suse.de>
33658 PR middle-end/80533
33659 * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33660 stripping ARRAY_REFs from MEM_EXPR make sure we're not
33661 keeping a reference to a trailing array.
33663 2017-04-27 Richard Biener <rguenther@suse.de>
33665 PR middle-end/80539
33666 * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33667 being in loop-closed SSA form conservatively.
33668 (chrec_fold_multiply_poly_poly): Likewise.
33670 2017-04-27 Tamar Christina <tamar.christina@arm.com>
33672 PR middle-end/79665
33673 * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33674 CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33676 2017-04-27 Jakub Jelinek <jakub@redhat.com>
33679 * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33680 (aarch64_function_arg_alignment): Return unsigned int again, but still
33681 ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33682 (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33683 Don't emit -Wpsabi note.
33684 (aarch64_function_arg_boundary): Likewise.
33685 (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33688 2017-04-26 Nathan Sidwell <nathan@acm.org>
33690 * tree.h (crc32_unsigned_n): Declare.
33691 (crc32_unsigned, crc32_unsigned): Make inline.
33692 * tree.c (crc32_unsigned_bits): Replace with ...
33693 (crc32_unsigned_n): ... this.
33694 (crc32_unsigned, crc32_byte): Remove.
33695 (crc32_string): Remove unnecessary braces.
33697 2017-04-25 Jan Hubicka <hubicka@ucw.cz>
33699 * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33700 * ipa-inline-analysis.c (MAX_TIME): Remove.
33701 (account_size_time): Use sreal for time.
33702 (dump_inline_summary): Update.
33703 (estimate_function_body_sizes): Update.
33704 (estimate_edge_size_and_time): Update.
33705 (estimate_calls_size_and_time): Update.
33706 (estimate_node_size_and_time): Update.
33707 (inline_merge_summary): Update.
33708 (inline_update_overall_summary): Update.
33709 (estimate_time_after_inlining): Update.
33710 (inline_read_section): Update.
33711 (inline_write_summary): Update.
33712 * ipa-inline.c (compute_uninlined_call_time): Update.
33713 (compute_inlined_call_time): Update.
33714 (recursive_inlining): Update.
33715 (inline_small_functions): Update.
33716 (dump_overall_stats): Update.
33717 * ipa-inline.h: Include sreal.h.
33718 (size_time_entry): Turn time to sreal.
33719 (inline_summary): Turn self_time nad time to sreal.
33721 2017-04-25 Jan Hubicka <hubicka@ucw.cz>
33723 * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33725 (sreal::stream_out, sreal::stream_in): New.
33726 * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33728 2017-04-25 Jakub Jelinek <jakub@redhat.com>
33730 * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33733 2017-04-25 Uros Bizjak <ubizjak@gmail.com>
33736 * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33738 (dimode_scalar_chain::compute_convert_gain): Ditto.
33739 (dimode_scalar_chain::make_vector_copies): Ditto.
33740 (dimode_scalar_chain::convert_reg): Ditto.
33741 (dimode_scalar_chain::convert_insn): Ditto.
33742 * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33743 (VI248_AVX512BW_1): New mode iterator.
33744 (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33745 <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33748 2017-04-25 Martin Sebor <msebor@redhat.com>
33750 PR tree-optimization/80497
33751 * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33752 constants are representable in HOST_WIDE_INT.
33753 (parse_directive): Ditto.
33755 2017-04-25 Martin Sebor <msebor@redhat.com>
33758 * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33759 (new_zero_array): Adjust signature.
33760 (dom_info::dom_init): Used unsigned rather that size_t.
33761 (dom_info::dom_info): Same.
33763 2017-04-25 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
33764 Jakub Jelinek <jakub@redhat.com>
33767 * config/arm/arm.c: Include gimple.h.
33768 (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33769 returns negative, increment ncrn only if it returned positive.
33770 (arm_needs_doubleword_align): Return int instead of bool,
33771 ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33772 members, but if there is any such non-FIELD_DECL
33773 > PARM_BOUNDARY aligned decl, return -1 instead of false.
33774 (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33775 returns negative, increment nregs only if it returned positive.
33776 (arm_setup_incoming_varargs): Likewise.
33777 (arm_function_arg_boundary): Emit -Wpsabi note if
33778 arm_needs_doubleword_align returns negative, return
33779 DOUBLEWORD_ALIGNMENT only if it returned positive.
33781 2017-04-25 Marek Polacek <polacek@redhat.com>
33784 * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33785 first argument to type.
33787 2017-04-25 Bill Seurer <seurer@linux.vnet.ibm.com>
33790 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33791 type checks to test for compatibility instead of equality.
33793 2017-04-25 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
33794 Jakub Jelinek <jakub@redhat.com>
33797 * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33799 (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33800 struct. Ignore DECL_ALIGN of decls other than FIELD_DECL for
33801 the alignment computation, but return their maximum in warn_alignment.
33802 (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33803 Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33805 (aarch64_function_arg_boundary): Likewise. Simplify using MIN/MAX.
33806 (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33809 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33811 * config/arc/simdext.md (dmpyh): Fix typo.
33813 2017-04-25 Richard Biener <rguenther@suse.de>
33815 PR tree-optimization/80492
33816 * alias.c (compare_base_decls): Handle registers with asm
33817 specification conservatively.
33818 * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33819 compare_base_decls returning dont-know properly.
33821 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33823 * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33824 (legitimate_offset_address_p): New function.
33825 (arc_legitimate_address_p): Use above function.
33827 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33829 * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33831 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33833 * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33834 ACCH registers whenever they are available.
33836 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33838 * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33839 double regs fix when not used.
33841 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33843 * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
33845 (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
33846 (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
33848 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33850 * config/arc/arc.c (arc_output_addsi): Check for h-register class
33851 when emitting short ADD instructions.
33853 2017-04-25 Claudiu Zissulescu <claziss@synopsys.com>
33855 * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
33857 (cmpsi_cc_c_insn): Likewise.
33858 (cbranchsi4_scratch): Compute proper instruction length using
33859 compact_hreg_operand.
33860 * config/arc/predicates.md (compact_hreg_operand): New predicate.
33862 2017-04-25 Richard Biener <rguenther@suse.de>
33864 PR middle-end/80509
33865 * passes.c (pass_manager::pass_manager): Initialize
33866 m_name_to_pass_map.
33868 2017-04-25 Richard Biener <rguenther@suse.de>
33870 PR tree-optimization/79201
33871 * tree-ssa-sink.c (statement_sink_location): Handle calls.
33873 2017-04-25 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
33876 * config/s390/vector.md: Split MEM->GPR vector moves for
33877 non-s_operand addresses.
33879 2017-04-25 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
33882 * config/s390/predicates.md (reload_const_wide_int_operand): New
33884 * config/s390/s390.md ("movti"): Remove d/P alternative.
33885 ("movti_bigconst"): New pattern definition.
33887 2017-04-25 Dominik Vogt <vogt@linux.vnet.ibm.com>
33890 * s390-protos.h (s390_expand_cs_hqi): Removed.
33891 (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
33892 * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
33893 modes as well as CCZ1mode and CCZmode.
33894 (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
33895 signature of s390_emit_compare_and_swap.
33896 (s390_expand_cs_hqi): Likewise, make static.
33897 (s390_expand_cs_tdsi): Generate an explicit compare before trying
33898 compare-and-swap, in some cases.
33899 (s390_expand_cs): Wrapper function.
33900 (s390_expand_atomic_exchange_tdsi): New backend specific expander for
33902 (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
33903 * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
33904 patterns for small and large integers. Forbid symref memory operands.
33905 Move expander to s390.c. Require cc register.
33906 ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
33907 ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
33908 ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
33909 ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
33910 symref memory operands. Remove CC mode and call s390_match_ccmode
33912 ("atomic_exchange<mode>"): Allow and implement all integer modes.
33914 2017-04-25 Dominik Vogt <vogt@linux.vnet.ibm.com>
33916 * config/s390/s390.md (define_peephole2): New peephole to help
33917 combining the load-and-test pattern with volatile memory.
33919 2017-04-25 Dominik Vogt <vogt@linux.vnet.ibm.com>
33921 * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
33922 with CCZmode for TARGET_Z196.
33924 2017-04-25 Jakub Jelinek <jakub@redhat.com>
33926 PR rtl-optimization/80501
33927 * combine.c (make_compound_operation_int): Set subreg_code to SET
33928 even for AND with mask of the sign bit of mode.
33930 PR rtl-optimization/80500
33931 * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
33932 sum's initial value.
33934 2017-04-25 Julian Brown <julian@codesourcery.com>
33935 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
33937 * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
33939 2017-04-25 Marc Glisse <marc.glisse@inria.fr>
33941 * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
33943 2017-04-25 Julian Brown <julian@codesourcery.com>
33944 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
33946 * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
33947 (thunderx2t99_sha): New Reservation.
33949 2017-04-25 Julian Brown <julian@codesourcery.com>
33950 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
33952 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
33953 type for 1-element load.
33955 2017-04-24 Marc Glisse <marc.glisse@inria.fr>
33957 * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
33959 2017-04-24 Martin Jambor <mjambor@suse.cz>
33961 PR tree-optimization/80293
33962 * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
33963 char arrays not totally scalarizable if it is false.
33964 (analyze_all_variable_accesses): Pass correct value in the new
33965 parameter. Add a statistics counter.
33967 2017-04-24 Jan Hubicka <hubicka@ucw.cz>
33969 PR middle-end/79931
33970 * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
33972 2017-04-24 Richard Biener <rguenther@suse.de>
33974 PR tree-optimization/80494
33975 * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
33976 out for complex types.
33978 2017-04-24 Richard Biener <rguenther@suse.de>
33980 * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
33981 * tree-ssa-sccvn.c (print_scc): Print SCC size.
33982 (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
33983 (DFS): Adjust and never fail.
33984 (sccvn_dom_walker::fail): Remove.
33985 (sccvn_dom_walker::before_dom_children): Adjust.
33986 (run_scc_vn): Likewise and never fail.
33987 * tree-ssa-pre.c (pass_pre::execute): Adjust.
33988 (pass_fre::execute): Likewise.
33990 2017-04-24 Richard Biener <rguenther@suse.de>
33992 PR tree-optimization/79725
33993 * tree-ssa-sink.c (statement_sink_location): Return whether
33994 failure reason was zero uses. Move that check later.
33995 (sink_code_in_bb): Deal with zero uses by removing the stmt
33998 2017-04-24 Richard Biener <rguenther@suse.de>
34001 * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34002 pointer-based references.
34004 2017-04-24 Richard Biener <rguenther@suse.de>
34007 * pass_manager.h (pass_manager::operator new): Remove.
34008 (pass_manager::operator delete): Likewise.
34009 * passes.c (pass_manager::operator new): Remove.
34010 (pass_manager::operator delete): Likewise.
34011 (pass_manager::pass_manager): Zero individual pass members.
34013 2017-04-23 Uros Bizjak <ubizjak@gmail.com>
34016 * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34017 <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34018 Check "XEXP (src, 1)" operand here.
34019 <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34020 Check "XEXP (src, 1)" operand here.
34021 (dimode_scalar_chain::make_vector_copies): Detect count register
34022 of a shift instruction. Zero extend count register from QImode
34023 to DImode to satisfy vector shift pattern count operand predicate.
34024 Substitute vector shift count operand with a DImode copy.
34025 (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34028 2017-04-21 Uros Bizjak <ubizjak@gmail.com>
34030 * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34031 Remove UNSPEC_NOREX_MEM tag. Update corresponding peephole2 pattern.
34032 (*insvqi_1_mem_rex64): Move above insv<mode>_1. Remove
34033 UNSPEC_NOREX_MEM tag. Update corresponding peephole2 pattern.
34034 (UNSPEC_NOREX_MEM): Remove definition.
34036 2017-04-21 Richard Biener <rguenther@suse.de>
34038 PR tree-optimization/79547
34039 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34040 Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34041 bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34042 without any constraints.
34044 2017-04-21 Richard Biener <rguenther@suse.de>
34046 PR tree-optimization/78847
34047 * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34049 2017-04-21 Richard Biener <rguenther@suse.de>
34051 * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34052 (build_distinct_type_copy): Likewise.
34053 (build_variant_type_copy): Likewise.
34054 * tree.c (build_qualified_type): Pass down mem-stat info.
34055 (build_distinct_type_copy): Likewise.
34056 (build_variant_type_copy): Likewise.
34058 2017-04-21 Richard Biener <rguenther@suse.de>
34060 PR tree-optimization/80237
34061 * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34063 (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34064 for a simplified result.
34066 2016-04-21 Richard Biener <rguenther@suse.de>
34068 * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34069 sth as strict as a simple_iv but a chrec without symbols and an
34070 operand defined in the loop we are peeling (and not some subloop).
34071 (propagate_constants_for_unrolling): Propagate all constants.
34073 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
34076 * config/i386/i386.c (print_reg): Remove assert for disalowed
34077 regno values, call output_operand_lossage instead.
34079 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
34082 * config/i386/constraints.md (Yc): New register constraint.
34083 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34084 Use Yc constraint for alternative 2 of operand 0. Remove
34085 preferred_for_speed attribute.
34087 2017-04-20 Alexander Monakov <amonakov@ispras.ru>
34089 * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34090 lastprivate clauses in SIMT case.
34092 2017-04-20 Volker Reichelt <v.reichelt@netcologne.de>
34094 * doc/invoke.texi (-Wextra-semi): Document new warning option.
34096 2017-04-20 Richard Biener <rguenther@suse.de>
34098 PR tree-optimization/57796
34099 * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34100 as N scalar stores.
34101 (vect_model_load_cost): Cost gathers as N scalar loads.
34103 2017-04-20 Richard Biener <rguenther@suse.de>
34105 * ggc-page.c (ggc_allocated_p): Rename to ...
34106 (safe_lookup_page_table_entry): ... this and return the lookup
34108 (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34110 2017-04-20 Richard Biener <rguenther@suse.de>
34112 PR tree-optimization/80453
34113 * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34114 * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34115 from the conditions.
34116 (vn_phi_eq): Pass them down.
34117 (vn_phi_lookup): Record them.
34118 (vn_phi_insert): Likewise.
34120 2017-04-20 Matthew Fortune <matthew.fortune@imgtec.com>
34122 * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34123 uninitialized variable warning to avoid buffer overrun.
34125 2017-04-20 Alexander Monakov <amonakov@ispras.ru>
34128 * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34129 is suppressed for '{ 0 }' in C.
34131 2017-04-20 Jakub Jelinek <jakub@redhat.com>
34133 * BASE-VER: Set to 8.0.0.
34135 2017-04-20 Thomas Preud'homme <thomas.preudhomme@arm.com>
34137 * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34138 priority .init_array and .fini_array section with SECTION_NOTYPE
34141 2017-04-20 Jakub Jelinek <jakub@redhat.com>
34143 PR middle-end/80423
34144 * tree.h (build_array_type): Add typeless_storage default argument.
34145 * tree.c (type_cache_hasher::equal): Also compare
34146 TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34147 (build_array_type): Add typeless_storage argument, set
34148 TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34150 (build_nonshared_array_type): Adjust build_array_type_1 caller.
34151 (build_array_type): Likewise. Add typeless_storage argument.
34153 2017-04-19 Eric Botcazou <ebotcazou@adacore.com>
34154 Jakub Jelinek <jakub@redhat.com>
34156 PR tree-optimization/80426
34157 * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34158 operation on symbolic operands, also compute the overflow for the
34159 invariant part when the operation degenerates into a negation.
34161 2017-04-19 Jakub Jelinek <jakub@redhat.com>
34164 * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34165 Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34168 * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34170 2017-04-19 Georg-Johann Lay <avr@gjlay.de>
34173 * config/avr/avr.c (tree.h): Include it.
34174 (cgraph.h): Include it.
34175 (avr_encode_section_info): Don't warn for uninitialized progmem
34176 variable if it's just an alias.
34178 2017-04-19 Richard Biener <rguenther@suse.de>
34181 * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34182 when needed by AutoPGO.
34184 2017-04-19 Paulo J. Matos <paulo@matos-sorge.com>
34187 * doc/lto.texi: Remove an extra 'that'.
34189 2017-04-19 Segher Boessenkool <segher@kernel.crashing.org>
34191 PR rtl-optimization/80429
34192 * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34193 are only used in debug insns.
34195 2017-04-19 Eric Botcazou <ebotcazou@adacore.com>
34196 Vladimir Makarov <vmakarov@redhat.com>
34198 * config/sparc/predicates.md (input_operand): Add comment. Return
34199 true for any memory operand when LRA is in progress.
34200 * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34202 2017-04-18 Jeff Law <law@redhat.com>
34205 * mips.md ({return,simple_return}_internal): Do not overwrite
34208 2017-04-18 Jakub Jelinek <jakub@redhat.com>
34210 PR tree-optimization/80443
34211 * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34212 instead of adding 1, subtract -1 and similarly instead of subtracting
34215 2017-04-18 Richard Sandiford <richard.sandiford@arm.com>
34217 PR rtl-optimization/80357
34218 * haifa-sched.c (tmp_bitmap): New variable.
34219 (model_recompute): Handle duplicate use records.
34220 (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34221 (free_global_sched_pressure_data): Free it.
34223 2017-04-18 Bernd Edlinger <bernd.edlinger@hotmail.de>
34226 2017-02-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
34227 * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34228 (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34229 instead of SYSTEM_HEADER_DIR.
34231 2017-04-18 Jeff Law <law@redhat.com>
34233 PR middle-end/80422
34234 * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34235 predecessors after walking up the insn chain.
34237 2017-04-18 Jakub Jelinek <jakub@redhat.com>
34240 * dwarf2out.c (modified_type_die): Try harder not to emit internal
34241 sizetype type into debug info.
34243 2017-04-18 Michael Meissner <meissner@linux.vnet.ibm.com>
34246 * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34247 unneeded test for TARGET_UPPER_REGS_SF.
34248 * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34250 2017-04-18 Jakub Jelinek <jakub@redhat.com>
34253 * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34254 instead of gsi_after_labels.
34256 2017-04-18 Jeff Law <law@redhat.com>
34258 * regcprop.c (maybe_mode_change): Avoid creating copies of the
34262 2017-04-13 Jeff Law <law@redhat.com>
34263 * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34264 in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34266 2017-04-18 Georg-Johann Lay <avr@gjlay.de>
34269 * config/avr/avr.c (intl.h): Include it.
34270 (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34272 2017-04-18 Martin Liska <mliska@suse.cz>
34274 PR gcov-profile/78783
34275 * gcov-tool.c (gcov_output_files): Validate that destination
34276 file is either removed by the tool or by a user.
34278 2017-04-14 Andrew Burgess <andrew.burgess@embecosm.com>
34279 Guy Benyei <guybe@mellanox.com>
34281 * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34282 block, and do not negate it, the stored id is already negative.
34284 2017-04-14 Andrew Burgess <andrew.burgess@embecosm.com>
34286 * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34288 2017-04-14 Michael Meissner <meissner@linux.vnet.ibm.com>
34291 * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34292 masks of options that should be turned off if the VSX vector
34293 options are turned off.
34294 (OTHER_P8_VECTOR_MASKS): Likewise.
34295 (OTHER_VSX_VECTOR_MASKS): Likewise.
34296 * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34297 rs6000_disable_incompatible_switches to validate no type switches
34299 (rs6000_incompatible_switch): New function to disallow turning on
34300 other vector options if -mno-vsx, -mno-power8-vector, or
34301 -mno-power9-vector are specified.
34303 2017-04-14 Claudiu Zissulescu <claziss@synopsys.com>
34305 * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34307 2017-04-14 Claudiu Zissulescu <claziss@synopsys.com>
34309 * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34310 * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34311 * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34312 (ARG_POINTER_CFA_OFFSET): Likewise.
34314 2017-04-14 Claudiu Zissulescu <claziss@synopsys.com>
34316 * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34317 conditions to take advantage of various optimizations.
34319 2017-04-13 Jeff Law <law@redhat.com>
34321 * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34322 in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34323 (zero_extendsidi2_dext): Likewise.
34325 2017-04-13 Jakub Jelinek <jakub@redhat.com>
34328 * fold-const.c (fold_ternary_loc): Revert
34329 use op0 instead of fold_convert_loc (loc, type, arg0) part of
34332 2017-04-13 Vladimir Makarov <vmakarov@redhat.com>
34334 PR rtl-optimization/80343
34335 * lra-remat.c (update_scratch_ops): Assign original hard reg to
34336 new scratch pseudo.
34338 2017-04-13 Denis Khalikov <d.khalikov@partner.samsung.com>
34341 * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34342 to ubsan_encode_value.
34344 2017-04-13 Jeff Law <law@redhat.com>
34346 * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34347 appearing in DEBUG_INSNs.
34349 2017-04-13 Martin Liska <mliska@suse.cz>
34351 PR gcov-profile/80413
34352 * gcov-io.c (gcov_write_string): Copy to buffer just when
34353 allocated size is greater than zero.
34355 2017-04-13 Jakub Jelinek <jakub@redhat.com>
34358 * dwarf2out.c (decls_for_scope): Ignore declarations of
34359 current_function_decl in BLOCK_NONLOCALIZED_VARS.
34361 2017-04-12 Jan Hubicka <hubicka@ucw.cz>
34364 * ipa-visibility.c (non_local_p): Fix typos.
34365 (localize_node): When localizing symbol in same comdat group,
34366 dissolve the group only when we know external symbols are going
34368 (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34370 2017-04-12 Jakub Jelinek <jakub@redhat.com>
34372 PR tree-optimization/79390
34373 * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34374 order does not result in usable sequence, retry with reversed operand
34380 * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34381 to fold_build2_loc. Convert TREE_OPERAND (tem, 0) to type. Use
34382 op0 instead of fold_convert_loc (loc, type, arg0).
34384 2017-04-12 Jeff Law <law@redhat.com>
34386 * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34387 has a delay slot in the generated code.
34389 * config/cris/cris.md (cris_preferred_reload_class): Return
34390 GENNONACR_REGS rather than GENERAL_REGS.
34392 2017-04-12 Jakub Jelinek <jakub@redhat.com>
34395 * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34396 vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34397 signedness of the result type.
34399 2017-04-12 Richard Biener <rguenther@suse.de>
34400 Jeff Law <law@redhat.com>
34402 PR tree-optimization/80359
34403 * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34404 trim stores to TARGET_MEM_REFs.
34406 2017-04-12 Richard Biener <rguenther@suse.de>
34408 PR tree-optimization/79390
34409 * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34410 threading case even more.
34412 2017-04-12 Segher Boessenkool <segher@kernel.crashing.org>
34415 * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34416 for quad_address_p for TImode, instead of just not indexed_address.
34418 2017-04-12 Richard Biener <rguenther@suse.de>
34419 Bernd Edlinger <bernd.edlinger@hotmail.de>
34421 PR middle-end/79671
34422 * alias.c (component_uses_parent_alias_set_from): Handle
34423 TYPE_TYPELESS_STORAGE.
34424 (get_alias_set): Likewise.
34425 * tree-core.h (tree_type_common): Add typeless_storage flag.
34426 * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34427 * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34428 for types containing members with TYPE_TYPELESS_STORAGE.
34429 (place_field): Likewise.
34430 (layout_type): Likewise for ARRAY_TYPE.
34431 * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34432 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34433 TYPE_TYPELESS_STORAGE.
34434 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34436 2017-04-12 Jakub Jelinek <jakub@redhat.com>
34439 * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34440 first argument to type.
34442 2017-04-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
34446 * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34447 CONST0_RTX (mode) rather than const0_rtx where appropriate.
34448 (rs6000_expand_binop_builtin): Likewise.
34449 (rs6000_expand_ternop_builtin): Likewise; also add missing
34450 vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34451 vshasigma built-ins.
34452 * doc/extend.texi: Document that vec_xxpermdi's third argument
34453 must be a constant.
34455 2017-04-11 Uros Bizjak <ubizjak@gmail.com>
34457 * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34458 Use shift_const cost parameter when calculating gain of STV shifts.
34460 2017-04-11 Vladimir Makarov <vmakarov@redhat.com>
34462 PR rtl-optimization/70478
34463 * lra-constraints.c (process_alt_operands): Check memory for
34464 disfavoring memory insn operand.
34466 2017-04-11 Jakub Jelinek <jakub@redhat.com>
34468 PR middle-end/80100
34469 * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34470 left shift in unsigned HOST_WIDE_INT type.
34472 PR rtl-optimization/80385
34473 * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34474 (not (neg X)) into (plus X -1) for complex or non-integral modes.
34477 * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34478 if they have any depend clauses.
34480 2017-04-11 Martin Liska <mliska@suse.cz>
34483 * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34484 * ipa-split.c (split_function): Create a local comdat symbol
34485 if caller is in a comdat group.
34487 2017-04-11 Martin Liska <mliska@suse.cz>
34490 * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34493 2017-04-11 Martin Sebor <msebor@redhat.com>
34495 PR middle-end/80364
34496 * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34497 always use the int type. Use INTEGRAL_TYPE_P() rather than testing
34499 (directive::set_width, directive::set_precision, format_character):
34501 (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34504 2017-04-11 Richard Earnshaw <rearnsha@arm.com>
34507 * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34508 conflict, set target->arch_name instead of target->cpu_name.
34510 2017-04-11 Richard Biener <rguenther@suse.de>
34512 PR tree-optimization/80374
34513 * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34514 build_zero_cst, remove fold_convertible_p check again.
34516 2017-04-11 Martin Liska <mliska@suse.cz>
34519 * ubsan.c (instrument_object_size): Do not instrument register
34522 2017-04-11 Jakub Jelinek <jakub@redhat.com>
34525 * config/i386/i386-builtin-types.def
34526 (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34527 V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34528 V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34529 V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34530 V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34531 V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34532 V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34533 V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34534 V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34535 V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34536 V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34537 V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34538 V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34539 V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34540 V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34541 V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34542 V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34543 V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34544 * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34545 __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34546 __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34547 __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34548 __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34549 __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34550 __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34551 __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34552 __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34553 __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34554 __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34555 __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34556 __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34557 __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34558 __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34559 __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34560 __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34561 __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34562 __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34563 __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34564 __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34565 __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34566 __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34567 __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34568 __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34569 __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34570 __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34571 __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34573 * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34574 flag to second_arg_count, handle 4 argument function type _COUNT
34575 aliases, handle second_arg_count on second argument rather than last.
34577 2017-04-10 Jeff Law <law@redhat.com>
34579 PR tree-optimization/80374
34580 * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34581 record anything if we can not convert integer_zero_node to the
34584 2017-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
34587 * config/rs6000/rs6000.c (rs6000_option_override_internal):
34588 Enhance special handling given to the TARGET_P9_MINMAX option in
34589 relation to certain other options.
34591 2017-04-10 Bin Cheng <bin.cheng@arm.com>
34593 PR tree-optimization/80153
34594 * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34595 remove POINTER_PLUS_EXPR's base part directly, rather than through
34598 2017-04-10 Richard Biener <rguenther@suse.de>
34599 Bin Cheng <bin.cheng@arm.com>
34601 PR tree-optimization/80153
34602 * tree-affine.c (aff_combination_to_tree): Get base pointer from
34603 the first element of pointer type aff_tree. Build result expr in
34605 (add_elt_to_tree): Convert to type unconditionally. Remove other
34606 fold_convert calls.
34607 * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34608 (rewrite_use_nonlinear_expr): Check invariant using iv information.
34610 2017-04-10 Richard Biener <rguenther@suse.de>
34612 * tree-ssa-structalias.c (find_func_aliases): Properly handle
34615 2017-04-10 Vladimir Makarov <vmakarov@redhat.com>
34617 PR rtl-optimization/70478
34618 * lra-constraints.c (curr_small_class_check): New.
34619 (update_and_check_small_class_inputs): New.
34620 (process_alt_operands): Update curr_small_class_check. Disfavor
34621 alternative insn memory operands. Check available regs for small
34624 2017-03-31 Matthew Fortune <matthew.fortune@imgtec.com>
34627 * config/mips/mips.opt (-mvirt): Update description.
34628 * doc/invoke.texi (-mvirt): Likewise.
34630 2017-04-10 Richard Biener <rguenther@suse.de>
34632 PR middle-end/80362
34633 * fold-const.c (fold_binary_loc): Look at unstripped ops when
34634 looking for NEGATE_EXPR in -A / -B to A / B folding.
34636 2017-04-10 Martin Liska <mliska@suse.cz>
34638 PR gcov-profile/80224
34639 * gcov.c (print_usage): Fix usage string.
34640 (get_gcov_intermediate_filename): Remove.
34641 (output_gcov_file): Use both for normal and intermediate format.
34642 (generate_results): Do not initialize special file for
34643 intermediate format.
34645 2017-04-10 Richard Biener <rguenther@suse.de>
34647 PR tree-optimization/80304
34648 * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34651 2017-04-10 Nathan Sidwell <nathan@acm.org>
34654 * config/rs6000/rs6000.c (rs6000_vector_type): New.
34655 (rs6000_init_builtins): Use it.
34657 2016-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34659 * config/arm/arm.md (<mrc>): Add mode to SET source.
34660 (<mrrc>): Likewise.
34662 2017-04-10 Richard Biener <rguenther@suse.de>
34664 PR middle-end/80344
34665 * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34667 2017-04-10 Jakub Jelinek <jakub@redhat.com>
34670 * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34671 _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34672 _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34673 _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34674 _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34675 _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34676 _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34677 _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34678 _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34679 _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34680 _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34681 _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34682 _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34683 _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34684 _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34685 _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34686 _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34687 _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34688 _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34689 _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34690 _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34691 _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34692 _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34694 2017-04-08 Vladimir Makarov <vmakarov@redhat.com>
34696 PR rtl-optimization/70478
34697 * lra-constraints.c: Reverse the last patch.
34699 2017-04-08 Andreas Tobler <andreast@gcc.gnu.org>
34701 * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34702 Add comment for WCHAR_T.
34704 2017-04-08 Martin Liska <mliska@suse.cz>
34707 2017-04-07 Martin Liska <mliska@suse.cz>
34710 * ipa-split.c (split_function): Add function part to a same comdat
34713 2017-04-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
34716 * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34718 2017-04-07 Pat Haugen <pthaugen@us.ibm.com>
34720 * rs6000/rs6000.c (vec_load_pendulum): Rename...
34721 (vec_pairing): ...to this.
34722 (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34723 (rs6000_sched_init): Adjust for name change.
34724 (struct rs6000_sched_context): Likewise.
34725 (rs6000_init_sched_context): Likewise.
34726 (rs6000_set_sched_context): Likewise.
34728 2017-04-07 Jakub Jelinek <jakub@redhat.com>
34734 * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34736 * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34737 _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34738 _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34740 2017-04-07 Andreas Tobler <andreast@gcc.gnu.org>
34742 * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34744 2017-04-07 Vladimir Makarov <vmakarov@redhat.com>
34746 PR rtl-optimization/70703
34747 * ira-color.c (update_conflict_hard_regno_costs): Use
34748 int64_t instead of HOST_WIDE_INT.
34750 2017-04-07 Vladimir Makarov <vmakarov@redhat.com>
34752 PR rtl-optimization/70478
34753 * lra-constraints.c (process_alt_operands): Disfavor alternative
34754 insn memory operands.
34756 2017-04-07 Jeff Law <law@redhat.com>
34758 * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34759 CALL and NOTE_INSN_CALL_ARG_LOCATION.
34761 2017-04-07 Martin Liska <mliska@suse.cz>
34764 * config/aarch64/aarch64.c (aarch64_process_target_attr):
34765 Show error message instead of an ICE.
34767 2017-04-07 Martin Liska <mliska@suse.cz>
34770 * ipa-split.c (split_function): Add function part to a same comdat
34773 2017-04-07 Richard Biener <rguenther@suse.de>
34775 PR middle-end/80341
34776 * tree.c (get_unwidened): Also handle ! for_type case for
34778 * convert.c (do_narrow): Split out from ...
34779 (convert_to_integer_1): ... here. Do not pass final truncation
34780 type to get_unwidened for TRUNC_DIV_EXPR.
34782 2017-04-07 Richard Biener <rguenther@suse.de>
34784 * tree-affine.c (wide_int_ext_for_comb): Take type rather
34786 (aff_combination_const): Adjust.
34787 (aff_combination_scale): Likewise.
34788 (aff_combination_add_elt): Likewise.
34789 (aff_combination_add_cst): Likewise.
34790 (aff_combination_convert): Likewise.
34791 (add_elt_to_tree): Likewise. Remove unused argument.
34792 (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34794 2017-04-07 Sebastian Huber <sebastian.huber@embedded-brains.de>
34796 * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34798 * config/arm/arm.c (arm_default_short_enums): Use
34799 ARM_DEFAULT_SHORT_ENUMS.
34800 * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34802 2017-04-06 Jakub Jelinek <jakub@redhat.com>
34805 * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34806 members with redundant out-of-class redeclaration.
34808 2017-04-06 Uros Bizjak <ubizjak@gmail.com>
34811 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34812 * config/i386/i386.md (*zero_extendsidi2):
34813 Add (?*x,*x) and (?*v,*v) alternatives.
34815 2017-04-06 Uros Bizjak <ubizjak@gmail.com>
34818 * config/i386/i386.c (ix86_expand_builtin)
34819 <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34820 mode from insn data. Convert operands to insn operand mode.
34821 Copy operands that don't satisfy insn predicate to a register.
34823 2017-04-06 Sam Thursfield <sam.thursfield@codethink.co.uk>
34825 * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34828 2017-04-06 Richard Biener <rguenther@suse.de>
34830 PR tree-optimization/80334
34831 * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34832 preserve alignment of accesses.
34834 2017-04-06 Richard Biener <rguenther@suse.de>
34836 PR tree-optimization/80262
34837 * tree-sra.c (build_ref_for_offset): Preserve address-space
34839 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34840 Drop useless address-space information on MEM_REF offsets.
34842 2017-04-05 Andreas Schwab <schwab@linux-m68k.org>
34844 * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
34846 2017-04-05 Vladimir Makarov <vmakarov@redhat.com>
34848 PR rtl-optimization/70703
34849 * ira-color.c (update_conflict_hard_regno_costs): Use
34850 HOST_WIDE_INT instead of long.
34852 2017-04-05 Uros Bizjak <ubizjak@gmail.com>
34855 * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
34856 not defined for x86_64 target. Add -mmmx target option when __SSE2__
34858 * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
34859 for x86_64 target. Handle -m3dnowa option.
34861 2017-04-05 Vladimir Makarov <vmakarov@redhat.com>
34863 PR rtl-optimization/70703
34864 * ira-color.c (update_costs_from_allocno): Use the smallest mode.
34865 (update_conflict_hard_regno_costs): Use long instead of unsigned
34866 arithmetic for cost calculation.
34868 2017-04-05 Jakub Jelinek <jakub@redhat.com>
34869 Bernd Edlinger <bernd.edlinger@hotmail.de>
34872 * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
34875 2017-04-05 Eric Botcazou <ebotcazou@adacore.com>
34878 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
34879 ptr_mode with Pmode throughout.
34880 * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
34881 into probe_stack_range and use DImode.
34883 2017-04-05 Dominik Vogt <vogt@linux.vnet.ibm.com>
34886 * config/s390/s390.c (s390_register_info_gprtofpr): Return if
34887 call_eh_return is true.
34889 2017-04-05 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
34891 * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
34892 Initialize last_match_fntype_index.
34894 2017-04-05 Jakub Jelinek <jakub@redhat.com>
34897 * tree-nvr.c: Include internal-fn.h.
34898 (pass_return_slot::execute): Ignore internal calls without
34901 2017-04-04 Jakub Jelinek <jakub@redhat.com>
34902 Richard Biener <rguenther@suse.de>
34905 * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
34906 captures used multiple times, except for the last use.
34907 * generic-match-head.c: Include gimplify.h.
34909 2017-04-04 Jakub Jelinek <jakub@redhat.com>
34911 PR tree-optimization/79390
34912 * target.h (struct noce_if_info): Declare.
34913 * targhooks.h (default_noce_conversion_profitable_p): Declare.
34914 * target.def (noce_conversion_profitable_p): New target hook.
34915 * ifcvt.h (struct noce_if_info): New type, moved from ...
34916 * ifcvt.c (struct noce_if_info): ... here.
34917 (noce_conversion_profitable_p): Renamed to ...
34918 (default_noce_conversion_profitable_p): ... this. No longer
34920 (noce_try_store_flag_constants, noce_try_addcc,
34921 noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
34922 noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
34923 instead of noce_conversion_profitable_p.
34924 * config/i386/i386.c: Include ifcvt.h.
34925 (ix86_option_override_internal): Don't override
34926 PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
34927 (ix86_noce_conversion_profitable_p): New function.
34928 (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
34929 * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
34930 * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
34931 * doc/tm.texi: Regenerated.
34933 2017-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
34935 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
34938 2017-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
34941 * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
34942 instructions for small multiply cores.
34944 2017-04-04 Jeff Law <law@redhat.com>
34946 * config/mips/mips.c (mips_multi_add): Zero initialize the newly
34948 (mips_expand_vec_perm_const): Initialize elements in orig_perm
34949 that are not set by the loop over the elements.
34951 2017-04-04 Jakub Jelinek <jakub@redhat.com>
34954 * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
34955 int mode, convert_modes it to mode as unsigned, otherwise use
34956 lowpart_subreg to mode rather than SImode.
34957 * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
34958 ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
34959 Use DImode instead of SImode for the shift count operand.
34960 * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
34963 2017-04-04 Richard Biener <rguenther@suse.de>
34965 PR middle-end/80281
34966 * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
34967 arithmetic done for the negate or the plus. Simplify.
34968 (A - (-B) -> A + B): Likewise.
34969 * fold-const.c (split_tree): Make sure to not negate pointers.
34971 2017-04-04 Segher Boessenkool <segher@kernel.crashing.org>
34973 PR rtl-optimization/60818
34974 * simplify-rtx.c (simplify_binary_operation_1): Do not replace
34975 a compare of comparisons with the thing compared if this results
34976 in a different machine mode.
34978 2017-04-03 Jonathan Wakely <jwakely@redhat.com>
34980 * alias.c (base_alias_check): Fix typo in comment.
34981 * cgraph.h (class ipa_polymorphic_call_context): Likewise.
34982 * cgraphunit.c (symbol_table::compile): Likewise.
34983 * collect2.c (maybe_run_lto_and_relink): Likewise.
34984 * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
34985 * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
34986 * config/avr/avr.c (avr_map_op_t): Likewise.
34987 * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
34988 * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
34989 * config/epiphany/epiphany.md (movcc): Likewise.
34990 * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
34991 * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
34993 * config/mips/mips.c (mips_save_restore_reg): Likewise.
34994 * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
34995 * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
34996 * config/sh/sh.c (sh_rtx_costs): Likewise.
34997 * fold-const.c (fold_truth_andor): Likewise.
34998 * genautomata.c (collapse_flag): Likewise.
34999 * gengtype.h (struct type::u::s): Likewise.
35000 * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35001 * input.c (FORMAT_AMOUNT): Likewise.
35002 * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35003 (known_aggs_to_agg_replacement_list): Likewise.
35004 * ipa-inline-analysis.c: Likewise.
35005 * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35006 * ipa-polymorphic-call.c
35007 (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35008 * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35009 * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35011 * modulo-sched.c (apply_reg_moves): Likewise.
35012 * omp-expand.c (build_omp_regions_1): Likewise.
35013 * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35014 * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35015 * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35016 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35017 * value-prof.c: Likewise.
35018 * var-tracking.c (val_reset): Likewise.
35020 2017-04-03 Richard Biener <rguenther@suse.de>
35022 PR tree-optimization/80275
35023 * fold-const.c (split_address_to_core_and_offset): Handle
35026 2017-04-03 Eric Botcazou <ebotcazou@adacore.com>
35028 * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35029 descriptors is at least equal to that of functions.
35031 2017-04-02 Uros Bizjak <ubizjak@gmail.com>
35033 * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35035 2017-04-02 Uros Bizjak <ubizjak@gmail.com>
35038 * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35039 (mov<IMOD4:mode>): New expander.
35040 (*mov<IMOD4:mode>_internal): New insn and split pattern.
35042 2017-03-31 Segher Boessenkool <segher@kernel.crashing.org>
35044 PR rtl-optimization/79405
35045 * fwprop.c (propagations_left): New variable.
35046 (forward_propagate_into): Decrement it.
35047 (fwprop_init): Initialize it.
35048 (fw_prop): If the variable has reached zero, stop propagating.
35049 (fwprop_addr): Ditto.
35051 2017-03-31 Jakub Jelinek <jakub@redhat.com>
35054 * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35055 a FUNCTION_DECL, pass it as decl instead of origin to
35058 2017-03-31 Alexander Monakov <amonakov@ispras.ru>
35060 * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35063 2017-03-31 Pat Haugen <pthaugen@us.ibm.com>
35066 * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35067 TARGET_VSX_SMALL_INTEGER.
35069 2017-03-31 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35071 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35072 reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35074 2017-03-31 Matthew Fortune <matthew.fortune@imgtec.com>
35076 * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35077 extraction from odd-numbered MSA register.
35079 2017-03-31 Jakub Jelinek <jakub@redhat.com>
35081 PR middle-end/80173
35082 * expmed.c (store_bit_field_1): Don't attempt to create
35083 a word subreg out of hard registers wider than word if they
35084 have HARD_REGNO_NREGS of 1 for their mode.
35086 PR middle-end/80163
35087 * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35088 conversions to integer types wider than word and pointer.
35091 * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35092 (rtx_equal_for_cselib_p): Pass 0 to it.
35093 * cselib.c (cselib_hasher::equal): Likewise.
35094 (rtx_equal_for_cselib_1): Add depth argument. If depth
35095 is 128, don't look up VALUE locs and punt. Increment
35096 depth in recursive calls when walking VALUE locs.
35098 2017-03-31 Bernd Edlinger <bernd.edlinger@hotmail.de>
35100 * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35101 (make_gcov_file_name): Use the canonical path name for generating
35103 (read_line): Fix handling of files with ascii null bytes.
35105 2017-03-30 Matthew Fortune <matthew.fortune@imgtec.com>
35107 * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35108 to initialise a vector register instead
35109 of using a const_int.
35111 2017-03-30 Jakub Jelinek <jakub@redhat.com>
35113 PR translation/80189
35114 * gimplify.c (omp_default_clause): Use %qs instead of %s in
35115 diagnostic messages.
35117 2017-03-30 Peter Bergner <bergner@vnet.ibm.com>
35120 * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35121 (dfp_diex_<mode>): Update mode of operand 1.
35122 * doc/extend.texi (dxex, dxexq): Document change to return type.
35123 (diex, diexq): Document change to argument type.
35125 2017-03-30 Martin Jambor <mjambor@suse.cz>
35128 * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35129 * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35130 it reflects the signature changes performed at the callee side.
35131 * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35132 to cgraph_build_function_type_skip_args.
35133 (build_function_decl_skip_args): Adjust call to the above function.
35135 2017-03-30 Jakub Jelinek <jakub@redhat.com>
35138 * config/i386/sse.md
35139 (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35140 register as dest whenever it is a MEM not rtx_equal_p to the
35141 corresponding dup operand, and when forcing into reg move the
35142 reg into the memory afterwards.
35143 (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35144 Likewise. Use <ssehalfvecmode> instead of <ssequartermode>
35145 for the force_reg mode.
35146 (avx512vl_vextractf128<mode>): Use register as dest either
35147 always when a MEM, or when it is a MEM not rtx_equal_p to the
35148 corresponding dup operand, or even not when it is a CONST_VECTOR
35149 depending on the mode and lo vs. hi.
35150 (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35152 (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35153 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35154 Likewise. Require that operands[2] is even.
35155 (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35156 Remove extraneous parens. Require that operands[2] is a multiple
35158 (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35159 operands[0] is a MEM if <mask_applied>, the predicates/constraints
35160 disallow memory then.
35162 2017-03-30 Richard Biener <rguenther@suse.de>
35164 PR tree-optimization/77498
35165 * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35166 to non-constants over backedges.
35168 2017-03-29 Segher Boessenkool <segher@kernel.crashing.org>
35170 PR rtl-optimization/80233
35171 * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35172 as last_combined_insn. Do not test for BARRIER_P separately.
35174 2017-03-29 Andreas Schwab <schwab@suse.de>
35177 * calls.c (prepare_call_address): Convert funexp to Pmode before
35178 copying to temp reg.
35180 2017-03-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35182 PR tree-optimization/80158
35183 * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35184 Handle possible future case of more than one alternate
35186 (replace_rhs_if_not_dup): Likewise.
35187 (replace_one_candidate): Likewise.
35189 2017-03-28 Vladimir Makarov <vmakarov@redhat.com>
35191 PR rtl-optimization/80193
35192 * ira.c (ira): Do not check allocation for LRA.
35194 2017-03-28 Alexander Monakov <amonakov@ispras.ru>
35196 * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35197 (nvptx_output_simt_exit): Declare.
35198 * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35199 cfun->machine->unisimt_location. Handle NULL unisimt_predicate.
35200 (init_softstack_frame): Move initialization of crtl->is_leaf to...
35201 (nvptx_declare_function_name): ...here. Emit declaration of local
35202 memory space buffer for omp_simt_enter insn.
35203 (nvptx_output_unisimt_switch): New.
35204 (nvptx_output_softstack_switch): New.
35205 (nvptx_output_simt_enter): New.
35206 (nvptx_output_simt_exit): New.
35207 * config/nvptx/nvptx.h (struct machine_function): New fields
35208 has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35209 * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35210 (UNSPECV_SIMT_EXIT): Ditto.
35211 (omp_simt_enter_insn): New insn.
35212 (omp_simt_enter): New expansion.
35213 (omp_simt_exit): New insn.
35214 * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35216 * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35217 (expand_GOMP_SIMT_ENTER_ALLOC): New.
35218 (expand_GOMP_SIMT_EXIT): New.
35219 * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35220 (GOMP_SIMT_ENTER_ALLOC): Ditto.
35221 (GOMP_SIMT_EXIT): Ditto.
35222 * target-insns.def (omp_simt_enter): New insn.
35223 (omp_simt_exit): Ditto.
35224 * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35226 (lower_rec_simd_input_clauses): Implement SIMT privatization.
35227 (lower_rec_input_clauses): Likewise.
35228 (lower_lastprivate_clauses): Handle SIMT privatization.
35230 * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35231 (ompdevlow_adjust_simt_enter): New.
35232 (find_simtpriv_var_op): New.
35233 (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35234 IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35236 * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35237 * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35238 (copy_decl_for_dup_finish): Ditto.
35240 * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35242 2017-03-28 Uros Bizjak <ubizjak@gmail.com>
35245 * config/i386/i386.c (ix86_option_override_internal): Always
35246 allow -mpreferred-stack-boundary=3 for 64-bit targets.
35248 2017-03-28 Bin Cheng <bin.cheng@arm.com>
35250 * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35252 2017-03-28 Bin Cheng <bin.cheng@arm.com>
35254 * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35255 mark new edge's irreducible flag accordign to it.
35256 (vect_do_peeling): Check loop preheader edge's irreducible flag
35257 and pass it to function slpeel_add_loop_guard.
35259 2017-03-28 Richard Sandiford <richard.sandiford@arm.com>
35261 PR tree-optimization/80218
35262 * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35263 Update block frequencies and counts.
35265 2017-03-28 Richard Biener <rguenther@suse.de>
35267 PR tree-optimization/78644
35268 * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35269 of a simplification result we may not use it at all.
35271 2017-03-28 Richard Biener <rguenther@suse.de>
35274 * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35275 without arguments, generate default definition of a SSA name.
35277 2017-03-28 Richard Biener <rguenther@suse.de>
35279 PR middle-end/80222
35280 * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35281 TYPE_REF_CAN_ALIAS_ALL references.
35282 * fold-const.c (fold_indirect_ref_1): Likewise.
35284 2017-03-28 Martin Liska <mliska@suse.cz>
35287 * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35288 thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35290 2017-03-28 Claudiu Zissulescu <claziss@synopsys.com>
35291 Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
35293 * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35294 (EXTRA_SPECS): Define.
35295 (SUBTARGET_EXTRA_SPECS): Likewise.
35296 (SUBTARGET_CPP_SPEC): Likewise.
35297 * config/arc/elf.h (EXTRA_SPECS): Renamed to
35298 SUBTARGET_EXTRA_SPECS.
35299 * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35301 2017-03-28 Claudiu Zissulescu <claziss@synopsys.com>
35303 * config/arc/simdext.md (vst64_insn): Update pattern.
35304 (vld32wh_insn): Likewise.
35305 (vld32wl_insn): Likewise.
35306 (vld64_insn): Likewise.
35307 (vld32_insn): Likewise.
35309 2017-03-28 Marek Polacek <polacek@redhat.com>
35312 * fold-const.c (fold_comparison): Use protected_set_expr_location
35313 instead of SET_EXPR_LOCATION.
35315 2017-03-28 Markus Trippelsdorf <markus@trippelsdorf.de>
35317 * tree.c (add_expr): Avoid name lookup warning.
35319 2017-03-27 Jeff Law <law@redhat.com>
35321 PR tree-optimization/80216
35322 * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35323 function name. Limit recursion depth.
35324 (record_temporary_equivalences): Corresponding changes.
35326 2017-03-27 Jonathan Wakely <jwakely@redhat.com>
35328 * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35331 2017-03-27 Jakub Jelinek <jakub@redhat.com>
35334 * reg-notes.def (REG_CFA_NOTE): Define. Use it for CFA related
35336 * cfgcleanup.c (reg_note_cfa_p): New array.
35337 (insns_have_identical_cfa_notes): New function.
35338 (old_insns_match_p): Don't cross-jump in between /f
35339 and non-/f instructions. If both i1 and i2 are frame related,
35340 verify all CFA notes, their order and content.
35342 2017-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
35345 * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35346 HImode and SImode with zero extend to DImode to one insn.
35347 (bswap<mode>2_extenddi): Likewise.
35348 (bswapsi2_extenddi): Likewise.
35349 (bswaphi2_extendsi): Likewise.
35350 (bswaphi2): Combine bswap HImode and SImode into one insn.
35351 Separate memory insns from swapping register.
35352 (bswapsi2): Likewise.
35353 (bswap<mode>2): Likewise.
35354 (bswaphi2_internal): Delete, no longer used.
35355 (bswapsi2_internal): Likewise.
35356 (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35357 store, and gpr<-gpr swap insns.
35358 (bswap<mode>2_store): Likewise.
35359 (bswaphi2_reg): Register only splitter, combine with the splitter.
35360 (bswaphi2 splitter): Likewise.
35361 (bswapsi2_reg): Likewise.
35362 (bswapsi2 splitter): Likewise.
35363 (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35364 the insns into load, store, and register/register insns.
35365 (bswapdi2_ldbrx): Likewise.
35366 (bswapdi2_load): Likewise.
35367 (bswapdi2_store): Likewise.
35368 (bswapdi2_reg): Likewise.
35370 2017-03-27 Gunther Nikl <gnikl@users.sourceforge.net>
35372 * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35373 (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35375 2017-03-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
35378 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35380 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35381 special handling for target option conflicts between dform
35382 options (-mpower9-dform, -mpower9-dform-vector,
35383 -mpower9-dform-scalar) and -mno-direct-move.
35385 2017-03-27 Richard Biener <rguenther@suse.de>
35387 PR tree-optimization/80181
35388 * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35390 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
35392 * config/arc/predicates.md (move_double_src_operand): Replace the
35393 call to move_double_src_operand with a call to address_operand.
35395 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
35397 * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35398 * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35399 * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35401 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
35403 * config/arc/predicates.md (long_immediate_loadstore_operand):
35404 Consider scaled addresses cases.
35406 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
35408 * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35409 restored when in interrupt.
35410 * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35411 doesn't have delay slot.
35413 2017-03-27 Richard Biener <rguenther@suse.de>
35416 * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35417 inlined thunk clones.
35419 2017-03-27 Jakub Jelinek <jakub@redhat.com>
35422 * asan.c (instrument_derefs): Copy over last operand from
35423 original COMPONENT_REF to the new COMPONENT_REF with
35424 DECL_BIT_FIELD_REPRESENTATIVE.
35425 * ubsan.c (instrument_object_size): Likewise.
35427 2017-03-27 Richard Biener <rguenther@suse.de>
35429 PR tree-optimization/80170
35430 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35431 sure DR/SCEV didnt fold in constants we do not see when looking
35432 at the reference base alignment.
35434 2017-03-27 Richard Biener <rguenther@suse.de>
35436 PR middle-end/80171
35437 * gimple-fold.c (fold_ctor_reference): Properly guard against
35438 NULL return value from canonicalize_constructor_val.
35440 2017-03-25 Uros Bizjak <ubizjak@gmail.com>
35443 * config/i386/i386.c (ix86_expand_builtin)
35444 <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35445 flags reg setting and flags reg using instructions.
35446 <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto. Use non-flags reg
35447 clobbering instructions to zero extend op2.
35449 2017-03-25 Gerald Pfeifer <gerald@pfeifer.com>
35451 * doc/install.texi (Configuration) <--with-aix-soname>:
35452 Update link to AIX ld.
35454 2017-03-25 Bernd Schmidt <bschmidt@redhat.com>
35456 PR rtl-optimization/80160
35457 PR rtl-optimization/80159
35458 * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35459 reg_alternate_class into account.
35461 2017-03-24 Vladimir Makarov <vmakarov@redhat.com>
35464 * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35465 to consider in curr_insn_transform.
35467 2017-03-24 Jakub Jelinek <jakub@redhat.com>
35469 * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35470 * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35471 and emit_mode_inner.
35473 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35475 * config/s390/s390-builtins.def: Add VXE builtins. Add a flags
35476 argument to the overloaded builtin variants. Use the new flag to
35477 deprecate certain builtin variants.
35478 * config/s390/s390-builtin-types.def: Add new builtin types.
35479 * config/s390/s390-builtins.h: Support new flags field for
35480 overloaded builtins.
35481 * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35482 (s390_macro_to_expand): Enable vector float data type.
35483 (s390_cpu_cpp_builtins_internal): Indicate support of the new
35484 builtins by incrementing the __VEC__ version number.
35485 (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35487 (s390_resolve_overloaded_builtin): Emit error messages depending
35488 on the builtin flags.
35489 * config/s390/s390.c (s390_expand_builtin): Support additional
35490 flags argument. Change error message to match the messages
35491 emitted in s390-c.c.
35492 * config/s390/s390.md: New UNSPEC_* constants.
35493 (op_type): Add new instruction types.
35494 * config/s390/vecintrin.h: Add new builtins and test data class
35496 * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35497 (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35498 (VEC_INEXACT, VEC_NOINEXACT): New constants.
35499 ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35500 ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35501 ("vec_mergel<mode>"): V_HW -> VEC_HW.
35503 ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35504 ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35505 ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35506 ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35508 ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35509 ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35510 ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35511 ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35513 ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35514 ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35515 ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35516 ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35517 ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35518 ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35519 ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35521 ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35522 ("vec_scatter_element<V_HW_4:mode>_DI")
35523 ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35524 ("vec_fpint<mode>", "vflls")
35525 ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35526 ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35527 ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35528 ("*vec_cmphe<mode>_cc"): ... these.
35530 ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35531 mode constant instead of magic value.
35533 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35535 * config/s390/s390.c (s390_expand_vec_compare): Support other
35536 vector floating point modes than just V2DF.
35537 (s390_expand_vcond): Likewise.
35538 (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35539 (s390_cannot_change_mode_class): Prevent mode changes between TF
35540 and V1TF in vector registers.
35541 * config/s390/s390.md (DF, SF): New mode attributes.
35542 ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35543 ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35544 SFmode support for VRs.
35545 * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35547 (VFT, VF_HW): New mode iterators.
35548 (vw, sdx): New mode attributes.
35549 ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35550 ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35551 ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35552 ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35553 ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35554 also the new vector floating point modes. Renaming to ...
35556 ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35557 ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35558 ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35559 ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35560 ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35561 ("vec_unordered<mode>"): ... these.
35563 ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35564 ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35565 ("*vec_extendv2df"): New insn definitions.
35567 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35569 * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35570 ("mulditi3_2", "*muldi3_sign"): New patterns.
35571 ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35572 rename the pattern definition.
35574 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35576 * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35578 ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35580 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35582 * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35583 instruction if possible.
35584 * config/s390/vector.md (vec_halfnumelts): New mode
35586 ("*vec_vllezlf<mode>"): New pattern.
35588 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35590 * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35591 ("popcountv4si2", "popcountv2di2"): Rename to ...
35592 ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35593 ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35595 ("popcount<mode>2_vxe"): New pattern.
35597 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35599 * common/config/s390/s390-common.c (processor_flags_table): Add
35601 * config.gcc: Add arch12.
35602 * config/s390/driver-native.c (s390_host_detect_local_cpu):
35603 Default to arch12 for unknown CPU model numbers.
35604 * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35605 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35606 PROCESSOR_max sanity check.
35607 * config/s390/s390-opts.h (enum processor_type): Add
35609 * config/s390/s390.c (processor_table): Add arch12.
35610 (s390_expand_builtin): Add check for B_VXE flag.
35611 (s390_issue_rate): Add PROCESSOR_ARCH12.
35612 (s390_get_sched_attrmask): Likewise.
35613 (s390_get_unit_mask): Likewise.
35614 (s390_sched_score): Enable z13 scheduling for arch12.
35615 (s390_sched_reorder): Likewise.
35616 (s390_sched_variable_issue): Likewise.
35617 * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35619 (s390_tune_attr): Use z13 scheduling also for arch12.
35620 (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35621 (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35622 (TARGET_VXE_P): New macros.
35623 * config/s390/s390.md: Add arch12 to cpu attribute. Add arch12
35624 and vxe to cpu_facility. Add arch12 and vxe to enabled attribute.
35625 * config/s390/s390.opt: Add arch12 as processor_type.
35627 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35629 * config/s390/s390.md
35630 ("fixuns_truncdddi2", "fixuns_trunctddi2")
35631 ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35632 ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35634 ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35635 Rename expanders to ...
35637 ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35638 ("fixuns_truncdddi2_emu"): ... these.
35640 ("fixuns_trunc<mode>si2_emu"): New expander.
35642 ("*fixuns_truncdfdi2_z13"): Rename to ...
35643 ("*fixuns_truncdfdi2_vx"): ... this.
35645 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35647 * config/s390/2964.md: Remove the single element vector compare
35648 instructions which are no longer used.
35649 * config/s390/s390.c (s390_select_ccmode): Remove handling of
35651 (s390_canonicalize_comparison): Remove handling of DFmode
35653 (s390_expand_vec_compare_scalar): Remove function.
35654 (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35655 * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35657 ("*cmp<mode>_ccs"): Add wfcdb instruction.
35659 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35661 * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35663 ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35664 will anyway by matched by mov<mode>_64dfp.
35666 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35668 * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35669 vlef/vstef. Add missing operand to vleif.
35671 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35673 * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35674 pair for all vector types with 64 bit elements.
35675 * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35676 * config/s390/vector.md (V_HW_64): ... here.
35677 (V_128_NOSINGLE): New mode iterator.
35678 ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35679 ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35680 ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35681 ("*vec_load_pairv2di"): Change to ...
35682 ("*vec_load_pair<mode>"): ... this one.
35684 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35686 * config/s390/constraints.md: Add comments.
35687 (jKK): Reject element sizes > 8 bytes.
35688 * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35690 * config/s390/s390.md: Add the s_operand checks formerly in
35691 s390_split_ok_p to various splitters where they are still
35693 * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35694 for 128 bit vectors. Plus two splitters.
35696 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35698 * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35701 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
35704 * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35705 error if the boundary argument is not constant.
35707 2017-03-24 Jakub Jelinek <jakub@redhat.com>
35709 PR rtl-optimization/80112
35710 * loop-doloop.c (doloop_condition_get): Don't check condition
35711 if cmp isn't SET with IF_THEN_ELSE src.
35713 2017-03-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35715 PR tree-optimization/80158
35716 * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35717 replacing a candidate statement, also replace it for the
35718 candidate's alternate interpretation.
35719 (replace_rhs_if_not_dup): Likewise.
35720 (replace_one_candidate): Likewise.
35722 2017-03-24 Richard Biener <rguenther@suse.de>
35724 PR tree-optimization/80167
35725 * graphite-isl-ast-to-gimple.c
35726 (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35728 (translate_isl_ast_to_gimple::get_rename): Likewise.
35730 2017-03-23 Kelvin Nilsen <kelvin@gcc.gnu.org>
35732 * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35733 handling of certain combinations of target options, including the
35734 combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35735 -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35737 2017-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
35740 * config/arm/arm.md (*load_multiple): Add reload_completed to
35741 matching condition.
35743 2017-03-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35744 Richard Biener <rguenth@suse.de>
35746 PR tree-optimization/79908
35747 PR tree-optimization/80136
35748 * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35749 been cast away, gimplify_and_add suffices.
35751 2017-03-23 Markus Trippelsdorf <markus@trippelsdorf.de>
35753 * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35755 2017-03-23 Richard Biener <rguenther@suse.de>
35757 PR tree-optimization/80032
35758 * gimplify.c (gimple_push_cleanup): Forced unconditional
35759 cleanups still have to go to the conditional_cleanups
35762 2017-03-22 Jakub Jelinek <jakub@redhat.com>
35764 PR tree-optimization/80072
35765 * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35767 (next_operand_entry_id): Change type to unsigned int.
35768 (sort_by_operand_rank): Make sure to return the right return value
35769 even if unsigned fields are bigger than INT_MAX.
35770 (struct oecount): Change cnt and id type to unsigned int.
35771 (oecount_hasher::equal): Formatting fix.
35772 (oecount_cmp): Make sure to return the right return value
35773 even if unsigned fields are bigger than INT_MAX.
35774 (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35777 * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35778 TREE_READONLY on result if writing it more than once.
35781 * doc/invoke.texi (-fsanitize=thread): Document that with
35782 -fnon-call-exceptions atomics are not able to throw
35786 * tsan.c: Include tree-eh.h.
35787 (instrument_builtin_call): Call maybe_clean_eh_stmt or
35788 maybe_clean_or_replace_eh_stmt where needed.
35789 (instrument_memory_accesses): Add cfg_changed argument.
35790 Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35791 if it returned true.
35792 (tsan_pass): Adjust caller. Return TODO_cleanup_cfg if cfg_changed.
35794 PR rtl-optimization/63191
35795 * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35796 wrapper function, moved the whole old content into ...
35797 (ix86_delegitimize_address_1): ... this. New inline function.
35798 (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35799 true as last argument instead of ix86_delegitimize_address.
35801 2017-03-22 Wilco Dijkstra <wdijkstr@arm.com>
35803 * config/aarch64/aarch64.c (generic_branch_cost): Copy
35804 cortexa57_branch_cost.
35806 2017-03-22 Wilco Dijkstra <wdijkstr@arm.com>
35808 * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35810 2017-03-21 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
35813 * doc/md.texi (Constraints): Document wA constraint.
35814 * config/rs6000/constraints.md (wA): New.
35815 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35816 (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35817 * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35818 * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35820 2017-03-22 Cesar Philippidis <cesar@codesourcery.com>
35823 * gimplify.c (is_oacc_declared): New function.
35824 (oacc_default_clause): Use it to set default flags for acc declared
35825 variables inside parallel regions.
35826 (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35827 declared variables.
35828 (gimplify_oacc_declare): Gimplify the declare clauses. Add the
35829 declare attribute to any decl as necessary.
35831 2017-03-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
35834 * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35835 (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35836 * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35837 (arm_arch_lpae): This.
35838 * config/arm/arm.c (arm_arch7ve): Rename into ...
35839 (arm_arch_lpae): This. Define it in term of isa_bit_lpae.
35840 * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35843 2017-03-22 Martin Liska <mliska@suse.cz>
35846 * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
35847 error message instead of an ICE.
35849 2017-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35851 * doc/extend.texi (6.11 Additional Floating Types): Revise.
35853 2017-03-21 Kelvin Nilsen <kelvin@gcc.gnu.org>
35855 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
35857 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35860 2017-03-21 Martin Sebor <msebor@redhat.com>
35862 * doc/extend.texi: Use "cannot" instead of "can't."
35863 * doc/hostconfig.texi: Same.
35864 * doc/install.texi: Same.
35865 * doc/invoke.texi: Same.
35866 * doc/loop.texi: Same.
35867 * doc/md.texi: Same.
35868 * doc/objc.texi: Same.
35869 * doc/rtl.texi: Same.
35870 * doc/tm.texi: Same.
35871 * doc/tm.texi.in: Same.
35872 * doc/trouble.texi: Same.
35874 2017-03-21 Alexandre Oliva <aoliva@redhat.com>
35877 * dwarf2out.c (struct checksum_attributes): Add at_alignment.
35878 (collect_checksum_attributes): Set it.
35879 (die_checksum_ordered): Use it.
35881 2017-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35883 PR tree-optimization/79908
35884 * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
35885 change: For a VA_ARG whose LHS has been cast away, use
35886 force_gimple_operand to construct the side effects.
35888 2017-03-21 David Malcolm <dmalcolm@redhat.com>
35890 PR translation/80001
35891 * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
35892 more amenable to translation.
35893 (oacc_loop_auto_partitions): Likewise.
35895 2017-03-21 Marek Polacek <polacek@redhat.com>
35896 Martin Sebor <msebor@redhat.com>
35898 PR tree-optimization/80109
35899 * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
35900 on INTEGRAL_TYPE_P.
35902 2017-03-21 Jakub Jelinek <jakub@redhat.com>
35903 Segher Boessenkool <segher@kernel.crashing.org>
35906 * combine.c (can_combine_p): Revert the 2017-03-20 change, only
35907 check reg_used_between_p between insn and one of succ or succ2
35908 depending on if succ is artificial insn not inserted into insn
35911 2017-03-21 Martin Liska <mliska@suse.cz>
35913 PR gcov-profile/80081
35914 * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
35915 * doc/gcc.texi: Include gcov-dump stuff.
35916 * doc/gcov-dump.texi: New file.
35918 2017-03-21 Toma Tabacu <toma.tabacu@imgtec.com>
35920 PR rtl-optimization/79150
35921 * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
35922 conditional jump, if the jump is the last insn of the loop.
35924 2017-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35925 Richard Biener <rguenth@suse.de>
35927 PR tree-optimization/79908
35928 * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35929 been cast away, use force_gimple_operand to construct the side
35932 2017-03-21 Martin Liska <mliska@suse.cz>
35934 PR libfortran/79956
35935 * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
35938 2017-03-21 Brad Spengler <spender@grsecurity.net>
35941 * plugin.c (htab_hash_plugin): New function.
35942 (add_new_plugin): Use it and adjust.
35943 (parse_plugin_arg_opt): Adjust.
35944 (init_one_plugin): Likewise.
35946 2017-03-21 Richard Biener <rguenther@suse.de>
35948 PR tree-optimization/80032
35949 * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
35950 if set force the cleanup to happen unconditionally.
35951 (gimplify_target_expr): Push inserted clobbers with force_uncond
35952 to avoid them being removed by control-dependent DCE.
35954 2017-03-21 Richard Biener <rguenther@suse.de>
35956 PR tree-optimization/80122
35957 * tree-inline.c (copy_bb): Do not expans va-arg packs or
35958 va_arg_pack_len when the inlined call stmt requires pack
35960 * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
35962 2017-03-21 Jakub Jelinek <jakub@redhat.com>
35965 * tsan.c (instrument_builtin_call): If the memory model argument
35966 is not a constant, assume it is valid.
35969 * fold-const.c (round_up_loc): Negate divisor in unsigned type to
35972 2017-03-20 Segher Boessenkool <segher@kernel.crashing.org>
35974 PR rtl-optimization/79910
35975 * combine.c (can_combine_p): Do not allow combining an I0 or I1
35976 if its dest is used by an insn before I2 (other than the combined
35977 insns themselves, which are properly handled already).
35979 2017-03-20 Segher Boessenkool <segher@kernel.crashing.org>
35982 2017-03-17 Bernd Schmidt <bschmidt@redhat.com>
35984 * combine.c (record_used_regs): New static function.
35985 (try_combine): Handle situations where there is an additional
35986 instruction between I2 and I3 which needs to have a LOG_LINK
35990 2017-03-17 Jim Wilson <jim.wilson@linaro.org>
35992 * combine.c (try_combine): Delete redundant i1 test. Call
35993 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
35995 2017-03-20 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
35998 * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
35999 alternatives 13/14.
36001 2017-03-20 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
36003 PR tree-optimization/80054
36004 * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36005 the optimization if a PHI or any of its arguments is not dominated
36006 by the candidate's basis. Use gphi* rather than gimple* as
36008 (replace_profitable_candidates): Clean up a gimple* variable that
36009 should be a gphi* variable.
36011 2017-03-20 Martin Sebor <msebor@redhat.com>
36014 * doc/extend.texi (attribute constructor): Document present limitation.
36016 2017-03-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
36019 * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36020 __POWER9_VECTOR__ #ifdef control, change template definition to
36021 use Power9-specific built-in function.
36022 (vec_any_eq): Likewise.
36023 * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36024 to control outcomes from this test.
36025 (vector_ae_<mode>p): For VEC_F modes, likewise.
36027 2017-03-20 Ian Lance Taylor <iant@google.com>
36029 * config/i386/i386.c (ix86_function_regparm): Save an extra
36030 register for -fsplit-stack with DECL_STATIC_CHAIN.
36032 2017-03-17 Palmer Dabbelt <palmer@dabbelt.com>
36035 * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36036 (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36038 2017-03-17 Palmer Dabbelt <palmer@dabbelt.com>
36040 * config/riscv/riscv.c (riscv_print_operand): Use "fence
36042 * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36045 2017-03-20 Marek Polacek <polacek@redhat.com>
36048 * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36050 2017-03-20 Richard Biener <rguenther@suse.de>
36052 PR tree-optimization/80113
36053 * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36054 allocate extra SSA name for PHI def.
36055 (add_close_phis_to_outer_loops): Likewise.
36056 (add_close_phis_to_merge_points): Likewise.
36057 (copy_loop_close_phi_args): Likewise.
36058 (copy_cond_phi_nodes): Likewise.
36060 2017-03-20 Martin Liska <mliska@suse.cz>
36062 PR middle-end/79753
36063 * tree-chkp.c (chkp_build_returned_bound): Do not build
36064 returned bounds for a LHS that's not a BOUNDED_P type.
36066 2017-03-20 Martin Liska <mliska@suse.cz>
36070 * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36071 COMPLEX_CST and VECTOR_CST.
36073 2017-03-20 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
36076 * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36077 target operand. A new splitter adds the clobber statement in case
36078 the target operand is dead anyway.
36080 2017-03-19 Gerald Pfeifer <gerald@pfeifer.com>
36082 * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36083 to age-old versions of binutils and glibc.
36085 2017-03-18 Segher Boessenkool <segher@kernel.crashing.org>
36087 * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36089 2017-03-18 Gerald Pfeifer <gerald@pfeifer.com>
36091 * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36093 2017-03-18 Gerald Pfeifer <gerald@pfeifer.com>
36095 * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36096 requirement for binutils 2.13.
36098 2017-03-17 Jim Wilson <jim.wilson@linaro.org>
36100 * combine.c (try_combine): Delete redundant i1 test. Call
36101 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36103 2017-03-17 Palmer Dabbelt <palmer@dabbelt.com
36105 * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36106 riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36108 <riscv64-*-elf>: Re-arrange section
36109 <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36110 <riscv32-*-linux>: Likewise.
36111 <riscv64-*-elf>: Likewise
36112 <riscv64-*-linux>: Likewise.
36114 2017-03-17 Richard Earnshaw <rearnsha@arm.com>
36117 * aarch64.opt(verbose-cost-dump): Fix typo.
36119 2017-03-17 Pat Haugen <pthaugen@us.ibm.com>
36122 * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36123 for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36125 2017-03-17 Bernd Schmidt <bschmidt@redhat.com>
36127 * reload.c (find_reloads): When reloading a nonoffsettable address,
36128 use RELOAD_OTHER for it and its address reloads.
36130 PR rtl-optimization/79910
36131 * combine.c (record_used_regs): New static function.
36132 (try_combine): Handle situations where there is an additional
36133 instruction between I2 and I3 which needs to have a LOG_LINK
36136 2017-03-17 Jeff Law <law@redhat.com>
36138 PR tree-optimization/71437
36139 * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36140 conditional in the hash table first.
36141 (vrp_dom_walker::before_dom_children): Extract condition from
36142 ASSERT_EXPR. Record condition, its inverion and any implied
36143 conditions as well.
36145 2017-03-17 Marek Polacek <polacek@redhat.com>
36146 Markus Trippelsdorf <markus@trippelsdorf.de>
36148 PR tree-optimization/80079
36149 * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36152 2017-03-17 Richard Biener <rguenther@suse.de>
36154 PR middle-end/80075
36155 * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36156 Properly verify the LHS before the RHS possibly claims to be
36158 (stmt_could_throw_p): Hande gimple conds fully here. Clobbers
36161 2017-03-17 Martin Jambor <mjambor@suse.cz>
36163 * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36164 (List of -O2 options): Likewise.
36165 (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36168 2017-03-17 Tom de Vries <tom@codesourcery.com>
36170 * gcov-dump.c (print_usage): Print bug_report_url.
36172 2017-03-17 Richard Biener <rguenther@suse.de>
36174 PR middle-end/80050
36175 * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36176 (parser::peek): Likewise.
36178 2017-03-17 Richard Biener <rguenther@suse.de>
36180 PR tree-optimization/80048
36181 * sese.c (free_sese_info): Properly release rename_map and
36182 copied_bb_map elements.
36184 2017-03-16 Alexandre Oliva <aoliva@redhat.com>
36186 * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36187 Add linked-list forward and backlinks. Insert on
36188 construction, remove on destruction.
36189 (class pass_store_merging): Add m_stores_head field.
36190 (pass_store_merging::terminate_and_process_all_chains):
36191 Iterate over m_stores_head list.
36192 (pass_store_merging::terminate_all_aliasing_chains):
36194 (pass_store_merging::execute): Check for debug stmts first.
36195 Push new chains onto the m_stores_head stack.
36197 2017-03-16 Michael Meissner <meissner@linux.vnet.ibm.com>
36200 * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36201 SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36202 but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36204 2017-03-16 Jeff Law <law@redhat.com>
36206 PR tree-optimization/71437
36207 * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36208 member function. Implementation moved into after_dom_children
36209 member function and into the threader's thread_outgoing_edges
36211 (dom_opt_dom_walker::after_dom_children): Simplify by moving
36212 some code into new thread_outgoing_edges.
36213 * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36214 definition. Simplify marker handling (do it here). Assume we always
36215 have the available expression and the const/copies tables.
36216 (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36218 * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36219 * tree-vrp.c (equiv_stack): No longer file scoped.
36220 (vrp_dom_walker): New class.
36221 (vrp_dom_walker::before_dom_children): New member function.
36222 (vrp_dom_walker::after_dom_children): Likewise.
36223 (identify_jump_threads): Setup domwalker. Use it rather than
36224 walking edges in a random order by hand. Simplify setup/finalization.
36225 (finalize_jump_threads): Remove.
36226 (vrp_finalize): Do not call identify_jump_threads here.
36227 (execute_vrp): Do it here instead and call thread_through_all_blocks
36230 PR tree-optimization/71437
36231 * tree-ssa-dom.c (pfn_simplify): Add basic_block argument. All
36233 (simplify_stmt_for_jump_threading): Add basic_block argument. All
36235 (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36236 (dom_opt_dom_walker::thread_across_edge): Remove
36237 handle_dominating_asserts argument. All callers changed.
36238 (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36239 changes. Remove calls to lhs_of_dominating_assert. Other
36240 uses of handle_dominating_asserts turn into unconditional code
36241 (simplify_control_stmt_condition_1): Likewise.
36242 (simplify_control_stmt_condition): Likewise.
36243 (thread_through_normal_block, thread_across_edge): Likewise.
36244 * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36245 * tree-vrp.c (lhs_of_dominating_assert): Move here. Return original
36246 object if it is not an SSA_NAME.
36247 (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36248 before calling into the VRP specific simplifiers.
36249 (identify_jump_threads): Remove handle_dominating_asserts
36252 2017-03-16 Jakub Jelinek <jakub@redhat.com>
36255 * tree-diagnostic.c (default_tree_printer): No longer static.
36256 * tree-diagnostic.h (default_tree_printer): New prototype.
36258 2017-03-16 Tamar Christina <tamar.christina@arm.com>
36260 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36261 Change ins into fmov.
36263 2017-03-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36265 * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36266 * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36267 Use h_con constraint for operand 1.
36268 (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36269 (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36271 2017-03-15 Jeff Law <law@redhat.com>
36273 PR tree-optimization/71437
36274 * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36275 (record_temporary_equivalences): Use it.
36277 PR tree-optimization/71437
36278 * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36279 tree-ssa-scopedtables.
36280 (lookup_avail_expr, build_and_record_new_cond): Likewise.
36281 (record_conditions, record_cond, vuse_eq): Likewise.
36282 (record_edge_info): Adjust to API tweak of record_conditions.
36283 (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36284 (record_temporary_equivalences, optimize_stmt): Likewise.
36285 (eliminate_redundant_computations): Likewise.
36286 (record_equivalences_from_stmt): Likewise.
36287 * tree-ssa-scopedtables.c: Include options.h and params.h.
36288 (vuse_eq): New function, moved from tree-ssa-dom.c
36289 (build_and_record_new_cond): Likewise.
36290 (record_conditions): Likewise. Accept vector of conditions rather
36291 than edge_equivalence structure for first argument.
36292 for the first argument.
36293 (avail_exprs_stack::lookup_avail_expr): New member function, moved
36294 from tree-ssa-dom.c.
36295 (avail_exprs_stack::record_cond): Likewise.
36296 * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36297 from tree-ssa-dom.c.
36298 (avail_exprs_stack): Add new member functions lookup_avail_expr
36300 (record_conditions): Declare.
36302 2017-03-15 Vladimir Makarov <vmakarov@redhat.com>
36305 * lra-constraints.c (process_alt_operands): Increase reject for
36306 reloading an input/output operand.
36308 2017-03-15 Michael Meissner <meissner@linux.vnet.ibm.com>
36311 * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36312 insns to convert from signed/unsigned char/short to IEEE 128-bit
36314 (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36316 2017-03-15 Uros Bizjak <ubizjak@gmail.com>
36319 * config/i386/i386.c (ix86_vector_duplicate_value): Create
36320 subreg of inner mode for values already in registers.
36322 2017-03-15 Bernd Schmidt <bschmidt@redhat.com>
36324 * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36325 iteration reg is used after the loop.
36327 2017-03-14 Martin Sebor <msebor@redhat.com>
36329 PR tree-optimization/79800
36330 * gimple-ssa-sprintf.c (format_floating: Add argument. Handle
36331 precision in negative-positive range.
36332 (format_floating): Call non-const overload with adjusted precision.
36334 2017-03-14 Michael Meissner <meissner@linux.vnet.ibm.com>
36337 * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36340 2017-03-14 Martin Sebor <msebor@redhat.com>
36342 PR middle-end/80020
36343 * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36344 * builtins.def (aligned_alloc): Use it.
36347 * Makefile.in (GTFILES): Add calls.c.
36348 * calls.c: Include "gt-calls.h".
36350 2017-03-14 Bernd Schmidt <bschmidt@redhat.com>
36352 PR rtl-optimization/79728
36353 * regs.h (struct target_regs): New field
36354 x_contains_allocatable_regs_of_mode.
36355 (contains_allocatable_regs_of_mode): New macro.
36356 * reginfo.c (init_reg_sets_1): Initialize it, and change
36357 contains_reg_of_mode so it includes global regs as well.
36358 * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36359 rather than contains_regs_of_mode.
36361 2017-03-14 Martin Liska <mliska@suse.cz>
36363 * doc/invoke.texi: Document options that can't be combined with
36364 -fcheck-pointer-bounds.
36366 2017-03-14 Martin Liska <mliska@suse.cz>
36368 PR middle-end/79831
36369 * doc/invoke.texi (-Wchkp): Document the option.
36371 2017-03-14 Martin Liska <mliska@suse.cz>
36373 * Makefile.in: Install gcov-dump.
36375 2017-03-14 Martin Liska <mliska@suse.cz>
36377 * multiple_target.c (expand_target_clones): Bail out for
36378 an invalid attribute.
36380 2017-03-14 Richard Biener <rguenther@suse.de>
36382 * alias.c (struct alias_set_entry): Pack properly.
36383 * cfgloop.h (struct loop): Likewise.
36384 * cse.c (struct set): Likewise.
36385 * ipa-utils.c (struct searchc_env): Likewise.
36386 * loop-invariant.c (struct invariant): Likewise.
36387 * lra-remat.c (struct cand): Likewise.
36388 * recog.c (struct change_t): Likewise.
36389 * rtl.h (struct address_info): Likewise.
36390 * symbol-summary.h (function_summary): Likewise.
36391 * tree-loop-distribution.c (struct partition): Likewise.
36392 * tree-object-size.c (struct object_size_info): Likewise.
36393 * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36394 * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36395 * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36396 * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36397 * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36398 * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36399 (struct _stmt_vec_info): Likewise.
36401 2017-03-14 Martin Liska <mliska@suse.cz>
36404 * multiple_target.c (create_dispatcher_calls): Check that
36405 a target can create a function dispatcher.
36407 2017-03-14 Martin Liska <mliska@suse.cz>
36410 * multiple_target.c (expand_target_clones): Drop local.local
36411 flag for default implementation.
36413 2017-03-14 Richard Biener <rguenther@suse.de>
36415 PR tree-optimization/80030
36416 * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36418 2017-03-13 Kito Cheng <kito.cheng@gmail.com>
36420 * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36421 gcc_fallthrough() instead of __attribute__((fallthrough));
36423 2017-03-13 Gerald Pfeifer <gerald@pfeifer.com>
36425 * doc/gcc.texi: Remove "up" link to (DIR).
36426 * doc/gccint.texi: Ditto.
36428 2017-03-13 Gerald Pfeifer <gerald@pfeifer.com>
36430 * doc/install.texi (Specific) <avr>: Remove reference to
36433 2017-03-13 Jeff Law <law@redhat.com>
36435 * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36436 attribute rather than comments.
36438 * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36439 match_scratch operand is highest.
36441 2017-03-13 Martin Liska <mliska@suse.cz>
36443 PR middle-end/78339
36444 * ipa-pure-const.c (warn_function_noreturn): If the declarations
36445 is a CHKP clone, use original declaration.
36447 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
36449 * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36450 (arc_conditional_register_usage): Use a different allocation order
36451 when optimizing for size.
36452 * common/config/arc/arc-common.c (arc_option_optimization_table):
36453 Section anchors default on when optimizing for size.
36455 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
36457 * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36459 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
36461 * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36462 * config/arc/arc.md (cpu_facility): Add cd variant.
36463 (*movqi_insn): Add code density variant.
36464 (*movhi_insn): Likewise.
36465 (*movqi_insn): Likewise.
36466 (*addsi3_mixed): Likewise.
36467 (subsi3_insn): Likewise.
36469 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
36471 * config/arc/arc.md (movsi_cond_exec): Update constraint.
36473 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
36475 * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36476 expressions with MINUS and UNARY ops.
36478 2017-03-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36481 * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36483 (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36484 between vec_select and vector argument.
36485 (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36486 (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36487 (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36488 (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36489 (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36490 (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36492 2017-03-13 Richard Biener <rguenther@suse.de>
36495 * params.def (vect-max-peeling-for-alignment): Fix typo.
36497 2017-03-12 Gerald Pfeifer <gerald@pfeifer.com>
36499 * doc/install.texi (Specific) <mips-*-*>: Remove description of
36500 issue that only occurred with binutils below 2.18.
36502 2017-03-12 Gerald Pfeifer <gerald@pfeifer.com>
36504 * doc/install.texi (Specific) <cris-axis-elf>: No longer
36505 refer to binutils 2.11/2.12 minimum.
36507 2017-03-12 Gerald Pfeifer <gerald@pfeifer.com>
36509 * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36510 ftp.kernel.org and simplify binutils requirement.
36512 2017-03-11 Gerald Pfeifer <gerald@pfeifer.com>
36514 * doc/invoke.texi (Warning Options): Fix spelling of link-time
36516 (Optimize Options): Ditto. Also remove redundancy.
36518 2017-03-10 David Malcolm <dmalcolm@redhat.com>
36520 PR translation/79848
36521 * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36523 * ipa-pure-const.c (suggest_attribute): Likewise. Convert _
36524 to G_ to avoid double translation.
36526 2017-03-10 David Malcolm <dmalcolm@redhat.com>
36528 PR translation/79923
36529 * auto-profile.c (get_combined_location): Convert leading
36530 character of diagnostics to lower case and remove trailing period.
36531 (read_profile): Likewise for various diagnostics.
36532 * config/arm/arm.c (arm_option_override): Remove trailing period
36533 from various diagnostics.
36534 * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36535 (msp430_expand_delay_cycles): Likewise.
36537 2017-03-10 David Malcolm <dmalcolm@redhat.com>
36540 * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36541 full command-line argument, rather than just "str".
36542 (aarch64_validate_march): Likewise.
36543 (aarch64_validate_mtune): Likewise.
36545 2017-03-10 Bernd Schmidt <bschmidt@redhat.com>
36547 PR rtl-optimization/78911
36548 * lra-assigns.c (must_not_spill_p): New function.
36549 (spill_for): Use it.
36551 2017-03-10 Jakub Jelinek <jakub@redhat.com>
36553 PR tree-optimization/79981
36554 * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36555 ATOMIC_COMPARE_EXCHANGE ifn result.
36556 (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36557 IFN_ATOMIC_COMPARE_EXCHANGE.
36559 2017-03-10 David Malcolm <dmalcolm@redhat.com>
36562 * opts.c (parse_sanitizer_options): Add missing question mark to
36563 "did you mean" message.
36565 2017-03-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
36567 * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36569 (VMULEUH_UNS): Likewise.
36570 (VMULOUB_UNS): Likewise.
36571 (VMULOUH_UNS): Likewise.
36572 * config/rs6000/rs6000.c (builtin_function_type): Remove
36573 references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36575 2017-03-10 David Malcolm <dmalcolm@redhat.com>
36578 * read-rtl-function.c (function_reader::read_rtx_operand): Update
36579 x with result of extra_parsing_for_operand_code_0.
36580 (function_reader::extra_parsing_for_operand_code_0): Convert
36581 return type from void to rtx, returning x. When reading
36582 SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36583 larger size containing struct block_symbol.
36585 2017-03-10 Segher Boessenkool <segher@kernel.crashing.org>
36587 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36588 -mfloat128-hardware without -m64.
36590 2017-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
36593 * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36594 entries to the case statement that marks unsigned arguments to
36595 overloaded functions.
36597 2017-03-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
36599 * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36600 two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36602 2017-03-10 Pat Haugen <pthaugen@us.ibm.com>
36605 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36606 TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36608 2017-03-10 Martin Liska <mliska@suse.cz>
36612 * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36613 * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36616 2017-03-10 Olivier Hainque <hainque@adacore.com>
36618 * tree-switch-conversion (array_value_type): Start by resetting
36619 candidate type to it's main variant.
36621 2017-03-10 Jakub Jelinek <jakub@redhat.com>
36623 PR rtl-optimization/79909
36624 * combine.c (try_combine): Use simplify_replace_rtx on individual
36625 CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36626 of the whole CALL_INSN_FUNCTION_USAGE.
36628 PR tree-optimization/79972
36629 * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36630 get_range_info on SSA_NAMEs. Formatting fixes.
36632 2017-03-10 Richard Biener <rguenther@suse.de>
36633 Jakub Jelinek <jakub@redhat.com>
36635 PR tree-optimization/77975
36636 * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36637 edge to be constant.
36638 (get_val_for): For constant x return it. Formatting fix.
36639 (loop_niter_by_eval): Avoid pointless looping if the next iteration
36640 would use the same bases as the current one.
36642 2017-03-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
36644 * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36645 instead of vec_select for V1TImode.
36646 * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36648 (VSX_LE_128): Add V1TI to this mode iterator.
36649 (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36650 (*vsx_le_perm_store_<mode>): Likewise.
36651 (pre-reload splitter for VSX stores): Likewise.
36652 (post-reload splitter for VSX stores): Likewise.
36653 (*vsx_xxpermdi2_le_<mode>): Likewise.
36654 (*vsx_lxvd2x2_le_<mode>): Likewise.
36655 (*vsx_stxvd2x2_le_<mode>): Likewise.
36657 2017-03-09 Michael Eager <eager@eagercon.com>
36659 Correct failures with --enable-checking=yes,rtl.
36661 * config/microblaze/microblaze.c (microblaze_expand_shift):
36662 Replace GET_CODE test with CONST_INT_P and INTVAL test with
36663 test for const0_rtx.
36664 * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36665 lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36667 2017-03-09 Richard Biener <rguenther@suse.de>
36669 PR tree-optimization/79977
36670 * graphite-scop-detection.c (scop_detection::merge_sese):
36671 Handle the case of extra exits to blocks dominating the entry.
36673 2017-03-09 Toma Tabacu <toma.tabacu@imgtec.com>
36675 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36678 2017-03-09 Vladimir Makarov <vmakarov@redhat.com>
36680 PR rtl-optimization/79949
36681 * lra-constraints.c (process_alt_operands): Check memory when
36682 trying to predict a cycle. Print about the overall increase.
36684 2017-03-09 Richard Biener <rguenther@suse.de>
36686 PR middle-end/79971
36687 * gimple-expr.c (useless_type_conversion_p): Preserve
36688 TYPE_SATURATING for fixed-point types.
36690 2017-03-09 Richard Biener <rguenther@suse.de>
36693 * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36694 alignment of BLKmode params.
36696 2017-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36699 * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36700 (VALL_NO_V2Q): Likewise.
36702 * config/aarch64/aarch64-simd.md
36703 (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36705 (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36706 VALL_NO_V2Q mode iterator.
36707 (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36709 2017-03-09 Martin Liska <mliska@suse.cz>
36711 PR tree-optimization/79631
36712 * tree-chkp-opt.c (chkp_is_constant_addr): Call
36713 tree_int_cst_sign_bit just for INTEGER constants.
36715 2017-03-09 Martin Liska <mliska@suse.cz>
36719 * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36722 2017-03-09 Marek Polacek <polacek@redhat.com>
36725 * tree.c (inchash::add_expr): Handle TREE_VEC.
36727 2017-03-09 Martin Liska <mliska@suse.cz>
36730 (chkp_narrow_size_and_offset): New function.
36731 (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36732 (void chkp_parse_bit_field_ref): New function.
36733 (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36734 (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36736 2017-03-09 Martin Liska <mliska@suse.cz>
36739 * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36740 (chkp_find_bounds_1): Remove gcc_unreachable.
36742 2017-03-09 Jakub Jelinek <jakub@redhat.com>
36745 * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36746 BUILT_IN_SYNC*, determine the access type from the size suffix and
36747 always build a MEM_REF with that type. Handle forgotten
36748 BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36751 * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36752 _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36753 _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36754 _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36755 _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36756 _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36757 _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36758 _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36759 _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36760 _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36761 _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36762 _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36763 _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36764 _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36765 _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36766 _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36767 _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36768 _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36769 _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36770 _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36771 _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36772 _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36773 _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36774 _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36775 _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36776 _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36777 _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36778 _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36779 _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36780 _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36781 definitions outside of __OPTIMIZE__ guarded section.
36784 * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36785 _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36786 _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36787 _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36790 2017-03-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
36792 * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36793 ("vfenez<mode>"): Add missing constraints.
36795 2017-03-08 Martin Sebor <msebor@redhat.com>
36798 * config/nds32/nds32.c (nds32_option_override):
36799 Fix misspelled diagnostic.
36801 2017-03-08 Jakub Jelinek <jakub@redhat.com>
36804 * gimplify.c (gimplify_omp_for): Replace index var in outer
36805 taskloop statement with an artificial variable and add
36806 OMP_CLAUSE_PRIVATE clause for it.
36808 2017-03-08 Richard Biener <rguenther@suse.de>
36810 PR tree-optimization/79955
36811 * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36812 for accesses that are completely outside of the variable.
36814 2017-03-08 Andrew Haley <aph@redhat.com>
36816 PR tree-optimization/79943
36817 * tree-ssa-loop-split.c (compute_new_first_bound): When
36818 calculating the new upper bound, (END-BEG) should be added, not
36821 2017-03-08 Jakub Jelinek <jakub@redhat.com>
36823 * config/avr/avr.md (setmemhi): Make sure match_dup
36824 operand number comes before match_scratch.
36826 2017-03-08 Richard Biener <rguenther@suse.de>
36828 PR tree-optimization/79920
36829 * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36830 with ncopies == 1 to ...
36831 (vect_transform_slp_perm_load): ... here. Properly compute
36832 all element loads by iterating VF times over the group. Do
36833 not handle ncopies (computed in a broken way) in
36834 vect_create_mask_and_perm.
36836 2017-03-08 Jakub Jelinek <jakub@redhat.com>
36839 * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36840 is a uniform vector, use uniform_vector_p return value instead of
36841 building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36843 2017-03-07 Marek Polacek <polacek@redhat.com>
36845 PR middle-end/79809
36846 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
36847 (alloca_call_type): Likewise.
36849 2017-03-07 Martin Liska <mliska@suse.cz>
36851 * gcov.c (process_args): Put comment to correct location.
36853 2017-03-07 Martin Liska <mliska@suse.cz>
36855 PR middle-end/68270
36856 * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
36857 Use array_at_struct_end_p instead of DECL_CHAIN (field).
36858 (chkp_narrow_bounds_for_field): Likewise.
36859 (chkp_parse_array_and_component_ref): Pass one more argument to
36862 2017-03-07 Richard Biener <rguenther@suse.de>
36864 * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
36867 2017-03-07 Segher Boessenkool <segher@kernel.crashing.org>
36869 * config/i386/i386.c (ix86_local_alignment): Align most aggregates
36870 of 16 bytes and more to 16 bytes, not those of 16 bits and more.
36872 2017-03-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36875 * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
36876 to end of description.
36877 (PARAM_MAX_STORES_TO_MERGE): Likewise.
36879 2017-03-07 Jakub Jelinek <jakub@redhat.com>
36881 PR rtl-optimization/79901
36882 * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
36884 (*avx512f_<code><mode>3<mask_name>): ... this.
36885 (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
36886 iterator instead of VI8_AVX2_AVX512BW.
36888 PR rtl-optimization/79901
36889 * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
36890 min/max expander, expand it using expand_vec_cond_expr.
36893 * ubsan.c (ubsan_encode_value): Call mark_addressable on the
36896 2017-03-06 Jakub Jelinek <jakub@redhat.com>
36899 * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
36900 to void * for PCH reasons.
36901 * dwarf2out.c (output_loc_operands, output_die): Cast
36902 v.val_vec.array to unsigned char *.
36904 2017-03-06 John David Anglin <danglin@gcc.gnu.org>
36907 * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
36910 2017-03-06 Vladimir Makarov <vmakarov@redhat.com>
36912 PR rtl-optimization/79571
36913 * lra-constraints.c (process_alt_operands): Calculate static
36914 reject and subtract it from overall when only addresses will be
36917 2017-03-06 Julia Koval <julia.koval@intel.com>
36920 * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
36921 incoming stack boundary to 128 for 64-bit targets.
36923 2017-03-06 Richard Biener <rguenther@suse.de>
36925 PR tree-optimization/79894
36926 * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
36927 to NULL after folding it.
36929 2017-03-06 Richard Biener <rguenther@suse.de>
36931 PR tree-optimization/79824
36932 * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
36933 check disabling peeling for gaps.
36935 2017-03-06 Toma Tabacu <toma.tabacu@imgtec.com>
36937 * doc/sourcebuild.texi (Effective-Target Keywords, Environment
36938 attributes): Document gettimeofday.
36940 2017-03-06 Robin Dapp <rdapp@linux.vnet.ibm.com>
36942 * config/s390/s390.c (s390_option_override_internal): Set
36943 PARAM_MIN_VECT_LOOP_BOUND
36945 2017-03-06 Robin Dapp <rdapp@linux.vnet.ibm.com>
36947 * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
36948 * config/s390/s390.md: Likewise.
36950 2017-03-06 Jakub Jelinek <jakub@redhat.com>
36953 * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
36954 (<avx2_avx512>_perm<mode>): Rename to ...
36955 (avx2_perm<mode>): ... this. Use VI8F_256 iterator instead
36957 (<avx512>_perm<mode>_mask): Rename to ...
36958 (avx512vl_perm<mode>_mask): ... this. Use VI8F_256 iterator instead
36960 (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
36961 (avx2_perm<mode>_1<mask_name): ... this. Use VI8F_256 iterator
36962 instead of VI8F_256_512.
36963 (avx512f_perm<mode>): New define_expand.
36964 (avx512f_perm<mode>_mask): Likewise.
36965 (avx512f_perm<mode>_1<mask_name>): New define_insn.
36966 (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
36968 2017-03-06 Prachi Godbole <prachi.godbole@imgtec.com>
36970 * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
36971 msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
36973 (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
36975 2017-03-06 Martin Liska <mliska@suse.cz>
36978 * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
36979 when having a SSA NAME w/o VAR_DECL assigned to it.
36981 2017-03-06 Prachi Godbole <prachi.godbole@imgtec.com>
36983 * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
36984 msa_dpsub_<su>_d): Fix MODE for vec_select.
36986 2017-03-06 Prachi Godbole <prachi.godbole@imgtec.com>
36988 * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
36990 * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
36992 2017-03-06 Richard Biener <rguenther@suse.de>
36994 * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
36995 * plugin.c (register_plugin_info): Likewise.
36996 * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
36998 2017-03-05 Jakub Jelinek <jakub@redhat.com>
37000 * config/i386/sse.md (sse_storehps, sse_storelps,
37001 avx_<castmode><avxsizesuffix>_<castmode>,
37002 avx512f_<castmode><avxsizesuffix>_<castmode>,
37003 avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37004 in condition that at least one operand is not a MEM.
37006 2017-03-03 Jakub Jelinek <jakub@redhat.com>
37008 PR middle-end/79805
37009 * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37010 ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37012 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37013 gimple_call_nothrow_p flag based on whether original builtin can throw.
37014 If it can, emit following stmts on the fallthrough edge.
37015 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37016 don't create new bb if inserting just debug stmts on the edge, try to
37017 insert them on the fallthru bb or just reset debug stmts.
37019 2017-03-03 Segher Boesssenkool <segher@kernel.crashing.org>
37022 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37023 restore recog_data (including the operand rtxes inside it) around
37024 the call to get_insn_template.
37026 2017-03-03 Martin Sebor <msebor@redhat.com>
37028 PR tree-optimization/79699
37029 * context.c (context::~context): Free MPFR caches to avoid
37030 a memory leak on program exit.
37032 2017-03-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
37034 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37035 Use wide_int::ulow () instead of .elt (0).
37037 2017-03-03 Uros Bizjak <ubizjak@gmail.com>
37039 * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37040 (*pushxf): Limit oF constraint to 32bit targets and add oC
37041 constraint for 64bit targets.
37042 (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37043 (*pushdf): Change rmF constraint to rmC.
37045 2017-03-03 Martin Liska <mliska@suse.cz>
37047 * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37048 Remove unused variable.
37050 2017-03-03 Jakub Jelinek <jakub@redhat.com>
37053 * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37054 is a memory operand, increase num_memory.
37055 (ix86_expand_args_builtin): Likewise.
37057 2017-03-03 Jan Hubicka <jh@suse.cz>
37060 * ipa-devirt.c (maybe_record_node): Properly handle
37061 __cxa_pure_virtual visibility.
37063 2017-03-03 Martin Liska <mliska@suse.cz>
37065 PR tree-optimization/79803
37066 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37068 (pass_loop_prefetch::execute): Disabled optimization if an
37069 assumption about L1 cache size is not met.
37071 2017-03-03 Martin Liska <mliska@suse.cz>
37073 PR rtl-optimization/79574
37074 * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37075 (hash_scan_set): Likewise.
37076 (dump_hash_table): Likewise.
37077 (hoist_code): Likewise.
37079 2017-03-03 Richard Biener <rguenther@suse.de>
37081 * fixed-value.c (fixed_from_string): Restore use of elt (1)
37082 in place of uhigh ().
37083 (fixed_convert_from_real): Likewise.
37085 2017-03-03 Uros Bizjak <ubizjak@gmail.com>
37088 * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37090 2017-03-03 Richard Biener <rguenther@suse.de>
37092 PR middle-end/79818
37093 * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37094 TYPE_OVERFLOW_UNDEFINED check.
37096 2017-03-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
37098 * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37100 (vector_ae_<mode>_p): Likewise.
37101 (vector_nez_<mode>_p): Likewise.
37102 (vector_ne_v2di_p): Likewise.
37103 (vector_ae_v2di_p): Likewise.
37104 (vector_ne_<mode>_p): Likewise.
37105 * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37107 (vsx_tsqrt<mode>2_fe): Likewise.
37109 2017-03-02 Uros Bizjak <ubizjak@gmail.com>
37112 * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37114 2017-03-02 Jakub Jelinek <jakub@redhat.com>
37116 PR rtl-optimization/79780
37117 * cprop.c (one_cprop_pass): When second and further conditional trap
37118 in a single basic block is turned into an unconditional trap, turn it
37119 into a deleted note to avoid RTL verification failures.
37121 2017-03-02 Richard Biener <rguenther@suse.de>
37123 * fold-const.c (const_binop): Use ulow () instead of elt (0).
37125 2017-03-02 Richard Biener <rguenther@suse.de>
37127 PR tree-optimization/79345
37129 * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37130 param and abort the walk, returning -1 if it is hit.
37131 (walk_aliased_vdefs): Take a limit param and pass it on.
37132 * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37133 defaulting to 0 and return a signed int.
37134 * tree-ssa-uninit.c (struct check_defs_data): New struct.
37135 (check_defs): New helper.
37136 (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37137 about uninitialized memory.
37138 * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37139 bogus uninitialized warning.
37140 (fixed_convert_from_real): Likewise.
37142 2017-03-02 Bin Cheng <bin.cheng@arm.com>
37144 PR tree-optimization/66768
37145 * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37146 iv_use if base object can't be determined.
37148 2017-03-02 Jakub Jelinek <jakub@redhat.com>
37150 PR tree-optimization/79345
37151 * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37152 * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37153 (get_pattern_stats): Initialize it.
37154 * genemit.c (gen_expand): Verify match_scratch numbers come after
37155 match_operand/match_dup numbers.
37156 * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37157 match_scratch numbers.
37158 * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37160 * config/s390/s390.md (trunctdsd2): Likewise.
37162 2017-03-02 Richard Biener <rguenther@suse.de>
37164 * wide-int.h (wide_int_storage::operator=): Implement in terms
37167 2017-03-02 Richard Biener <rguenther@suse.de>
37169 PR tree-optimization/79777
37170 * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37171 the to insert expression to sth existing.
37173 2017-03-01 Martin Sebor <msebor@redhat.com>
37175 PR middle-end/79692
37176 * gimple-ssa-sprintf.c
37177 (directive::known_width_and_precision): New function.
37178 (format_integer): Use it.
37179 (get_mpfr_format_length): Consider the full range of precision
37180 when computing %g output with the # flag. Set the likely byte
37181 count to 3 rather than 1 when precision is indeterminate.
37182 (format_floating): Correct the lower bound of precision.
37184 2017-03-01 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
37186 * doc/invoke.texi: Document default code model for 64-bit Linux.
37188 2017-03-01 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
37191 * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37192 udiv rather than div since input pattern is unsigned.
37194 2017-03-01 Uros Bizjak <ubizjak@gmail.com>
37196 * config/i386/i386.c (print_reg): Warn for values of
37197 unsupported size in integer register.
37199 2017-03-01 Michael Meissner <meissner@linux.vnet.ibm.com>
37202 * config/rs6000/predicates.md (current_file_function_operand): Do
37203 not allow self calls to be local if the function is replaceable.
37205 2017-03-01 Kelvin Nilsen <kelvin@gcc.gnu.org>
37208 * config/rs6000/altivec.h (vec_ctz and others): Change the
37209 preprocessor macro that controls conditional compilation from
37210 _ARCH_PWR9 to __POWER9_VECTOR__.
37211 (vec_all_ne): Change parameterization of __altivec_scalar_pred
37212 macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37213 control (instead of _ARCH_PWR9 control) so that template
37214 definition uses power9-specific function.
37215 (vec_any_eq): Likewise.
37216 (vec_all_ne): Change macro definition to use a power9-specific
37217 expansion under #ifdef __POWER9_VECTOR__ control (instead of
37218 _ARCH_PWR9 control).
37219 (vec_any_eq) Likewise.
37220 * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37221 expansion for CMPNEF to remove support for xvcmpnesp instruction.
37222 (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37223 support for xvcmpnedp instruction.
37224 (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37225 macro expansion so that Power9 implementation of vec_all_ne does
37226 not use the AltiVec predicate framework.
37227 (VCMPNEH_P): Likewise.
37228 (VCMPNEW_P): Likewise.
37229 (VCMPNED_P): Likewise.
37230 (VCMPNEFP_P): Likewise.
37231 (VCMPNEDP_P): Likewise.
37232 (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37233 implementation of vec_any_eq to not use AltiVec predicate
37235 (VCMPAEH_P): Likewise.
37236 (VCMPAEW_P): Likewise.
37237 (VCMPAED_P): Likewise.
37238 (VCMPAEFP_P): Likewise.
37239 (VCMPAEDP_P): Likewise.
37240 (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37241 BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37242 not use the AltiVec predicate framework.
37243 (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37244 of vec_any_eq to not use AltiVec predicate framework.
37245 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37246 support for predefined __POWER9_VECTOR__ macro to indicate that
37247 Power9 instruction selection is enabled.
37248 (altivec_overloaded_builtins): Remove extraneous
37249 ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37250 function argument types RS6000_BTI_bool_V16QI and
37251 RS6000_BTI_bool_V16QI. Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37252 entry for overloaded function argument types RS6000_BTI_bool_V4SI
37253 andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB. Remove
37254 two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37255 P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37256 Power9 for implementations of vec_cmpne. Change the signature for
37257 all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37258 (representing vec_all_ne) to remove the previously described first
37259 argument of type RS6000_BTI_INTSI, as this was an artifact of
37260 reliance on the AltiVec predicate framework, which is no longer
37261 used in the implementation of these functions. Add
37262 P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37263 function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37264 since, unlike the AltiVec predicate framework implementation, we
37265 do not share function descriptors between vec_alle and vec_anyeq.
37266 (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37267 set of modes that receive special treatment even when
37268 TARGET_P9_VECTOR is true. The special treatment emits code that
37269 does not depend on Power9 instructions.
37270 * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37271 define_expand to not rely on AltiVec predicate framework.
37272 (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37274 (vector_ne_v2di_p): Change this define_expand to not rely on
37275 AltiVec predicate framework.
37276 (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37278 (vector_ne_<mode>_p): Change this define_expand to not rely on
37279 AltiVec predicate framework.
37280 (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37282 * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37283 (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37284 define_insn pattern.
37285 (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37286 define_insn pattern because the xvcmpne<VSs>. instruction is not
37288 (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37289 instruction is not supported.
37291 2017-03-01 Jakub Jelinek <jakub@redhat.com>
37293 * config/nvptx/nvptx.c: Include intl.h.
37295 2017-03-01 Martin Jambor <mjambor@suse.cz>
37298 * ipa-prop.h (ipa_bits): Removed field known.
37299 (ipa_jump_func): Removed field vr_known. Changed fields bits and m_vr
37300 to pointers. Adjusted their comments to warn about their sharing.
37301 (ipcp_transformation_summary): Change bits to a vector of pointers.
37302 (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37303 (ipa_get_ipa_bits_for_value): Declare.
37304 * tree-vrp.h (value_range): Mark as GTY((for_user)).
37305 * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37306 (ipa_bits_hash_table): Likewise.
37307 (ipa_vr_ggc_hash_traits): Likewise.
37308 (ipa_vr_hash_table): Likewise.
37309 (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37310 being pointers and vr_known being removed.
37311 (ipa_set_jf_unknown): Likewise.
37312 (ipa_get_ipa_bits_for_value): New function.
37313 (ipa_set_jfunc_bits): Likewise.
37314 (ipa_get_value_range): New overloaded functions.
37315 (ipa_set_jfunc_vr): Likewise.
37316 (ipa_compute_jump_functions_for_edge): Use the above functions to
37317 construct bits and vr parts of jump functions.
37318 (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37319 ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37321 (ipcp_grow_transformations_if_necessary): Also allocate
37322 ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37324 (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37325 them. Fix too long lines.
37326 (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37327 vr_known being removed.
37328 (ipa_read_jump_function): Use new setter functions to construct bits
37329 and vr parts of jump functions or set them to NULL.
37330 (write_ipcp_transformation_info): Adjust for bits being pointers.
37331 (read_ipcp_transformation_info): Likewise.
37332 (ipcp_update_bits): Likewise. Fix excessively long lines a trailing
37334 Include gt-ipa-prop.h.
37335 * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37337 (ipcp_store_bits_results): Likewise.
37338 (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37339 Do not write to existing jump functions but use a temporary instead.
37341 2017-03-01 Jakub Jelinek <jakub@redhat.com>
37344 * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37345 attempt to use its first operand as BIT_FIELD_REF base.
37347 2017-03-01 Richard Biener <rguenther@suse.de>
37349 PR middle-end/79721
37350 * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37351 interpolating formula in wrapping arithmetic.
37352 (chrec_apply): Convert chrec_evaluate return value to wanted type.
37354 2017-03-01 Jakub Jelinek <jakub@redhat.com>
37356 PR tree-optimization/79734
37357 * tree-vect-generic.c (expand_vector_condition): Optimize
37358 AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37359 Handle VEC_COND_EXPR where comparison has different inner width from
37360 type's inner width.
37362 2017-02-28 Sandra Loosemore <sandra@codesourcery.com>
37364 * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37365 markup, and similar issues. Remove @opindex entries for things
37366 that aren't options. Add missing -mmpy-option entries.
37368 2017-02-28 Jakub Jelinek <jakub@redhat.com>
37370 PR tree-optimization/79737
37371 * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37372 a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37373 tmpbuf[byte_size - 1]. Call natice_encode_expr with byte_size - 1
37374 instead of byte_size. Formatting fix.
37375 (shift_bytes_in_array_right): Formatting fix.
37377 2017-02-28 Eric Botcazou <ebotcazou@adacore.com>
37380 * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37381 condition on optimize for the leaf function test.
37383 2017-02-28 Martin Liska <mliska@suse.cz>
37386 * read-rtl-function.c (function_reader::handle_unknown_directive):
37387 Bail out when one uses -flto.
37389 2017-02-28 Martin Liska <mliska@suse.cz>
37391 * common.opt: Replace space with tabular for options of <number>
37393 * config/i386/i386.opt: Show <number> value for
37394 -mlarge-data-threshold.
37395 * opts.c (print_filtered_help): Do not display number in hexadecimal
37398 2017-02-28 Martin Liska <mliska@suse.cz>
37400 * common.opt: Fix --help=option -Q for options which are of
37403 2017-02-28 Uros Bizjak <ubizjak@gmail.com>
37405 * config/i386/i386.c (print_reg): Error out for values
37406 of 8-bit size in invalid integer register.
37408 2017-02-28 Martin Sebor <msebor@redhat.com>
37410 PR tree-optimization/79691
37411 * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37413 2017-02-28 Jakub Jelinek <jakub@redhat.com>
37416 * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37417 gcc_unreachable with output_operand_lossage.
37419 2017-02-28 Richard Biener <rguenther@suse.de>
37421 PR tree-optimization/79740
37422 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37424 (visit_nary_op): Insert the nary into the hashtable if we
37425 pattern-matched sth.
37426 * tree-ssa-pre.c (eliminate_insert): Robustify.
37428 2017-02-28 Richard Biener <rguenther@suse.de>
37430 PR middle-end/79731
37431 * fold-const.c (decode_field_reference): Reject out-of-bound
37434 2017-02-28 Jakub Jelinek <jakub@redhat.com>
37436 * config/i386/i386.c: Include intl.h.
37437 (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37438 instead of just cond ? "..." : "...".
37439 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37440 * coverage.c (read_counts_file): Likewise.
37441 * omp-offload.c: Include intl.h.
37442 (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37443 of just cond ? "..." : "...".
37444 * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37445 of just cond ? "..." : "...".
37447 2017-02-28 Richard Earnshaw <rearnsha@arm.com>
37450 * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37452 * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37453 'tune for' CPU name.
37454 * config/arm/arm-cpu-data.h: Regenerated.
37456 2017-02-28 Richard Biener <rguenther@suse.de>
37458 PR tree-optimization/79732
37459 * tree-inline.c (expand_call_inline): Do not shadow var.
37461 2017-02-28 Richard Biener <rguenther@suse.de>
37463 PR tree-optimization/79723
37464 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37465 address-space properly.
37467 2017-02-28 Thomas Schwinge <thomas@codesourcery.com>
37469 * doc/optinfo.texi (Optimization groups): Fix option used for
37471 * doc/invoke.texi (-fopt-info): Document "omp".
37472 * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37473 (OPTGROUP_ALL): Add OPTGROUP_OMP.
37474 * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37475 * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37476 * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37478 * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP. Adjust
37480 * dumpfile.c (optgroup_options): Instead of "openmp", associate
37481 OPTGROUP_OMP with "omp".
37483 2017-02-27 Pat Haugen <pthaugen@us.ibm.com>
37486 * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37487 for arithmetic shift of unsigned V2DI.
37489 2017-02-27 Claudiu Zissulescu <claziss@synopsys.com>
37491 * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37492 arc/linux.h headers.
37493 * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37494 (LINK_SPEC): Likewise.
37495 (ARC_TLS_EXTRA_START_SPEC): Likewise.
37496 (EXTRA_SPECS): Likewise.
37497 (STARTFILE_SPEC): Likewise.
37498 (ENDFILE_SPEC): Likewise.
37499 (LIB_SPEC): Likewise.
37500 (TARGET_SDATA_DEFAULT): Likewise.
37501 (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37502 (MULTILIB_DEFAULTS): Likewise.
37503 (DWARF2_UNWIND_INFO): Likewise.
37504 * config/arc/big.h: New file.
37505 * config/arc/elf.h: Likewise.
37506 * config/arc/linux.h: Likewise.
37507 * config/arc/t-uClibc: Remove.
37509 2017-02-27 Bin Cheng <bin.cheng@arm.com>
37511 PR tree-optimization/77536
37512 * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37513 (tree_transform_and_unroll_loop): Use above function to compute the
37514 estimated niter of unrolled loop and use it when scaling profile.
37515 Also use count info rather than frequency if it's non-zero.
37516 * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37517 * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37518 (vect_transform_loop): Call above function.
37520 2017-02-27 Richard Biener <rguenther@suse.de>
37522 PR tree-optimization/45397
37523 * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37524 * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37525 (visit_nary_op): Add pattern matching for CSEing sign-changed
37526 or truncated operations with wider ones.
37528 2017-02-27 Richard Biener <rguenther@suse.de>
37530 PR tree-optimization/79690
37531 * tree-vect-stmts.c (vectorizable_store): Use vector type
37532 built from the DR with address-space.
37534 2017-02-26 Gerald Pfeifer <gerald@pfeifer.com>
37536 * doc/invoke.texi (Optimize Options): Refine the description
37537 of asan-use-after-return.
37539 2017-02-25 Alan Modra <amodra@gmail.com>
37541 PR rtl-optimization/79584
37542 * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37543 base, not ad->base_term, the reg within base. Remove assertion
37544 that ad->base == ad->base_term. Replace gen_int_mode using
37545 bogus mode with const0_rtx.
37547 2017-02-25 Jakub Jelinek <jakub@redhat.com>
37549 PR middle-end/79396
37550 * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37551 FMA_EXPR like tcc_binary or tcc_unary.
37553 * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37556 * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37558 (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37559 (output_loc_operands): Handle DW_OP_call_ref and
37560 DW_OP_GNU_variable_value.
37561 (struct variable_value_struct): New type.
37562 (struct variable_value_hasher): Likewise.
37563 (variable_value_hash): New variable.
37564 (string_types): Remove.
37565 (copy_loc_descr): New function.
37566 (add_loc_descr_to_each): Clarify comment. Use copy_loc_descr.
37567 (prepend_loc_descr_to_each): New function.
37568 (add_loc_list): Fix comment typo. Use prepend_loc_descr_to_each
37569 instead of add_loc_descr_to_each if the first argument is single
37570 location list and the second has multiple.
37571 (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37572 (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37573 when looking for variable value which doesn't have other location info.
37574 (loc_list_from_tree): Formatting fix.
37575 (gen_array_type_die): Simplify DW_AT_string_length handling.
37576 (adjust_string_types): Remove.
37577 (gen_subprogram_die): Don't call adjust_string_types nor test/set
37578 string_types. Call resolve_variable_values.
37579 (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37580 (resolve_addr_in_expr): Likewise. Add A argument.
37581 (copy_deref_exprloc): Remove deref argument. Adjust for the
37582 original expression being DW_OP_GNU_variable_value with optionally
37583 DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37584 optionally after it.
37585 (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37586 (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37587 callers. Set remove_AT_byte_size if removing DW_AT_string_length.
37588 (variable_value_hasher::hash, variable_value_hasher::equal): New
37590 (resolve_variable_value_in_expr, resolve_variable_value,
37591 resolve_variable_values, note_variable_value_in_expr,
37592 note_variable_value): New functions.
37593 (dwarf2out_early_finish): Call note_variable_value on all toplevel
37596 2017-02-24 Jakub Jelinek <jakub@redhat.com>
37599 * opts.h (handle_generated_option): Add GENERATED_P argument.
37600 * opts-common.c (handle_option): Adjust function comment.
37601 (handle_generated_option): Add GENERATED_P argument, pass it to
37603 (control_warning_option): Pass false to handle_generated_option
37605 * opts.c (maybe_default_option): Pass true to handle_generated_option
37607 * optc-gen.awk: Likewise.
37609 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
37611 * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37612 a REG, look at the REG it is a SUBREG of.
37613 (splitter for cmpeqsi_t): Ditto.
37615 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
37617 * config/pa/pa.c (pa_combine_instructions): Do not share RTL. Make
37618 the special USEs with the pattern of the insn, not the insn itself.
37620 2017-02-24 Matthew Fortune <matthew.fortune@imgtec.com>
37623 * doc/invoke.texi: Document -mload-store-pairs.
37625 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
37626 Sandra Loosemore <sandra@codesourcery.com>
37628 * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37629 argument isn't a CONST_INT.
37630 (nios2_alternate_compare_const): Assert op is a CONST_INT.
37631 (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37632 (nios2_validate_compare): Bypass alternate compare logic if *op2
37633 is not a CONST_INT.
37634 (ldstwm_operation_p): Return false if first_base is not a REG or
37635 if first_offset is not a CONST_INT.
37637 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
37639 * config/cris/cris.md: Use correct operand in a define_peephole2.
37641 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
37643 * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37645 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
37647 * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37648 this_insn if it is an INSN or JUMP_INSN.
37649 (force_offsettable): Look at base, not at addr.
37650 * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37651 on things that aren't necessarily CONST_INTs.
37653 2017-02-24 Uros Bizjak <ubizjak@gmail.com>
37655 * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37656 -mfpmath=sse is the default also for x86-32 targets with SSE2
37657 instruction set when @option{-ffast-math} is enabled
37659 2017-02-24 Jeff Law <law@redhat.com>
37661 PR rtl-optimizatoin/79286
37662 * ira.c (update_equiv_regs): Drop may_trap_p exception to
37665 2017-02-24 Richard Biener <rguenther@suse.de>
37667 PR tree-optimization/79389
37668 * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37671 2017-02-24 Aldy Hernandez <aldyh@redhat.com>
37673 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37674 function comment to reflect reality.
37675 (loop_exits_before_overflow): Fix typo in function description.
37677 2017-02-24 Richard Biener <rguenther@suse.de>
37679 PR tree-optimization/79389
37680 * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37681 properly that a threading opportunity exists. Detect conditional
37682 copy/constant propagation opportunities.
37684 2017-02-23 Eric Botcazou <ebotcazou@adacore.com>
37686 * config/visium/visium.md (type): Add trap.
37687 (b): New mode attribute.
37688 (*btst): Rename into...
37689 (*btst<mode>): ...this and adjust.
37690 (*cbranchsi4_btst_insn): Rename into...
37691 (*cbranch<mode>4_btst_insn): ...this and adjust.
37692 (trap): New define_insn.
37694 2017-02-23 Jakub Jelinek <jakub@redhat.com>
37696 PR tree-optimization/79389
37697 * ifcvt.c (struct noce_if_info): Add rev_cond field.
37698 (noce_reversed_cond_code): New function.
37699 (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37700 reversed_comparison_code. Formatting fix.
37701 (noce_try_store_flag): Test rev_cond != NULL in addition to
37702 reversed_comparison_code.
37703 (noce_try_store_flag_constants): Likewise.
37704 (noce_try_store_flag_mask): Likewise.
37705 (noce_try_addcc): Use rev_cond if non-NULL instead of
37706 reversed_comparison_code.
37707 (noce_try_cmove_arith): Likewise. Formatting fixes.
37708 (noce_try_minmax, noce_try_abs): Clear rev_cond.
37709 (noce_find_if_block): Initialize rev_cond.
37710 (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37711 instead of false as last argument never attempt to reverse it
37714 2017-02-23 Bin Cheng <bin.cheng@arm.com>
37716 PR tree-optimization/79663
37717 * tree-predcom.c (combine_chains): Process refs in reverse order
37718 only for ZERO length chains, and add explaining comment.
37720 2017-02-23 Jeff Law <law@redhat.com>
37722 PR tree-optimization/79578
37723 * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37724 in call to operand_equal_p.
37726 2017-01-23 Dominique d'Humieres <dominiq@lps.ens.fr>
37729 * config/i386/cpuid.h: Fix another undefined behavior.
37731 2017-02-23 Richard Biener <rguenther@suse.de>
37733 PR tree-optimization/79683
37734 * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37735 vector types for data-refs.
37737 2017-02-23 Martin Liska <mliska@suse.cz>
37739 * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37741 2017-02-23 Jakub Jelinek <jakub@redhat.com>
37743 PR middle-end/79665
37744 * internal-fn.c (get_range_pos_neg): Moved to ...
37745 * tree.c (get_range_pos_neg): ... here. No longer static.
37746 * tree.h (get_range_pos_neg): New prototype.
37747 * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37748 are known to be in between 0 and signed maximum inclusive, try to
37749 expand both unsigned and signed divmod and use the cheaper one from
37752 2017-02-22 Jeff Law <law@redhat.com>
37754 PR tree-optimization/79578
37755 * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37756 to compare base operands.
37758 2017-02-22 Segher Boessenkool <segher@kernel.crashing.org>
37761 * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37762 gpc_reg_operand instead of fpr_reg_operand.
37764 2017-02-22 Sameera Deshpande <sameera.deshpande@imgtec.com>
37766 * config/mips/mips.c (mips_return_in_memory): Force FP
37767 vector types to be returned in memory for o32 ABI.
37769 2017-02-22 Jakub Jelinek <jakub@redhat.com>
37771 * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37772 instead of DW_TAG_member for static data member declarations and don't
37773 set no_linkage_name for static inline data members.
37774 (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37777 2017-02-22 Martin Liska <mliska@suse.cz>
37779 * doc/invoke.texi: Replace inequality signs with square brackets
37782 2017-02-22 Matthew Fortune <matthew.fortune@imgtec.com>
37785 * lra-constraints.c (simplify_operand_subreg): Handle
37786 WORD_REGISTER_OPERATIONS targets.
37788 2017-02-22 Jakub Jelinek <jakub@redhat.com>
37791 * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37792 and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37793 elimination by swapping fld*.
37795 2017-02-22 Richard Biener <rguenther@suse.de>
37797 PR tree-optimization/79673
37798 * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37799 convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37800 irrelevant address-space qualifiers and avoiding a
37801 ADDR_SPACE_CONVERT_EXPR from fold_convert.
37803 2017-02-22 Richard Biener <rguenther@suse.de>
37805 PR tree-optimization/79666
37806 * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37807 to not symbolically negate if that may introduce undefined
37810 2017-02-22 Martin Liska <mliska@suse.cz>
37813 * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37814 * data-streamer-out.c (streamer_write_gcov_count_stream):
37816 * value-prof.c (stream_out_histogram_value): Make assert more
37817 precise based on type of counter.
37819 2017-02-21 Uros Bizjak <ubizjak@gmail.com>
37822 * config/i386/i386.md (standard_x87sse_constant_load splitter):
37823 Use nonimmediate_operand instead of memory_operand for operand 1.
37824 (float-extend standard_x87sse_constant_load splitter): Ditto.
37826 2017-02-21 Jeff Law <law@redhat.com>
37828 PR tree-optimization/79621
37829 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37830 blocks with edges to themselves.
37832 2017-02-21 Jakub Jelinek <jakub@redhat.com>
37835 * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37836 is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37837 Use gimple_call_builtin_p.
37840 * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37841 on temporarily removed DEBUG_INSNs.
37843 PR tree-optimization/79649
37844 * tree-loop-distribution.c (classify_partition): Give up on
37845 non-generic address space loads/stores.
37847 2017-02-21 Aldy Hernandez <aldyh@redhat.com>
37849 * doc/loop.texi (Loop manipulation): Remove nonexistent
37850 tree_ssa_loop_version from the documentation.
37851 * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
37853 2017-02-21 Jakub Jelinek <jakub@redhat.com>
37856 * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
37857 make_reg_eh_region_note_nothrow_nononlocal on call_insn.
37858 * config/rs6000/rs6000.c: Include except.h.
37859 (rs6000_expand_split_stack_prologue): Call
37860 make_reg_eh_region_note_nothrow_nononlocal on the call insn.
37862 2017-02-21 Martin Jambor <mjambor@suse.cz>
37865 * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
37866 have been analyzed.
37868 2017-02-21 Martin Jambor <mjambor@suse.cz>
37870 * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
37871 for backward compatibility only.
37872 * doc/invoke.texi (Option Summary): Remove all references to
37873 -fipa-cp-alignment.
37875 2017-02-21 Matthew Fortune <matthew.fortune@imgtec.com>
37879 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
37881 * lra-constraints.c (curr_insn_transform): Handle
37882 WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37884 2017-02-21 Martin Liska <mliska@suse.cz>
37886 * config/i386/i386.opt: Replace -masm-dialect with -masm.
37888 2017-02-21 Thomas Schwinge <thomas@codesourcery.com>
37890 PR translation/79638
37891 * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
37893 2017-02-21 Eric Botcazou <ebotcazou@adacore.com>
37896 * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
37897 (arm_function_ok_for_sibcall): Return false for an indirect call by
37898 descriptor if all the argument registers are used.
37899 (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
37900 alignment of the function.
37902 2017-02-21 Jakub Jelinek <jakub@redhat.com>
37904 PR tree-optimization/61441
37905 * simplify-rtx.c (simplify_const_unary_operation): For
37906 -fsignaling-nans and sNaN operand, return NULL_RTX rather than
37907 the sNaN unmodified.
37909 2017-02-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
37911 * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
37912 (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
37913 instead of SYSTEM_HEADER_DIR.
37915 2017-02-20 Gerald Pfeifer <gerald@pfeifer.com>
37916 Martin Liška <mliska@suse.cz>
37918 * doc/invoke.texi (use-after-scope-direct-emission-threshold):
37919 Fix typos and grammar, use active voice, and clarify.
37921 2017-02-20 Marek Polacek <polacek@redhat.com>
37923 PR middle-end/79537
37924 * gimplify.c (gimplify_expr): Handle unused *&&L;.
37927 * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
37929 2017-02-20 Jakub Jelinek <jakub@redhat.com>
37932 * config/i386/i386.c (ix86_expand_builtin): Handle
37933 OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
37934 ix86_builtins_isa[fcode].isa as a requirement of those
37935 flags and any other flag in the bitmask.
37936 (ix86_init_mmx_sse_builtins): Use 0 instead of
37937 ~OPTION_MASK_ISA_64BIT as mask.
37938 * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
37939 __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
37940 __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
37941 __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
37943 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
37946 * lra-constraints.c (split_reg): Check requested split mode
37947 is supported by the register.
37949 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
37951 * lra-constraints.c (simplify_operand_subreg): Remove early
37954 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
37957 * lra-constraints.c (curr_insn_transform): Tighten condition
37958 for converting SUBREG reloads from OP_OUT to OP_INOUT.
37960 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
37963 * lra-constraints.c (curr_insn_transform): Handle
37964 WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37966 2017-02-19 Uros Bizjak <ubizjak@gmail.com>
37969 2016-05-30 Uros Bizjak <ubizjak@gmail.com>
37971 * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
37973 2017-02-19 Jonathan Wakely <jwakely@redhat.com>
37976 * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
37979 2017-02-19 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
37981 * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
37984 2017-02-18 Jakub Jelinek <jakub@redhat.com>
37986 * final.c (last_columnnum, override_columnnum): New variables.
37987 (final_start_function): Set last_columnnum, pass it to begin_prologue
37988 hook and pass 0 to dwarf2out_begin_prologue.
37989 (final_scan_insn): Update override_columnnum. Pass last_columnnum
37990 to source_line debug hook.
37991 (notice_source_line): Compute last_columnnum and for debug_column_info
37992 return true on column changes.
37993 * debug.h (struct gcc_debug_hooks): Add column argument to
37994 source_line and begin_prologue hooks.
37995 (debug_nothing_int_charstar_int_bool): Remove prototype.
37996 (debug_nothing_int_int_charstar,
37997 debug_nothing_int_int_charstar_int_bool): New prototypes.
37998 (dwarf2out_begin_prologue): Add column argument.
37999 * debug.c (do_nothing_debug_hooks): Adjust source_line and
38000 begin_prologue hooks.
38001 (debug_nothing_int_charstar_int_bool): Remove.
38002 (debug_nothing_int_int_charstar,
38003 debug_nothing_int_int_charstar_int_bool): New functions.
38004 * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38005 through to dwarf2out_source_line.
38006 (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38007 (dwarf2out_source_line): Add column argument, emit it if requested.
38008 * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38010 * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38011 * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38012 * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38013 through to dwarf2out_begin_prologue.
38014 (vmsdbgout_source_line): Add column argument, pass it through to
38015 dwarf2out_source_line.
38016 * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38017 dbxout_source_line caller.
38018 (dbxout_source_line): Add column argument.
38020 * common.opt (gno-column-info, gcolumn-info): New options.
38021 * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38022 (check_die): Also test for multiple DW_AT_decl_column attributes.
38023 (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38024 DW_AT_decl_column if requested.
38025 (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38027 (gen_variable_die): Likewise.
38028 (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38029 * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38032 * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38033 * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38034 (ix86_handle_option): Handle OPT_m3dnowa.
38035 * doc/invoke.texi (-m3dnowa): Document.
38036 * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38037 -m3dnowa instead of -m3dnow -march=athlon.
38040 * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38041 instead of gcc_assert for K, r and R code checks. Formatting fixes.
38043 2017-02-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
38046 * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38047 support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38048 * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38049 generator for vsx_xxpermdi_<mode>_be.
38050 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38051 force big-endian semantics.
38052 (vsx_xxpermdi_<mode>_be): New define_expand with same
38053 implementation as previous version of vsx_xxpermdi_<mode>.
38055 2017-02-17 Jakub Jelinek <jakub@redhat.com>
38057 PR tree-optimization/79327
38058 * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38059 variable, its initialization and use.
38061 2017-02-17 Julia Koval <julia.koval@intel.com>
38063 * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38064 (OPTION_MASK_ISA_PKU_UNSET): New.
38065 (ix86_handle_option): Handle -mrdpid.
38066 * config/i386/cpuid.h (bit_RDPID): New.
38067 * config/i386/driver-i386.c (host_detect_local_cpu):
38068 Detect RDPID feature.
38069 * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38070 * config/i386/i386-c.c (ix86_target_macros_internal):
38072 * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38073 (ix86_valid_target_attribute_inner_p): Add "rdpid".
38074 (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38075 * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38076 * config/i386/i386.md (define_insn "rdpid"): New.
38077 * config/i386/i386.opt Add -mrdpid.
38078 * config/i386/immintrin.h (_rdpid_u32): New.
38080 2017-02-17 Vladimir Makarov <vmakarov@redhat.com>
38082 PR rtl-optimization/79541
38083 * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38084 instead of transforming it into USE.
38086 2017-02-17 Segher Boessenkool <segher@kernel.crashing.org>
38088 * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38089 If HONOR_SNANS (SFmode) force the input to a register.
38090 (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38091 (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38092 an frsp or similar insn.
38094 2017-02-17 Martin Liska <mliska@suse.cz>
38096 PR rtl-optimization/79577
38097 * params.def (selsched-max-sched-times): Increase minimum to 1.
38099 2017-02-17 Martin Liska <mliska@suse.cz>
38101 PR rtl-optimization/79574
38102 * gcse.c (want_to_gcse_p): Prevent integer overflow.
38104 2017-02-17 Martin Liska <mliska@suse.cz>
38106 PR tree-optimization/79529
38107 * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38108 ssa_defined_default_def_p to handle cases which are implicitly
38110 * tree-ssa.c (ssa_defined_default_def_p): New function.
38111 (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38112 which are implicitly defined.
38113 * tree-ssa.h (ssa_defined_default_def_p): Declare.
38115 2017-02-17 Richard Biener <rguenther@suse.de>
38117 PR middle-end/79576
38118 * params.def (max-ssa-name-query-depth): Limit to 10.
38120 2017-02-17 Richard Biener <rguenther@suse.de>
38122 PR tree-optimization/79552
38123 * tree-ssa-structalias.c (visit_loadstore): Properly verify
38126 2017-02-17 Richard Biener <rguenther@suse.de>
38129 * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38131 2017-02-17 Marek Polacek <polacek@redhat.com>
38133 PR middle-end/79536
38134 * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38135 (fold_negate_expr): New wrapper.
38137 2017-02-16 Sandra Loosemore <sandra@codesourcery.com>
38139 * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]:
38140 Correct terminology and de-emphasize pre-standard behavior.
38142 2017-02-16 Alan Modra <amodra@gmail.com>
38144 PR rtl-optimization/79286
38145 * ira.c (def_dominates_uses): New function.
38146 (update_equiv_regs): Don't create an equivalence for insns that
38147 may trap where the register def does not dominate the use.
38149 2017-02-16 Vladimir Makarov <vmakarov@redhat.com>
38151 PR rtl-optimization/78127
38152 * lra.c (lra): Call lra_eliminate before finish the loop after
38155 2017-02-16 Richard Biener <rguenther@suse.de>
38157 * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38159 * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38160 (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38161 * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38162 (isl_val_int_from_wi): New function.
38163 (extract_affine_gmp): Rename to ...
38164 (extract_affine_wi): ... this, take a widest_int.
38165 (extract_affine_int): Just wrap extract_affine_wi.
38166 (add_param_constraints): Use isl_val_int_from_wi.
38167 (add_loop_constraints): Likewise, and extract_affine_wi.
38169 2017-02-15 Jeff Law <law@redhat.com>
38171 PR middle-end/79521
38172 * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38173 ira_init_register_move_cost_if_necessary.
38175 2017-02-15 Martin Sebor <msebor@redhat.com>
38177 PR middle-end/32003
38178 * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38179 removed in a prior commit.
38181 2017-02-15 Bin Cheng <bin.cheng@arm.com>
38183 PR tree-optimization/79347
38184 * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38185 counters during peeling.
38187 2017-02-15 Thomas Schwinge <thomas@codesourcery.com>
38189 * Makefile.in (site.exp): Remove "set ISLVER".
38191 2017-02-15 Jakub Jelinek <jakub@redhat.com>
38194 * real.c (real_from_integer): Call real_convert even for decimal.
38196 2017-02-15 Dominik Vogt <vogt@linux.vnet.ibm.com>
38199 * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38201 2017-02-14 Andrew Pinski <apinski@cavium.com>
38203 * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38204 cores and change the partno/implementer to be correct.
38205 (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38206 the 'B" as the implementer.
38207 * config/aarch64/aarch64-tune.md: Regenerate.
38209 2017-02-14 Carl Love <cel@us.ibm.com>
38211 * config/rs6000/rs6000.c: Add case statement entry to make the
38212 xvcvuxdsp built-in argument unsigned.
38213 * config/rs6000/vsx.md: Fix the source and return operand types so they
38214 match the instruction definitions from the ISA document. Fix typo
38215 in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38218 2017-02-14 Vladimir Makarov <vmakarov@redhat.com>
38221 * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38222 member early_clobber_alts.
38223 * lra-lives.c (reg_early_clobber_p): New.
38224 (process_bb_lives): Use it.
38225 * lra.c (new_insn_reg): New arg early_clobber_alts. Use it.
38226 (debug_operand_data): Initialize early_clobber_alts.
38227 (setup_operand_alternative): Set up early_clobber_alts.
38228 (collect_non_operand_hard_regs): Ditto. Pass early clobber
38229 alternatives to new_insn_reg.
38230 (add_regs_to_insn_regno_info): Add arg early_clobber_alts. Use
38232 (lra_update_insn_regno_info): Pass the new arg.
38234 2017-02-14 Jakub Jelinek <jakub@redhat.com>
38236 PR middle-end/79505
38237 * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38238 (new_oacc_loop_raw): Don't clear already cleared fields.
38241 * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38242 _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38243 _mm512_prefetch_i64gather_ps): New inline functions and macros.
38245 2017-02-14 Uros Bizjak <ubizjak@gmail.com>
38248 * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38250 2017-02-14 H.J. Lu <hongjiu.lu@intel.com>
38253 * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38254 the extra instruction to the right place to store 128-bit constant
38257 2017-02-14 Martin Sebor <msebor@redhat.com>
38259 PR middle-end/79448
38260 * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38261 warning for strings of unknown length.
38263 2017-02-13 Segher Boessenkool <segher@kernel.crashing.org>
38265 * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38267 2017-02-14 Jeff Law <law@redhat.com>
38270 * ira-costs.c (scan_one_insn): Initialize register move costs
38271 for pseudos seen in USE/CLOBBER insns.
38273 PR tree-optimization/79095
38274 * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38275 if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38276 (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38277 if the operands are known to be not equal, then the resulting range
38279 (intersect_ranges): If the new range is ~[0,0] and the old range is
38280 wide, then prefer ~[0,0].
38281 * tree-vrp.c (overflow_comparison_p_1): New function.
38282 (overflow_comparison_p): New function.
38283 * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38284 if NAME is used in an overflow test.
38285 (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38286 overflow check that can be expressed as an equality test, then adjust
38287 ops to be that equality test.
38289 2017-02-14 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
38291 * config/s390/s390-builtin-types.def: Remove flags argument.
38292 * config/s390/s390.c (s390_init_builtins): Likewise.
38294 2017-02-14 Martin Liska <mliska@suse.cz>
38296 * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38297 vector. Fix trailing white spaces.
38299 2017-02-14 James Greenhalgh <james.greenhalgh@arm.com>
38301 * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38304 2017-02-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
38306 PR rtl-optimization/68664
38307 * config/arm/arm.c (arm_sched_can_speculate_insn):
38308 New function. Declare prototype.
38309 (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38311 2017-02-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
38313 PR rtl-optimization/68664
38314 * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38316 (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38318 2017-02-14 Amit Pawar <amit.pawar@amd.com>
38320 * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38321 max skip bytes for function, loop and jump.
38323 2017-02-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
38325 * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38326 ABS_EXPR for gimple dump.
38328 2017-02-14 Jakub Jelinek <jakub@redhat.com>
38331 * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38333 PR tree-optimization/79408
38334 * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38335 case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38336 (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38337 also if rhs1 is INTEGER_CST.
38339 2017-02-14 Richard Biener <rguenther@suse.de>
38341 PR middle-end/79432
38342 * tree-into-ssa.c (insert_phi_nodes): When the function can
38343 have abnormal edges rewrite SSA names with broken use-def
38344 dominance out of SSA and register them for PHI insertion.
38346 2017-02-13 Martin Sebor <msebor@redhat.com>
38348 PR middle-end/79496
38349 * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38350 clearing info.nowrite flag when snprintf size argument is a range.
38352 2017-02-13 Jakub Jelinek <jakub@redhat.com>
38354 * cprop.c (cprop_jump): Add missing space in string literal.
38355 * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38356 (get_constraint_for_component_ref): Likewise.
38357 * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38358 * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38359 * lra-constraints.c (process_alt_operands): Likewise.
38360 * ipa-inline.c (inline_small_functions): Likewise.
38361 * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38362 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38363 * trans-mem.c (diagnose_tm_1_op): Likewise.
38364 * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38365 (grid_parallel_clauses_gridifiable): Likewise.
38367 * config/nvptx/mkoffload.c (process): Add space in between
38370 * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38371 "MOD4_SSE_REGS" and "ALL_REGS".
38373 * spellcheck.c (test_data): Add , in between "foo" and "food".
38375 2017-02-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
38378 * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38379 boundary crossing check and subsequent code generation agree.
38381 2017-02-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
38383 * config/aarch64/aarch64.c (has_memory_op): Delete.
38384 (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38387 2017-02-13 Jakub Jelinek <jakub@redhat.com>
38389 PR rtl-optimization/79388
38390 PR rtl-optimization/79450
38391 * combine.c (distribute_notes): When removing TEM_INSN for which
38392 corresponding dest has last value recorded, invalidate that last
38395 2017-02-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
38397 * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38398 of explicit '@'. Add missing assembly comment marker on branch costs
38401 2017-02-13 Nathan Sidwell <nathan@acm.org>
38403 * gengtype-lex.l (<in_struct>): Add '/'.
38405 2017-02-13 Martin Liska <mliska@suse.cz>
38408 * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38410 2017-02-13 Richard Biener <rguenther@suse.de>
38412 * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38414 * configure: Re-generate.
38415 * config.in: Likewise.
38416 * graphite-dependences.c: Simplify as if
38417 HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38418 * graphite-isl-ast-to-gimple.c: Likewise.
38419 * graphite-optimize-isl.c: Likewise.
38420 * graphite-poly.c: Likewise.
38421 * graphite-sese-to-poly.c: Likewise.
38422 * graphite.h: Likewise.
38423 * toplev.c: Include isl/version.h and use isl_version () for
38424 printing the ISL version.
38425 * doc/install.texi: Update ISL requirement.
38427 2017-02-12 Gerald Pfeifer <gerald@pfeifer.com>
38429 * doc/standards.texi (Standards): Update reference to
38432 2017-02-12 Gerald Pfeifer <gerald@pfeifer.com>
38434 * doc/extend.texi (Named Address Spaces): sourceware.org now
38436 * doc/install.texi (Binaries): Ditto.
38439 2017-02-11 Sandra Loosemore <sandra@codesourcery.com>
38441 * doc/cpp.texi: Replace "stringify"/"stringification" with C
38442 standard terminology "stringize"/"stringizing" throughout.
38443 * doc/cppinternals.texi: Likewise.
38445 2017-02-11 Sandra Loosemore <sandra@codesourcery.com>
38447 * doc/extend.texi: Fix some spelling mistakes and typos.
38448 * doc/invoke.texi: Likewise.
38450 2017-02-11 Jan Hubicka <hubicka@ucw.cz>
38453 * params.def (inline-min-speedup) Change from 10 to 8.
38455 2017-02-11 Jakub Jelinek <jakub@redhat.com>
38457 * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38460 2017-02-11 Jan Hubicka <hubicka@ucw.cz>
38463 * ipa-inline-analysis.c (get_minimal_bb): New function.
38464 (record_modified): Use it.
38465 (remap_edge_change_prob): Handle also ancestor functions.
38467 2017-02-11 Gerald Pfeifer <gerald@pfeifer.com>
38469 * doc/contrib.texi (Contributors): Remove broken link into
38470 the Mauve CVS repository.
38472 2017-02-11 Jakub Jelinek <jakub@redhat.com>
38474 PR middle-end/79454
38475 * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38476 result computation whenever lhs doesn't have vector mode, not
38477 just when it has BLKmode.
38479 2017-02-10 Gerald Pfeifer <gerald@pfeifer.com>
38481 * doc/makefile.texi (profiledbootstrap): Refer to the
38482 installation instructions only in textual form.
38484 2017-02-10 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
38487 * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38489 2017-02-10 Gerald Pfeifer <gerald@pfeifer.com>
38491 * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38492 (Specific): Update mingw-w64 reference.
38494 (Specific): Remove broken link to Renesas RX processor.
38496 2017-02-10 Richard Biener <rguenther@suse.de>
38498 * toplev.c (process_options): Do not mention obsolete graphite
38499 options when printing sorry message about missing graphite support.
38500 Mention -floop-nest-optimize.
38502 2017-02-10 Christophe Lyon <christophe.lyon@linaro.org>
38504 * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38505 (vtst_p16): Likewise.
38506 (vtstq_p8): Likewise.
38507 (vtstq_p16): Likewise.
38509 (vtstq_p64): Likewise.
38510 * config/arm/arm_neon.h (vgetq_lane_p64): New.
38511 (vset_lane_p64): New.
38512 (vsetq_lane_p64): New.
38514 2017-02-10 Jakub Jelinek <jakub@redhat.com>
38516 PR tree-optimization/79411
38517 * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38518 stmt operands are SSA_NAMEs used in abnormal phis.
38519 (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38522 2017-02-09 Jan Hubicka <hubicka@ucw.cz>
38525 * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38528 2017-02-09 Jan Hubicka <hubicka@ucw.cz>
38530 * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38532 2017-02-09 Jakub Jelinek <jakub@redhat.com>
38534 * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38538 * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38539 not arbitrary TREE_CONSTANT.
38542 * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38543 "omp declare target link" attribute unless is_global_var.
38544 * omp-offload.c (find_link_var_op): Likewise.
38546 2017-02-09 Nathan Sidwell <nathan@codesourcery.com>
38547 Chung-Lin Tang <cltang@codesourcery.com>
38549 * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38551 (gimplify_adjust_omp_clauses): Don't delete TILE.
38552 (gimplify_omp_for): Deal with TILE.
38553 * internal-fn.c (expand_GOACC_TILE): New function.
38554 * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38556 * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38557 (expand_oacc_collapse_init): Add LOC paramter. Initialize tile
38559 (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38560 avoid DIV for outermost collapse var.
38561 (expand_oacc_for): Insert tile element loop as needed. Adjust.
38562 Remove out of date comments, fix whitespace.
38563 * omp-general.c (omp_extract_for_data): Deal with tiling.
38564 * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38565 adjust OLF_DIM_BASE value.
38566 (struct omp_for_data): Add tiling field.
38567 * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38568 (lower_oacc_head_mark): Add OLF_TILE as appropriate. Ensure 2 levels
38569 for auto loops. Remove default auto determining, moved to
38570 oacc_loop_fixed_partitions.
38571 * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38572 stmts, add e_mask field.
38573 (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38574 (oacc_thread_numbers): Use oacc_dim_call.
38575 (oacc_xform_tile): New.
38576 (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38577 (finish_oacc_loop): Adjust for ifns vector.
38578 (oacc_loop_discover_walk): Append loop abstraction sites to list,
38579 add case for GOACC_TILE fns.
38580 (oacc_loop_xform_loop): Delete.
38581 (oacc_loop_process): Iterate over call list directly, and add
38582 handling for GOACC_TILE fns.
38583 (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38585 (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38586 vector partitioning to outer loops. Assign 2 partitions to loops
38587 when available. Add TILE handling.
38588 (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38589 (execite_oacc_device_lower): Process GOACC_TILE fns,
38590 ignore unknown specs.
38591 * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38592 * tree.c (omp_clause_num_ops): Adjust TILE ops.
38593 * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38595 2017-02-09 Gerald Pfeifer <gerald@pfeifer.com>
38597 * configure.ac (ACX_BUGURL): Update.
38598 * configure: Regenerate.
38600 2017-02-09 Richard Biener <rguenther@suse.de>
38602 PR tree-optimization/69823
38603 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38604 Properly enumerate all BBs in the region. Use auto_vec/auto_bitmap.
38606 2017-02-09 Andrew Burgess <andrew.burgess@embecosm.com>
38608 * config/arc/arc-c.def: Add __NPS400__ definition.
38609 * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38610 (TARGET_NPS400): Define.
38612 2017-02-09 Andrew Burgess <andrew.burgess@embecosm.com>
38614 * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38616 (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38617 pointer, arch_info.
38618 (arc_cpu_types): Fill the arch_info field with a pointer into the
38619 arc_arch_types table.
38620 (arc_selected_cpu): Declare.
38621 * config/arc/arc.c (arc_selected_cpu): Make global.
38622 (arc_selected_arch): Delete.
38623 (arc_base_cpu): Delete.
38624 (arc_override_options): Remove references to deleted variables,
38625 update access to arch information.
38626 (ARC_OPT): Update access to arch information.
38627 (ARC_OPTX): Likewise.
38628 * config/arc/arc.h (arc_base_cpu): Remove declaration.
38629 (TARGET_ARC600): Update access to arch information.
38630 (TARGET_ARC601): Likewise.
38631 (TARGET_ARC700): Likewise.
38632 (TARGET_EM): Likewise.
38633 (TARGET_HS): Likewise.
38634 * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38637 2017-02-08 Pat Haugen <pthaugen@us.ibm.com>
38640 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38641 condition/operands for integer GE/LE/GEU/LEU operations.
38643 2017-02-08 Segher Boessenkool <segher@kernel.crashing.org>
38645 PR translation/79397
38646 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38649 2017-02-08 Martin Jambor <mjambor@suse.cz>
38652 * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38653 whether allocation happened.
38654 (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38655 nothing was allocated.
38657 2017-02-08 Jakub Jelinek <jakub@redhat.com>
38659 PR tree-optimization/79408
38660 * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38661 constant, but SSA_NAME with a known integer range, use the minimum
38662 of that range instead of op1 to determine if modulo can be replaced
38663 with its first operand.
38665 2016-02-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
38667 * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38669 2017-02-08 Richard Biener <rguenther@suse.de>
38671 PR tree-optimization/71824
38672 * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38673 Check all loops contained in the merged region.
38675 2017-02-07 Andrew Pinski <apinski@cavium.com>
38677 * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38679 2017-02-07 Andrew Pinski <apinski@cavium.com>
38681 * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38682 (thunderxt88): Likewise.
38683 (thunderxt81): Disable LSE and change v8.1 to v8.
38684 (thunderxt83): Likewise.
38686 2017-02-07 Jakub Jelinek <jakub@redhat.com>
38687 Richard Biener <rguenther@suse.de>
38689 PR middle-end/79399
38690 * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38691 type from int to size_t.
38692 * ira-costs.c (struct_costs_size): Change type from int to size_t.
38694 2017-02-07 Jakub Jelinek <jakub@redhat.com>
38696 PR rtl-optimization/79386
38697 * cprop.c (bypass_conditional_jumps): Initialize
38698 bypass_last_basic_block already before splitting bbs after
38699 unconditional traps...
38700 (bypass_conditional_jumps): ... rather than here.
38703 * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38704 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38705 *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38706 fix -masm=intel patterns.
38708 2017-02-07 Richard Biener <rguenther@suse.de>
38710 PR tree-optimization/79256
38711 PR middle-end/79278
38712 * builtins.c (get_object_alignment_2): Use min_align_of_type
38713 to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38714 and ADJUST_FIELD_ALIGN.
38716 * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38718 * doc/tm.texi: Regenerate.
38719 * stor-layout.c (layout_decl): Adjust.
38720 (update_alignment_for_field): Likewise.
38721 (place_field): Likewise.
38722 (min_align_of_type): Likewise.
38723 * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38724 * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38725 * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38726 * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38727 * config/frv/frv.c (frv_adjust_field_align): Likewise.
38728 * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38729 * config/i386/i386.c (x86_field_alignment): Likewise.
38730 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38731 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38732 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38733 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38734 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38735 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38739 2017-01-30 Richard Biener <rguenther@suse.de>
38741 PR tree-optimization/79256
38742 * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38743 BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38746 2017-02-07 Toma Tabacu <toma.tabacu@imgtec.com>
38748 * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38749 argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38750 builtins to SImode and emit a zero-extend, if necessary.
38752 2017-02-06 Palmer Dabbelt <palmer@dabbelt.com>
38754 * docs/invoke.texi (RISC-V Options): Alphabetize.
38756 2017-02-06 Palmer Dabbelt <palmer@dabbelt.com>
38758 * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38761 2017-02-06 Palmer Dabbelt <palmer@dabbelt.com>
38763 * config/riscv/riscv.c: New file.
38764 * common/config/riscv/riscv-common.c: Likewise.
38765 * config.gcc: Likewise.
38766 * config/riscv/constraints.md: Likewise.
38767 * config/riscv/elf.h: Likewise.
38768 * config/riscv/generic.md: Likewise.
38769 * config/riscv/linux.h: Likewise.
38770 * config/riscv/multilib-generator: Likewise.
38771 * config/riscv/peephole.md: Likewise.
38772 * config/riscv/pic.md: Likewise.
38773 * config/riscv/predicates.md: Likewise.
38774 * config/riscv/riscv-builtins.c: Likewise.
38775 * config/riscv/riscv-c.c: Likewise.
38776 * config/riscv/riscv-ftypes.def: Likewise.
38777 * config/riscv/riscv-modes.def: Likewise.
38778 * config/riscv/riscv-opts.h: Likewise.
38779 * config/riscv/riscv-protos.h: Likewise.
38780 * config/riscv/riscv.h: Likewise.
38781 * config/riscv/riscv.md: Likewise.
38782 * config/riscv/riscv.opt: Likewise.
38783 * config/riscv/sync.md: Likewise.
38784 * config/riscv/t-elf-multilib: Likewise.
38785 * config/riscv/t-linux: Likewise.
38786 * config/riscv/t-linux-multilib: Likewise.
38787 * config/riscv/t-riscv: Likewise.
38788 * configure.ac: Likewise.
38789 * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38790 Waterman as RISC-V maintainers.
38791 * doc/install.texi: Add RISC-V entries.
38792 * doc/invoke.texi: Add RISC-V options section.
38793 * doc/md.texi: Add RISC-V constraints section.
38794 * configure: Regenerated.
38796 2017-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
38799 * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38800 false values to be constant vectors with all 0 or all 1 bits set.
38801 (vcondu<mode><mode>): Likewise.
38802 * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38804 (fpmask_comparison_operator): Update comment.
38805 (vecint_comparison_operator): New predicate.
38806 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38807 vector conditionals when the true and false values are constant
38808 vectors with all 0 bits or all 1 bits set.
38810 2017-02-06 Martin Sebor <msebor@redhat.com>
38812 PR tree-optimization/79376
38813 * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38815 2017-02-06 Uros Bizjak <ubizjak@gmail.com>
38817 * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38818 explicit subreg RTX with operand 1. Use VECTOR_MODE_P predicate
38819 to simplify split condition.
38821 2017-02-06 Jakub Jelinek <jakub@redhat.com>
38823 * omp-expand.c (oxpand_omp_atomic_fetch_op,
38824 expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38827 2017-02-06 Segher Boessenkool <segher@kernel.crashing.org>
38829 PR rtl-optimization/68664
38830 * target.def (can_speculate_insn): New hook.
38831 * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38832 * doc/tm.texi: Regenerate.
38833 * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38834 * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38835 (rs6000_sched_can_speculate_insn): New function.
38837 2017-02-06 Jakub Jelinek <jakub@redhat.com>
38839 PR tree-optimization/79284
38840 * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38841 * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38842 vectorizable_mask_load_store, vectorizable_operation,
38843 vect_is_simple_cond, get_same_sized_vectype): Use it instead
38844 of comparing TREE_CODE of a type against BOOLEAN_TYPE.
38845 * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
38846 vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
38847 * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
38848 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
38849 Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
38850 is_gimple_assign (stmt). Replace another such test with
38851 is_gimple_assign (stmt).
38853 2017-02-06 Georg-Johann Lay <avr@gjlay.de>
38856 * config/avr/avr.c (rtl-iter.h): Include it.
38857 (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
38858 (avr_legitimate_combined_insn): ...and implementation.
38860 2017-02-06 Dominik Vogt <vogt@linux.vnet.ibm.com>
38862 * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
38863 * config/s390/s390.c (s390_const_operand_ok)
38864 (s390_canonicalize_comparison, s390_extract_part)
38865 (s390_single_part, s390_contiguous_bitmask_nowrap_p)
38866 (s390_contiguous_bitmask_p, s390_rtx_costs)
38867 (legitimize_pic_address): Likewise.
38868 * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
38869 * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
38870 ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
38871 ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
38872 * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
38874 2017-02-06 Georg-Johann Lay <avr@gjlay.de>
38876 * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
38877 REGNO($0) == REGNO($1).
38879 2017-02-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
38881 * config/s390/linux.h(SIZE_TYPE): Add comment.
38883 2017-02-06 Julian Brown <julian@codesourcery.com>
38884 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
38885 Virendra Pathak <virendra.pathak@broadcom.com>
38887 * config/aarch64/aarch64-cores.def: Change the scheduler
38889 * config/aarch64/aarch64.md: Include thunderx2t99.md.
38890 * config/aarch64/thunderx2t99.md: New file.
38892 2017-02-05 Gerald Pfeifer <gerald@pfeifer.com>
38894 * doc/standards.texi (Go Language): Update link to language
38897 2017-02-05 Jan Hubicka <hubicka@ucw.cz>
38899 * tree-eh.c (lower_resx): Sanitize profile.
38900 (cleanup_empty_eh_move_lp): Likewise.
38902 2017-02-05 Jan Hubicka <hubicka@ucw.cz>
38905 * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
38907 * cfgloopmanip.h (loop_version): Update prototype.
38908 * modulo-sched.c (sms_schedule): Update call of loop_version.
38909 * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
38910 * tree-parloops.c (gen_parallel_loop): Likewise.
38911 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
38912 * tree-ssa-loop-split.c (split_loop): Likewise.
38913 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
38914 * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
38916 2017-02-05 Martin Liska <mliska@suse.cz>
38919 * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
38921 (print_operand_address): Initialize a struct to zero.
38923 2017-02-05 Gerald Pfeifer <gerald@pfeifer.com>
38925 * doc/contrib.texi (Contributors): Refer to Hans Boehm's
38926 garbage collector only in textual form.
38928 2017-02-05 Gerald Pfeifer <gerald@pfeifer.com>
38930 * doc/extend.texi (x86 specific memory model extensions for
38931 transactional memory): Simplify a phrase.
38933 2017-02-05 Eric Botcazou <ebotcazou@adacore.com>
38936 * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
38937 'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
38938 (atomic_storedi_1): Likewise.
38940 2017-02-04 Jakub Jelinek <jakub@redhat.com>
38942 PR tree-optimization/79338
38943 * tree-parloops.c (gather_scalar_reductions): Don't call
38944 vect_analyze_loop_form for loop->inner before destroying loop's
38947 2017-02-03 Martin Sebor <msebor@redhat.com>
38949 PR tree-optimization/79327
38950 * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
38951 when precision has resulted in leading zeros.
38952 (format_integer): Adjust the likely counter to assume an unknown
38953 argument that may be zero is non-zero.
38955 2017-02-03 Jason Merrill <jason@redhat.com>
38958 * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
38959 avoid copying non-taken branch.
38961 2017-02-03 Jakub Jelinek <jakub@redhat.com>
38963 PR tree-optimization/79340
38964 * tree-vect-loop.c (vectorizable_reduction): Release
38965 vec_defs elements after safe_splicing them into other vectors.
38968 PR tree-optimization/79327
38969 * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
38970 true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
38972 (format_integer): Use wide_int_to_tree instead of build_int_cst
38973 + to_?hwi. If argmin is NULL, just set argmin and argmax to
38974 TYPE_{MIN,MAX}_VALUE of argtype. Simplify and fix computation
38975 of shortest and longest sequence.
38977 2017-02-03 Uros Bizjak <ubizjak@gmail.com>
38979 * config/i386/i386.c (dimode_scalar_chain::convert_reg):
38980 Use pextrd for TARGET_SSE4_1 when creating scalar copy.
38982 2017-02-03 Walter Lee <walt@tilera.com>
38985 * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
38986 after initial stackframe link reg save.
38987 * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
38989 2017-02-03 Jakub Jelinek <jakub@redhat.com>
38992 * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
38993 wu for stxssp alternative.
38995 2017-02-03 Martin Sebor <msebor@redhat.com>
38997 PR tree-optimization/79352
38998 * gimple-fold.c (get_range_strlen): Add argument.
38999 (get_range_strlen): Change return type to bool.
39000 (get_maxval_strlen): Pass in a dummy argument.
39001 * gimple-fold.h (get_range_strlen): Change return type to bool.
39002 * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39003 * tree.h (array_at_struct_end_p): Add argument.
39004 * tree.c (array_at_struct_end_p): Handle it.
39006 2017-02-03 Martin Liska <mliska@suse.cz>
39009 * multiple_target.c (create_dispatcher_calls): Redirect edge
39010 from a caller of a dispatcher.
39011 (expand_target_clones): Make the clones local.
39012 (ipa_target_clone): Do both target clones and resolvers.
39013 (ipa_dispatcher_calls): Remove the pass.
39014 (pass_dispatcher_calls::gate): Likewise.
39015 (make_pass_dispatcher_calls): Likewise.
39016 * passes.def (pass_target_clone): Put as very first IPA early
39019 2017-02-03 Martin Liska <mliska@suse.cz>
39021 * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39022 in case of a function with ifunc attribute.
39024 2017-02-03 Martin Liska <mliska@suse.cz>
39026 * cgraph.c (cgraph_node::dump): Dump function version info.
39027 * symtab.c (symtab_node::dump_base): Add missing new line.
39029 2017-02-02 Jan Hubicka <hubicka@ucw.cz>
39031 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39032 (ifcombine_ifandif): Use it.
39034 2017-02-03 Martin Liska <mliska@suse.cz>
39036 * doc/invoke.texi: Document default value for
39037 use-after-scope-direct-emission-threshold.
39039 2017-02-03 Martin Liska <mliska@suse.cz>
39041 PR tree-optimization/79339
39042 * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39043 (format_floating): Likewise.
39045 2017-02-03 Martin Liska <mliska@suse.cz>
39048 * ipa-prop.c (ipa_node_params_t::insert): Remove current
39050 (ipa_node_params_t::remove): Likewise.
39051 * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39052 initialization from removed ipa_node_params_t::insert.
39053 (ipa_node_params::~ipa_node_params): Move from removed
39054 ipa_node_params_t::release.
39055 * symbol-summary.h (symbol_summary::m_released): New member.
39056 Do not release a summary twice. Do not allow to call finalizer
39057 for types of a summary that live in GGC memory.
39059 2017-02-02 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
39061 * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39064 2017-02-02 Martin Sebor <msebor@redhat.com>
39066 PR middle-end/79275
39067 * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39068 (format_string): Tighten up the range of output for non-constant
39069 strings and correct the expected range for wide non-constant strings.
39071 2017-02-02 Martin Sebor <msebor@redhat.com>
39073 * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39075 PR middle-end/32003
39076 * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39078 (-fdump-tree-@var): Add to index and document how to come up
39079 with pass-specific option and dump file names.
39080 (-fdump-passes): Clarify where to look for output.
39082 2017-02-02 Jan Hubicka <hubicka@ucw.cz>
39084 PR middle-end/77445
39085 * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39086 statistics of the analyzed path; allow threading for speed when
39087 any of BBs along the path are optimized for speed.
39089 2017-02-02 Eric Botcazou <ebotcazou@adacore.com>
39091 PR middle-end/78468
39092 * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39093 settings of the virtual registers.
39096 2016-08-23 Dominik Vogt <vogt@linux.vnet.ibm.com>
39098 * explow.c (get_dynamic_stack_size): Take known alignment of stack
39099 pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39102 2017-02-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
39104 * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39105 ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39107 2017-02-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
39109 * config/s390/s390.md: Add missing comments with the expanded
39111 * config/s390/vector.md: Likewise.
39112 * config/s390/vx-builtins.md: Likewise.
39114 2017-02-02 Jakub Jelinek <jakub@redhat.com>
39117 * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39118 (fixuns_trunc<mode>di2): ... this, remove previous expander. Put all
39119 conditions on a single line.
39121 2017-02-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
39123 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39124 __S390_VX__ to __VX__.
39126 2017-02-01 Andrew Pinski <apinski@cavium.com>
39128 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39129 stmt_info to record_stmt_cost.
39130 (vect_get_known_peeling_cost): Pass stmt_info if known to
39132 * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39133 cpu_vector_cost field into
39134 scalar_int_stmt_cost and scalar_fp_stmt_cost. Split vec_stmt_cost
39135 field into vec_int_stmt_cost and vec_fp_stmt_cost.
39136 * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39137 splitting of scalar_stmt_cost and vec_stmt_cost.
39138 (thunderx_vector_cost): Likewise.
39139 (cortexa57_vector_cost): LIkewise.
39140 (exynosm1_vector_cost): Likewise.
39141 (xgene1_vector_cost): Likewise.
39142 (thunderx2t99_vector_cost): Improve after the splitting of the two
39144 (aarch64_builtin_vectorization_cost): Update for the splitting of
39145 scalar_stmt_cost and vec_stmt_cost.
39147 2017-02-01 Torvald Riegel <triegel@redhat.com>
39148 Richard Henderson <rth@redhat.com>
39150 * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39151 conditional on existance of a fast atomic load.
39152 * optabs-query.c (can_atomic_load_p): New function.
39153 * optabs-query.h (can_atomic_load_p): Declare it.
39154 * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39155 no fast atomic load is available for the particular size of access.
39156 (expand_atomic_compare_and_swap): Likewise.
39157 (expand_atomic_load): Likewise.
39158 (expand_atomic_store): Likewise.
39159 (expand_atomic_fetch_op): Likewise.
39160 * testsuite/lib/target-supports.exp
39161 (check_effective_target_sync_int_128): Remove x86 because it provides
39162 no fast atomic load.
39163 (check_effective_target_sync_int_128_runtime): Likewise.
39165 2017-02-01 Richard Biener <rguenther@suse.de>
39167 * graphite.c: Include tree-vectorizer.h for find_loop_location.
39168 (graphite_transform_loops): Provide opt-info for optimized nests.
39169 * tree-parloop.c (parallelize_loops): Provide opt-info for
39170 parallelized loops.
39172 2017-02-01 Richard Biener <rguenther@suse.de>
39174 PR middle-end/79315
39175 * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39176 was not set before.
39178 2017-02-01 Richard Biener <rguenther@suse.de>
39180 PR tree-optimization/71824
39181 * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39182 Verify the loops are valid in the merged SESE region.
39183 (scop_detection::can_represent_loop_1): Check analyzing the
39184 evolution of the number of iterations in the region succeeds.
39186 2017-01-31 Ian Lance Taylor <iant@golang.org>
39188 * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39189 REG_ARGS_SIZE note to 32-bit push insns and call insn.
39191 2017-01-31 David Malcolm <dmalcolm@redhat.com>
39193 PR preprocessor/79210
39194 * input.c (get_substring_ranges_for_loc): Replace line_width
39195 assertion with error-handling.
39197 2017-01-31 Richard Biener <rguenther@suse.de>
39199 PR tree-optimization/77318
39200 * graphite-sese-to-poly.c (extract_affine): Fix assert.
39201 (create_pw_aff_from_tree): Take loop parameter.
39202 (add_condition_to_pbb): Pass loop of the condition to
39203 create_pw_aff_from_tree.
39205 2017-01-31 Jakub Jelinek <jakub@redhat.com>
39207 * config/s390/s390.c (s390_asan_shadow_offset): New function.
39208 (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39210 2017-01-31 Michael Meissner <meissner@linux.vnet.ibm.com>
39214 * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39216 (convert_int_to_float128): Likewise.
39217 * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39218 (convert_int_to_float128): Likewise.
39219 * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39220 (UNSPEC_IEEE128_CONVERT): Likewise.
39221 (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39222 rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39223 Use local variables for IBM extended format.
39224 (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39225 (fix_trunc<mode>si2_fprs): Likewise.
39226 (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39227 (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39228 (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39229 to know that we can now have integers of all sizes in vector
39231 (fix<uns>_<mode>di2_hw): Likewise.
39232 (float<uns>_<mode>si2_hw): Likewise.
39233 (fix_<mode>si2_hw): Likewise.
39234 (fixuns_<mode>si2_hw): Likewise.
39235 (float<uns>_<mode>di2_hw): Likewise.
39236 (float_<mode>di2_hw): Likewise.
39237 (float_<mode>si2_hw): Likewise.
39238 (floatuns_<mode>di2_hw): Likewise.
39239 (floatuns_<mode>si2_hw): Likewise.
39240 (xscvqp<su>wz_<mode>): Delete, no longer used.
39241 (xscvqp<su>dz_<mode>): Likewise.
39242 (xscv<su>dqp_<mode>): Likewise.
39243 (ieee128_mfvsrd_64bit): Likewise.
39244 (ieee128_mfvsrd_32bit): Likewise.
39245 (ieee128_mfvsrwz): Likewise.
39246 (ieee128_mtvsrw): Likewise.
39247 (ieee128_mtvsrd_64bit): Likewise.
39248 (ieee128_mtvsrd_32bit): Likewise.
39250 2017-01-31 Martin Liska <mliska@suse.cz>
39253 * ipa-prop.c (ipa_free_all_node_params): Call release method
39254 instead of ~sumbol_summary to not to trigger double times
39257 2017-01-31 Aldy Hernandez <aldyh@redhat.com>
39259 PR tree-optimization/71691
39260 * bitmap.h (class auto_bitmap): New.
39261 * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39262 is_maybe_undefined instead of ssa_undefined_value_p.
39264 2017-01-31 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
39266 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39267 __S390_ARCH_LEVEL__ to __ARCH__.
39269 2017-01-31 Jakub Jelinek <jakub@redhat.com>
39271 PR tree-optimization/79267
39272 * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39273 if should_remove_lhs_p is true.
39275 2017-01-30 Alexandre Oliva <aoliva@redhat.com>
39278 * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39279 (add_alignment_attribute): New.
39280 (base_type_die): Add alignment attribute.
39281 (subrange_type_die): Likewise.
39282 (modified_type_die): Likewise.
39283 (gen_array_type_die): Likewise.
39284 (gen_descr_array_type_die: Likewise.
39285 (gen_enumeration_type_die): Likewise.
39286 (gen_subprogram_die): Likewise.
39287 (gen_variable_die): Likewise.
39288 (gen_field_die): Likewise.
39289 (gen_ptr_to_mbr_type_die): Likewise.
39290 (gen_struct_or_union_type_die): Likewise.
39291 (gen_subroutine_type_die): Likewise.
39292 (gen_typedef_die): Likewise.
39293 (base_type_cmp): Compare alignment attribute.
39295 2017-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
39298 * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39299 (setb_unsigned) New pattern for setb with CCUNS.
39300 * config/rs6000/rs6000.c (expand_block_compare): Use a different
39301 subfc./subfe sequence to avoid overflow problems. Generate a
39302 shorter sequence with cmpld/setb for power9.
39303 * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39304 for generating subfc. instruction.
39305 (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39306 now uses this instruction.
39308 2017-01-30 Ian Lance Taylor <iant@google.com>
39311 * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39312 for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39314 2017-01-30 Martin Sebor <msebor@redhat.com>
39316 * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39317 Move constant to the right of a relational operator.
39318 (get_mpfr_format_length, format_character, format_string): Ditto.
39319 (should_warn_p, maybe_warn): Same.
39321 * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39323 2017-01-30 Maxim Ostapenko <m.ostapenko@samsung.com>
39326 * asan.c (get_translation_unit_decl): Remove function.
39327 (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39329 2017-01-30 Martin Liska <mliska@suse.cz>
39331 PR gcov-profile/79259
39332 * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39333 -fprofile-generate.
39335 2017-01-30 Martin Liska <mliska@suse.cz>
39338 * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39339 Initialize variables with NULL value.
39341 2017-01-30 Richard Earnshaw <rearnsh@arm.com>
39344 * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39346 * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39348 2017-01-30 Richard Biener <rguenther@suse.de>
39350 PR tree-optimization/79276
39351 * tree-vrp.c (process_assert_insertions): Properly adjust common
39352 when removing a duplicate.
39354 2017-01-30 Richard Biener <rguenther@suse.de>
39356 PR tree-optimization/79256
39357 * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39358 BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39360 * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39362 2017-01-30 Dominik Vogt <vogt@linux.vnet.ibm.com>
39365 * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39366 ("*r<noxa>sbg_<mode>_sll_bitmask")
39367 ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39368 ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39369 Use contiguous_bitmask_nowrap_operand.
39371 2017-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
39374 * config/rs6000/altivec.h (vec_xl): Revise #define.
39375 (vec_xst): Likewise.
39377 2017-01-27 Uros Bizjak <ubizjak@gmail.com>
39379 * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39381 2017-01-27 Bernd Schmidt <bschmidt@redhat.com>
39383 PR rtl-optimization/79194
39384 * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39385 traps before call to bypass_conditional_jumps.
39387 2017-01-27 Vladimir Makarov <vmakarov@redhat.com>
39389 PR tree-optimization/71374
39390 * lra-constraints.c (check_conflict_input_operands): New.
39391 (match_reload): Use it.
39393 2017-01-27 Vladimir Makarov <vmakarov@redhat.com>
39396 * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39397 account to calculate conflict_set.
39399 2017-01-27 Bin Cheng <bin.cheng@arm.com>
39401 PR rtl-optimization/78559
39402 * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39403 other_insn in combine.
39405 2017-01-27 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
39407 * builtin-types.def: Use unsigned_char_type_node for BT_UINT8. Use
39408 uint16_type_node for BT_UINT16.
39410 2017-01-27 David Malcolm <dmalcolm@redhat.com>
39412 * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39413 "RTL Tests" to menu.
39414 (GIMPLE Tests): New node.
39415 (RTL Tests): New node.
39417 2017-01-27 Richard Biener <rguenther@suse.de>
39419 PR tree-optimization/79245
39420 * tree-loop-distribution.c (distribute_loop): Apply cost
39421 modeling also to detected patterns.
39423 2017-01-27 Richard Biener <rguenther@suse.de>
39425 PR tree-optimization/71433
39426 * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39427 (compare_assert_loc): New function.
39428 (process_assert_insertions): Sort and optimize assert locations
39429 to remove duplicates and push down identical assertions on
39430 edges to their destination block.
39432 2017-01-27 Richard Biener <rguenther@suse.de>
39434 PR tree-optimization/79244
39435 * tree-vrp.c (remove_range_assertions): Forcefully propagate
39436 out SSA names even if abnormal.
39438 2017-01-27 Jakub Jelinek <jakub@redhat.com>
39440 * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39441 * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39442 instead of MPFR_RNDN.
39444 2017-01-27 Richard Earnshaw <rearnsha@arm.com>
39447 * arm.c (arm_option_override): Don't call build_target_option_node
39448 until after doing all option overrides.
39449 (arm_valid_target_attribute_tree): Likewise.
39451 2017-01-27 Martin Liska <mliska@suse.cz>
39453 * doc/invoke.texi (-fprofile-arcs): Document profiling support
39454 for {cd}tors and C++ {cd}tors.
39456 2017-01-27 Dominik Vogt <vogt@linux.vnet.ibm.com>
39458 * config/s390/s390.md ("*setmem_long_and")
39459 ("*setmem_long_and_31z"): Use zero_extend instead of and.
39461 2017-01-26 Martin Sebor <msebor@redhat.com>
39463 * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39466 2017-01-26 Martin Sebor <msebor@redhat.com>
39468 * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39469 HAVE_DFmode before using XFmode or DFmode.
39470 (parse_directive): Avoid using the z length modifier to avoid
39471 the ISO C++98 does not support the ‘z’ gnu_printf length modifier.
39473 PR middle-end/78703
39474 * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39475 to accept adjustment as an array.
39476 (get_int_range): New function.
39477 (struct directive): Make width and prec arrays.
39478 (directive::set_width, directive::set_precision): Call get_int_range.
39479 (format_integer, format_floating): Handle width and precision ranges.
39480 (format_string, parse_directive): Same.
39482 2017-01-26 Jakub Jelinek <jakub@redhat.com>
39485 * dwarf2out.c (generate_skeleton_bottom_up): For children with
39486 comdat_type_p set, just clone them, but keep the children in the
39490 * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39491 which have direct callers with -fvar-tracking-assignments enabled
39493 (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39494 inside of type units.
39496 2017-01-26 Martin Sebor <msebor@redhat.com>
39498 PR middle-end/78703
39499 * gimple-ssa-sprintf.c (struct result_range): Add likely and
39501 (struct format_result): Replace number_chars, number_chars_min,
39502 and number_chars_max with a single member of struct result_range.
39504 (format_result::operator+=): Adjust.
39505 (struct fmtresult): Remove bounded. Handle likely and unlikely
39507 (fmtresult::adjust_for_width_or_precision): New function.
39508 (fmtresult:type_max_digits): New function.
39509 (bytes_remaining): Handle likely and unlikely counters.
39510 (min_bytes_remaining): Remove.
39511 (format_percent): Simplify.
39512 (format_integer, format_floating): Set likely and unlikely counters.
39513 (get_string_length, format_character, format_string): Same.
39514 (format_plain, should_warn_p): New function.
39515 (maybe_warn): Call should_warn_p. Update diagnostic messages
39516 and handle those for all directives, including plain strings.
39517 (format_directive): Handle likely and unlikely counters.
39518 Remove unnecessary quoting from diagnostics. Add an informational
39520 (add_bytes): Remove.
39521 (pass_sprintf_length::compute_format_length): Simplify.
39522 (try_substitute_return_value): Handle likely and unlikely counters.
39524 2017-01-26 Carl Love <cel@us.ibm.com>
39526 * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39527 bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39529 2017-01-26 Vladimir Makarov <vmakarov@redhat.com>
39532 * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39533 endianess for subregs into account.
39534 * lra-constraints.c (lra_constraints): Do risky transformations
39535 always on the first iteration.
39536 * lra-lives.c (check_pseudos_live_through_calls): Add arg
39537 last_call_used_reg_set.
39538 (process_bb_lives): Define and use last_call_used_reg_set.
39539 * lra.c (lra): Always continue after lra_constraints on the first
39542 2017-01-26 Kirill Yukhin <kirill.yukhin@gmail.com>
39544 * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39546 * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39548 2017-01-26 Jakub Jelinek <jakub@redhat.com>
39550 * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39551 _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39552 * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39553 _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39554 * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39555 __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39556 OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39557 * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39558 (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39560 2017-01-26 Marek Polacek <polacek@redhat.com>
39563 * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39564 for the third operand.
39566 2017-01-26 Jakub Jelinek <jakub@redhat.com>
39568 PR middle-end/79236
39569 * omp-low.c (struct omp_context): Add simt_stmt field.
39570 (scan_omp_for): Return omp_context *.
39571 (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39572 context to the _simt_ SIMD stmt.
39573 (lower_omp_for): For combined SIMD with sibling _simt_
39574 SIMD, make sure to use the same decls in _looptemp_
39575 clauses as in the sibling.
39577 2017-01-26 David Sherwood <david.sherwood@arm.com>
39579 PR middle-end/79212
39580 * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39583 2017-01-26 Jakub Jelinek <jakub@redhat.com>
39586 * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39587 emit fld b; fld a; if possible.
39589 * brig-builtins.def: Update copyright years.
39590 * config/arm/arm_acle_builtins.def: Update copyright years.
39592 2017-01-25 Michael Meissner <meissner@linux.vnet.ibm.com>
39595 * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39596 constraint instead of o for the stxsd instruction.
39598 2017-01-25 Carl Love <cel@us.ibm.com>
39600 * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39601 of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39603 2017-01-25 Jonathan Wakely <jwakely@redhat.com>
39605 * doc/invoke.texi (C++ Dialect Options): Fix typo.
39607 2017-01-25 Richard Biener <rguenther@suse.de>
39609 PR tree-optimization/69264
39610 * target.def (vector_alignment_reachable): Improve documentation.
39611 * doc/tm.texi: Regenerate.
39612 * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39614 * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39615 earlier changes with respect to TYPE_USER_ALIGN.
39616 (vector_alignment_reachable_p): Likewise. Improve dumping.
39618 2016-01-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
39621 * config/arm/arm.md (xordi3): Force constant operand into a register
39624 2016-01-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
39626 * doc/invoke.texi (-fstore-merging): Correct default optimization
39627 levels at which it is enabled.
39628 (-O): Move -fstore-merging from list to...
39631 2017-01-25 Richard Biener <rguenther@suse.de>
39634 * omp-expand.c: Include debug.h.
39635 (expand_omp_taskreg): Make sure to generate early debug before
39636 outlining anything from a function.
39637 (expand_omp_target): Likewise.
39638 (grid_expand_target_grid_body): Likewise.
39640 2017-01-25 Maxim Ostapenko <m.ostapenko@samsung.com>
39643 * asan.c (get_translation_unit_decl): New function.
39644 (asan_add_global): Extract modules file name from globals
39645 TRANSLATION_UNIT_DECL name.
39647 2017-01-24 Eric Botcazou <ebotcazou@adacore.com>
39650 * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39651 for long calls with APCS frame and VFP.
39653 2017-01-24 David Malcolm <dmalcolm@redhat.com>
39655 * cfg.c (original_copy_tables_initialized_p): New function.
39656 * cfg.h (original_copy_tables_initialized_p): New decl.
39657 * cfgrtl.c (relink_block_chain): Guard the call to
39658 free_original_copy_tables with a call to
39659 original_copy_tables_initialized_p.
39660 * cgraph.h (symtab_node::native_rtl_p): New decl.
39661 * cgraphunit.c (symtab_node::native_rtl_p): New function.
39662 (symtab_node::needed_p): Don't assert for early assembly output
39663 for __RTL functions.
39664 (cgraph_node::finalize_function): Set "force_output" for __RTL
39666 (cgraph_node::analyze): Bail out early for __RTL functions.
39667 (analyze_functions): Update assertion to support __RTL functions.
39668 (cgraph_node::expand): Bail out early for __RTL functions.
39669 * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39671 * function.h (struct function): Update comment for field
39673 * gimple-expr.c: Include "tree-pass.h".
39674 (gimple_has_body_p): Return false for __RTL functions.
39675 * Makefile.in (OBJS): Add run-rtl-passes.o.
39676 * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39678 (gcc::pass_manager::get_clean_slate): New accessor.
39679 * passes.c: Include "insn-addr.h".
39680 (should_skip_pass_p): Add logging. Update logic for running
39681 "expand" to be compatible with both __GIMPLE and __RTL. Guard
39682 property-provider override so it is only done for gimple passes.
39684 (skip_pass): New function.
39685 (execute_one_pass): Call skip_pass when skipping passes.
39686 * read-md.c (md_reader::read_char): Support filtering
39687 the input to a subset of line numbers.
39688 (md_reader::md_reader): Initialize fields
39689 m_first_line and m_last_line.
39690 (md_reader::read_file_fragment): New function.
39691 * read-md.h (md_reader::read_file_fragment): New decl.
39692 (md_reader::m_first_line): New field.
39693 (md_reader::m_last_line): New field.
39694 * read-rtl-function.c (function_reader::create_function): Only
39695 create cfun if it doesn't already exist. Set PROP_rtl on cfun's
39696 curr_properties. Set DECL_INITIAL to a dummy block.
39697 (read_rtl_function_body_from_file_range): New function.
39698 * read-rtl-function.h (read_rtl_function_body_from_file_range):
39700 * run-rtl-passes.c: New file.
39701 * run-rtl-passes.h: New file.
39703 2017-01-24 Jeff Law <law@redhat.com>
39705 * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39708 2017-01-24 Bin Cheng <bin.cheng@arm.com>
39710 PR tree-optimization/79159
39711 * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39712 (record_nonwrapping_iv): Improve boundary using above function if no
39713 value range information.
39715 2017-01-24 Pekka Jääskeläinen <pekka@parmance.com>
39716 Martin Jambor <mjambor@suse.cz>
39718 * brig-builtins.def: New file.
39719 * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39720 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39721 (DEF_HSAIL_SAT_BUILTIN): Likewise.
39722 (DEF_HSAIL_INTR_BUILTIN): Likewise.
39723 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39724 * builtin-types.def (BT_INT8): New.
39725 (BT_INT16): Likewise.
39726 (BT_UINT8): Likewise.
39727 (BT_UINT16): Likewise.
39728 (BT_FN_ULONG): Likewise.
39729 (BT_FN_UINT_INT): Likewise.
39730 (BT_FN_UINT_ULONG): Likewise.
39731 (BT_FN_UINT_LONG): Likewise.
39732 (BT_FN_UINT_PTR): Likewise.
39733 (BT_FN_ULONG_PTR): Likewise.
39734 (BT_FN_INT8_FLOAT): Likewise.
39735 (BT_FN_INT16_FLOAT): Likewise.
39736 (BT_FN_UINT32_FLOAT): Likewise.
39737 (BT_FN_UINT16_FLOAT): Likewise.
39738 (BT_FN_UINT8_FLOAT): Likewise.
39739 (BT_FN_UINT64_FLOAT): Likewise.
39740 (BT_FN_UINT16_UINT32): Likewise.
39741 (BT_FN_UINT32_UINT16): Likewise.
39742 (BT_FN_UINT16_UINT16_UINT16): Likewise.
39743 (BT_FN_INT_PTR_INT): Likewise.
39744 (BT_FN_UINT_PTR_UINT): Likewise.
39745 (BT_FN_LONG_PTR_LONG): Likewise.
39746 (BT_FN_ULONG_PTR_ULONG): Likewise.
39747 (BT_FN_VOID_UINT64_UINT64): Likewise.
39748 (BT_FN_UINT8_UINT8_UINT8): Likewise.
39749 (BT_FN_INT8_INT8_INT8): Likewise.
39750 (BT_FN_INT16_INT16_INT16): Likewise.
39751 (BT_FN_INT_INT_INT): Likewise.
39752 (BT_FN_UINT_FLOAT_UINT): Likewise.
39753 (BT_FN_FLOAT_UINT_UINT): Likewise.
39754 (BT_FN_ULONG_UINT_UINT): Likewise.
39755 (BT_FN_ULONG_UINT_PTR): Likewise.
39756 (BT_FN_ULONG_ULONG_ULONG): Likewise.
39757 (BT_FN_UINT_UINT_UINT): Likewise.
39758 (BT_FN_VOID_UINT_PTR): Likewise.
39759 (BT_FN_UINT_UINT_PTR: Likewise.
39760 (BT_FN_UINT32_UINT64_PTR): Likewise.
39761 (BT_FN_INT_INT_UINT_UINT): Likewise.
39762 (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39763 (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39764 (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39765 (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39766 (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39767 (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39768 (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39769 (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39770 (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39771 (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39772 (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39773 * doc/frontends.texi: List BRIG FE.
39774 * doc/install.texi (Testing): Add BRIG tesring requirements.
39775 * doc/invoke.texi (Overall Options): Mention BRIG.
39776 * doc/standards.texi (Standards): Doucment BRIG HSA version.
39778 2017-01-24 Richard Biener <rguenther@suse.de>
39780 PR translation/79208
39781 * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39783 2017-01-24 Martin Jambor <mjambor@suse.cz>
39786 * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39787 * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39788 and known_contexts.
39790 2017-01-24 Aldy Hernandez <aldyh@redhat.com>
39792 PR middle-end/79123
39793 * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39794 casts from signed to unsigned really don't have a range.
39796 2017-01-24 Markus Trippelsdorf <markus@trippelsdorf.de>
39798 * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39799 GMP_RNDx for compatiblity.
39801 2017-01-24 Martin Liska <mliska@suse.cz>
39804 * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39805 that would prevent us to call alloca with -1 as argument.
39807 2017-01-24 Jakub Jelinek <jakub@redhat.com>
39809 * dwarf2out.c (output_compilation_unit_header, output_file_names):
39810 Avoid -Wformat-security warning.
39812 2017-01-23 Andrew Pinski <apinski@cavium.com>
39814 * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39817 2017-01-23 Martin Sebor <msebor@redhat.com>
39819 PR middle-end/78703
39820 * gimple-ssa-sprintf.c (warn_level): New global.
39821 (format_integer): Use it here and throughout the rest of the file.
39822 Use the same switch to compute sign as base.
39823 (maybe_warn): New function.
39824 (format_directive): Factor out warnings into maybe_warn.
39825 Add debugging output. Use warn_level.
39826 (add_bytes): Use warn_level.
39827 (pass_sprintf_length::compute_format_length): Add debugging output.
39828 (try_substitute_return_value): Same.
39829 (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39831 PR middle-end/78703
39832 * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39833 (struct fmtresult, format_integer, format_floating): Adjust.
39834 (fmtresult::fmtresult): Set max correctly in two argument ctor.
39835 (get_string_length, format_string,format_directive): Same.
39836 (pass_sprintf_length::compute_format_length): Same.
39837 (try_substitute_return_value): Simplify slightly.
39839 PR middle-end/78703
39840 * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39841 (fmtresult::operator+=): Outlined.
39842 (struct fmtresult): Add ctors.
39843 (struct conversion_spec): Rename...
39844 (struct directive): ...to this. Add and remove data members.
39845 (directive::set_width, directive::set_precision): New functions.
39846 (format_percent): Use fmtresult ctor.
39847 (get_width_and_precision): Remove.
39848 (format_integer): Make naming changes. Avoid computing width and
39850 (format_floating): Same. Adjust indentation.
39851 (format_character, format_none): New functions.
39852 (format_string): Moved character handling to format_character.
39853 (format_directive): Remove arguments, change return type.
39854 (parse_directive): New function.
39855 (pass_sprintf_length::compute_format_length): Move directive
39856 parsing to parse_directive.
39858 2017-01-23 Jakub Jelinek <jakub@redhat.com>
39860 * tree.h (assign_assembler_name_if_neeeded): Rename to ...
39861 (assign_assembler_name_if_needed): ... this.
39862 * tree.c (assign_assembler_name_if_neeeded): Rename to ...
39863 (assign_assembler_name_if_needed): ... this.
39864 (free_lang_data_in_cgraph): Adjust callers.
39865 * cgraphunit.c (cgraph_node::analyze): Likewise.
39866 * omp-expand.c (expand_omp_taskreg, expand_omp_target):
39869 2017-01-23 Richard Biener <rguenther@suse.de>
39871 PR tree-optimization/79088
39872 PR tree-optimization/79188
39873 * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
39874 resetting loop bounds after last path deletion. Reset loop
39875 bounds of the target loop, make code match the comments.
39876 * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
39877 Make sure loops need no fixups.
39879 2017-01-23 Kelvin Nilsen <kelvin@gcc.gnu.org>
39881 * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
39882 exponent support with double type for first argument.
39883 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
39884 type returned by __builtin_vec_extract_sig,
39885 __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
39886 functions from "vector int" to "vector unsigned int" or from
39887 "vector long long int" to "vector unsigned long long int".
39888 Changed type returned by __builtin_vec_extract_exp,
39889 __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
39890 functions from "vector int" to "vector unsigned int" or from
39891 "vector long long int" to "vector unsigned long long int".
39892 Changed return type of __builtin_vec_test_data_class,
39893 __builtin_vec_test_data_class_sp, and
39894 __builtin_vec_test_data_class_dp from "vector int" to
39895 "vector bool int" or from "vector long long int" to "vector bool
39896 long long int" and changed second argument type from "unsigned
39897 int" to "int". Added new overloaded function forms "vector float
39898 __builtin_vec_insert_exp (vector float, vector unsigned int)" and
39899 "vector float __builtin_vec_insert_exp_sp (vector float, vector
39900 unsigned int)" and "vector double __builtin_vec_insert_exp (vector
39901 double, vector unsigned long long int)" and "vector double
39902 __builtin_vec_insert_exp_dp (vector double, vector unsigned long
39903 long int)". Changed return type of
39904 __builtin_scalar_test_data_class and
39905 __builtin_scalar_test_data_class_sp and
39906 __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
39907 int" and changed second argument from "unsigned int" to "int".
39908 Changed type returned by __builtin_scalar_test_neg,
39909 __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
39910 from "int" to "bool int". Added new overloaded function form
39911 "double __builtin_scalar_insert_exp (double, unsigned long long int)".
39912 * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
39913 exponent double-precision with floating point first argument.
39914 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
39915 documentation of scalar_test_data_class, scalar_test_neg,
39916 scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
39917 vector_extract_exp, vec_extract_sig, vec_insert_exp, and
39918 vec_test_data_class built-in functions to reflect refinements in
39919 their type signatures.
39921 2017-01-23 Andreas Tobler <andreast@gcc.gnu.org>
39923 * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
39925 (aarch64_elf_asm_destructor): Likewise.
39927 2017-01-23 Bernd Schmidt <bschmidt@redhat.com>
39929 PR rtl-optimization/78634
39930 * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
39931 (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
39932 * ifcvt.c (noce_try_cmove): Add missing cost check.
39934 PR rtl-optimization/71724
39935 * combine.c (if_then_else_cond): Look for situations where it is
39936 beneficial to undo the work of one of the recursive calls.
39938 2017-01-23 Bin Cheng <bin.cheng@arm.com>
39940 PR tree-optimization/70754
39941 * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
39942 (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE. Insert
39943 combined stmt before it if not NULL.
39944 (combine_chains): Process refs reversely and compute dominance point
39947 2017-01-23 Martin Liska <mliska@suse.cz>
39949 PR tree-optimization/79196
39950 * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
39951 (fold_strstr_to_strncmp): ... this. Fold the pattern to strncmp
39953 (strlen_optimize_stmt): Call the renamed function.
39955 2017-01-23 Michael Matz <matz@suse.de>
39957 PR tree-optimization/78384
39958 * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
39960 2017-01-23 Richard Biener <rguenther@suse.de>
39962 PR tree-optimization/79186
39963 * tree-vrp.c (register_new_assert_for): Make sure we've seen
39964 both incoming edges before moving an assert.
39966 2017-01-23 Martin Jambor <mjambor@suse.cz>
39968 * ipa-prop.c (load_from_param_1): Removed.
39969 (load_from_unmodified_param): Bits from load_from_param_1 put back
39971 (load_from_param): Removed.
39972 (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
39973 with stmt. Reverted back to use of load_from_unmodified_param.
39975 2017-01-23 Martin Jambor <mjambor@suse.cz>
39978 * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
39979 (ipa_node_params): Annotate with GTY((for_user)). Make descriptors
39980 field a pointer to garbage collected vector, mark lattices and
39981 ipcp_orig_node with GTY((skip)).
39982 (ipa_get_param_count): Adjust to descriptors being a pointer.
39983 (ipa_get_param): Likewise.
39984 (ipa_get_type): Likewise.
39985 (ipa_get_param_move_cost): Likewise.
39986 (ipa_set_param_used): Likewise.
39987 (ipa_get_controlled_uses): Likewise.
39988 (ipa_set_controlled_uses): Likewise.
39989 (ipa_is_param_used): Likewise.
39990 (ipa_node_params_t): Move into garbage collector. New methods insert
39992 (ipa_node_params_sum): Annotate wth GTY(()).
39993 (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
39995 (ipa_load_from_parm_agg): Adjust declaration.
39996 * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
39997 * ipa-profile.c (ipa_profile): Likewise.
39998 * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
39999 (ipa_populate_param_decls): Make descriptors parameter garbage
40001 (ipa_dump_param): Adjust to descriptors being a pointer.
40002 (ipa_alloc_node_params): Likewise.
40003 (ipa_initialize_node_params): Likewise.
40004 (load_from_param_1): Make descriptors parameter garbage collected.
40005 (load_from_unmodified_param): Likewise.
40006 (load_from_param): Likewise.
40007 (ipa_load_from_parm_agg): Likewise.
40008 (ipa_node_params::~ipa_node_params): Removed.
40009 (ipa_free_all_node_params): Remove call to delete operator.
40010 (ipa_node_params_t::insert): New.
40011 (ipa_node_params_t::remove): Likewise.
40012 (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40013 copy known_csts and known_contexts vectors.
40014 (ipa_read_node_info): Adjust to descriptors being a pointer.
40015 (ipcp_modif_dom_walker): Make m_descriptors field garbage
40017 (ipcp_transform_function): Make descriptors variable garbage
40020 2017-01-23 Andrew Senkevich <andrew.senkevich@intel.com>
40022 * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40023 * config/i386/avx512dqintrin.h: Ditto.
40024 * config/i386/avx512fintrin.h: Ditto.
40025 * config/i386/i386.c: Handle new builtins.
40026 * config/i386/i386-builtin.def: Add new builtins.
40027 * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40028 (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40030 2017-01-23 Jakub Jelinek <jakub@redhat.com>
40031 Martin Liska <mliska@suse.cz>
40033 * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40034 * asan.c (asan_expand_poison_ifn): Support stores and use
40035 appropriate ASAN report function.
40036 * internal-fn.c (expand_ASAN_POISON_USE): New function.
40037 * internal-fn.def (ASAN_POISON_USE): Declare.
40038 * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40039 (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40040 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40041 ASAN_POISON calls w/o LHS.
40042 * tree-ssa.c (execute_update_addresses_taken): Create clobber
40043 for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40044 from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40045 * gimplify.c (asan_poison_variables): Add attribute
40046 use_after_scope_memory to variables that really needs to live
40048 * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40049 having the attribute.
40051 2017-01-23 Martin Liska <mliska@suse.cz>
40053 * asan.c (create_asan_shadow_var): New function.
40054 (asan_expand_poison_ifn): Likewise.
40055 * asan.h (asan_expand_poison_ifn): New declaration.
40056 * internal-fn.c (expand_ASAN_POISON): Likewise.
40057 * internal-fn.def (ASAN_POISON): New builtin.
40058 * sanopt.c (pass_sanopt::execute): Expand
40059 asan_expand_poison_ifn.
40060 * tree-inline.c (copy_decl_for_dup_finish): Make function
40062 * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40063 * tree-ssa.c (is_asan_mark_p): New function.
40064 (execute_update_addresses_taken): Rewrite local variables
40065 (identified just by use-after-scope as addressable) into SSA.
40067 2017-01-22 Gerald Pfeifer <gerald@pfeifer.com>
40069 * doc/install.texi (Specific): opensource.apple.com uses https
40070 now. Remove trailing slash.
40072 2017-01-22 Gerald Pfeifer <gerald@pfeifer.com>
40074 * README.Portability: Remove note on an Irix compatibility issue.
40076 2017-01-22 Dimitry Andric <dim@FreeBSD.org>
40078 * gcov.c (INCLUDE_ALGORITHM): Define.
40079 (INCLUDE_VECTOR): Define.
40080 No longer include <vector> and <algorithm> directly.
40082 2017-01-21 Gerald Pfeifer <gerald@pfeifer.com>
40084 * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40086 * doc/invoke.texi (Code Gen Options): Ditto.
40088 2017-01-21 Jan Hubicka <hubicka@ucw.cz>
40091 * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40093 2017-01-21 Bernd Schmidt <bschmidt@redhat.com>
40095 rtl-optimization/79125
40096 * cprop.c (local_cprop_pass): Handle cases where we make an
40097 unconditional trap.
40099 2017-01-20 Segher Boessenkool <segher@kernel.crashing.org>
40103 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40104 read from, for big endian.
40106 2017-01-20 Jiong Wang <jiong.wang@arm.com>
40108 * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40109 register pauth builtins for LP64 only.
40111 2017-01-20 Marek Polacek <polacek@redhat.com>
40114 * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40117 2017-01-20 Alexander Monakov <amonakov@ispras.ru>
40119 * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40121 * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40122 * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40123 * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40125 2017-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
40128 * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40129 in big-endian mode when they are not a single duplicated value.
40131 2017-01-20 Richard Biener <rguenther@suse.de>
40133 * BASE-VER: Bump to 7.0.1.
40135 2017-01-20 Alexander Monakov <amonakov@ispras.ru>
40137 * omp-low.c (omplow_simd_context): New struct. Use it...
40138 (lower_rec_simd_input_clauses): ...here and...
40139 (lower_rec_input_clauses): ...here to hold common data. Adjust all
40140 references to idx, lane, max_vf, is_simt.
40142 2017-01-20 Graham Markall <graham.markall@embecosm.com>
40144 * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40147 2017-01-20 Martin Jambor <mjambor@suse.cz>
40149 * hsa.h: Renaed to hsa-common.h. Adjusted a comment.
40150 * hsa.c: Renaed to hsa-common.c. Change include of gt-hsa.h to
40152 * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40153 (GTFILES): Rename hsa.c to hsa-common.c.
40154 * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40155 * hsa-dump.c: Likewise.
40156 * hsa-gen.c: Likewise.
40157 * hsa-regalloc.c: Likewise.
40158 * ipa-hsa.c: Likewise.
40159 * omp-expand.c: Likewise.
40160 * omp-low.c: Likewise.
40161 * toplev.c: Likewise.
40163 2017-01-20 Marek Polacek <polacek@redhat.com>
40166 * doc/invoke.texi: Document -Wduplicated-branches.
40167 * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40168 COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40169 POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40170 STATEMENT_LIST, and RETURN_EXPR. For non-pure non-const functions
40171 return 0 only when not OEP_LEXICOGRAPHIC.
40172 (fold_build_cleanup_point_expr): Use the expression
40173 location when building CLEANUP_POINT_EXPR.
40174 * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40175 * tree.c (add_expr): Handle error_mark_node.
40177 2017-01-20 Martin Liska <mliska@suse.cz>
40180 * tree-profile.c (init_ic_make_global_vars): Do not call
40182 (gimple_init_gcov_profiler): Likewise.
40184 2017-01-20 Martin Liska <mliska@suse.cz>
40187 * cgraph.h (maybe_create_reference): Remove argument and
40189 * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40191 * ipa-cp.c (create_specialized_node): Likewise.
40192 * symtab.c (symtab_node::maybe_create_reference): Handle
40193 VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40195 2017-01-20 Martin Liska <mliska@suse.cz>
40197 * read-rtl-function.c (function_reader::create_function): Use
40198 build_decl instread of build_decl_stat.
40200 2017-01-20 Andrew Senkevich <andrew.senkevich@intel.com>
40202 * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40203 * config/i386/avx512dqintrin.h: Ditto.
40204 * config/i386/avx512fintrin.h: Ditto.
40205 * config/i386/i386-builtin-types.def: Add new types.
40206 * config/i386/i386.c: Handle new types.
40207 * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40208 (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40209 (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40210 (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40211 (__builtin_ia32_kshiftridi): New.
40212 * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40214 2017-01-19 Segher Boessenkool <segher@kernel.crashing.org>
40218 * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40219 define to rs6000_init_stack_protect_guard.
40220 (rs6000_init_stack_protect_guard): New function.
40222 2017-01-19 Matthew Fortune <matthew.fortune@imgtec.com>
40223 Yunqiang Su <yunqiang.su@imgtec.com>
40225 * config.gcc (supported_defaults): Add madd4.
40226 (with_madd4): Add validation.
40227 (all_defaults): Add madd4.
40228 * config/mips/mips.opt (mmadd4): New option.
40229 * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40231 (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40233 (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40234 (ISA_HAS_FUSED_MADD4): Likewise.
40235 * doc/invoke.texi (-mmadd4): Document the new option.
40236 * doc/install.texi (--with-madd4): Document the new option.
40238 2017-01-19 Jiong Wang <jiong.wang@arm.com>
40240 * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40241 entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40242 AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40243 (aarch64_init_pauth_hint_builtins): New.
40244 (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40245 (aarch64_expand_builtin): Expand new builtins.
40247 2017-01-19 Jiong Wang <jiong.wang@arm.com>
40249 * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40250 * combine-stack-adj.c (no_unhandled_cfa): Handle
40251 REG_CFA_TOGGLE_RA_MANGLE.
40252 * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40253 * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40254 info for return address signing.
40255 (aarch64_expand_epilogue): Likewise.
40257 2017-01-19 Jiong Wang <jiong.wang@arm.com>
40259 * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40260 * config/aarch64/aarch64-protos.h
40261 (aarch64_return_address_signing_enabled): New declaration.
40262 * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40264 (aarch64_expand_prologue): Sign return address before it's pushed onto
40266 (aarch64_expand_epilogue): Authenticate return address fetched from
40268 (aarch64_override_options): Sanity check for ILP32 and ISA level.
40269 (aarch64_attributes): New function attributes for "sign-return-address".
40270 * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40271 UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40272 ("*do_return"): Generate combined instructions according to key index.
40273 ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40274 * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40276 (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40277 * config/aarch64/aarch64.opt (msign-return-address=): New.
40278 * doc/extend.texi (AArch64 Function Attributes): Documents
40279 "sign-return-address=".
40280 * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40282 2017-01-19 Matthew Fortune <matthew.fortune@imgtec.com>
40284 * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40285 overall option summary.
40287 2017-01-19 Jiong Wang <jiong.wang@arm.com>
40289 * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40290 * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40291 AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40292 * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40294 2017-01-19 Michael Meissner <meissner@linux.vnet.ibm.com>
40296 * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40297 -mpower9-minmax by default for -mcpu=power9.
40298 (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40299 128-bit floating point.
40301 2017-01-20 Alan Modra <amodra@gmail.com>
40303 * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40304 optimizing for size.
40306 2017-01-20 Alan Modra <amodra@gmail.com>
40309 * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40310 for strcmp and strncmp from corresponding builtin decl.
40312 2017-01-19 Uros Bizjak <ubizjak@gmail.com>
40314 * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40315 instead of i386/rtems-64.h.
40316 * config/i386/rtems-64.h: Remove.
40318 2017-01-19 Uros Bizjak <ubizjak@gmail.com>
40322 2013-11-05 Uros Bizjak <ubizjak@gmail.com>
40324 * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40326 2017-01-19 Tamar Christina <tamar.christina@arm.com>
40328 * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40329 Change int to HOST_WIDE_INT.
40330 * config/aarch64/aarch64-protos.h
40331 (aarch64_simd_gen_const_vector_dup): Likewise.
40332 * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40334 2017-01-19 David Malcolm <dmalcolm@redhat.com>
40336 * langhooks-def.h (lhd_type_for_size): New decl.
40337 (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40338 * langhooks.c (lhd_type_for_size): New function, taken from
40341 2017-01-19 Pat Haugen <pthaugen@us.ibm.com>
40343 * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40344 define_bypass for CR latency.
40345 (power9-cracked-alu): Update bypass latency and remove power9-branch.
40346 (power9-alu2): Add define_bypass for CR latency.
40348 (power9-mul): Update insn latency.
40349 (power9-mul-compare): Update insn latency, bypass latency and remove
40352 2016-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
40354 * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40356 * config/aarch64/aarch64.md
40357 (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40358 aarch64_nopcrelative_literal_loads.
40359 (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40361 2017-01-19 Chenghua Xu <paul.hua.gm@gmail.com>
40363 * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40364 TARGET_LOONGSON_3A.
40365 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40367 2017-01-19 Doug Gilmore <doug.gilmore@imgtec.com>
40370 * config.gcc (supported_defaults): Add lxc1-sxc1.
40371 (with_lxc1_sxc1): Add validation.
40372 (all_defaults): Add lxc1-sxc1.
40373 * config/mips/mips.opt (mlxc1-sxc1): New option.
40374 * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40376 (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40377 __mips_no_lxc1_sxc1.
40378 (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40379 * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40380 * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40382 2017-01-19 Richard Biener <rguenther@suse.de>
40384 PR tree-optimization/72488
40385 * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40386 sure to restore SSA info.
40387 * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40389 2017-01-19 Richard Earnshaw <rearnsha@arm.com>
40391 PR rtl-optimization/79121
40392 * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40393 of the inner type when shifting an extended value.
40395 2017-01-17 Jan Hubicka <hubicka@ucw.cz>
40398 * symtab.c (symtab_node::equal_address_to): Fix comparing of
40399 interposable aliases.
40401 2017-01-18 Peter Bergner <bergner@vnet.ibm.com>
40404 * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40405 Use the evmergelohi instruction.
40406 (mov_si<mode>_e500_subreg4_2_le): Likewise.
40407 (mov_sitf_e500_subreg8_2_be): Likewise.
40408 (mov_sitf_e500_subreg12_2_le): Likewise.
40409 (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40410 (mov_si<mode>_e500_subreg4_2_be): Likewise.
40411 (mov_sitf_e500_subreg8_2_le): Likewise.
40412 (mov_sitf_e500_subreg12_2_be): Likewise.
40414 2017-01-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
40416 * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40417 attribute from vecsimple to vecperm.
40418 (altivec_vbpermq2): Likewise.
40420 2017-01-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
40423 * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40425 2017-01-18 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
40426 * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40427 * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40428 strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40429 case where N arg is SIZE_MAX.
40430 * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40431 (cmpstrsi): Add pattern.
40433 2017-01-18 Michael Meissner <meissner@linux.vnet.ibm.com>
40435 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40436 __builtin_vec_revb builtins.
40437 * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40438 built-in functions to support generation of the ISA 3.0 XXBR<x>
40439 vector byte reverse instructions.
40440 (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40441 (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40442 (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40443 (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40444 (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40445 (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40446 (P9V_BUILTIN_VEC_REVB): Likewise.
40447 * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40448 generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40449 (p9_xxbrq_v16qi): Likewise.
40450 (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40451 (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40452 (p9_xxbrh_v8hi): Likewise.
40453 * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40454 * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40455 vec_revb built-in functions.
40457 2017-01-18 Uros Bizjak <ubizjak@gmail.com>
40459 PR rtl-optimization/78952
40460 * config/i386/i386.md (any_extract): New code iterator.
40461 (*insvqi_2): Use any_extract for source operand.
40462 (*insvqi_3): Use any_shiftrt for source operand.
40464 2017-01-18 Wilco Dijkstra <wdijkstr@arm.com>
40466 * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40468 (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40470 2017-01-18 Matthias Klose <doko@ubuntu.com>
40472 * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40474 2016-01-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
40476 * config/rs6000/altivec.h (vec_bperm): Change #define.
40477 * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40478 (altivec_vbpermq2): New define_insn.
40479 (altivec_vbpermd): Likewise.
40480 * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40481 function interface.
40482 (VBPERMD): Likewise.
40483 (VBPERM): New polymorphic function interface.
40484 * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40485 Add entries for P9V_BUILTIN_VEC_VBPERM.
40486 * doc/extend.texi: Add interfaces for vec_bperm.
40488 2017-01-18 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
40490 * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40491 first letter of error messages.
40492 (s390_resolve_overloaded_builtin): Likewise.
40493 * config/s390/s390.c (s390_expand_builtin): Likewise.
40494 (s390_invalid_arg_for_unprototyped_fn): Likewise.
40495 (s390_valid_target_attribute_inner_p): Likewise.
40496 * config/s390/s390.md ("tabort"): Likewise.
40498 2017-01-18 Toma Tabacu <toma.tabacu@imgtec.com>
40500 * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40501 (ISA_AVOID_DIV_HILO): New macro.
40502 (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40503 (ISA_HAS_DDIV): Likewise.
40505 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
40507 * doc/invoke.texi (fabi-version): Correct number of occurrences.
40509 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
40511 * doc/invoke.texi (fabi-version): Spelling fix.
40513 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
40516 * doc/invoke.texi (fabi-version): Mention mangling fix for
40519 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
40522 * doc/invoke.texi (fabi-version): Document discriminator mangling.
40524 2017-01-17 Segher Boessenkool <segher@kernel.crashing.org>
40527 * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40528 * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40530 * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40532 (stack_protect_test): Ditto.
40533 * config/rs6000/rs6000.opt (mstack-protector-guard=,
40534 mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40536 * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40537 Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40538 -mstack-protector-guard-offset=.
40539 (RS/6000 and PowerPC Options): Ditto.
40541 2017-01-17 Uros Bizjak <ubizjak@gmail.com>
40543 * config/i386/i386.h (MASK_CLASS_P): New define.
40544 * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40545 there are no registers from different register sets also when
40546 mask registers are used. Update function comment.
40547 * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40548 to (*k/*r) and (*k/*km) alternatives.
40550 2017-01-17 Wilco Dijkstra <wdijkstr@arm.com>
40552 * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40553 * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40554 (EH_RETURN_HANDLER_RTX): New define.
40555 * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40556 Force frame pointer in EH return functions.
40557 (aarch64_expand_epilogue): Add barrier for eh_return.
40558 (aarch64_final_eh_return_addr): Remove.
40559 (aarch64_eh_return_handler_rtx): New function.
40560 * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40562 (aarch64_eh_return_handler_rtx): New prototype.
40564 2017-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
40566 * config/rs6000/altivec.h (vec_rlmi): New #define.
40567 (vec_vrlnm): Likewise.
40568 (vec_rlnm): Likewise.
40569 * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40570 (UNSPEC_VRLNM): Likewise.
40571 (VIlong): New mode iterator.
40572 (altivec_vrl<VI_char>mi): New define_insn.
40573 (altivec_vrl<VI_char>nm): Likewise.
40574 * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40576 (VRLDNM): Likewise.
40577 (RLNM): New polymorphic function entry.
40578 (VRLWMI): New monomorphic function entry.
40579 (VRLDMI): Likewise.
40580 (RLMI): New polymorphic function entry.
40581 * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40582 new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40583 * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40586 2017-01-17 Jakub Jelinek <jakub@redhat.com>
40589 * dwarf2out.c (field_byte_offset): Restore the
40590 PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40591 and DECL_FIELD_BIT_OFFSET. Use fold_build2 instead of build2 + fold.
40592 (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40595 2017-01-17 Eric Botcazou <ebotcazou@adacore.com>
40598 * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40600 2017-01-17 Jakub Jelinek <jakub@redhat.com>
40603 * dwarf2out.c (add_data_member_location_attribute): For constant
40604 offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40605 instead of DW_AT_data_member_location, DW_AT_bit_offset and
40606 DW_AT_byte_size attributes.
40608 2017-01-17 Eric Botcazou <ebotcazou@adacore.com>
40610 * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40611 after forcing to constant memory when the code model is medium.
40613 2017-01-17 Julia Koval <julia.koval@intel.com>
40616 * config/i386/avx512fintrin.h
40617 (_mm512_i32gather_ps): Change __addr type to void const*.
40618 (_mm512_mask_i32gather_ps): Ditto.
40619 (_mm512_i32gather_pd): Ditto.
40620 (_mm512_mask_i32gather_pd): Ditto.
40621 (_mm512_i64gather_ps): Ditto.
40622 (_mm512_mask_i64gather_ps): Ditto.
40623 (_mm512_i64gather_pd): Ditto.
40624 (_mm512_mask_i64gather_pd): Ditto.
40625 (_mm512_i32gather_epi32): Ditto.
40626 (_mm512_mask_i32gather_epi32): Ditto.
40627 (_mm512_i32gather_epi64): Ditto.
40628 (_mm512_mask_i32gather_epi64): Ditto.
40629 (_mm512_i64gather_epi32): Ditto.
40630 (_mm512_mask_i64gather_epi32): Ditto.
40631 (_mm512_i64gather_epi64): Ditto.
40632 (_mm512_mask_i64gather_epi64): Ditto.
40633 (_mm512_i32scatter_ps): Change __addr type to void*.
40634 (_mm512_mask_i32scatter_ps): Ditto.
40635 (_mm512_i32scatter_pd): Ditto.
40636 (_mm512_mask_i32scatter_pd): Ditto.
40637 (_mm512_i64scatter_ps): Ditto.
40638 (_mm512_mask_i64scatter_ps): Ditto.
40639 (_mm512_i64scatter_pd): Ditto.
40640 (_mm512_mask_i64scatter_pd): Ditto.
40641 (_mm512_i32scatter_epi32): Ditto.
40642 (_mm512_mask_i32scatter_epi32): Ditto.
40643 (_mm512_i32scatter_epi64): Ditto.
40644 (_mm512_mask_i32scatter_epi64): Ditto.
40645 (_mm512_i64scatter_epi32): Ditto.
40646 (_mm512_mask_i64scatter_epi32): Ditto.
40647 (_mm512_i64scatter_epi64): Ditto.
40648 (_mm512_mask_i64scatter_epi64): Ditto.
40649 * config/i386/avx512pfintrin.h
40650 (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40651 (_mm512_mask_prefetch_i32gather_ps): Ditto.
40652 (_mm512_mask_prefetch_i64gather_pd): Ditto.
40653 (_mm512_mask_prefetch_i64gather_ps): Ditto.
40654 (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40655 (_mm512_prefetch_i32scatter_ps): Ditto.
40656 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40657 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40658 (_mm512_prefetch_i64scatter_pd): Ditto.
40659 (_mm512_prefetch_i64scatter_ps): Ditto.
40660 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40661 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40662 * config/i386/avx512vlintrin.h
40663 (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40664 (_mm_mmask_i32gather_ps): Ditto.
40665 (_mm256_mmask_i32gather_pd): Ditto.
40666 (_mm_mmask_i32gather_pd): Ditto.
40667 (_mm256_mmask_i64gather_ps): Ditto.
40668 (_mm_mmask_i64gather_ps): Ditto.
40669 (_mm256_mmask_i64gather_pd): Ditto.
40670 (_mm_mmask_i64gather_pd): Ditto.
40671 (_mm256_mmask_i32gather_epi32): Ditto.
40672 (_mm_mmask_i32gather_epi32): Ditto.
40673 (_mm256_mmask_i32gather_epi64): Ditto.
40674 (_mm_mmask_i32gather_epi64): Ditto.
40675 (_mm256_mmask_i64gather_epi32): Ditto.
40676 (_mm_mmask_i64gather_epi32): Ditto.
40677 (_mm256_mmask_i64gather_epi64): Ditto.
40678 (_mm_mmask_i64gather_epi64): Ditto.
40679 (_mm256_i32scatter_ps): Change __addr type to void*.
40680 (_mm256_mask_i32scatter_ps): Ditto.
40681 (_mm_i32scatter_ps): Ditto.
40682 (_mm_mask_i32scatter_ps): Ditto.
40683 (_mm256_i32scatter_pd): Ditto.
40684 (_mm256_mask_i32scatter_pd): Ditto.
40685 (_mm_i32scatter_pd): Ditto.
40686 (_mm_mask_i32scatter_pd): Ditto.
40687 (_mm256_i64scatter_ps): Ditto.
40688 (_mm256_mask_i64scatter_ps): Ditto.
40689 (_mm_i64scatter_ps): Ditto.
40690 (_mm_mask_i64scatter_ps): Ditto.
40691 (_mm256_i64scatter_pd): Ditto.
40692 (_mm256_mask_i64scatter_pd): Ditto.
40693 (_mm_i64scatter_pd): Ditto.
40694 (_mm_mask_i64scatter_pd): Ditto.
40695 (_mm256_i32scatter_epi32): Ditto.
40696 (_mm256_mask_i32scatter_epi32): Ditto.
40697 (_mm_i32scatter_epi32): Ditto.
40698 (_mm_mask_i32scatter_epi32): Ditto.
40699 (_mm256_i32scatter_epi64): Ditto.
40700 (_mm256_mask_i32scatter_epi64): Ditto.
40701 (_mm_i32scatter_epi64): Ditto.
40702 (_mm_mask_i32scatter_epi64): Ditto.
40703 (_mm256_i64scatter_epi32): Ditto.
40704 (_mm256_mask_i64scatter_epi32): Ditto.
40705 (_mm_i64scatter_epi32): Ditto.
40706 (_mm_mask_i64scatter_epi32): Ditto.
40707 (_mm256_i64scatter_epi64): Ditto.
40708 (_mm256_mask_i64scatter_epi64): Ditto.
40709 (_mm_i64scatter_epi64): Ditto.
40710 (_mm_mask_i64scatter_epi64): Ditto.
40711 * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40712 (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40713 (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40714 (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40715 (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40716 (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40717 (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40718 (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40719 (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40720 (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40721 (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40722 (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40723 (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40724 (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40725 (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40726 (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40727 (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40728 (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40729 (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40730 (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40731 (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40732 (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40733 (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40734 (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40735 (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40736 (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40737 (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40738 (V16SF_V16SF_PCVOID_V16SI_HI_INT, V8DF_V8DF_PCVOID_V8SI_QI_INT)
40739 (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40740 (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40741 (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40742 (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40743 (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40744 (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40745 (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40746 (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40747 (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40748 (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40749 (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40750 (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40751 (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40752 (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40753 (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40754 (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40755 (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40756 (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40757 (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40758 (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40759 (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40760 (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40761 (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40762 (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40763 (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40764 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40765 definitions accordingly.
40767 2017-01-17 Kito Cheng <kito.cheng@gmail.com>
40768 Kuan-Lin Chen <kuanlinchentw@gmail.com>
40771 * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40774 2017-01-17 Vladimir Makarov <vmakarov@redhat.com>
40777 * ira-conflicts.c (ira_build_conflicts): Update total conflict
40778 hard regs for inner regno.
40780 2017-01-17 Martin Liska <mliska@suse.cz>
40783 * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40784 assumption and add comment.
40786 2017-01-17 Nathan Sidwell <nathan@acm.org>
40788 * ipa-visibility.c (localize_node): New function, broken out of ...
40789 (function_and_variable_visibility): ... here. Call it.
40791 2017-01-17 Jan Hubicka <hubicka@ucw.cz>
40793 PR middle-end/77445
40794 * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40795 correctly set frequency of oudgoing edge.
40796 (duplicate_thread_path): Fix profile updating.
40798 2017-01-17 Jakub Jelinek <jakub@redhat.com>
40801 * configure.ac: Add GCC_BASE_VER.
40802 * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40803 version from BASE-VER file.
40804 (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40805 (gcc.o): Depend on $(BASEVER).
40806 * common.opt (dumpfullversion): New option.
40807 * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40808 * doc/invoke.texi: Document -dumpfullversion.
40809 * doc/install.texi: Document --with-gcc-major-version-only.
40810 * configure: Regenerated.
40812 2017-01-17 Richard Biener <rguenther@suse.de>
40814 PR tree-optimization/71433
40815 * tree-vrp.c (register_new_assert_for): Merge same asserts
40816 on all incoming edges.
40817 (process_assert_insertions_for): Handle insertions at the
40820 2017-01-17 Gerald Pfeifer <gerald@pfeifer.com>
40822 * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40823 * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40825 2017-01-17 Kaz Kojima <kkojima@gcc.gnu.org>
40828 * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40831 2017-01-17 Alan Modra <amodra@gmail.com>
40834 * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40835 * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40836 symbolic stack limit when pic.
40838 2017-01-16 Martin Sebor <msebor@redhat.com>
40840 PR tree-optimization/78608
40841 * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40843 2017-01-16 Jeff Law <law@redhat.com>
40846 2016-12-02 Tadek Kijkowski <tkijkowski@gmail.com>
40847 * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
40848 for several include directories that may be relative to sysroot.
40849 * config/i386/x-mingw32 (gplus_includedir): Define.
40850 (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
40851 (native_system_includedir): Likewise.
40852 * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
40853 override if TARGET_SYSTEM_ROOT is defined.
40854 (NATIVE_SYSTEM_HEADER_DIR): Likewise.
40856 PR tree-optimization/79090
40857 PR tree-optimization/33562
40858 PR tree-optimization/61912
40859 PR tree-optimization/77485
40860 * tree-ssa-dse.c (compute_trims): Accept STMT argument. Dump STMT
40861 and computed trims into the dump file.
40863 2017-01-17 Uros Bizjak <ubizjak@gmail.com>
40865 * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
40867 2017-01-16 Jakub Jelinek <jakub@redhat.com>
40870 * gimplify.c (gimplify_init_constructor): If want_value and
40871 object == lhs, unshare lhs to avoid invalid tree sharing. Formatting
40875 * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
40876 sequence. Formatting fixes.
40877 (doloop_optimize): Formatting fixes.
40880 * gcc.c (debug_level_greater_than_spec_func): New function.
40881 (static_spec_functions): Add debug-level-gt spec function.
40882 (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
40884 * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40885 * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40886 * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
40887 gpubnames, ggnu-pubnames, gno-record-gcc-switches,
40888 grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
40889 gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
40891 2017-01-16 Uros Bizjak <ubizjak@gmail.com>
40893 * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
40894 QImode fixups to general and mask registers only.
40896 2017-01-16 Carl Love <cel@us.ibm.com>
40898 * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
40899 for built-in functions
40900 vector signed char vec_nabs (vector signed char)
40901 vector signed short vec_nabs (vector signed short)
40902 vector signed int vec_nabs (vector signed int)
40903 vector signed long long vec_nabs (vector signed long long)
40904 vector float vec_nabs (vector float)
40905 vector double vec_nabs (vector double)
40906 * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
40908 * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
40909 * config/rs6000/altivec.h: New define for vec_nabs built-in function.
40910 * doc/extend.texi: Update the documentation file for the new built-in
40913 2017-01-16 Martin Sebor <msebor@redhat.com>
40915 * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
40918 2017-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
40920 * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
40921 UNSPEC_VSX__XXSPLTD to require special splat handling.
40923 2017-01-16 David Malcolm <dmalcolm@redhat.com>
40926 * system.h: Poison strndup.
40928 2017-01-16 Alan Modra <amodra@gmail.com>
40931 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
40934 2017-01-16 Georg-Johann Lay <avr@gjlay.de>
40936 * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
40938 2017-01-15 Uros Bizjak <ubizjak@gmail.com>
40940 * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
40941 call recog here. Assert that INSN_CODE (insn) is non-negative.
40943 2017-01-15 Segher Boessenkool <segher@kernel.crashing.org>
40946 * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
40948 * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
40949 in the currently scheduled RTL fragment.
40951 2017-01-15 Segher Boessenkool <segher@kernel.crashing.org>
40953 PR rtl-optimization/78751
40954 * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
40957 2017-01-14 Jeff Law <law@redhat.com>
40959 PR tree-optimization/79090
40960 * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
40961 variable length stores.
40962 (compute_trims): Delete dead assignment to *trim_tail.
40963 (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
40966 2017-01-14 Bernd Schmidt <bschmidt@redhat.com>
40968 PR rtl-optimization/78626
40969 PR rtl-optimization/78727
40970 * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
40971 of a block, and split such blocks after everything else is finished.
40973 2017-01-14 Alan Modra <amodra@gmail.com>
40976 * combine.c (recog_for_combine_1): Set INSN_CODE before calling
40977 target legitimate_combined_insn.
40978 * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
40979 (rs6000_legitimate_combined_insn): New function.
40980 * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
40982 (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
40983 (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
40984 (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
40986 2017-01-14 Gerald Pfeifer <gerald@pfeifer.com>
40988 * doc/frontends.texi (G++ and GCC): Remove references to Java.
40990 2017-01-13 Jeff Law <law@redhat.com>
40992 PR tree-optimization/33562
40993 PR tree-optimization/61912
40994 PR tree-optimization/77485
40995 * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
40997 (delete_dead_assignment): Likewise.
40998 (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
40999 statement to delete_dead_call and delete_dead_assignment.
41001 2017-01-13 David Malcolm <dmalcolm@redhat.com>
41004 * substring-locations.c (format_warning_va): Strengthen case 1 so
41005 that both endpoints of the substring must be within the format
41006 range for just the substring to be printed.
41008 2017-01-13 Uros Bizjak <ubizjak@gmail.com>
41010 * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41011 * config/i386/i386.c (ix86_target_string): Add missing options
41012 to isa_opts and reorder options by implied ISAs. Rename isa_opts2 to
41013 isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41014 flags_other and ix86_target_other to flags2_other. Display unknown
41016 (ix86_valid_target_attribute_inner_p): Add missing options and
41017 reorder options by implied ISAs, as in ix86_target_string.
41019 2017-01-13 Richard Sandiford <richard.sandiford@arm.com>
41021 * hash-table.h (hash_table::too_empty_p): New function.
41022 (hash_table::expand): Use it.
41023 (hash_table::traverse): Likewise.
41024 (hash_table::empty_slot): Use sizeof (value_type) instead of
41025 sizeof (PTR) to convert bytes to elements. Shrink the table
41026 if the current size is excessive for the current number of
41029 2017-01-13 Richard Sandiford <richard.sandiford@arm.com>
41031 * ira-costs.c (record_reg_classes): Break from the inner loop
41032 early once alt_fail is known to be true. Update outer loop
41033 handling accordingly.
41035 2017-01-13 Jeff Law <law@redhat.com>
41037 * tree-ssa-dse.c (decrement_count): New function.
41038 (increment_start_addr, maybe_trim_memstar_call): Likewise.
41039 (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41040 when we know the partially dead statement is a mem* function.
41042 PR tree-optimization/61912
41043 PR tree-optimization/77485
41044 * tree-ssa-dse.c: Include expr.h.
41045 (maybe_trim_constructor_store): New function.
41046 (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41048 PR tree-optimization/33562
41049 PR tree-optimization/61912
41050 PR tree-optimization/77485
41051 * doc/invoke.texi: Document new dse-max-object-size param.
41052 * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41053 * tree-ssa-dse.c: Include params.h.
41054 (dse_store_status): New enum.
41055 (initialize_ao_ref_for_dse): New, partially extracted from
41057 (valid_ao_ref_for_dse, normalize_ref): New.
41058 (setup_live_bytes_from_ref, compute_trims): Likewise.
41059 (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41060 (maybe_trim_partially_dead_store): Likewise.
41061 (maybe_trim_complex_store): Likewise.
41062 (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41063 Track what bytes live from the original store. Return tri-state
41064 for dead, partially dead or live.
41065 (dse_dom_walker): Add constructor, destructor and new private members.
41066 (delete_dead_call, delete_dead_assignment): New extracted from
41068 (dse_optimize_stmt): Make a member of dse_dom_walker.
41069 Use initialize_ao_ref_for_dse.
41071 PR tree-optimization/33562
41072 PR tree-optimization/61912
41073 PR tree-optimization/77485
41074 * sbitmap.h (bitmap_count_bits): Prototype.
41075 (bitmap_clear_range, bitmap_set_range): Likewise.
41076 * sbitmap.c (bitmap_clear_range): New function.
41077 (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41079 2017-01-13 Martin Liska <mliska@suse.cz>
41082 * function.c (set_cfun): Add new argument force.
41083 * function.h (set_cfun): Likewise.
41084 * ipa-inline-transform.c (inline_call): Use the function when
41085 strict alising from is dropped for function we inline to.
41087 2017-01-13 Richard Biener <rguenther@suse.de>
41089 * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41090 for dumping GIMPLE INTEGER_CSTs.
41092 2017-01-13 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
41094 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41095 to 201112L since C++17.
41097 2017-01-13 Maxim Ostapenko <m.ostapenko@samsung.com>
41100 * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41101 if -fsanitize=kernel-address is present.
41103 2017-01-13 Richard Biener <rguenther@suse.de>
41105 * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41106 as _Literal ( type ) number in case usual suffixes do not
41107 preserve all information.
41109 2017-01-13 Richard Biener <rguenther@suse.de>
41111 PR tree-optimization/77283
41112 * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41113 and ssa-iterators.h.
41114 (is_feasible_trace): Implement a cost model based on joiner
41117 2017-01-12 Michael Meissner <meissner@linux.vnet.ibm.com>
41120 * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41121 char or short to __float128/_Float128 directly.
41123 2017-01-12 Martin Sebor <msebor@redhat.com>
41125 to -Wformat-overflow.
41126 * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41127 (min_bytes_remaining): Same.
41128 (get_string_length): Same.
41129 (format_string): Same.
41130 (format_directive): Same.
41132 (pass_sprintf_length::handle_gimple_call): Same.
41134 2017-01-12 Jakub Jelinek <jakub@redhat.com>
41136 * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41137 info.nowrite calls with no lhs that can't throw. Return bool
41138 whether gsi_remove has been called or not.
41139 (pass_sprintf_length::handle_gimple_call): Return bool whether
41140 try_substitute_return_value called gsi_remove. Formatting fix.
41141 (pass_sprintf_length::execute): Don't use gsi_remove if
41142 handle_gimple_call returned true.
41145 * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41146 be removed due to side-effects, don't remove following barrier nor
41147 turn the successor edge into fallthru edge.
41149 2017-01-12 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
41152 * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41153 element-reversing loads and stores as not swappable.
41155 2017-01-12 Nathan Sidwell <nathan@acm.org>
41156 Nicolai Stange <nicstange@gmail.com>
41158 * combine.c (try_combine): Don't ignore result of overlap checking
41159 loop. Combine overlap & asm check into single loop.
41161 2017-01-12 Richard Biener <rguenther@suse.de>
41163 * tree-pretty-print.c (dump_generic_node): Provide -gimple
41164 variant for MEM_REF. Sanitize INTEGER_CST for -gimple.
41166 2017-01-12 Richard Biener <rguenther@suse.de>
41168 * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41169 and TS_TARGET_OPTION directly derive from TS_BASE.
41170 * tree-core.h (tree_optimization_option): Derive from tree_base.
41171 (tree_target_option): Likewise.
41173 2017-01-11 Uros Bizjak <ubizjak@gmail.com>
41175 * config/i386/i386.c (memory_address_length): Increase len
41176 only when rip_relative_addr_p returns false.
41178 2017-01-11 Julia Koval <julia.koval@intel.com>
41180 * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41181 (OPTION_MASK_ISA_SGX_SET): New.
41182 (ix86_handle_option): Handle OPT_msgx.
41183 * config.gcc: Added sgxintrin.h.
41184 * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41185 * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41186 * config/i386/i386.c (ix86_target_string): Add -msgx.
41188 (ix86_option_override_internal): Handle new options.
41189 (ix86_valid_target_attribute_inner_p): Add sgx.
41190 * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41191 * config/i386/i386.opt: Add msgx.
41192 * config/i386/sgxintrin.h: New file.
41193 * config/i386/x86intrin.h: Add sgxintrin.h.
41195 2017-01-11 Jakub Jelinek <jakub@redhat.com>
41198 * fold-const.c (maybe_nonzero_address): Return 1 for function
41200 (tree_single_nonzero_warnv_p): Don't handle function local objects
41204 * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41207 2017-01-11 David Malcolm <dmalcolm@redhat.com>
41210 * opts.c: Include "spellcheck.h"
41211 (struct string_fragment): New struct.
41212 (struct edit_distance_traits<const string_fragment &>): New
41214 (get_closest_sanitizer_option): New function.
41215 (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41217 2017-01-11 Jakub Jelinek <jakub@redhat.com>
41219 * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41221 (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41222 DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41223 (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41224 (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41225 for initial die_offset if dwarf_split_debug_info.
41226 (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41227 initial next_die_offset if dwo_id is non-NULL. Don't emit padding
41229 (output_skeleton_debug_sections): Formatting fix. Use
41230 DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41231 DWARF_COMPILE_UNIT_HEADER_SIZE. Don't emit padding.
41233 2017-01-11 Wilco Dijkstra <wdijkstr@arm.com>
41235 * config/arm/cortex-a53.md: Add bypasses for
41236 cortex_a53_r2f_cvt.
41237 (cortex_a53_r2f): Only use for transfers.
41238 (cortex_a53_f2r): Likewise.
41239 (cortex_a53_r2f_cvt): Add reservation for conversions.
41240 (cortex_a53_f2r_cvt): Likewise.
41242 2017-01-11 Tamar Christina <tamar.christina@arm.com>
41244 * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41245 to all inlined functions, change static to extern.
41247 2017-01-11 Christophe Lyon <christophe.lyon@linaro.org>
41250 * config/arm/arm.c (legitimize_pic_address): Handle reference to
41252 (arm_assemble_integer): Likewise.
41254 2017-01-11 Richard Earnshaw <rearnsha@arm.com>
41256 * config.gcc: Use new awk script to check CPU, FPU and architecture
41257 parameters for --with-... options.
41258 * config/arm/parsecpu.awk: New file
41259 * config/arm/arm-cpus.in: New file.
41260 * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41262 * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41264 * config/arm/t-arm: Update dependency rules.
41265 * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41266 of processing .def files.
41267 * config/arm/genopt.sh: Deleted.
41268 * config/arm/gentune.sh: Deleted.
41269 * config/arm/arm-cores.def: Deleted.
41270 * config/arm/arm-arches.def: Deleted.
41271 * config/arm/arm-fpus.def: Deleted.
41272 * config/arm/arm-tune.md: Regenerated.
41273 * config/arm/arm-tables.opt: Regenerated.
41274 * config/arm/arm-cpu.h: New generated file.
41275 * config/arm/arm-cpu-data.h: New generated file.
41276 * config/arm/arm-cpu-cdata.h: New generated file.
41278 2017-01-11 Maxim Ostapenko <m.ostapenko@samsung.com>
41281 * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41283 (input_varpool_node): Unpack dynamically_initialized bit.
41285 2017-01-11 Eric Botcazou <ebotcazou@adacore.com>
41287 PR rtl-optimization/79032
41288 * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41289 the alignment of the adjusted memory reference against that of MODE,
41290 instead of the alignment of the original memory reference.
41292 2017-01-11 Martin Jambor <mjambor@suse.cz>
41294 * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41296 * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41297 decorated functions.
41299 2017-01-11 Richard Biener <rguenther@suse.de>
41301 * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41302 set range/nonnull info for PHI results. Do not set it on
41303 stmts marked for removal.
41305 2017-01-10 Eric Botcazou <ebotcazou@adacore.com>
41307 * expr.c (store_field): In the bitfield case, fetch the return value
41308 from the registers before applying a single big-endian adjustment.
41309 Always do a final load for a BLKmode value not larger than a word.
41311 2017-01-10 David Malcolm <dmalcolm@redhat.com>
41314 * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41315 that we correctly handle column numbers greater than
41316 LINE_MAP_MAX_COLUMN_NUMBER.
41318 2017-01-10 Martin Sebor <msebor@redhat.com>
41320 PR middle-end/78245
41321 * gimple-ssa-sprintf.c (get_destination_size): Call
41322 {init,fini}object_sizes.
41323 * tree-object-size.c (addr_object_size): Adjust.
41324 (pass_through_call): Adjust.
41325 (pass_object_sizes::execute): Adjust.
41326 * tree-object-size.h (fini_object_sizes): Declare.
41328 2017-01-10 Martin Sebor <msebor@redhat.com>
41330 PR tree-optimization/78775
41331 * builtins.c (get_size_range): Move...
41332 * calls.c: ...to here.
41333 (alloc_max_size): Accept zero argument.
41334 (operand_signed_p): Remove.
41335 (maybe_warn_alloc_args_overflow): Call get_size_range.
41336 * calls.h (get_size_range): Declare.
41338 2017-01-10 Joe Seymour <joe.s@somniumtech.com>
41340 * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41341 from TI's devices.csv file as of September 2016.
41342 * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41344 2017-01-10 Sandra Loosemore <sandra@codesourcery.com>
41346 * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41347 * doc/invoke.texi: Likewise.
41348 * doc/md.texi: Likewise.
41349 * doc/objc.texi: Likewise.
41351 2017-01-10 Joshua Conner <joshconner@google.com>
41353 * config/arm/fuchsia-elf.h: New file.
41354 * config/fuchsia.h: New file.
41355 * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41356 (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41358 * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41360 2016-01-10 Richard Biener <rguenther@suse.de>
41362 PR tree-optimization/79034
41363 * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41364 Propagate out degenerate PHIs in the joiner.
41366 2017-01-10 Martin Liska <mliska@suse.cz>
41368 * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41369 (sort_congruence_classes_by_decl_uid): Likewise.
41370 (sort_congruence_class_groups_by_decl_uid): Likewise.
41371 (sem_item_optimizer::merge_classes): Sort class, groups in these
41372 classes and members in the groups by DECL_UID of declarations.
41373 This would make merge operations stable.
41375 2017-01-10 Martin Liska <mliska@suse.cz>
41377 * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41378 usage of m_classes_vec.
41379 (sem_item_optimizer::~sem_item_optimizer): Likewise.
41380 (sem_item_optimizer::get_group_by_hash): Likewise.
41381 (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41382 (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41383 (sem_item_optimizer::verify_classes): Likewise.
41384 (sem_item_optimizer::process_cong_reduction): Likewise.
41385 (sem_item_optimizer::dump_cong_classes): Likewise.
41386 (sem_item_optimizer::merge_classes): Likewise.
41387 * ipa-icf.h (congruence_class_hash): Rename from
41388 congruence_class_group_hash. Remove declaration of m_classes_vec.
41390 2017-01-10 Andrew Senkevich <andrew.senkevich@intel.com>
41392 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41393 OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41394 * config.gcc: Add avx512vpopcntdqintrin.h.
41395 * config/i386/avx512vpopcntdqintrin.h: New.
41396 * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41397 * config/i386/i386-builtin-types.def: Add new types.
41398 * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41399 __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41400 __builtin_ia32_vpopcountq_v8di_mask): New.
41401 * config/i386/i386-c.c (ix86_target_macros_internal): Define
41402 __AVX512VPOPCNTDQ__.
41403 * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41404 (PTA_AVX512VPOPCNTDQ): Define.
41405 * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41406 TARGET_AVX512VPOPCNTDQ_P): Define.
41407 * config/i386/i386.opt: Add mavx512vpopcntdq.
41408 * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41409 * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41411 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
41413 PR middle-end/77484
41414 * predict.def (PRED_CALL): Set to 67.
41416 2017-01-09 Eric Botcazou <ebotcazou@adacore.com>
41418 * expr.c (store_field): In the bitfield case, if the value comes from
41419 a function call and is of an aggregate type returned in registers, do
41420 not modify the field mode; extract the value in all cases if the mode
41421 is BLKmode and the size is not larger than a word.
41423 2017-01-09 Dominique d'Humieres <dominiq@lps.ens.fr>
41426 * config/i386/cpuid.h: Fix undefined behavior.
41428 2017-01-04 Jeff Law <law@redhat.com>
41430 PR tree-optimization/79007
41431 PR tree-optimization/67955
41432 * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41433 conservative for pt.null when flag_non_call_exceptions is on.
41435 2017-01-09 Jakub Jelinek <jakub@redhat.com>
41437 PR translation/79019
41438 PR translation/79020
41439 * params.def (PARAM_INLINE_MIN_SPEEDUP,
41440 PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41441 PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41443 * config/avr/avr.opt (maccumulate-args): Likewise.
41444 * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41445 * common.opt (freport-bug): Likewise.
41446 * cif-code.def (CIF_FINAL_ERROR): Likewise.
41447 * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41448 * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41449 translatable string.
41450 * config/i386/i386.c (function_value_32): Likewise.
41451 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41452 * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41454 * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41455 * common/config/msp430/msp430-common.c (msp430_handle_option):
41457 * symtab.c (symtab_node::verify_base): Likewise.
41458 * opts.c (set_debug_level): Likewise.
41459 * tree.c (verify_type_variant): Likewise. Fix typo in comment.
41460 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41461 missing whitespace to translatable strings.
41462 * config/avr/avr.md (bswapsi2): Fix typo in comment.
41463 * config/sh/superh.h: Likewise.
41464 * config/i386/xopintrin.h: Likewise.
41465 * config/i386/znver1.md: Likewise.
41466 * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41467 * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41468 * double-int.h (struct double_int): Likewise.
41469 * double-int.c (div_and_round_double): Likewise.
41470 * wide-int.cc: Likewise.
41471 * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41472 * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41473 * cfgcleanup.c (crossjumps_occured): Renamed to ...
41474 (crossjumps_occurred): ... this.
41475 (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41478 PR tree-optimization/78899
41479 * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41480 returning bool return struct loop *, NULL for failure and the new
41482 (versionable_outer_loop_p): Don't version outer loop if it has
41483 dont_vectorized bit set.
41484 (tree_if_conversion): When versioning outer loop, ensure
41485 tree_if_conversion is performed also on the inner loop of the
41486 non-vectorizable outer loop copy.
41487 * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix. Fold
41488 LOOP_VECTORIZED in inner loop of the scalar outer loop and
41489 prevent vectorization of it.
41490 (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41491 the outer loop vectorization of the non-scalar version is attempted
41492 before vectorization of the inner loop in scalar version. If
41493 outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41494 vectorization of its inner loop.
41495 * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41496 has 2 inner loops, rename also on edges from bb whose single pred
41497 is outer_loop->header. Fix typo in function comment.
41499 2017-01-09 Martin Sebor <msebor@redhat.com>
41502 * asan.c (asan_emit_stack_protection): Increase local buffer size
41503 to avoid snprintf truncation warning.
41505 2017-01-09 Andrew Pinski <apinski@cavium.com>
41507 * config/aarch64/aarch64-cores.def: Add thunderx2t99. Change vulcan
41508 to reference thunderx2t99 for the tuning structure
41509 * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41511 (thunderx2t99_extra_costs): This.
41512 * config/aarch64/aarch64-tune.md: Regenerate.
41513 * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41514 (vulcan_addrcost_table): This.
41515 (vulcan_regmove_cost): Rename to ...
41516 (thunderx2t99_regmove_cost): This.
41517 (vulcan_vector_cost): Rename to ...
41518 (thunderx2t99_vector_cost): this.
41519 (vulcan_branch_cost): Rename to ...
41520 (thunderx2t99_branch_cost): This.
41521 (vulcan_tunings): Rename to ...
41522 (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41523 * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41525 2017-01-09 Martin Jambor <mjambor@suse.cz>
41529 * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41530 * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41531 (propagate_vr_accross_jump_function): Use the above function for all
41532 value range computations for pass-through jump functions and type
41533 converasion from explicit value range values.
41534 (ipcp_propagate_stage): Do not attempt to deduce types of formal
41535 parameters from TYPE_ARG_TYPES.
41536 * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41537 (ipa_write_node_info): Stream type of the actual argument.
41538 (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41540 2017-01-09 Martin Liska <mliska@suse.cz>
41543 * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41544 (lookup_compiler): Do not show error message with have_E.
41546 2017-01-09 Jakub Jelinek <jakub@redhat.com>
41548 PR tree-optimization/78938
41549 * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41550 where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41551 BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41552 {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR. Formatting
41555 2017-01-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
41557 * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41560 2017-01-09 Richard Biener <rguenther@suse.de>
41562 PR tree-optimization/78997
41563 * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41564 name condition properly.
41566 2017-01-09 Richard Biener <rguenther@suse.de>
41569 * dwarf2out.c (is_cxx): New overload with context.
41570 (is_naming_typedef_decl): Use it.
41572 2017-01-08 Sandra Loosemore <sandra@codesourcery.com>
41574 * invoke.texi (Option Summary): Correct spacing in option lists
41575 and add line breaks to fix over-long lines.
41577 2017-01-08 Sandra Loosemore <sandra@codesourcery.com>
41579 PR middle-end/17660
41581 * extend.texi (Common Variable Attributes): Add xref to GCC
41582 Internals manual to explain mode attribute keywords.
41584 2017-01-08 Sandra Loosemore <sandra@codesourcery.com>
41587 * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41588 and Preprocessor Options.
41589 (Options for Linking): Document -pthread here....
41590 (RS/6000 and PowerPC Options): ...not here.
41591 (Solaris 2 Options): ...or here.
41592 * doc/cppopts.texi: Document -pthread.
41594 2017-01-08 Martin Sebor <msebor@redhat.com>
41596 PR middle-end/77708
41597 * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41598 * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41599 New member functions.
41600 (format_directive): Used them.
41602 (pass_sprintf_length::handle_gimple_call): Same.
41603 * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41604 to avoid truncation for any argument.
41605 (extract_affine_mul): Same.
41606 * tree.c (get_file_function_name): Same.
41608 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
41610 PR middle-end/77484
41611 * predict.def (PRED_INDIR_CALL): Set to 86.
41613 2017-01-07 Sandra Loosemore <sandra@codesourcery.com>
41615 PR preprocessor/54124
41616 * doc/cppopts.texi: Reformat -d subtable to list the full name
41617 of the options. Add cross-reference to the docs for the general
41618 compiler -d options.
41619 * doc/invoke.texi (Developer Options): Add cross-reference to the
41620 preprocessor-specific -d option documentation.
41622 2017-01-07 Sandra Loosemore <sandra@codesourcery.com>
41624 PR preprocessor/13498
41625 * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41626 redudant material, and reflect new command-line options.
41627 (System Headers): Likewise.
41629 2017-01-07 Sandra Loosemore <sandra@codesourcery.com>
41631 * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41632 -isystem, and -idirafter. Copy-edit.
41633 * doc/cppopts.texi: Copy-edit. Remove contradiction about
41634 default for -ftrack-macro-expansion. Delete obsolete and
41635 badly-formatted implementation details about -fdebug-cpp output.
41636 * doc/cppwarnopts.texi: Copy-edit.
41638 2017-01-07 David Malcolm <dmalcolm@redhat.com>
41641 * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41642 that the transition from a max line width >= 1<<10 to narrower
41643 lines works correctly.
41645 2017-01-07 Alexandre Oliva <aoliva@redhat.com>
41647 * doc/options.texi (PerFunction): New.
41648 * opt-functions.awk (switch_flags): Map both Optimization and
41649 PerFunction to CL_OPTIMIZATION.
41650 * opth-gen.awk: Test for PerFunction flag along with
41652 * optc-save-gen.awk: Likewise. Introduce var_opt_hash and set
41653 it only when the latter is present. Skip those that don't in
41654 the hash function generator.
41655 * common.opt (fvar-tracking): Mark as PerFunction instead of
41657 (fvar-tracking-assignments): Likewise.
41658 (fvar-tracking-assignments-toggle): Likewise.
41659 (fvar-tracking-uninit): Likewise.
41661 2017-01-07 Jakub Jelinek <jakub@redhat.com>
41663 PR translation/79018
41664 * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41667 2017-01-06 Mikael Pettersson <mikpelinux@gmail.com>
41670 * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41671 * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41672 TARGET_LONG_JUMP_TABLE_OFFSETS.
41673 * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41674 * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41675 * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41676 (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41677 * config/m68k/m68k.md (tablejump expander): Likewise.
41678 (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41679 TARGET_LONG_JUMP_TABLE_OFFSETS.
41680 (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41681 * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41683 2017-01-06 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41684 David Holsgrove <david.holsgrove@xilinx.com>
41686 * common/config/microblaze/microblaze-common.c
41687 (TARGET_EXCEPT_UNWIND_INFO): Remove.
41688 * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41690 * config/microblaze/microblaze.c (microblaze_must_save_register)
41691 (microblaze_expand_epilogue, microblaze_return_addr): Handle
41693 (microblaze_eh_return): New function.
41694 * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41695 (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41696 (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41697 * config/microblaze/microblaze.md (eh_return): New pattern.
41699 2017-01-06 Jakub Jelinek <jakub@redhat.com>
41701 * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41702 GCC_DIAGNOSTIC_STRINGIFY): Define.
41704 * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41706 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
41708 * config/arm/arm.md (<mcrr>): New.
41710 * config/arm/arm.c (arm_arch5te): New.
41711 (arm_option_override): Set arm_arch5te.
41712 (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41714 * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41715 (arm_mcrr_qualifiers): ... this. New.
41716 (MRRC_QUALIFIERS): Define to...
41717 (arm_mrrc_qualifiers): ... this. New.
41718 * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41720 * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41721 * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41722 (MRRCI, mrrc, MRRC): New.
41723 * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41724 VUNSPEC_MRRC2): New.
41726 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
41728 * config/arm/arm.md (<mcr>): New.
41730 * config/arm/arm.c (arm_coproc_builtin_available): Add
41731 support for mcr, mrc, mcr2 and mrc2.
41732 * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41733 (arm_mcr_qualifiers): ... this. New.
41734 (MRC_QUALIFIERS): Define to ...
41735 (arm_mrc_qualifiers): ... this. New.
41736 (MCR_QUALIFIERS): Define to ...
41737 (arm_mcr_qualifiers): ... this. New.
41738 * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41740 * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41741 * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41742 * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41743 VUNSPEC_MRC2): New.
41745 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
41747 * config/arm/arm.md (*ldc): New.
41751 * config/arm/arm.c (arm_coproc_builtin_available): Add
41752 support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41753 (arm_coproc_ldc_stc_legitimate_address): New.
41754 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41755 'qualifier_const_pointer'.
41756 (LDC_QUALIFIERS): Define to...
41757 (arm_ldc_qualifiers): ... this. New.
41758 (STC_QUALIFIERS): Define to...
41759 (arm_stc_qualifiers): ... this. New.
41760 * config/arm/arm-protos.h
41761 (arm_coproc_ldc_stc_legitimate_address): New.
41762 * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41763 __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41764 * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41765 stc2, stcl, stc2l): New.
41766 * config/arm/constraints.md (Uz): New.
41767 * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41768 * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41769 VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41770 VUNSPEC_STC2L): New.
41772 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
41774 * config/arm/arm.md (<cdp>): New.
41775 * config/arm/arm.c (neon_const_bounds): Rename this ...
41776 (arm_const_bounds): ... this.
41777 (arm_coproc_builtin_available): New.
41778 * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41779 (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41780 (CDP_QUALIFIERS): Define to...
41781 (arm_cdp_qualifiers): ... this. New.
41783 (arm_expand_builtin_args): Add case for 6 arguments.
41784 * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41785 (arm_const_bounds): ... this.
41786 (arm_coproc_builtin_available): New.
41787 * config/arm/arm_acle.h (__arm_cdp): New.
41789 * config/arm/arm_acle_builtins.def (cdp): New.
41791 * config/arm/iterators.md (CDPI,CDP,cdp): New.
41792 * config/arm/neon.md: Rename all 'neon_const_bounds' to
41793 'arm_const_bounds'.
41794 * config/arm/types.md (coproc): New.
41795 * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41796 * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41797 * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41798 arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41800 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
41802 * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41803 (UBINOP_QUALIFIERS): New.
41805 (acle_builtin_data): New. Change comment.
41806 (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41807 ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41808 ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41809 arm_acle_builtins.def.
41810 (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41811 (arm_init_acle_builtins): New.
41812 (CRC32_BUILTIN): Remove.
41813 (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41814 crc32cb, crc32ch and crc32cw.
41815 (arm_init_crc32_builtins): Remove.
41816 (arm_init_builtins): Use arm_init_acle_builtins rather
41817 than arm_init_crc32_builtins.
41818 (arm_expand_acle_builtin): New.
41819 (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41820 * config/arm/arm_acle_builtins.def: New.
41822 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
41824 * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41825 (arm_builtin_datum): ... this.
41826 (arm_init_neon_builtin): Rename to ...
41827 (arm_init_builtin): ... this. Add a new parameters PREFIX
41828 and USE_SIG_IN_NAME.
41829 (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41830 'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41831 'arm_builtin_datum'.
41832 (arm_init_vfp_builtins): Likewise.
41833 (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41834 'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41835 (arm_expand_neon_args): Rename to ...
41836 (arm_expand_builtin_args): ... this. Rename builtin_arg
41837 enum values and differentiate between ARG_BUILTIN_MEMORY
41838 and ARG_BUILTIN_NEON_MEMORY.
41839 (arm_expand_neon_builtin_1): Rename to ...
41840 (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41841 values, arm_expand_builtin_args and add bool parameter NEON.
41842 (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41843 (arm_expand_vfp_builtin): Likewise.
41844 (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
41846 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
41848 PR middle-end/77484
41849 * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
41850 * predict.c (tree_estimate_probability_bb): Reverse direction of
41851 polymorphic call predictor.
41853 2017-01-06 David Malcolm <dmalcolm@redhat.com>
41855 * passes.c (execute_one_pass): Split out pass-skipping logic into...
41856 (determine_pass_name_match): ...this new function and...
41857 (should_skip_pass_p): ...this new function.
41859 2017-01-06 Nathan Sidwell <nathan@acm.org>
41861 * ipa-visibility.c (function_and_variable_visibility): Reformat
41862 comments and long lines. Remove extrneous if.
41863 * symtab.c (symtab_node::make_decl_local): Fix code format.
41864 (symtab_node::set_section_for_node): Fix comment typo.
41866 2017-01-06 Martin Liska <mliska@suse.cz>
41869 * lra-constraints.c: Rename invariant to lra_invariant.
41870 * predict.c (set_even_probabilities): Initialize e to NULL.
41872 2017-01-05 Martin Sebor <msebor@redhat.com>
41874 PR tree-optimization/78910
41875 * gimple-ssa-sprintf.c (tree_digits): Add an argument.
41876 (format_integer): Correct off-by-one error in the handling
41877 of precision with negative numbers in signed conversions..
41879 2017-01-05 Eric Botcazou <ebotcazou@adacore.com>
41881 * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
41883 2017-01-05 Jakub Jelinek <jakub@redhat.com>
41885 PR tree-optimization/71016
41886 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
41887 factor_out_conditional_conversion. Formatting fix.
41888 (factor_out_conditional_conversion): Add cond_stmt argument.
41889 If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
41890 cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
41893 2017-01-05 David Malcolm <dmalcolm@redhat.com>
41895 * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
41896 read-rtl-function.o, and selftest-rtl.o.
41897 * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
41898 (selftest::aarch64_test_loading_full_dump): New function.
41899 (selftest::aarch64_run_selftests): New function.
41900 (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
41901 selftest::aarch64_run_selftests.
41902 * config/i386/i386.c
41903 (selftest::ix86_test_loading_dump_fragment_1): New function.
41904 (selftest::ix86_test_loading_call_insn): New function.
41905 (selftest::ix86_test_loading_full_dump): New function.
41906 (selftest::ix86_test_loading_unspec): New function.
41907 (selftest::ix86_run_selftests): Call the new functions.
41908 * emit-rtl.c (maybe_set_max_label_num): New function.
41909 * emit-rtl.h (maybe_set_max_label_num): New decl.
41910 * function.c (instantiate_decls): Guard call to
41911 instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
41912 * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
41914 * gensupport.c (gen_reader::gen_reader): Pass "false"
41915 for new "compact" param of rtx_reader.
41916 * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
41917 rather than an empty string for NULL strings.
41918 * read-md.c: Potentially include config.h rather than bconfig.h.
41919 Wrap include of errors.h with #ifdef GENERATOR_FILE.
41920 (have_error): New global, copied from errors.c.
41921 (md_reader::read_name): Rename to...
41922 (md_reader::read_name_1): ...this, adding "out_loc" param,
41923 and converting "missing name or number" to returning false, rather
41925 (md_reader::read_name): Reimplement in terms of read_name_1.
41926 (md_reader::read_name_or_nil): New function.
41927 (md_reader::read_string): Handle "(nil)" by returning NULL.
41928 (md_reader::md_reader): Add new param "compact".
41929 (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
41930 (md_reader::read_file): New method.
41931 * read-md.h (md_reader::md_reader): Add new param "compact".
41932 (md_reader::read_file): New method.
41933 (md_reader::is_compact): New accessor.
41934 (md_reader::read_name): Convert return type from void to file_location.
41935 (md_reader::read_name_or_nil): New decl.
41936 (md_reader::read_name_1): New decl.
41937 (md_reader::m_compact): New field.
41938 (noop_reader::noop_reader): Pass "false" for new "compact" param
41940 (rtx_reader::rtx_reader): Add new "compact" param.
41941 (rtx_reader::read_rtx_operand): Make virtual and convert return
41942 type from void to rtx.
41943 (rtx_reader::read_until): New decl.
41944 (rtx_reader::handle_any_trailing_information): New virtual function.
41945 (rtx_reader::postprocess): New virtual function.
41946 (rtx_reader::finalize_string): New virtual function.
41947 (rtx_reader::m_in_call_function_usage): New field.
41948 (rtx_reader::m_reuse_rtx_by_id): New field.
41949 * read-rtl-function.c: New file.
41950 * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
41951 * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
41952 (selftest::verify_three_block_rtl_cfg): New decl.
41953 * read-rtl-function.h: New file.
41954 * read-rtl.c: Potentially include config.h rather than bconfig.h.
41955 For host, include function.h, memmodel.h, and emit-rtl.h.
41956 (one_time_initialization): New function.
41957 (struct compact_insn_name): New struct.
41958 (compact_insn_names): New array.
41959 (find_code): Handle insn codes in compact dumps.
41960 (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
41961 (bind_subst_iter_and_attr): Likewise.
41962 (add_condition_to_string): Likewise.
41963 (add_condition_to_rtx): Likewise.
41964 (apply_attribute_uses): Likewise.
41965 (add_current_iterators): Likewise.
41966 (apply_iterators): Likewise.
41967 (initialize_iterators): Guard usage of apply_subst_iterator with
41968 #ifdef GENERATOR_FILE.
41969 (read_conditions): Wrap with #ifdef GENERATOR_FILE.
41970 (md_reader::read_mapping): Likewise.
41971 (add_define_attr_for_define_subst): Likewise.
41972 (add_define_subst_attr): Likewise.
41973 (read_subst_mapping): Likewise.
41974 (check_code_iterator): Likewise.
41975 (rtx_reader::read_rtx): Likewise. Move one-time initialization
41977 (one_time_initialization): New function.
41978 (rtx_reader::read_until): New method.
41979 (read_flags): New function.
41980 (parse_reg_note_name): New function.
41981 (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
41982 Handle reuse_rtx ids.
41983 Wrap iterator lookup within #ifdef GENERATOR_FILE.
41984 Add parsing support for RTL dumps, mirroring the special-cases in
41985 print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
41986 values, and calling handle_any_trailing_information.
41987 (rtx_reader::read_rtx_operand): Convert return type from void
41988 to rtx, returning return_rtx. Handle case 'e'. Call
41989 finalize_string on XSTR and XTMPL fields.
41990 (rtx_reader::read_nested_rtx): Handle dumps in which trailing
41991 "(nil)" values were omitted. Call the postprocess vfunc on the
41993 (rtx_reader::rtx_reader): Add new "compact" param and pass to base
41994 class ctor. Initialize m_in_call_function_usage. Call
41995 one_time_initialization.
41996 * rtl-tests.c (selftest::test_uncond_jump): Call
41997 set_new_first_and_last_insn.
41998 * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
41999 * selftest-rtl.c: New file.
42000 * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42001 (selftest::get_insn_by_uid): New decl.
42002 * selftest-run-tests.c (selftest::run_tests): Call
42003 read_rtl_function_c_tests.
42004 * selftest.h (selftest::read_rtl_function_c_tests): New decl.
42005 * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42008 2017-01-05 Uros Bizjak <ubizjak@gmail.com>
42010 * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42011 operands in a special way. Assert that pos+len <= mode precision.
42013 2017-01-05 Jakub Jelinek <jakub@redhat.com>
42015 * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42016 3 argument Alias with unlimited for the negative form.
42017 (fno-vect-cost-model): Removed.
42019 2017-01-05 Martin Liska <mliska@suse.cz>
42021 * hsa-gen.c (gen_hsa_divmod): New function.
42022 (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42024 2017-01-05 Martin Liska <mliska@suse.cz>
42027 * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42030 2017-01-05 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
42032 * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42033 small constant length operands.
42035 2017-01-05 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
42037 * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42038 between loop iterations.
42040 2017-01-05 Martin Liska <mliska@suse.cz>
42043 * gimplify.c (gimplify_decl_expr): Compare to
42044 asan_poisoned_variables instread of checking flags.
42045 (gimplify_target_expr): Likewise.
42046 (gimplify_expr): Likewise.
42047 (gimplify_function_tree): Conditionally initialize
42048 asan_poisoned_variables.
42050 2017-01-04 Jeff Law <law@redhat.com>
42052 PR tree-optimizatin/78812
42053 * rtl.h (contains_mem_rtx_p): Prototype.
42054 * ifcvt.c (containts_mem_rtx_p): Move from here to...
42055 * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42056 * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42057 and prune MEMs that are not at the toplevel of a SET_SRC rtx. Look
42058 through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42060 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
42062 * input.c (assert_char_at_range): Default-initialize actual_range.
42064 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
42066 * df-scan.c (df_ref_create_structure): Make regno unsigned,
42067 to match the caller.
42069 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
42071 * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42072 insns after final jump in test to emit dummy move.
42074 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
42076 * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42077 * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42079 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
42081 * multiple_target.c (create_dispatcher_calls): Init e_next.
42082 * tree-ssa-loop-split.c (split_loop): Init border.
42083 * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42086 2017-01-04 Michael Meissner <meissner@linux.vnet.ibm.com>
42091 * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42092 (altivec_register_operand): Do not return true if the operand
42093 contains a SUBREG mixing SImode and SFmode.
42094 (vsx_register_operand): Likewise.
42095 (vsx_reg_sfsubreg_ok): New predicate.
42096 (vfloat_operand): Do not return true if the operand contains a
42097 SUBREG mixing SImode and SFmode.
42098 (vint_operand): Likewise.
42099 (vlogical_operand): Likewise.
42100 (gpc_reg_operand): Likewise.
42101 (int_reg_operand): Likewise.
42102 * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42103 * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42104 determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42106 (rs6000_emit_move_si_sf_subreg): New helper function.
42107 (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42108 fixup SUBREGs involving SImode and SFmode.
42109 * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42110 numbers for the new peephole2 optimization.
42111 (peephole2 for SFmode unions): New peephole2 to optimize cases in
42112 the GLIBC math library that do AND/IOR/XOR operations on single
42113 precision floating point.
42114 * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42115 target macros to say whether we need to avoid SUBREGs mixing
42117 (TARGET_ALLOW_SF_SUBREG): Likewise.
42118 * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42119 (UNSPEC_SI_FROM_SF): Likewise.
42120 (iorxor): Change spacing.
42121 (and_ior_xor): New iterator for AND, IOR, and XOR.
42122 (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42123 (movdi_from_sf_zero_ext): Likewise.
42124 (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42125 instead of gpc_reg_operand. Add SImode/SFmode SUBREG support.
42126 (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42127 (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42128 (fms<mode>4): Likewise.
42129 (fnma<mode>4): Likewise.
42130 (fnms<mode>4): Likewise.
42131 (nfma<mode>4): Likewise.
42132 (nfms<mode>4): Likewise.
42134 2017-01-04 Marek Polacek <polacek@redhat.com>
42137 * doc/invoke.texi: Document -Wpointer-compare.
42139 2017-01-04 Jakub Jelinek <jakub@redhat.com>
42141 * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42144 * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42145 descriptions for -gdwarf-5 and emit them as uleb128 instead of
42148 2017-01-04 Kelvin Nilsen <kelvin@gcc.gnu.org>
42151 * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42152 documentation of the powerpc_popcntb_ok attribute.
42153 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42154 code to issue warning messages if a requested CPU configuration is
42155 not supported by the binary (assembler and loader) toolchain.
42156 (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42157 made to define a built-in function that has been disabled.
42158 (paired_init_builtins): Add assertion to prevent ICE if attempt is
42159 made to define a built-in function that has been disabled.
42160 (altivec_init_builtins): Add comment explaining why definition
42161 of the DST built-in functions is not preceded by an assertion
42162 check. Add assertions to prevent ICE if attempts are made to
42163 define an altivec predicate or an abs* built-in function that has
42165 (htm_init_builtins): Add comment explaining why definition of the
42166 htm built-in functions is not preceded by an assertion check.
42168 2017-01-04 Jeff Law <law@redhat.com>
42170 PR tree-optimizatin/67955
42171 * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42172 Allow any SSA_VAR_P as the base objects. Use integer_zerop. Verify
42173 the points-to solution does not include pt_null. Use DECL_PT_UID
42176 2017-01-04 Uros Bizjak <ubizjak@gmail.com>
42178 * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42179 Use gen_int_mode instead of gen_lopwart for const_int operands.
42181 2017-01-04 Jakub Jelinek <jakub@redhat.com>
42183 PR tree-optimization/71563
42184 * match.pd: Simplify X << Y into X if Y is known to be 0 or
42185 out of range value - has low bits known to be zero.
42187 2017-01-04 Alan Modra <amodra@gmail.com>
42189 * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42190 * configure: Regenerate.
42191 * config.in: Regenerate.
42193 2017-01-04 Jakub Jelinek <jakub@redhat.com>
42196 * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42197 a substring of the message, but strcmp with the whole message. Ifdef
42198 ENABLE_NLS, translate the message first using dgettext.
42200 2017-01-03 Jeff Law <law@redhat.com>
42202 PR tree-optimizatin/78856
42203 * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42204 (mark_threaded_blocks): Remove code to truncate thread paths that
42205 cross multiple loop headers. Instead invalidate the cached loop
42206 iteration information and handle case of a thread path walking
42207 into an irreducible region.
42209 2017-01-03 Michael Meissner <meissner@linux.vnet.ibm.com>
42212 * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42213 assertions. Add support for doing the signbit if the IEEE 128-bit
42214 floating point value is in a GPR.
42215 * config/rs6000/rs6000.md (Fsignbit): Delete.
42216 (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42217 Update the length attribute if the value is in a GPR.
42218 (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42219 the sign or zero extension instruction, since the value is always 0/1.
42220 (signbit<mode>2_dm2): Delete using <Fsignbit>.
42223 * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42224 extracting SImode to a GPR register so that we can generate a
42225 store, limit the vector to be in a traditional Altivec register
42226 for the vextuwrx instruction.
42228 2017-01-03 Ian Lance Taylor <iant@google.com>
42230 * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42232 2017-01-03 Martin Sebor <msebor@redhat.com>
42234 PR tree-optimization/78696
42235 * gimple-ssa-sprintf.c (format_floating): Correct handling of
42236 precision. Use MPFR for %f for greater fidelity. Correct handling
42238 (pass_sprintf_length::compute_format_length): Set width and precision
42239 specified by asrerisk to void_node for vararg functions.
42240 (try_substitute_return_value): Adjust dump output.
42242 2017-01-03 David Edelsohn <dje.gcc@gmail.com>
42244 * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42246 2017-01-03 Eric Botcazou <ebotcazou@adacore.com>
42248 * doc/invoke.texi (SPARC options): Document -mlra as the default.
42249 * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42250 -mlra/-mno-lra was passed to the compiler.
42252 2017-01-03 James Cowgill <James.Cowgill@imgtec.com>
42254 PR rtl-optimization/65618
42255 * emit-rtl.c (try_split): Move initialization of "before" and
42256 "after" to just before the call to emit_insn_after_setloc.
42258 2017-01-03 Gerald Pfeifer <gerald@pfeifer.com>
42260 * doc/md.texi (Standard Names): Remove reference to Java frontend.
42262 2017-01-03 Pierre-Marie de Rodat <derodat@adacore.com>
42264 * dwarf2out.c (gen_enumeration_type_die): When
42265 -gno-strict-dwarf, add a DW_AT_encoding attribute.
42267 2017-01-03 Jakub Jelinek <jakub@redhat.com>
42269 PR tree-optimization/78965
42270 * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42271 Change first argument from const call_info & to call_info &. For %n
42272 set info.nowrite to false.
42274 PR middle-end/78901
42275 * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42276 possibly throwing calls.
42278 * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42279 and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42280 and fns handling, rather than in a separate case SSA_NAME.
42282 2017-01-02 Jeff Law <law@redhat.com>
42284 * config/darwin-driver.c (darwin_driver_init): Const-correctness
42285 fixes for first_period and second_period variables.
42287 2017-01-02 Uros Bizjak <ubizjak@gmail.com>
42290 * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42291 (*insvqi_1): New insn pattern.
42292 (*insvqi_1_mem_rex64): Ditto.
42293 (*insvqi_2): Ditto.
42294 (*insvqi_3): Rename from *insvqi.
42296 (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42298 2017-01-02 Gerald Pfeifer <gerald@pfeifer.com>
42300 * doc/cfg.texi (Edges): Remove reference to Java.
42301 (Maintaining the CFG): Ditto.
42303 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
42305 PR middle-end/77674
42306 * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42307 transparent aliases.
42309 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
42311 PR middle-end/77484
42312 * predict.def (PRED_CALL): Update hitrate.
42313 (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42314 * predict.c (tree_estimate_probability_bb): Split CALL predictor
42315 into direct/indirect/polymorphic variants.
42317 2017-01-01 Jakub Jelinek <jakub@redhat.com>
42319 Update copyright years.
42321 * gcc.c (process_command): Update copyright notice dates.
42322 * gcov-dump.c (print_version): Ditto.
42323 * gcov.c (print_version): Ditto.
42324 * gcov-tool.c (print_version): Ditto.
42325 * gengtype.c (create_file): Ditto.
42326 * doc/cpp.texi: Bump @copying's copyright year.
42327 * doc/cppinternals.texi: Ditto.
42328 * doc/gcc.texi: Ditto.
42329 * doc/gccint.texi: Ditto.
42330 * doc/gcov.texi: Ditto.
42331 * doc/install.texi: Ditto.
42332 * doc/invoke.texi: Ditto.
42334 Copyright (C) 2017 Free Software Foundation, Inc.
42336 Copying and distribution of this file, with or without modification,
42337 are permitted in any medium without royalty provided the copyright
42338 notice and this notice are preserved.