1 2018-10-23 Richard Biener <rguenther@suse.de>
3 * tree-vrp.c (add_assert_info): Guard dump_printf with
5 * gimple-ssa-evrp-analyze.c
6 (evrp_range_analyzer::record_ranges_from_incoming_edge):
7 Use value_range::ignore_equivs_equal_p.
9 2018-10-23 Richard Biener <rguenther@suse.de>
11 PR tree-optimization/87105
12 PR tree-optimization/87608
13 * passes.def (pass_all_early_optimizations): Add early phi-opt
15 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
16 addition to debug stmts.
17 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
18 and abs replacement early.
19 * tree-cfg.c (gimple_empty_block_p): Likewise.
21 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
24 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
25 specified to configure.
26 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
28 2018-10-23 Richard Biener <rguenther@suse.de>
30 PR tree-optimization/87700
31 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
33 2018-10-23 Jakub Jelinek <jakub@redhat.com>
36 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
37 second argument from __mmask16 to __mmask8.
38 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
39 _mm_mask_packs_epi32): Likewise.
40 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
42 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
44 2018-10-23 Richard Biener <rguenther@suse.de>
46 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
48 2018-10-23 Richard Biener <rguenther@suse.de>
50 PR tree-optimization/86144
51 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
54 2018-10-23 Richard Biener <rguenther@suse.de>
56 PR tree-optimization/87693
57 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
58 the case we do not find the taken edge.
60 2018-10-22 Paul Koning <ni1d@arrl.net>
62 * symtab.c (symtab_node::increase_alignment): Correct max
65 2018-10-22 Yury Gribov <tetra2005@gmail.com>
67 PR tree-optimization/87633
68 * match.pd: Do not generate unordered integer comparisons.
70 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
72 PR rtl-optimization/87600
73 * combine.c: Add include of expr.h.
74 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
76 (make_more_copies): New function, add a copy to a new pseudo after
77 the moves from hard registers into pseudos.
78 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
79 later. Call make_more_copies.
81 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
83 * lra-constraints.c (process_alt_operands): New local array,
84 matching_early_clobber. Check matching_early_clobber before
85 decrementing reject, and set matching_early_clobber after.
87 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
90 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
91 call output_addr_const and hope for the best.
93 2018-10-22 Richard Biener <rguenther@suse.de>
95 * gimple-ssa-evrp-analyze.c
96 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
97 smarter about what ranges to use.
98 * tree-vrp.c (add_assert_info): Dump here.
99 (register_edge_assert_for_2): Instead of here at multiple but
102 * gcc.dg/tree-ssa/evrp12.c: New testcase.
103 * gcc.dg/predict-6.c: Adjust.
104 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
105 * gcc.dg/tree-ssa/vrp02.c: Likewise.
106 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
108 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
109 Richard Biener <rguenther@suse.de>
111 * bitmap.h: Update data structure documentation, including a
112 description of bitmap views as either linked-lists or splay trees.
113 (struct bitmap_element_def): Update comments for splay tree bitmaps.
114 (struct bitmap_head_def): Likewise.
115 (bitmap_list_view, bitmap_tree_view): New prototypes.
116 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
118 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
119 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
120 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
121 released bitmap element here.
122 (bitmap_element_free): Remove.
123 (bitmap_elt_clear_from): Work on splay tree bitmaps.
124 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
125 this function similar ones such that linked-list bitmap implementation
126 functions are grouped.
127 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
128 and moved for grouping.
129 (bitmap_list_insert_element_after): Renamed from
130 bitmap_elt_insert_after, and moved for grouping.
131 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
132 (bitmap_tree_link_left, bitmap_tree_link_right,
133 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
134 bitmap_tree_link_element, bitmap_tree_unlink_element,
135 bitmap_tree_find_element): New functions for splay-tree bitmap
137 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
138 Renamed and moved, see above entries.
139 (bitmap_tree_listify_from): New function to convert part of a splay
140 tree bitmap to a linked-list bitmap.
141 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
142 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
143 (bitmap_find_bit): Remove.
144 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
145 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
146 Handle splay tree bitmaps.
147 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
148 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
149 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
150 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
151 bitmap_intersect_compl_p, bitmap_ior_and_compl,
152 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
153 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
154 corresponding changes to use linked-list specific bitmap_element
155 manipulation functions as applicable for efficiency.
156 (bitmap_tree_to_vec): New function.
157 (debug_bitmap_elt_file): New function split out from ...
158 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
159 (bitmap_print): Likewise.
161 PR tree-optimization/63155
162 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
164 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
167 2018-10-22 Martin Liska <mliska@suse.cz>
169 PR tree-optimization/87686
171 2018-08-29 Martin Liska <mliska@suse.cz>
173 * tree-switch-conversion.c (switch_conversion::expand):
174 Strenghten assumption about gswitch statements.
176 2018-10-22 Martin Liska <mliska@suse.cz>
178 * ipa-icf.c (sem_item::compare_attributes): Remove.
179 (sem_item::compare_referenced_symbol_properties): Use
180 attribute_list_equal instead.
181 (sem_function::equals_wpa): Likewise.
182 * ipa-icf.h: Remove compare_attributes.
184 2018-10-22 Richard Biener <rguenther@suse.de>
187 * mem-stats.h (mem_usage::operator==): Fix pasto.
189 2018-10-22 Richard Biener <rguenther@suse.de>
191 PR tree-optimization/87640
192 * tree-vrp.c (set_value_range_with_overflow): Decompose
194 (extract_range_from_binary_expr_1): Adjust.
196 2018-10-22 Martin Jambor <mjambor@suse.cz>
198 * tree-eh.h (stmt_could_throw_p): Add function parameter.
199 (stmt_can_throw_external): Likewise.
200 (stmt_can_throw_internal): Likewise.
201 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
202 (lower_eh_constructs_2): Likewise.
203 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
204 (stmt_can_throw_external): Likewise.
205 (stmt_can_throw_internal): Likewise.
206 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
207 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
208 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
209 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
210 (pass_lower_eh_dispatch::execute): Pass cfun to
211 stmt_can_throw_external.
212 (cleanup_empty_eh): Likewise.
213 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
214 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
215 stmt_can_throw_external instead of pushing it to cfun.
216 (symbol_table::create_edge): Likewise.
217 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
218 stmt_can_throw_internal.
219 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
220 to stmt_could_throw_p.
221 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
222 stmt_can_throw_internal.
223 (pass_store_merging::execute): Likewise.
224 * gimple-ssa-strength-reduction.c
225 (find_candidates_dom_walker::before_dom_children): Pass cfun to
227 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
228 stmt_can_throw_internal.
229 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
230 to stmt_can_throw_external.
231 (check_stmt): Pass cfun to stmt_could_throw_p.
232 (check_stmt): Pass cfun to stmt_can_throw_external.
233 (pass_nothrow::execute): Likewise.
234 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
235 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
236 stmt_can_throw_internal.
237 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
238 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
239 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
240 * tree-complex.c (expand_complex_libcall): Pass cfun to
241 stmt_could_throw_p and to stmt_can_throw_internal.
242 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
243 * tree-inline.c (copy_edges_for_bb): Likewise.
244 (maybe_move_debug_stmts_to_successors): Likewise.
245 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
247 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
248 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
249 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
250 stmt_can_throw_internal.
251 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
252 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
254 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
255 stmt_can_throw_internal.
256 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
257 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
259 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
260 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
261 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
262 (convert_mult_to_fma_1): Likewise.
263 (convert_to_divmod): Likewise.
264 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
265 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
266 * tree-ssa-propagate.c
267 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
268 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
269 (maybe_optimize_range_tests): Likewise.
270 (linearize_expr_tree): Likewise.
271 (reassociate_bb): Likewise.
272 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
273 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
274 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
275 (handle_char_store): Likewise.
276 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
277 stmt_can_throw_internal.
278 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
280 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
281 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
282 (vectorizable_simd_clone_call): Likewise.
283 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
284 (gimple_stringop_fixed_value): Likewise.
286 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
288 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
289 literal pool references.
290 (s390_check_qrst_address): Adapt to the new behavior of
291 s390_loadrelative_operand_p ().
293 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
296 * config/i386/sse.md (*andnot<mode>3_bcst): New.
298 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
301 * config/i386/sse.md (*<code><mode>3_bcst): New.
303 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
306 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
307 V4DI, V16SI and V8DI.
308 (*sub<mode>3<mask_name>_bcst): New.
309 (*add<mode>3<mask_name>_bcst): Likewise.
311 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
314 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
315 __builtin_ia32_vfnmsubpd512_mask.
316 (_mm512_mask_fnmsub_round_pd): Likewise.
317 (_mm512_fnmsub_pd): Likewise.
318 (_mm512_mask_fnmsub_pd): Likewise.
319 (_mm512_maskz_fnmsub_round_pd): Use
320 __builtin_ia32_vfnmsubpd512_maskz.
321 (_mm512_maskz_fnmsub_pd): Likewise.
322 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
323 (_mm512_mask_fnmsub_round_ps): Likewise.
324 (_mm512_fnmsub_ps): Likewise.
325 (_mm512_mask_fnmsub_ps): Likewise.
326 (_mm512_maskz_fnmsub_round_ps): Use
327 __builtin_ia32_vfnmsubps512_maskz.
328 (_mm512_maskz_fnmsub_ps): Likewise.
329 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
330 __builtin_ia32_vfnmsubpd256_mask.
331 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
332 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
333 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
334 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
335 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
336 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
337 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
338 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
339 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
340 __builtin_ia32_vfnmsubpd.
341 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
342 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
343 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
344 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
345 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
346 * config/i386/i386-builtin.def: Add
347 __builtin_ia32_vfnmsubpd256_mask,
348 __builtin_ia32_vfnmsubpd256_maskz,
349 __builtin_ia32_vfnmsubpd128_mask,
350 __builtin_ia32_vfnmsubpd128_maskz,
351 __builtin_ia32_vfnmsubps256_mask,
352 __builtin_ia32_vfnmsubps256_maskz,
353 __builtin_ia32_vfnmsubps128_mask,
354 __builtin_ia32_vfnmsubps128_maskz,
355 __builtin_ia32_vfnmsubpd512_mask,
356 __builtin_ia32_vfnmsubpd512_maskz,
357 __builtin_ia32_vfnmsubps512_mask,
358 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
359 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
360 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
361 __builtin_ia32_vfnmsubpd256.
362 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
363 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
364 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
366 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
368 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
370 (fmai_vmfnmsub_<mode><round_name>): Likewise.
372 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
375 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
376 __builtin_ia32_vfnmaddpd512_mask.
377 (_mm512_mask_fnmadd_round_pd): Likewise.
378 (_mm512_fnmadd_pd): Likewise.
379 (_mm512_mask_fnmadd_pd): Likewise.
380 (_mm512_maskz_fnmadd_round_pd): Use
381 __builtin_ia32_vfnmaddpd512_maskz.
382 (_mm512_maskz_fnmadd_pd): Likewise.
383 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
384 (_mm512_mask_fnmadd_round_ps): Likewise.
385 (_mm512_fnmadd_ps): Likewise.
386 (_mm512_mask_fnmadd_ps): Likewise.
387 (_mm512_maskz_fnmadd_round_ps): Use
388 __builtin_ia32_vfnmaddps512_maskz.
389 (_mm512_maskz_fnmadd_ps): Likewise.
390 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
391 __builtin_ia32_vfnmaddpd256_mask.
392 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
393 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
394 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
395 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
396 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
397 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
398 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
399 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
400 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
401 __builtin_ia32_vfnmaddpd.
402 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
403 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
404 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
405 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
406 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
407 * config/i386/i386-builtin.def: Add
408 __builtin_ia32_vfnmaddpd256_mask,
409 __builtin_ia32_vfnmaddpd256_maskz,
410 __builtin_ia32_vfnmaddpd128_mask,
411 __builtin_ia32_vfnmaddpd128_maskz,
412 __builtin_ia32_vfnmaddps256_mask,
413 __builtin_ia32_vfnmaddps256_maskz,
414 __builtin_ia32_vfnmaddps128_mask,
415 __builtin_ia32_vfnmaddps128_maskz,
416 __builtin_ia32_vfnmaddpd512_mask,
417 __builtin_ia32_vfnmaddpd512_maskz,
418 __builtin_ia32_vfnmaddps512_mask,
419 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
420 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
421 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
422 __builtin_ia32_vfnmaddpd256.
423 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
424 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
425 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
427 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
429 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
431 (fmai_vmfnmadd_<mode><round_name>): Likewise.
433 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
436 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
437 __builtin_ia32_vfmsubpd512_mask.
438 (_mm512_mask_fmsub_round_pd): Likewise.
439 (_mm512_fmsub_pd): Likewise.
440 (_mm512_mask_fmsub_pd): Likewise.
441 (_mm512_maskz_fmsub_round_pd): Use
442 __builtin_ia32_vfmsubpd512_maskz.
443 (_mm512_maskz_fmsub_pd): Likewise.
444 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
445 (_mm512_mask_fmsub_round_ps): Likewise.
446 (_mm512_fmsub_ps): Likewise.
447 (_mm512_mask_fmsub_ps): Likewise.
448 (_mm512_maskz_fmsub_round_ps): Use
449 __builtin_ia32_vfmsubps512_maskz.
450 (_mm512_maskz_fmsub_ps): Likewise.
451 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
452 __builtin_ia32_vfmsubpd256_mask.
453 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
454 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
455 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
456 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
457 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
458 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
459 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
460 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
461 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
462 __builtin_ia32_vfmsubpd.
463 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
464 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
465 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
466 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
467 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
468 * config/i386/i386-builtin.def: Add
469 __builtin_ia32_vfmsubpd256_mask,
470 __builtin_ia32_vfmsubpd256_maskz,
471 __builtin_ia32_vfmsubpd128_mask,
472 __builtin_ia32_vfmsubpd128_maskz,
473 __builtin_ia32_vfmsubps256_mask,
474 __builtin_ia32_vfmsubps256_maskz,
475 __builtin_ia32_vfmsubps128_mask,
476 __builtin_ia32_vfmsubps128_maskz,
477 __builtin_ia32_vfmsubpd512_mask,
478 __builtin_ia32_vfmsubpd512_maskz,
479 __builtin_ia32_vfmsubps512_mask,
480 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
481 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
482 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
483 __builtin_ia32_vfmsubpd256.
484 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
485 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
486 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
488 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
490 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
492 (fmai_vmfmsub_<mode><round_name>): Likewise.
494 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
496 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
497 Remove plus. Renamed to ...
498 (*sub<mode>3<mask_name>_bcst): This.
499 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
500 (*add<mode>3<mask_name>_bcst): This.
502 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
505 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
507 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
510 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
511 (_mm_or_epi32): Likewise.
512 (_mm256_xor_epi32): Likewise.
513 (_mm_xor_epi32): Likewise.
514 (_mm256_or_epi64): Likewise.
515 (_mm_or_epi64): Likewise.
516 (_mm256_xor_epi64): Likewise.
517 (_mm_xor_epi64): Likewise.
519 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
522 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
524 2018-10-20 Jakub Jelinek <jakub@redhat.com>
527 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
529 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
531 * doc/ux.texi: Move @section directly after @node.
533 2018-10-19 Jakub Jelinek <jakub@redhat.com>
537 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
538 depend closely nested inside of loop with ordered clause with
541 2018-10-19 David Malcolm <dmalcolm@redhat.com>
543 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
544 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
545 * doc/ux.texi: New file.
547 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
549 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
550 be the first CR field allocated.
552 2018-10-19 Richard Biener <rguenther@suse.de>
555 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
556 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
558 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
562 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
563 (*add<mode>3<mask_name>_bcst_2): Likewise.
565 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
568 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
569 Replace nonimmediate_operand with register_operand.
570 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
572 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
575 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
577 PR rtl-optimization/87596
578 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
579 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
580 for instructions in FROM..TO range.
582 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
584 * cfgexpand.c (expand_one_var): Use specific wording in error message
585 for non-local frame variables.
586 * stor-layout.c (layout_decl): Do not issue a warning for them.
588 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
590 * haifa-sched.c (priority): Add force_recompute parameter.
591 (apply_replacement): Call priority () with force_recompute = true.
592 (restore_pattern): Likewise.
594 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
596 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
597 HOST_BITS_PER_WIDE_INT.
598 (test_vector_ops_duplicate): Likewise.
600 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
603 * config/i386/sse.md (VF_AVX512): New.
604 (avx512bcst): Likewise.
605 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
607 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
609 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
612 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
614 * doc/invoke.texi (-dumpversion): Improve grammar.
615 (-dumpfullversion): Make more consistent with -dumpversion.
617 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
619 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
620 Set JUMP_LABEL to the jump insn.
621 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
622 Predict emitted jump and add label to jump insn.
624 2018-10-18 David Malcolm <dmalcolm@redhat.com>
626 PR tree-optimization/87562
627 * input.c (get_substring_ranges_for_loc): Use
628 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
629 getting the linemap for the endpoint. Verify that it's either
630 in the same linemap as the start point's spelling location, or
631 at least in the same file.
633 2018-10-18 Richard Biener <rguenther@suse.de>
635 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
636 feed width-specific load/store costs through ix86_vec_cost.
637 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
639 (bdver_cost): Likewise.
640 (znver1_cost): Likewise.
641 (btver1_cost): Likewise.
642 (btver2_cost): Likewise.
644 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
646 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
647 to simplify subreg of vec_merge.
649 2018-10-18 Richard Biener <rguenther@suse.de>
651 * config/i386/i386.c: Fix costing of vector FMA.
653 2018-10-18 Richard Biener <rguenther@suse.de>
655 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
657 (ix86_builtin_vectorization_cost): For vec_construct properly
658 cost insertion into SSE regs.
659 (...): Adjust calls to ix86_vec_cost.
661 2018-10-18 Richard Biener <rguenther@suse.de>
665 2018-02-07 Richard Biener <rguenther@suse.de>
667 PR tree-optimization/84204
668 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
671 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
674 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
676 (test_vector_ops_duplicate): Add test for a scalar subreg of a
677 VEC_MERGE of a VEC_DUPLICATE.
679 2018-10-17 Joseph Myers <joseph@codesourcery.com>
681 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
682 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
683 * doc/standards.texi (C Language): Document C2X.
684 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
685 (rl78_option_override): Handle "GNU C2X" language name.
687 2018-10-17 Joseph Myers <joseph@codesourcery.com>
689 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
690 Document C17 as published in 2018.
692 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
695 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
696 bail out if both sides do not have the same storage order.
698 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
700 * bitmap.c (bitmap_head::dump): New.
701 * bitmap.h (bitmap_head): Add dump().
702 * gimple-ssa-evrp-analyze.c
703 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
704 (evrp_range_analyzer::set_ssa_range_info): Same.
705 (evrp_range_analyzer::record_ranges_from_phis): Same.
706 (evrp_range_analyzer::record_ranges_from_stmt): Same.
707 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
708 * gimple-ssa-sprintf.c (get_int_range): Same.
709 (format_integer): Same.
710 (sprintf_dom_walker::handle_gimple_call): Same.
711 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
712 (ipcp_vr_lattice::top_p): Same.
713 (ipcp_vr_lattice::bottom_p): Same.
714 (ipcp_vr_lattice::set_to_bottom): Same.
715 (ipa_vr_operation_and_type_effects): Same.
716 (propagate_vr_across_jump_function): Same.
717 (ipcp_store_vr_results): Same.
718 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
719 (ipa_print_node_jump_functions_for_edge): Same.
720 (ipa_get_value_range): Same.
721 (ipa_compute_jump_functions_for_edge): Same.
722 (ipa_write_jump_function): Same.
723 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
724 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
726 * vr-values.c (set_value_range_to_nonnegative): Same.
727 (set_value_range_to_truthvalue): Same.
728 (vr_values::get_value_range): Same.
729 (vr_values::set_defs_to_varying): Same.
730 (vr_values::update_value_range): Same.
731 (symbolic_range_based_on_p): Same.
732 (vr_values::op_with_boolean_value_range_p): Same.
733 (vr_values::extract_range_for_var_from_comparison_expr): Same.
734 (vr_values::extract_range_from_ssa_name): Same.
735 (vr_values::extract_range_from_binary_expr): Same.
736 (vr_values::extract_range_from_unary_expr): Same.
737 (vr_values::extract_range_from_cond_expr): Same.
738 (vr_values::extract_range_from_comparison): Same.
739 (vr_values::check_for_binary_op_overflow): Same.
740 (vr_values::extract_range_basic): Same.
741 (vr_values::extract_range_from_assignment): Same.
742 (compare_ranges): Same.
743 (compare_range_with_value): Same.
744 (vr_values::adjust_range_with_scev): Same.
745 (vrp_valueize): Same.
746 (vrp_valueize_1): Same.
747 (vr_values::get_vr_for_comparison): Same.
748 (vr_values::compare_name_with_value): Same.
749 (vr_values::compare_names): Same.
750 (vr_values::vrp_evaluate_conditional): Same.
751 (find_case_label_ranges): Same.
752 (vr_values::vrp_visit_switch_stmt): Same.
753 (vr_values::extract_range_from_phi_node): Same.
754 (vr_values::simplify_div_or_mod_using_ranges): Same.
755 (vr_values::simplify_bit_ops_using_ranges): Same.
756 (test_for_singularity): Same.
757 (range_fits_type_p): Same.
758 (vr_values::simplify_cond_using_ranges_1): Same.
759 (vr_values::simplify_switch_using_ranges): Same.
760 (vr_values::simplify_float_conversion_using_ranges): Same.
761 (vr_values::two_valued_val_range_p): Same.
762 (vr_values::add_equivalence): Move to value_range::equiv_add.
763 * vr-values.h (vr_values::add_equivalence): Remove.
764 (VR_INITIALIZER): Remove.
765 * tree-vrp.c (value_range::set): New.
766 (value_range::equiv_add): New.
767 (value_range::value_range): New.
768 (value_range::deep_copy): New.
769 (value_range::check): New.
770 (value_range::equal_p): New.
771 (value_range::ignore_equivs_equal_p): New.
772 (value_range::operator==): New.
773 (value_range::operator!=): New.
774 (value_range::symbolic_p): New.
775 (value_range::numeric_p): New.
776 (value_range::set_undefined): New.
777 (value_range::set_varying): New.
778 (value_range::may_contain_p): New.
779 (value_range::equiv_clear): New.
780 (value_range::singleton_p): New.
781 (value_range::intersect): New.
782 (value_range::dump): New.
783 (value_range::set_and_canonicalize): New.
784 (set_value_range): Adjust for value_range API.
785 (set_value_range_to_undefined): Same.
786 (set_value_range_to_varying): Same.
787 (set_and_canonicalize_value_range): Same.
788 (set_value_range_to_nonnull): Same.
789 (set_value_range_to_null): Same.
790 (range_is_null): Same.
791 (range_is_nonnull): Same.
792 (range_int_cst_p): Same.
793 (range_int_cst_singleton_p): Same.
794 (symbolic_range_p): Same.
795 (range_includes_zero_p): Same.
796 (value_range_constant_singleton): Same.
797 (vrp_set_zero_nonzero_bits): Same.
798 (ranges_from_anti_range): Same.
799 (extract_range_into_wide_ints): Same.
800 (extract_range_from_multiplicative_op): Same.
801 (set_value_range_with_overflow): Same.
802 (extract_range_from_binary_expr_1): Same.
803 (extract_range_from_unary_expr): Same.
804 (dump_value_range): Same.
805 (debug_value_range): Same.
806 (vrp_prop::check_array_ref): Same.
807 (vrp_prop::check_mem_ref): Same.
808 (vrp_prop::vrp_initialize): Same.
809 (vrp_prop::visit_stmt): Same.
810 (intersect_ranges): Same.
811 (vrp_prop::visit_phi): Same.
812 (vrp_prop::vrp_finalize): Same.
813 (determine_value_range_1): Same.
814 (determine_value_range): Same.
815 (vrp_intersect_ranges_1): Rename to...
816 (vrp_intersect_1): this.
817 (vrp_intersect_ranges): Rename to...
818 (value_range::intersect_helper): ...this.
819 (vrp_meet_1): Rename to...
820 (value_range::union_helper): ...this.
821 (vrp_meet): Rename to...
822 (value_range::union_): ...this.
823 (copy_value_range): Remove.
824 * tree-vrp.h (struct value_range): Rewrite into a proper class.
825 (value_range::vrtype): New.
826 (value_range::type): New.
827 (value_range::equiv): New.
828 (value_range::min): New.
829 (value_range::max): New.
830 (value_range::varying_p): New.
831 (value_range::undefined_p): New.
832 (value_range::null_p): New.
833 (value_range::equiv_add): New.
834 (copy_value_range): Remove.
836 2018-10-17 David Malcolm <dmalcolm@redhat.com>
838 * Makefile.in (SELFTEST_TARGETS): New.
839 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
840 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
841 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
843 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
844 (selftest-c++-gdb, selftest-c++-valgrind): Move to
846 * configure: Regenerate.
847 * configure.ac (selftest_languages): New.
849 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
851 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
852 overflow wraps argument.
853 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
854 wide_int_range_multiplicative_op.
855 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
856 overflow wraps argument.
857 (wide_int_range_multiplicative_op): Same.
858 (wide_int_range_lshift): Same.
859 (wide_int_range_div): Same.
860 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
861 (wide_int_range_lshift): Same.
862 (wide_int_range_div): Same.
864 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
866 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
867 use sign as argument.
868 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
869 wide_int_range_shift_undefined_p.
871 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
873 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
875 (@despeculate_copy<ALLI_TI:mode>): ... This.
876 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
879 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
881 * config.gcc: Obsolete *-*-solaris2.10*.
882 * doc/install.texi (Specific, *-*-solaris2*): Document it.
884 2018-10-12 Jeff Law <law@redhat.com>
886 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
887 reg + sym +- const_int addressing modes.
889 2018-10-15 David Malcolm <dmalcolm@redhat.com>
891 * common.opt (fdiagnostics-minimum-margin-width=): New option.
892 * diagnostic-show-locus.c (layout::layout): Apply the minimum
894 (layout::start_annotation_line): Only print up to 3 of the
895 margin character, to avoid touching the left-hand side.
896 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
897 minimum margin width, as set by test_diagnostic_context's ctor.
898 (selftest::test_fixit_insert_containing_newline): Likewise.
899 (selftest::test_fixit_insert_containing_newline_2): Likewise.
900 (selftest::test_line_numbers_multiline_range): Clear
902 * diagnostic.c (diagnostic_initialize): Initialize
904 * diagnostic.h (struct diagnostic_context): Add field
906 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
907 * opts.c (common_handle_option): Handle
908 OPT_fdiagnostics_minimum_margin_width_.
909 * selftest-diagnostic.c
910 (selftest::test_diagnostic_context::test_diagnostic_context):
911 Initialize min_margin_width to 6.
912 * toplev.c (general_init): Initialize global_dc->min_margin_width.
914 2018-10-15 David Malcolm <dmalcolm@redhat.com>
916 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
917 Fix usage of "error_at_rich_loc" in the comment.
919 2018-10-15 Renlin Li <renlin.li@arm.com>
922 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
923 if-conversioned loop when it contains ifn with types not
924 supported by backend.
925 * internal-fn.c (expand_direct_optab_fn): Add an assert.
926 (direct_internal_fn_supported_p): New helper function.
927 * internal-fn.h (direct_internal_fn_supported_p): Declare.
929 2018-10-15 Jakub Jelinek <jakub@redhat.com>
932 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
933 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
934 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
936 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
938 PR tree-optimization/87022
939 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
940 bits in dist vector rather than the first one.
942 2018-10-15 Richard Biener <rguenther@suse.de>
945 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
946 (visit_loadstore): When a used restrict tag escaped verify that
947 the points-to solution of "other" pointers do not include
949 (compute_dependence_clique): If a used restrict tag escaped
950 communicated that down to visit_loadstore.
952 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
954 * config/s390/s390.c (s390_expand_vec_init): Force vector element
955 into reg if it isn't a general operand.
957 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
960 * config/i386/sse.md (*vec_dupv2di): Add register source to
963 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
966 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
967 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
968 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
969 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
971 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
973 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
974 (notice_args_size): Set it in the current trace if no insn that can
975 throw internally has been seen yet.
976 (connect_traces): When connecting args_size between traces, allow the
977 incoming values not to match if there is an insn setting it before the
978 first insn that can throw internally; in that case, force the creation
979 of a CFI note on this latter insn.
981 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
983 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
984 the base class, because using the injected-class-name was not clearly
985 specified until DR 176.
987 2018-10-12 Paul Koning <ni1d@arrl.net>
989 * config/pdp11/pdp11.md (doloop_end): New expander.
990 (doloop_end_insn): renamed from "doloop_end".
991 (addqi3): New pattern.
992 (subqi3): New pattern.
993 * config/pdp11/predicates.md (incdec_operand): New predicate.
995 2018-10-12 Yury Gribov <tetra2005@gmail.com>
998 * real.c (format_helper::can_represent_integral_type_p): New function
999 * real.h (format_helper::can_represent_integral_type_p): Ditto.
1000 * match.pd: New pattern.
1002 2018-10-12 Alexandre Oliva <oliva@adacore.com>
1004 * configure.ac: Introduce --enable-large-address-aware
1005 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
1006 * doc/install.texi: Document it.
1007 * configure, config.in: Rebuilt.
1008 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
1009 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
1010 (LINK_SPEC): Insert it.
1011 * config/i386/mingw-264.h: Likewise.
1013 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
1015 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
1017 PR rtl-optimization/87600
1018 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
1020 2018-10-12 Paul Koning <ni1d@arrl.net>
1022 * doc/md.texi (doloop_end): Document that the pattern code may
1023 need to check operand mode.
1025 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
1027 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
1028 to zero-extend between int and floating-point registers.
1029 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
1030 ldp into floating-point registers. Add type and arch attributes.
1031 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
1032 Use f_loads for type attribute.
1034 2018-10-11 Martin Sebor <msebor@redhat.com>
1036 * doc/extend.texi (attribute packed): Correct typos.
1038 2018-10-11 Martin Sebor <msebor@redhat.com>
1040 * doc/extend.texi (attribute flatten): Mention interaction with
1043 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
1046 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1047 Set new_decl virtual flag to zero.
1049 2018-10-11 Martin Sebor <msebor@redhat.com>
1052 * doc/extend.texi (attribute format_arg): Discuss using multiple
1053 attributes on a single function.
1055 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
1057 PR tree-optimization/86829
1058 * match.pd (sin (atan (x))): New simplification rules.
1059 (cos (atan (x))): Likewise.
1060 * real.c (build_sinatan_real): New function.
1061 * real.h (build_sinatan_real): Prototype.
1063 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
1065 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
1067 (fold_mergeeo_helper): New helper function.
1068 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
1069 intrinsics. Correct some whitespace indentation issues.
1071 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
1074 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
1075 Use HOST_WIDE_INT_1U for shift.
1077 2018-10-11 Doug Rupp <rupp@adacore.com>
1078 Olivier Hainque <hainque@adacore.com>
1080 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
1081 Pass --relax to the linker for RTPs.
1082 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
1084 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
1085 Jan Hubicka <jh@suse.cz>
1086 Martin Jambor <mjambor@suse.cz>
1088 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
1089 the same elements are repeated rather than printing all of them.
1090 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
1091 "repeated" elements.
1092 * read-rtl-function.c (test_loading_repeat): New function.
1093 (read_rtl_function_c_tests): Call test_loading_repeat.
1094 * rtl-tests.c (test_dumping_repeat): New function.
1095 (rtl_tests_c_tests): Call test_dumping_repeat.
1097 2018-10-11 Richard Biener <rguenther@suse.de>
1099 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
1100 bdver?_cost): Unify to ...
1101 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
1102 * config/i386/i386.c (processor_cost_table): Adjust.
1104 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
1107 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1108 the thunk when expanding to GIMPLE.
1110 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
1112 * varasm.c (mergeable_string_section): Don't try to move zero-length
1113 strings to the merge section.
1115 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
1118 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1120 2018-10-10 Jakub Jelinek <jakub@redhat.com>
1123 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1124 to special_args set.
1126 2018-10-10 Richard Biener <rguenther@suse.de>
1128 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1129 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1130 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1131 and recursing and pattern terminating the recursion on SSE
1132 vector width using ix86_expand_reduc.
1133 (reduc_sminmax_scal_<mode>): Split into part reducing to half
1134 width and recursing and SSE2 vector variant doing the final
1135 reduction with ix86_expand_reduc.
1136 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1137 with terminating the recursion at AVX level, splitting that
1140 2018-10-09 David Malcolm <dmalcolm@redhat.com>
1142 * genmatch.c (error_cb): Rename to...
1143 (diagnostic_cb): ...this, converting int params to enums.
1144 (fatal_at): Update for renaming.
1145 (warning_at): Likewise.
1147 * input.c (selftest::ebcdic_execution_charset::apply):
1148 Update for renaming of...
1149 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1151 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1152 converting level and reason to enums.
1153 (class selftest::lexer_error_sink): Rename to...
1154 (class selftest::lexer_test_options): ...this, renaming field
1155 "m_errors" to "m_diagnostics".
1156 (selftest::lexer_test_options::apply): Update for renaming of...
1157 (selftest::lexer_test_options::on_error): ...this, renaming to...
1158 (selftest::lexer_test_options::on_diagnostic): ...this
1159 converting level and reason to enums.
1160 (selftest::test_lexer_string_locations_raw_string_unterminated):
1161 Update for renamings.
1162 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
1165 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
1167 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1168 * config/rs6000/pmmintrin.h: New file.
1170 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
1172 PR tree-optimization/86659
1173 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1175 2018-10-09 Richard Biener <rguenther@suse.de>
1177 PR tree-optimization/63155
1178 * tree-ssa-structalias.c: Include tree-ssa.h.
1179 (get_constraint_for_ssa_var): For undefs return nothing_id.
1180 (find_func_aliases): Cleanup PHI handling.
1182 2018-10-09 Richard Biener <rguenther@suse.de>
1184 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1187 2018-10-09 Martin Liska <mliska@suse.cz>
1189 * asan.c (asan_emit_stack_protection): If a stack variable
1190 is located in a same file as current function, then emit
1191 line info into variable definition string.
1193 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
1195 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1198 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
1200 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1203 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
1205 PR tree-optimization/86659
1206 * gimple-match.h (struct gimple_match_op): Add reverse field.
1207 (gimple_match_op::set_op): New overloaded method.
1208 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1209 the REF_REVERSE_STORAGE_ORDER flag on the value.
1210 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
1211 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
1213 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
1216 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
1217 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
1219 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
1222 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
1223 Defined with __builtin_ia32_vfmaddsubpd512_mask.
1225 2018-10-08 Richard Biener <rguenther@suse.de>
1227 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
1228 cost the same as AVX128 ones.
1230 2018-10-08 Paul Koning <ni1d@arrl.net>
1232 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
1233 (expand_block_move): New function.
1234 * config/pdp11/pdp11.c (output_block_move): Remove.
1235 (expand_block_move): New function.
1236 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
1237 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
1238 (*movmemhi1): Remove.
1240 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
1242 * config/s390/2827.md: Increase latencies for some FP instructions.
1244 2018-10-08 Richard Biener <rguenther@suse.de>
1246 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1248 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
1249 * tree-vectorizer.h (dump_stmt_cost): Adjust.
1250 (add_stmt_cost): Dump return value of the hook.
1252 2018-10-08 Richard Biener <rguenther@suse.de>
1254 PR tree-optimization/63155
1255 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
1256 (ssa_propagation_engine::ssa_propagate): Remove redundant
1257 bitmap bit clearing.
1259 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
1261 PR rtl-optimization/86939
1262 PR rtl-optimization/87479
1263 * ira.h (non_conflicting_reg_copy_p): New prototype.
1264 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
1265 (make_hard_regno_dead): Don't add conflicts for register
1266 ignore_reg_for_conflicts.
1267 (make_object_dead): Likewise.
1268 (non_conflicting_reg_copy_p): New function.
1269 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
1270 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
1271 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
1272 (make_hard_regno_dead): Don't add conflicts for register
1273 ignore_reg_for_conflicts. Remove special conflict handling of
1274 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
1275 check_pic_pseudo_p and update callers.
1276 (mark_pseudo_dead): Don't add conflicts for register
1277 ignore_reg_for_conflicts.
1278 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
1280 2018-10-05 Andrew Waterman <andrew@sifive.com>
1281 Jim Wilson <jimw@sifive.com>
1283 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
1284 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
1285 new pattern using HONOR_SNANS that emits one extra instruction.
1287 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
1289 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
1290 patterns): Merge SI and DI patterns to a GPR pattern.
1291 (unnamed define_insn and define_split for record form of that): Merge
1292 to a single define_insn_and_split pattern.
1294 2018-10-05 David Malcolm <dmalcolm@redhat.com>
1297 * input.c (expand_location_to_spelling_point): Add param "aspect"
1298 and use rather than hardcoding LOCATION_ASPECT_CARET.
1299 (get_substring_ranges_for_loc): Handle the case of a single token
1300 within a macro expansion.
1301 * input.h (expand_location_to_spelling_point): Add "aspect" param,
1302 defaulting to LOCATION_ASPECT_CARET.
1304 2018-10-05 Paul Koning <ni1d@arrl.net>
1306 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
1307 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
1308 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1309 (pdp11_guard_type): New function.
1311 2018-10-05 Paul Koning <ni1d@arrl.net>
1313 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
1314 * config/pdp11/pdp11.opt (mfloat32): Remove.
1316 * doc/invoke.texi (pdp11 -mfloat32): Remove:
1317 (pdp11 -mfloat64): Remove.
1319 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
1321 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
1322 (*cmp<mode>_cc_i387): Ditto.
1323 (*cmpu<mode>_cc_i387): Ditto.
1324 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1325 * config/i386/i386.c (ix86_expand_fp_compare): Remove
1327 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
1328 Emit x86_sahf_1 pattern.
1329 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
1330 (ix86_expand_carry_flag_compare): Ditto.
1332 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
1334 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
1335 to reg_or_0_operand. Add "C" constraint.
1336 (*cmpxf_cc_i387): Ditto.
1337 (*cmp<mode>_i387): Change operand 2 predicate
1338 to nonimm_or_0_operand. Add "C" constraint.
1339 (*cmp<mode>_cc_i387): Ditto.
1340 (*cmp<mode>_0_i387): Remove insn pattern.
1341 (*cmp<mode>_0_cc_i387): Ditto.
1343 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
1345 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
1346 * config/i386/predicates.md (nonimm_or_0_operand): Rename
1347 from vector_move_operand. Update all uses.
1349 2018-10-05 Martin Sebor <msebor@redhat.com>
1351 PR tree-optimization/87490
1352 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
1355 2018-10-05 Richard Biener <rguenther@suse.de>
1357 PR tree-optimization/63155
1358 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
1359 vertical space in dumpfiles.
1360 * tree-ssa-propagate.h
1361 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
1362 * tree-ssa-propagate.c (cfg_blocks_back): New global.
1363 (ssa_edge_worklist_back): Likewise.
1364 (curr_order): Likewise.
1365 (cfg_blocks_get): Remove abstraction.
1366 (cfg_blocks_add): Likewise.
1367 (cfg_blocks_empty_p): Likewise.
1368 (add_ssa_edge): Add to current or next worklist based on
1370 (add_control_edge): Likewise.
1371 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
1373 (ssa_propagation_engine::ssa_propagate): ... here. Unify
1374 iteration from CFG and SSA edge worklist so we process
1375 everything in RPO order, prioritizing forward progress
1377 (ssa_prop_init): Allocate new worklists, do not dump
1379 (ssa_prop_fini): Free new worklists.
1381 2018-10-05 Richard Biener <rguenther@suse.de>
1383 * tree-core.h (tree_block::abstract_flag): Remove.
1384 (tree_block::block_num): Make full 32bits.
1385 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
1386 * tree.h (BLOCK_ABSTRACT): Remove.
1387 * dwarf2out.c (gen_lexical_block_die): Remove dead code
1388 resulting from BLOCK_ABSTRACT being always false.
1389 (gen_inlined_subroutine_die): Likewise.
1390 (gen_block_die): Likewise.
1391 * tree.c (block_ultimate_origin): Likewise.
1392 * tree-pretty-print.c (dump_block_node): Remove code dealing
1393 with BLOCK_ABSTRACT.
1394 * tree-ssa-live.c (dump_scope_block): Likewise.
1395 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
1396 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
1398 2018-10-05 Richard Biener <rguenther@suse.de>
1400 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
1401 is asked for initialize mode to the component mode of the
1404 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
1407 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
1408 assembler for -mavx.
1409 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
1411 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
1414 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
1416 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
1417 to be DImode. When using mffscrn, force the operand to a register.
1419 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
1421 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
1422 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
1423 X87MODEF mode iterator.
1424 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
1425 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
1426 X87MODEF mode iterator.
1428 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
1430 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
1431 -Wno-prio-ctor-dtor.
1433 2018-10-04 David Malcolm <dmalcolm@redhat.com>
1435 * Makefile.in (OBJS): Add opt-problem.o.
1436 * dump-context.h: Include "selftest.h.
1437 (selftest::temp_dump_context): New forward decl.
1438 (class dump_context): Make friend of class
1439 selftest::temp_dump_context.
1440 (dump_context::dump_loc_immediate): New decl.
1441 (class dump_pretty_printer): Move here from dumpfile.c.
1442 (class temp_dump_context): Move to namespace selftest.
1443 (temp_dump_context::temp_dump_context): Add param
1444 "forcibly_enable_dumping".
1445 (selftest::verify_dumped_text):
1446 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
1447 (selftest::verify_item):
1448 (ASSERT_IS_TEXT): Move here from dumpfile.c.
1449 (ASSERT_IS_TREE): Likewise.
1450 (ASSERT_IS_GIMPLE): Likewise.
1451 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
1453 (dump_context::dump_loc_immediate): ...this new function.
1454 (class dump_pretty_printer): Move to dump-context.h.
1455 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
1456 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
1457 (temp_dump_context::temp_dump_context): Move to "selftest"
1458 namespace. Add param "forcibly_enable_dumping", and use it to
1459 conditionalize the use of m_pp;
1460 (selftest::verify_dumped_text): Make non-static.
1461 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
1462 (selftest::verify_item): Make non-static.
1463 (ASSERT_IS_TEXT): Move to dump-context.h.
1464 (ASSERT_IS_TREE): Likewise.
1465 (ASSERT_IS_GIMPLE): Likewise.
1466 (selftest::test_capture_of_dump_calls): Pass "true" for new
1467 param of temp_dump_context.
1468 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
1469 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
1471 * opt-problem.cc: New file.
1472 * opt-problem.h: New file.
1473 * optinfo-emit-json.cc
1474 (selftest::test_building_json_from_dump_calls): Pass "true" for
1475 new param of temp_dump_context.
1476 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
1477 (optinfo::emit_for_opt_problem): New function.
1478 (optinfo::emit): Clarity which emit_item is used.
1479 * optinfo.h (optinfo::get_dump_location): New accessor.
1480 (optinfo::emit_for_opt_problem): New decl.
1481 (optinfo::emit): Make const.
1482 * selftest-run-tests.c (selftest::run_tests): Call
1483 selftest::opt_problem_cc_tests.
1484 * selftest.h (selftest::opt_problem_cc_tests): New decl.
1485 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
1486 bool to opt_result, converting fprintf messages to
1487 opt_result::failure_at calls. Add "stmt" param for use by the
1489 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
1490 (runtime_alias_check_p): Convert return type from bool to
1491 opt_result, converting dump_printf calls to
1492 opt_result::failure_at, using the statement DDR_A for their
1494 (find_data_references_in_stmt): Convert return type from bool to
1495 opt_result, converting "return false" to opt_result::failure_at
1497 * tree-data-ref.h: Include "opt-problem.h".
1498 (dr_analyze_innermost): Convert return type from bool to opt_result,
1499 and add a const gimple * param.
1500 (find_data_references_in_stmt): Convert return type from bool to
1502 (runtime_alias_check_p): Likewise.
1503 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
1504 dr_analyze_innermost.
1505 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
1506 Convert return type from bool to opt_result, adding a message for
1507 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
1508 (vect_analyze_data_ref_dependence): Convert return type from bool
1509 to opt_result. Change sense of return type from "false"
1510 effectively meaning "no problems" to "false" meaning a problem,
1511 so that "return false" becomes "return opt_result::success".
1512 Convert "return true" calls to opt_result::failure_at, using
1513 the location of statement A rather than vect_location.
1514 (vect_analyze_data_ref_dependences): Convert return type from bool
1516 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
1517 calls to opt_result::failure_at, using the stmt location rather
1519 (vect_verify_datarefs_alignment): Convert return type from bool
1521 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
1522 into multiple more-tightly-scoped copies.
1523 (vect_analyze_data_refs_alignment): Convert return type from bool
1525 (vect_analyze_data_ref_accesses): Likewise, converting a
1526 "return false" to a "return opt_result::failure_at", adding a
1528 (vect_prune_runtime_alias_test_list): Convert return type from
1529 bool to opt_result, converting dump_printf_loc to
1530 opt_result::failure_at. Add a %G to show the pertinent statement,
1531 and use the stmt's location rather than vect_location.
1532 (vect_find_stmt_data_reference): Convert return type from
1533 bool to opt_result, converting dump_printf_loc to
1534 opt_result::failure_at, using stmt's location.
1535 (vect_analyze_data_refs): Convert return type from bool to
1536 opt_result. Convert "return false" to "return
1537 opt_result::failure_at", adding messages as needed.
1538 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
1539 type from bool to opt_result.
1540 (vect_determine_vf_for_stmt): Likewise.
1541 (vect_determine_vectorization_factor): Likewise, converting
1542 dump_printf_loc to opt_result::failure_at, using location of phi
1543 rather than vect_location.
1544 (vect_analyze_loop_form_1): Convert return type from bool to
1545 opt_result, converting dump_printf_loc calls, retaining the use of
1547 (vect_analyze_loop_form): Convert return type from loop_vec_info
1548 to opt_loop_vec_info.
1549 (vect_analyze_loop_operations): Convert return type from bool to
1550 opt_result, converting dump_printf_loc calls, using the location
1551 of phi/stmt rather than vect_location where available. Convert
1552 various "return false" to "return opt_result::failure_at" with
1553 "unsupported phi" messages.
1554 (vect_get_datarefs_in_loop): Convert return type from bool to
1555 opt_result. Add a message for the
1556 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
1557 (vect_analyze_loop_2): Convert return type from bool to
1558 opt_result. Ensure "ok" is set to a opt_result::failure_at before
1559 each "goto again;", adding new messages where needed.
1560 Add "unsupported grouped {store|load}" messages.
1561 (vect_analyze_loop): Convert return type from loop_vec_info to
1563 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
1565 * tree-vect-stmts.c (process_use): Likewise, converting
1566 dump_printf_loc call and using stmt location, rather than
1568 (vect_mark_stmts_to_be_vectorized): Likeise.
1569 (vect_analyze_stmt): Likewise, adding a %G.
1570 (vect_get_vector_types_for_stmt): Convert return type from bool to
1571 opt_result, converting dump_printf_loc calls and using stmt
1572 location, rather than vect_location.
1573 (vect_get_mask_type_for_stmt): Convert return type from tree to
1574 opt_tree, converting dump_printf_loc calls and using stmt location.
1575 * tree-vectorizer.c: Include "opt-problem.h.
1576 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
1577 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
1578 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
1579 enabled, use it to report at the top level "couldn't vectorize
1580 loop" followed by the problem.
1581 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
1582 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
1584 (vect_analyze_stmt): Likewise.
1585 (vect_get_vector_types_for_stmt): Likewise.
1586 (tree vect_get_mask_type_for_stmt): Likewise.
1587 (vect_analyze_data_ref_dependences): Likewise.
1588 (vect_enhance_data_refs_alignment): Likewise.
1589 (vect_analyze_data_refs_alignment): Likewise.
1590 (vect_verify_datarefs_alignment): Likewise.
1591 (vect_analyze_data_ref_accesses): Likewise.
1592 (vect_prune_runtime_alias_test_list): Likewise.
1593 (vect_find_stmt_data_reference): Likewise.
1594 (vect_analyze_data_refs): Likewise.
1595 (vect_analyze_loop): Convert return type from loop_vec_info to
1597 (vect_analyze_loop_form): Likewise.
1598 (vect_analyze_slp): Convert return type from bool to opt_result.
1600 2018-10-04 David Malcolm <dmalcolm@redhat.com>
1602 * doc/invoke.texi (-fopt-info): Document new "internals"
1604 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
1605 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
1607 (optinfo_verbosity_options): Add "internals".
1608 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
1609 (dump_context::apply_dump_filter_p): New member function.
1610 (dump_context::dump_loc): Use apply_dump_filter_p rather than
1611 explicitly masking the dump_kind.
1612 (dump_context::begin_scope): Increment the scope depth first. Use
1613 apply_dump_filter_p rather than explicitly masking the dump_kind.
1614 (dump_context::emit_item): Use apply_dump_filter_p rather than
1615 explicitly masking the dump_kind.
1616 (dump_dec): Likewise.
1617 (dump_hex): Likewise.
1618 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
1619 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
1620 (opt_info_switch_p): Update handling of default
1621 MSG_OPTIMIZED_LOCATIONS to cope with default of
1622 MSG_PRIORITY_USER_FACING.
1623 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
1624 masking the dump_kind.
1625 (selftest::test_capture_of_dump_calls): Update test_dump_context
1626 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
1627 than MSG_ALL. Generalize scope test to be run at all four
1628 combinations of with/without MSG_PRIORITY_USER_FACING and
1629 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
1630 for each of the two values.
1631 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
1632 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
1633 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
1634 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
1635 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
1636 with MSG_PRIORITY_*.
1637 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
1638 dump messages as MSG_PRIORITY_USER_FACING.
1639 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
1640 about the interaction with MSG_PRIORITY_*.
1642 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1644 * varasm.c (output_constant): Add new parameter merge_strings.
1645 Make strings properly zero terminated in merge string sections.
1646 (mergeable_string_section): Don't fail if the last char is non-zero.
1647 (assemble_variable_contents): Handle merge string sections.
1648 (assemble_variable): Likewise.
1649 (assemble_constant_contents): Likewise.
1650 (output_constant_def_contents): Likewise.
1651 (output_constructor_array_range,
1652 output_constructor_regular_field): Adjust call to output_constant.
1653 (output_object_block): Adjust call to assemble_constant_contents
1654 and assemble_variable_contents.
1656 2018-10-04 Martin Liska <mliska@suse.cz>
1659 * cgraphunit.c (process_function_and_variable_attributes):
1660 Warn about a function with alias attribute and a body.
1662 2018-10-04 Martin Liska <mliska@suse.cz>
1665 * multiple_target.c (redirect_to_specific_clone): New function.
1666 (ipa_target_clone): Use it.
1667 * tree-inline.c: Fix comment.
1669 2018-10-04 David Malcolm <dmalcolm@redhat.com>
1671 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
1673 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
1674 (gcc::dump_manager::register_pass): New member function, adapted
1675 from loop body in gcc::pass_manager::register_pass, adding a
1676 call to update_dfi_for_opt_info.
1677 (gcc::dump_manager::opt_info_enable_passes): Store the
1678 -fopt-info options into the new fields. Move the loop
1680 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
1682 * dumpfile.h (struct opt_pass): New forward decl.
1683 (gcc::dump_manager::register_pass): New decl.
1684 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
1685 (class gcc::dump_manager): Add fields "m_optgroup_flags",
1686 "m_optinfo_flags", and "m_optinfo_filename".
1687 * passes.c (gcc::pass_manager::register_pass): Move all of the
1688 dump-handling code to gcc::dump_manager::register_pass.
1690 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
1692 PR rtl-optimization/87466
1693 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
1694 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
1695 * doc/tm.texi: Regenerate.
1696 * ira-lives.c (process_bb_node_lives): Use the new target hook.
1697 * lra-lives.c (process_bb_lives): Likewise.
1698 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
1701 2018-10-04 Tamar Christina <tamar.christina@arm.com>
1703 * params.c (add_params): Fix initialization.
1705 2018-10-04 Martin Liska <mliska@suse.cz>
1707 PR gcov-profile/84107
1708 * tree-profile.c (init_ic_make_global_vars):
1709 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
1710 Come up with new ic_tuple* variables. Emit
1711 __gcov_indirect_call{,_topn} variables.
1712 (gimple_gen_ic_profiler): Access the variable
1714 (gimple_gen_ic_func_profiler): Access
1715 __gcov_indirect_call.callee field.
1716 (gimple_init_gcov_profiler): Use ptr_type_node.
1717 * value-prof.c (gimple_ic): Use ptr_type_node.
1719 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1721 PR tree-optimization/85787
1722 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
1723 into this function and add support for detecting multiple phis.
1724 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
1726 2018-10-04 Martin Liska <mliska@suse.cz>
1729 * ipa-inline.c (inline_to_all_callers_1):
1730 Call ultimate_alias_target for node being inlined.
1732 2018-10-03 Jeff Law <law@redhat.com>
1734 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
1736 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
1737 * tree.h (get_typenode_from_name): Prototype.
1739 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
1741 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
1742 Change operand 2 predicate to nonimmediate_operand.
1743 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1745 2018-10-03 Martin Sebor <msebor@redhat.com>
1746 Jeff Law <law@redhat.com>
1748 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
1750 (get_string_length): Detect unterminated arrays.
1751 (format_string): Same.
1752 (format_directive): Warn about unterminated arrays.
1753 (handle_gimple_call): Mark statements with no_warning as needed.
1755 2018-10-03 Jim Wilson <jimw@sifive.com>
1757 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
1758 also define __riscv_abi_rve. Delete trailing white space.
1760 2018-10-03 Paul Koning <ni1d@arrl.net>
1762 Enable LRA register allocator for PDP11.
1763 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
1766 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
1767 * config/pdp11/pdp11.opt (-mlra): New option.
1768 * doc/invoke.texi (PDP-11 Options): Document -mlra.
1770 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
1772 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
1773 (*<absneg:code>extend<mode>xf2): Ditto.
1775 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
1777 PR tree-optimization/87415
1778 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
1781 2018-10-02 Jeff Law <law@redhat.com>
1783 * gimple-fold.c (get_range_strlen): Only set *nonstr when
1784 an unterminated string is discovered. Bubble up range
1785 even for unterminated strings.
1786 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
1787 indicates the string was not terminated via NONSTR.
1789 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
1791 * tree-vrp.c (extract_range_from_unary_expr): Special case all
1792 pointer conversions.
1793 Do not do anything special for anti-ranges.
1795 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
1797 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
1798 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
1800 2018-10-03 Martin Liska <mliska@suse.cz>
1802 PR gcov-profile/86109
1803 * coverage.c (coverage_begin_function): Do not
1804 mark lambdas as artificial.
1805 * tree-core.h (struct GTY): Remove tm_clone_flag
1806 and introduce new lambda_function.
1807 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
1809 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1812 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
1813 P8_VECTOR and VSX are enabled.
1815 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
1817 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
1818 0x3907 as CPU model number.
1820 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
1822 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
1823 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
1824 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
1825 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
1826 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
1827 * config/s390/s390.md: Likewise. Rename also the cpu attribute
1828 value from arch12 to z14.
1830 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
1832 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
1834 (isinf<mode>2): Ditto.
1836 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
1838 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
1839 before emitting fxam. Perform calculations in XFmode.
1841 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
1843 * match.pd (((X /[ex] A) +- B) * A): New transformation.
1845 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
1848 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
1849 * tree.c (signed_or_unsigned_type_for): Handle complex.
1851 2018-10-02 Jeff Law <law@redhat.com>
1853 * gimple-fold.c (get_range_strlen): Remove dead code.
1855 2018-10-02 Martin Sebor <msebor@redhat.com>
1856 Jeff Law <law@redhat.com>
1858 * builtins.c (unterminated_array): Add new arguments.
1859 If argument is not terminated, bubble up size and exact
1861 (expand_builtin_strnlen): Detect, avoid expanding
1862 and diagnose unterminated arrays.
1863 (c_strlen): Fill in offset of start of unterminated strings.
1864 * builtins.h (unterminated_array): Update prototype.
1866 2018-10-02 Richard Biener <rguenther@suse.de>
1868 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
1869 of haddv4df, first reduce to SSE width and exploit the fact
1870 that we only need element zero with the reduction result.
1871 (reduc_plus_scal_v2df): Likewise.
1873 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
1875 * dojump.h (do_jump): Delete.
1876 (do_jump_1): Likewise.
1877 (split_comparison): Move around.
1878 * dojump.c (do_jump): Make static.
1879 (do_jump_1): Likewise.
1880 (jumpifnot): Move around.
1881 (jumpifnot_1): Likewise.
1883 (jumpif_1): Likewise.
1884 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
1886 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
1888 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
1889 insns in the delay slot and add_insn_after for the jump insn.
1891 2018-10-02 Richard Biener <rguenther@suse.de>
1893 * tree-inline.c (expand_call_inline): Use the location of
1894 the callee declaration for the inline-entry marker.
1895 * final.c (notice_source_line): Remove special-casing of
1896 NOTE_INSN_INLINE_ENTRY.
1898 2018-10-01 Carl Love <cel@us.ibm.com>
1901 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
1902 (__builtin_mtfsb0): New.
1903 (__builtin_mtfsb1): New.
1904 ( __builtin_set_fpscr_rn): New.
1905 (__builtin_set_fpscr_drn): New.
1906 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
1907 (rs6000_expand_set_fpscr_rn_builtin): Add.
1908 (rs6000_expand_set_fpscr_drn_builtin): Add.
1909 (rs6000_expand_builtin): Add case statement entries for
1910 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
1911 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
1912 RS6000_BUILTIN_MFFSL.
1913 (rs6000_init_builtins): Add ftype initialization and def_builtin
1914 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1915 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1916 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1917 rs6000_mffscdrn): Add define_insn.
1918 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1919 * doc/extend.texi: Add documentation for the builtins.
1921 2018-10-01 Richard Biener <rguenther@suse.de>
1923 PR tree-optimization/87465
1924 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1925 causing branch miscounts.
1927 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1929 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1930 aarch64_option_default_param): New.
1931 (params.h): Include.
1932 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1933 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1934 stack-clash protection validation code.
1936 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1938 * params.c (validate_param): New.
1939 (add_params): Use it.
1940 (set_param_value): Refactor param validation into validate_param.
1941 (diagnostic.h): Include.
1942 * diagnostic.h (diagnostic_ready_p): New.
1944 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1946 * params.c (set_param_value):
1947 Add index of parameter being validated.
1948 * common/common-target.def (option_validate_param): New.
1949 * common/common-targhooks.h (default_option_validate_param): New.
1950 * common/common-targhooks.c (default_option_validate_param): New.
1951 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1952 * doc/tm.texi: Regenerate.
1954 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1957 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1958 Add validation for stack-clash parameters and set defaults.
1960 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1963 * configure.ac: Add stack-clash-protection-guard-size.
1964 * doc/install.texi: Document it.
1965 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1966 * params.def: Update comment for guard-size.
1967 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1968 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1969 * configure: Regenerate.
1971 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1974 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1975 STACK_DYNAMIC_OFFSET): New.
1976 * config/aarch64/aarch64.c (aarch64_layout_frame):
1977 Update outgoing args size.
1978 (aarch64_stack_clash_protection_alloca_probe_range,
1979 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1981 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1984 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1986 * target.def (stack_clash_protection_alloca_probe_range): New.
1987 (stack_clash_protection_final_dynamic_probe): Remove.
1988 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1989 (default_stack_clash_protection_final_dynamic_probe): Remove.
1990 * targhooks.c: Likewise.
1991 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1992 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1993 * doc/tm.texi: Regenerate.
1995 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1998 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1999 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
2000 aarch64_clamp_to_uimm12_shift): New.
2001 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
2002 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
2004 2018-10-01 Tamar Christina <tamar.christina@arm.com>
2007 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
2009 2018-10-01 Jeff Law <law@redhat.com>
2010 Richard Sandiford <richard.sandiford@linaro.org>
2011 Tamar Christina <tamar.christina@arm.com>
2014 * config/aarch64/aarch64.md
2015 (probe_stack_range): Add k (SP) constraint.
2016 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
2017 STACK_CLASH_MAX_UNROLL_PAGES): New.
2018 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
2019 stack probes for stack clash.
2020 (aarch64_allocate_and_probe_stack_space): New.
2021 (aarch64_expand_prologue): Use it.
2022 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
2023 (aarch64_sub_sp): Add emit_move_imm optional param.
2025 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
2027 PR tree-optimization/87261
2028 * match.pd: Remove trailing whitespace.
2029 Add (x & y) | ~(x | y) -> ~(x ^ y),
2030 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
2032 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
2034 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
2036 (ashlsi3_insn): Update instruction constraints.
2037 (ashrsi3_insn): Likewise.
2038 (rotrsi3): Likewise.
2039 (add_shift): Likewise.
2040 * config/arc/constraints.md (Csz): New 32 bit constraint. It
2041 avoids placing in the limm field small constants which, otherwise,
2042 could end into a small instruction.
2044 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
2046 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
2047 destination register is not odd-even.
2048 (umaddsidi4_split): Likewise.
2050 2018-10-01 Richard Biener <rguenther@suse.de>
2052 * tree-inline.c (expand_call_inline): Store origin of fn
2053 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
2054 * tree.c (block_ultimate_origin): Simplify and do some
2057 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
2059 * config/i386/mmx.md (EMMS): New int iterator.
2060 (emms): New int attribute.
2061 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
2062 EMMS int iterator. Explicitly declare clobbers.
2063 (mmx_emms): Remove expander.
2065 * config/i386/predicates.md (emms_operation): Remove predicate.
2066 (vzeroall_pattern): New predicate.
2067 (vzeroupper_pattern): Rename from vzeroupper_operation.
2068 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
2069 vzeroupper_pattern and vzeroall_pattern predicates.
2071 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
2073 PR rtl-optimization/86939
2074 * ira-lives.c (make_hard_regno_born): Rename from this...
2075 (make_hard_regno_live): ... to this. Remove update to conflict
2076 information. Update function comment.
2077 (make_hard_regno_dead): Add conflict information update. Update
2079 (make_object_born): Rename from this...
2080 (make_object_live): ... to this. Remove update to conflict information.
2081 Update function comment.
2082 (make_object_dead): Add conflict information update. Update function
2084 (mark_pseudo_regno_live): Call make_object_live.
2085 (mark_pseudo_regno_subword_live): Likewise.
2086 (mark_hard_reg_dead): Update function comment.
2087 (mark_hard_reg_live): Call make_hard_regno_live.
2088 (process_bb_node_lives): Likewise.
2089 * lra-lives.c (make_hard_regno_born): Rename from this...
2090 (make_hard_regno_live): ... to this. Remove update to conflict
2091 information. Remove now uneeded check_pic_pseudo_p argument.
2092 Update function comment.
2093 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
2094 to conflict information. Update function comment.
2095 (mark_pseudo_live): Remove update to conflict information. Update
2097 (mark_pseudo_dead): Add conflict information update.
2098 (mark_regno_live): Call make_hard_regno_live.
2099 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
2100 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
2102 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
2105 * config/i386/i386.c (construct_container): Use TImode for
2106 BLKmode values in 2 integer registers.
2108 2018-09-29 Jeff Law <law@redhat.com>
2110 * builtins.c (unterminated_array): Pass in c_strlen_data * to
2111 c_strlen rather than just a tree *.
2112 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2113 Update recursive calls appropriately. If caller did not provide a
2114 suitable data pointer, create a local one. When a non-terminated
2115 string is discovered, bubble up information about the string via the
2116 c_strlen_data object.
2117 * builtins.h (c_strlen): Update prototype.
2118 (c_strlen_data): New structure.
2119 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2120 For a type 2 call, if c_strlen indicates a non-terminated string
2121 use the length of the non-terminated string.
2122 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2124 2018-09-29 Jakub Jelinek <jakub@redhat.com>
2127 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2128 __m512d type for __A argument rather than __m512.
2130 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
2132 * match.pd (simple_comparison): Don't optimize if either operand is
2133 a function pointer when target needs function pointer canonicalization.
2135 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
2137 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2138 power5 .. power9 to remove indirection.
2139 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2140 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2141 ASM_CPU_476_SPEC): Delete.
2142 (ASM_CPU_SPEC): Adjust.
2143 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2144 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2146 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
2148 * config.in: Delete HAVE_AS_DCI.
2149 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2150 * config/rs6000/rs6000.h: Ditto.
2151 * configure.ac: Delete HAVE_AS_DCI.
2152 * configure: Regenerate.
2154 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
2156 * config.in (HAVE_AS_LWSYNC): Delete.
2157 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2158 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
2160 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2161 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
2163 * configure.ac: Delete HAVE_AS_LWSYNC.
2164 * configure: Regenerate.
2166 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
2167 Pierre-Marie de Rodat <derodat@adacore.com>
2169 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2170 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2171 (cgraph_node::create_thunk): Add indirect_offset parameter.
2172 (thunk_adjust): Likewise.
2173 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2174 and initialize the corresponding field with it.
2175 (cgraph_node::dump): Dump indirect_offset field.
2176 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2177 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2178 (thunk_adjust): Add indirect_offset parameter and deal with it.
2179 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2180 pass it to thunk_adjust. Do not call the target hook if it's non-zero
2181 or if the thunk is external or local. Fix formatting. Do not chain
2182 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
2183 if any, in the GIMPLE representation.
2184 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2185 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2186 (input_node): Read indirect_offset field.
2187 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2188 call to thunk_adjust.
2189 * tree-nested.c (struct nesting_info): Add thunk_p field.
2190 (create_nesting_tree): Set it.
2191 (convert_all_function_calls): Copy static chain from targets to thunks.
2192 (finalize_nesting_tree_1): Return early for thunks.
2193 (unnest_nesting_tree_1): Do not finalize thunks.
2194 (gimplify_all_functions): Do not gimplify thunks.
2196 2018-09-28 David Malcolm <dmalcolm@redhat.com>
2198 * opt-suggestions.c (option_proposer::build_option_suggestions):
2199 Release "option_values".
2201 2018-09-28 David Malcolm <dmalcolm@redhat.com>
2203 * coverage.c (get_coverage_counts): Convert problem-reporting dump
2204 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2205 * dumpfile.c (kind_as_string): New function.
2206 (dump_loc): Rather than a hardcoded prefix of "note: ", use
2207 kind_as_string to vary the prefix based on dump_kind.
2208 (selftest::test_capture_of_dump_calls): Update for above.
2210 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2212 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2213 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
2215 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2217 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
2218 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
2219 INVALID_REGNUM instead of FPSR_REG.
2220 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
2221 * config/i386/i386.md: Update comment of FP compares.
2222 (fldenv): Do not clobber FPSR_REG.
2224 2018-09-28 Richard Biener <rguenther@suse.de>
2226 * tree.h (BLOCK_ORIGIN): New.
2227 * omp-expand.c (grid_expand_target_grid_body): Assign
2228 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
2229 * tree-inline.c (remap_block): Likewise.
2230 * auto-profile.c (get_function_decl_from_block): Simplify
2231 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
2232 * langhooks.c (lhd_print_error_function): Likewise.
2233 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
2235 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2236 * tree.c (block_nonartificial_location): Likewise.
2237 (block_ultimate_origin): Likewise.
2238 * tree-pretty-print.c (percent_K_format): Likewise. Remove
2239 no longer needed LTO case.
2241 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
2242 Jan Hubicka <jh@suse.cz>
2243 Martin Jambor <mjambor@suse.cz>
2245 * simplify-rtx.c (simplify_merge_mask): New function.
2246 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
2247 same masks are used in op1 or op2.
2248 (test_vec_merge): New function.
2249 (test_vector_ops): Call test_vec_merge.
2251 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
2253 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
2254 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
2255 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
2256 (cypress_costs): Set it.
2257 (supersparc_costs): Likewise.
2258 (hypersparc_costs): Likewise.
2259 (leon_cost): Likewise.
2260 (leon3_costs): Likewise.
2261 (sparclet_costs): Likewise.
2262 (ultrasparc_costs): Likewise.
2263 (ultrasparc_costs): Likewise.
2264 (niagara_costs): Likewise.
2265 (niagara2_costs): Likewise.
2266 (niagara3_costs): Likewise.
2267 (niagara4_costs): Likewise.
2268 (niagara7_costs): Likewise.
2269 (m8_costs): Likewise.
2270 (TARGET_CAN_FOLLOW_JUMP): Define.
2271 (pass_work_around_errata::gate): Minor tweak.
2272 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
2273 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
2274 Automaitcally clear MASK_FSMULD mask for V7 processors.
2275 (sparc_can_follow_jump): New static function.
2276 (output_ubranch): Deal with CROSSING_JUMP_P.
2277 (sparc_use_sched_lookahead): Rewrite using switch statement.
2278 (sparc_issue_rate): Reorder.
2279 (sparc_branch_cost): New function.
2281 2018-09-27 Martin Sebor <msebor@redhat.com>
2283 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
2284 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
2285 (int_fits_type_p): Same.
2287 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
2289 * config/i386/i386.md (FPCR_REG): Remove.
2290 (UNSPEC_FLDCW): Remove.
2291 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
2292 (x86_fldcw_1): Remove insn pattern.
2293 (fnstenv): Do not clobber FPCR_REG.
2295 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
2296 (CALL_USED_REGISTERS): Ditto.
2297 (REG_ALLOC_ORDER): Ditto.
2298 (REG_CLASS_CONTENTS): Ditto.
2299 (HI_REGISTER_NAMES): Ditto.
2300 (ADDITIONAL_REGISTER_NAMES): Use defines instead
2301 of numerical constants.
2302 * config/i386/i386.c (regclass_map): Remove fpsr register.
2303 (dbx_register_map): Ditto.
2304 (dbx64_register_map): Ditto.
2305 (svr4_dbx_register_map): Ditto.
2306 (print_reg): Do not handle FPCR_REG.
2308 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
2311 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
2312 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
2313 Delete, always treat as true.
2314 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2315 Ditto. Simplify remaining code.
2316 * config/powerpcspe/powerpcspe.h: Ditto.
2317 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
2318 Simplify remaining code.
2319 (rs6000_expand_builtin): Ditto.
2320 * config/rs6000/rs6000.h: Ditto.
2321 * configure.ac: Ditto.
2322 * configure: Regenerate.
2324 2018-09-27 Martin Liska <mliska@suse.cz>
2326 * coverage.c (get_coverage_counts): Revert the formatting
2327 of missing profile opt info.
2329 2018-09-27 Richard Biener <rguenther@suse.de>
2333 * dwarf2out.c (set_block_origin_self): Do not mark outermost
2334 block as we do not output that.
2335 (gen_inlined_subroutine_die): Elide the originally outermost
2336 block, matching what we do for concrete instances.
2337 (decls_for_scope): Add parameter specifying whether to recurse
2340 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
2341 Tom de Vries <tom@codesourcery.com>
2345 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
2346 STORE_FLAG_VALUE == 1.
2348 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
2350 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
2351 constant definitions.
2352 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
2353 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
2354 ("speculation_barrier"): New expander definition.
2356 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
2358 PR gcov-profile/86957
2359 * common.opt: New warning option -Wmissing-profile.
2360 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
2361 * doc/invoke.texi: Document -Wmissing-profile.
2363 2018-09-26 Jim Wilson <jimw@sifive.com>
2365 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
2366 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
2368 2018-09-26 Martin Sebor <msebor@redhat.com>
2370 * tree.c (zerop): Change return type to bool.
2371 (integer_zerop, integer_onep, integer_each_onep): Same.
2372 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2373 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2374 (real_onep, real_minus_onep, chain_index): Same.
2375 (print_type_hash_statistics, type_list_equal): Same.
2376 * tree.h (zerop): Same.
2377 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
2378 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2379 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2380 (real_onep, real_minus_onep, chain_index): Same.
2381 (print_type_hash_statistics, type_list_equal): Same.
2383 2018-09-26 Jim Wilson <jimw@sifive.com>
2385 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
2387 2018-09-26 Jakub Jelinek <jakub@redhat.com>
2390 * config/i386/i386.c: Include debug.h and dwarf2out.h.
2391 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2394 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
2396 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
2398 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
2400 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
2401 and FP_SECOND_SSE_REGS.
2402 (REG_CLASS_NAMES): Ditto.
2403 (REG_CLASS_CONTENTS): Ditto.
2404 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
2405 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
2406 (ix86_preferred_output_reload_class): Ditto.
2407 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
2408 clobber constraint to "=&f".
2409 (fix_truncdi_i387): Ditto.
2410 (lrintxfdi2): Ditto.
2411 (fistdi2_<rounding>): Ditto.
2412 (fpremxf4_i387): Change "=u" constraint to "=f".
2413 (fprem1xf4_i387): Ditto.
2415 (fptanxf4_i387): Ditto.
2416 (fxtractxf3_i387): Ditto.
2417 (fscalexf4_i387): Ditto.
2418 (atan2xf3): Change "u" constraint to "f".
2419 (fyl2xxf3_i387): Ditto.
2420 (fyl2xp1xf3_i387): Ditto.
2422 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
2425 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
2426 for removed I387_MASK_PM entity.
2429 2018-09-26 Jeff Law <law@redhat.com>
2431 2018-09-26 Alexey Neyman <stilor@att.net>
2433 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2434 headers are no longer pulled in by <isl/val.h>.
2436 2018-09-26 Richard Biener <rguenther@suse.de>
2439 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
2440 or concrete instance DIE to the tree. Create abstract origin
2441 attributes also for concrete instances.
2443 2018-09-26 Alexey Neyman <stilor@att.net>
2445 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2446 headers are no longer pulled in by <isl/val.h>.
2448 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
2450 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
2451 Use new helper functions.
2452 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
2453 Use new helper functions.
2454 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
2455 aarch_mm_needs_release): New declarations.
2456 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
2457 aarch_mm_needs_release): New.
2459 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
2461 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
2462 (arm32_output_mi_thunk): Deal with long calls.
2464 2018-09-26 Richard Biener <rguenther@suse.de>
2468 * tree-inline.c (expand_call_inline): When the location
2469 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2470 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2471 the inserted BLOCK to make inlined_function_outer_scope_p
2473 * dwarf2out.c (add_call_src_coords_attributes): Do not add
2474 coords for reserved locations.
2476 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
2478 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
2479 (*call_indirect_nonlocal_sysv<mode>): Ditto.
2480 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
2481 (*sibcall_nonlocal_sysv<mode>): Ditto.
2482 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
2483 (<bd>_<mode>): Ditto.
2484 (<bd>tf_<mode>): Ditto.
2486 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
2488 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
2489 control string as a list of templates instead of as C code.
2490 (*altivec_movti): Ditto.
2491 * config/rs6000/darwin.md (movdf_low_di): Ditto.
2493 2018-09-25 Jim Wilson <jimw@sifive.com>
2495 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
2496 when target symbol is weak.
2498 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
2501 * builtins.c (unterminated_array): Simplify.
2502 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
2503 where pointer arithmetic is safe.
2505 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
2508 * config/rs6000/altivec.md (altivec_vspltb): Use
2509 const_0_to_15_operand instead of u5bit_cint_operand.
2510 (*altivec_vspltb_internal): Ditto.
2511 (altivec_vspltb_direct): Ditto.
2512 (altivec_vsplth): Use const_0_to_7_operand instead of
2514 (*altivec_vsplth_internal): Ditto.
2515 (altivec_vsplth_direct): Ditto.
2516 (altivec_vspltw): Use const_0_to_3_operand instead of
2518 (*altivec_vspltw_internal): Ditto.
2519 (altivec_vspltw_direct): Ditto.
2520 (altivec_vspltsf): Ditto.
2521 (*altivec_vspltsf_internal): Ditto.
2522 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
2523 various splats with the proper size immediate. Reorder the various
2524 cases by ascending size of immediate, and put all such together.
2526 2018-09-25 Richard Biener <rguenther@suse.de>
2529 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
2531 (maybe_create_die_with_external_ref): Do not create
2532 DW_TAG_imported_unit here.
2533 (add_abstract_origin_attribute): Handle external BLOCK refs.
2534 (dwarf2out_abstract_function): Simplify LTO case.
2535 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
2536 rather than using maybe_create_die_with_external_ref.
2538 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
2541 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
2542 (frndintxf2_mask_pm_i387): Ditto.
2543 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
2544 Enable for !flag_trapping_math.
2545 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
2546 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
2547 Change operand 1 predicate to nonimmediate_operand.
2548 (attr "i387_cw"): Remove mask_pm.
2549 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
2550 (enum ix86_entity): Remove I387_MASK_PM.
2551 * config/i386/i386.c (ix86_i387_mode_needed): Do not
2552 handle I387_MASK_PM.
2553 (ix86_mode_needed): Ditto.
2554 (ix86_mode_after): Ditto.
2555 (ix86_mode_entry): Ditto.
2556 (ix86_mode_exit): Ditto.
2557 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
2559 2018-09-25 Jakub Jelinek <jakub@redhat.com>
2561 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
2562 to_update_switch_stmts to vNULL instead of calling create on them
2565 2018-09-25 Richard Biener <rguenther@suse.de>
2567 PR tree-optimization/87402
2568 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
2569 (visit_phi): Re-instantiate handling of supposed to be VARYING
2570 but non-VARYING backedge value.
2572 2018-09-25 Richard Biener <rguenther@suse.de>
2575 * dwarf2out.c (struct sym_off_pair): New.
2576 (external_die_map): New global.
2577 (lookup_decl_die): When in LTO create DIEs lazily from the
2579 (lookup_block_die): New function, create DIEs lazily in LTO.
2580 (equate_block_to_die): New function.
2581 (dwarf2out_die_ref_for_decl): During WPA get the association
2582 from the external DIE map.
2583 (dwarf2out_register_external_die): Record mapping into the
2585 (maybe_create_die_with_external_ref): New function split out from
2586 DIE generation part of old dwarf2out_register_external_die.
2587 (add_abstract_origin_attribute): Do not return the DIE. When
2588 in LTO reference externals directly.
2589 (dwarf2out_abstract_function): When in LTO ignore calls for
2590 decls with external DIEs (already present abstract instances).
2591 (gen_call_site_die): Adjust.
2592 (add_high_low_attributes): Likewise.
2593 (gen_lexical_block_die): Likewise.
2594 (gen_inlined_subroutine_die): Likewie.
2595 (gen_block_die): Likewise.
2596 (dwarf2out_inline_entry): Likewise.
2597 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
2600 2018-09-25 Martin Liska <mliska@suse.cz>
2602 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
2603 integers and not by a float value.
2605 2018-09-25 Martin Liska <mliska@suse.cz>
2608 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
2610 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
2611 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
2612 * godump.c (go_format_type): Remove extra parenthesis.
2614 2018-09-25 Martin Liska <mliska@suse.cz>
2616 * alias.c (set_dest_equal_p): Remove unused function.
2617 * config/i386/i386.c (def_builtin_pure2): Likewise.
2618 * diagnostic-show-locus.c (class layout): Remove
2620 (layout::layout): Likewise here.
2621 * dump-context.h (class temp_dump_context): Likewise.
2622 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
2623 (add_AT_loclistsptr): Likewise.
2624 (add_AT_offset): Likewise.
2625 (get_AT_hi_pc): Likewise.
2626 (is_comdat_die): Likewise.
2627 (type_is_enum): Likewise.
2628 (ceiling): Likewise.
2629 (add_AT_vms_delta): Likewise.
2630 (is_class_die): Likewise.
2631 * edit-context.c (class line_event): Remove unused field.
2632 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
2634 * ipa-cp.c (ipa_get_vr_lat): Likewise.
2635 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
2636 (ok_for_base_p_nonstrict): Likewise.
2637 * tree-chrec.c (is_not_constant_evolution): Likewise.
2638 (chrec_fold_poly_cst): Likewise.
2639 * tree-if-conv.c (has_pred_critical_p): Likewise.
2640 * tree-ssa-coalesce.c (print_exprs): Likewise.
2641 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
2642 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
2643 * tree-vrp.c (value_ranges_intersect_p): Likewise.
2644 (value_range_nonnegative_p): Likewise.
2646 2018-09-25 Martin Liska <mliska@suse.cz>
2648 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2649 Do not handle "GNU Pascal".
2650 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2652 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
2653 from documentation. Likewise.
2654 * dbxout.c (dbxout_range_type): Likewise.
2655 * doc/cpp.texi: Likewise.
2656 * doc/extend.texi: Likewise.
2657 * doc/frontends.texi: Likewise.
2658 * doc/invoke.texi: Remove Pascal entry.
2659 * tree.def (CLEANUP_POINT_EXPR): Likewise.
2660 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
2662 2018-09-25 Martin Liska <mliska@suse.cz>
2665 * doc/invoke.texi: Document all parameters and remove default
2668 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
2671 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
2672 contains HOST_WIDE_INTs when computing its size.
2674 2018-09-24 Jim Wilson <jimw@sifive.com>
2677 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
2679 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
2681 2018-09-24 Andrew Pinski <apinski@marvell.com>
2683 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2684 access prev before checking it for NULLness in the
2685 AARCH64_FUSE_CMP_BRANCH case.
2687 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
2690 * config/i386/i386.c (rest_of_insert_endbranch): Set
2691 endbr_queued_at_entrance to true and don't insert ENDBR if
2692 x86_function_profiler will be called.
2693 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2695 * config/i386/i386.h (machine_function): Add
2696 endbr_queued_at_entrance.
2698 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
2700 * genattrtab.c (mk_attr_alt): Use alternative_mask.
2701 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
2703 (check_attr_test): Use alternative_mask.
2704 (get_attr_value): Likewise.
2705 (compute_alternative_mask): Use alternative_mask and XWINT.
2706 (make_alternative_compare): Use alternative_mask.
2707 (attr_alt_subset_p): Use XWINT.
2708 (attr_alt_subset_of_compl_p): Likewise.
2709 (attr_alt_intersection): Use alternative_mask and XWINT.
2710 (attr_alt_union): Likewise.
2711 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
2712 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
2713 (simplify_test_exp): Use alternative_mask and XWINT.
2714 (write_test_expr): Use alternative_mask and XWINT, adjust bit
2715 number calculation to support 64 bits. Generate code that
2716 checks 64-bit masks.
2717 (main): Use alternative_mask.
2718 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
2720 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
2723 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
2724 RETURN+USE when returning via %r14.
2726 2018-09-24 Martin Liska <mliska@suse.cz>
2728 * gcov.c (output_lines): Print colorization legend
2729 for both flag_use_colors and flag_use_hotness_colors.
2732 2018-09-24 Martin Liska <mliska@suse.cz>
2734 * coverage.c (get_coverage_counts): Use warning_at
2735 with current_function_decl location. Use %qD in warning
2738 2018-09-24 Martin Liska <mliska@suse.cz>
2740 * memory-block.h (memory_block_pool::release): Annotate with
2741 valgrind that the memory is not accessible.
2743 2018-09-24 Martin Liska <mliska@suse.cz>
2746 * asan.c: Make asan_handled_variables extern.
2748 * cfgexpand.c (expand_stack_vars): Make sure
2749 a representative is unpoison if another
2750 variable in the partition is handled by
2751 use-after-scope sanitization.
2753 2018-09-24 Richard Biener <rguenther@suse.de>
2755 PR tree-optimization/63155
2756 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2757 the worklist when the edge of the respective argument isn't
2760 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
2762 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
2763 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
2764 (MASK_CLASS_P): Update for rename.
2765 (MAYBE_MASK_CLASS_P): Ditto.
2766 (REG_CLASS_NAMES): Update.
2767 (REG_CLASS_CONTENT): Update.
2768 * config/i386/i386.c (regclass_map): Update for MASK_REG
2769 and ALL_MASK_REGS rename.
2770 * config/i386/constraints.md (Yk): Update for rename.
2773 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
2775 * config/i386/i386.h (enum reg_class): Remove
2776 EVEX_SSE_REGS and MOD4_SSE_REGS.
2777 (REG_CLASS_NAMES): Update.
2778 (REG_CLASS_CONTENT): Update.
2779 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
2780 registers as ALL_SSE_REGS.
2781 (ix86_additional_allocno_class_p): Remove.
2782 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
2783 (ix86_register_priority): Lower priority of EVEX SSE registers.
2784 Use IN_RANGE macro where appropriate.
2785 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
2786 AVX-5124VNNIW checks.
2787 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
2788 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
2789 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
2790 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
2791 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
2792 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
2793 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
2794 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
2795 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
2796 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
2797 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
2798 * config/i386/constraints.md (Yh): Remove.
2800 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
2802 * config/i386/i386.c (regclass_map): Declare integer REX registers
2805 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
2807 * doc/service.texi (Service): Switch the fsf.org link to https.
2809 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
2812 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2813 Define to speculation_safe_value_not_needed.
2815 2018-09-21 Florian Weimer <fweimer@redhat.com>
2818 * doc/extend.texi (Common Function Attributes): Mention that
2819 noreturn suppresses tail call optimization.
2821 2018-09-21 Jeff Law <law@redhat.com>
2823 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
2824 vr_values::cleanup_edges_and_switches.
2825 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
2827 (identify_jump_threads): Remove EDGE_IGNORE handling.
2828 (execute_vrp): Move handling of to_remove_edges and
2829 to_update_switch_stmts into vr_values class member functions.
2830 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
2831 (to_update_switch_stmts): Likewise.
2832 * vr-values.c: Include cfghooks.h.
2833 (vr_values::vr_values): Initialize to_remove_edges and
2834 to_update_switch_stmts.
2835 (vr_values::~vr_values): Verify to_remove_edges and
2836 to_update_switch_stmts are empty.
2837 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
2838 (vr_values::cleanup_edges_and_switches): New member function.
2839 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
2840 function. Add new data members.
2842 2018-09-21 David Malcolm <dmalcolm@redhat.com>
2844 PR tree-optimization/87309
2845 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
2846 calls with pflags and alt_flags.
2847 (selftest::test_capture_of_dump_calls): Add test of interaction of
2848 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
2850 2018-09-21 Olivier Hainque <hainque@adacore.com>
2852 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
2854 2018-09-21 Olivier Hainque <hainque@adacore.com>
2856 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
2858 2018-09-21 Olivier Hainque <hainque@adacore.com>
2860 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
2861 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
2862 (VXWORKS_LIBS_RTP): Minor reordering.
2864 2018-09-21 Olivier Hainque <hainque@adacore.com>
2866 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
2867 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
2869 2018-09-21 Olivier Hainque <hainque@adacore.com>
2871 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
2872 (PTRDIFF_TYPE): Likewise.
2874 2018-09-21 Olivier Hainque <hainque@adacore.com>
2876 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
2877 triplet, similar to support for VxWorks7.
2878 * config/vxworks-dummy.h: Provide a default definition
2879 of TARGET_VXWORKS64 to 0.
2881 2018-09-21 Olivier Hainque <hainque@adacore.com>
2883 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
2884 * config/vxworks-dummy.h: here.
2886 2018-09-21 Olivier Hainque <hainque@adacore.com>
2888 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
2890 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
2891 Bo Zhou <zbo.zhou@hisilicon.com>
2893 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
2894 * config/aarch64/aarch64-tune.md: Regenerated.
2895 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
2896 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
2897 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
2899 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
2900 Julian Brown <julian@codesourcery.com>
2902 * builtins.c (get_builtin_sync_mem): Handle address spaces.
2904 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2906 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2907 if the call takes a static chain.
2909 2018-09-21 Martin Liska <mliska@suse.cz>
2911 * auto-profile.c (autofdo_source_profile::read): Do not
2913 (read_profile): Do not add working sets.
2914 (read_autofdo_file): Remove sum_all.
2915 (afdo_callsite_hot_enough_for_early_inline): Remove const
2917 * coverage.c (struct counts_entry): Remove gcov_summary.
2918 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2919 do not support GCOV_TAG_PROGRAM_SUMMARY.
2920 (get_coverage_counts): Remove summary and expected
2922 * coverage.h (get_coverage_counts): Likewise.
2923 * doc/gcov-dump.texi: Remove -w option.
2924 * gcov-dump.c (dump_working_sets): Remove.
2925 (main): Do not support '-w' option.
2926 (print_usage): Likewise.
2927 (tag_summary): Likewise.
2928 * gcov-io.c (gcov_write_summary): Do not dump
2930 (gcov_read_summary): Likewise.
2931 (gcov_histo_index): Remove.
2932 (gcov_histogram_merge): Likewise.
2933 (compute_working_sets): Likewise.
2934 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2936 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2937 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2938 (GCOV_HISTOGRAM_SIZE): Remove.
2939 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2940 (struct gcov_summary): Simplify rapidly just
2941 to runs and sum_max fields.
2942 (gcov_histo_index): Remove.
2943 (NUM_GCOV_WORKING_SETS): Likewise.
2944 (compute_working_sets): Likewise.
2945 * gcov-tool.c (print_overlap_usage_message): Remove
2946 trailing empty line.
2947 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2948 (output_lines): Remove program related line.
2949 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2950 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2952 (input_profile_summary): Do not read it.
2953 (merge_profile_summaries): And do not merge it.
2954 (input_symtab): Do not call removed function.
2955 * modulo-sched.c (sms_schedule): Do not print sum_max.
2956 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2957 removed when histogram method was invented.
2958 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2960 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2962 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2963 and dump selected value.
2964 * profile.c (add_working_set): Remove.
2965 (get_working_sets): Likewise.
2966 (find_working_set): Likewise.
2967 (get_exec_counts): Do not work with working sets.
2968 (read_profile_edge_counts): Do not inform as sum_max is removed.
2969 (compute_branch_probabilities): Likewise.
2970 (compute_value_histograms): Remove argument for call of
2971 get_coverage_counts.
2972 * profile.h: Do not make gcov_summary const.
2974 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
2976 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2978 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2980 PR tree-optimization/86990
2981 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2982 Check that the entire merged store group is made of constants only for
2985 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
2987 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2988 (VTABLE_VERIFICATION_SPEC): Likewise.
2989 (SANITIZER_EARLY_SPEC): Likewise.
2990 (SANITIZER_SPEC): Likewise.
2991 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2992 * doc/invoke.texi (Link Options): Document -r.
2994 2018-09-20 Richard Biener <rguenther@suse.de>
2997 * gimplify.c (gimplify_expr): Retain alignment of
2998 addressable lvalue in dereference.
3000 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
3003 * configure.ac: Check for .loc is_stmt support.
3004 * configure, config.in: Rebuilt.
3005 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
3008 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
3010 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
3011 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
3014 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
3016 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3018 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
3021 2018-09-20 Marek Polacek <polacek@redhat.com>
3023 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
3025 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
3027 PR tree-optimization/87288
3028 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
3029 into account when determining PEELING_FOR_NITERS.
3031 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
3033 PR tree-optimization/86877
3034 * tree-vect-loop.c (vect_analyze_loop_2): Call
3035 vect_verify_datarefs_alignment.
3037 2018-09-19 Marek Polacek <polacek@redhat.com>
3039 * doc/invoke.texi: Document -Wclass-conversion.
3041 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
3043 * config/pa/pa.c (pa_adjust_priority): Delete.
3044 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
3046 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3047 (atomic_storehi): Likewise.
3048 (atomic_storesi): Likewise.
3049 (atomic_loaddi): Restore compare and swap exchange loop code.
3051 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
3053 PR rtl-optimization/86902
3054 * combine.c (try_combine): When changing the CC mode used, don't change
3055 an unrelated mode in other_insn to that new CC mode.
3057 2018-09-19 David Malcolm <dmalcolm@redhat.com>
3059 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
3060 with %T in place of calls to dump_generic_expr.
3061 (prune_runtime_alias_test_list): Likewise.
3062 (create_runtime_alias_checks): Likewise.
3063 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
3064 (vect_analyze_data_ref_dependence): Likewise.
3065 (vect_slp_analyze_data_ref_dependence): Likewise.
3066 (vect_record_base_alignment): Likewise. Use %G in place of call
3067 to dump_gimple_stmt.
3068 (vect_compute_data_ref_alignment): Likewise.
3069 (verify_data_ref_alignment): Likewise.
3070 (vect_find_same_alignment_drs): Likewise.
3071 (vect_analyze_group_access_1): Likewise.
3072 (vect_analyze_data_ref_accesses): Likewise.
3073 (dependence_distance_ge_vf): Likewise.
3074 (dump_lower_bound): Likewise.
3075 (vect_prune_runtime_alias_test_list): Likewise.
3076 (vect_find_stmt_data_reference): Likewise.
3077 (vect_analyze_data_refs): Likewise.
3078 (vect_create_addr_base_for_vector_ref): Likewise.
3079 (vect_create_data_ref_ptr): Likewise.
3080 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
3081 (vect_can_advance_ivs_p): Likewise.
3082 (vect_update_ivs_after_vectorizer): Likewise.
3083 (vect_gen_prolog_loop_niters): Likewise.
3084 (vect_prepare_for_masked_peels): Likewise.
3085 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3086 (vect_determine_vectorization_factor): Likewise.
3087 (vect_is_simple_iv_evolution): Likewise.
3088 (vect_analyze_scalar_cycles_1): Likewise.
3089 (vect_analyze_loop_operations): Likewise.
3090 (report_vect_op): Likewise.
3091 (vect_is_slp_reduction): Likewise.
3092 (check_reduction_path): Likewise.
3093 (vect_is_simple_reduction): Likewise.
3094 (vect_create_epilog_for_reduction): Likewise.
3095 (vect_finalize_reduction:): Likewise.
3096 (vectorizable_induction): Likewise.
3097 (vect_transform_loop_stmt): Likewise.
3098 (vect_transform_loop): Likewise.
3099 (optimize_mask_stores): Likewise.
3100 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
3101 (vect_split_statement): Likewise.
3102 (vect_recog_over_widening_pattern): Likewise.
3103 (vect_recog_average_pattern): Likewise.
3104 (vect_determine_min_output_precision_1): Likewise.
3105 (vect_determine_precisions_from_range): Likewise.
3106 (vect_determine_precisions_from_users): Likewise.
3107 (vect_mark_pattern_stmts): Likewise.
3108 (vect_pattern_recog_1): Likewise.
3109 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3110 (vect_record_max_nunits): Likewise.
3111 (vect_build_slp_tree_1): Likewise.
3112 (vect_build_slp_tree_2): Likewise.
3113 (vect_print_slp_tree): Likewise.
3114 (vect_analyze_slp_instance): Likewise.
3115 (vect_detect_hybrid_slp_stmts): Likewise.
3116 (vect_detect_hybrid_slp_1): Likewise.
3117 (vect_slp_analyze_operations): Likewise.
3118 (vect_slp_analyze_bb_1): Likewise.
3119 (vect_transform_slp_perm_load): Likewise.
3120 (vect_schedule_slp_instance): Likewise.
3121 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3122 (vect_mark_stmts_to_be_vectorized): Likewise.
3123 (vect_init_vector_1): Likewise.
3124 (vect_get_vec_def_for_operand): Likewise.
3125 (vect_finish_stmt_generation_1): Likewise.
3126 (vect_check_load_store_mask): Likewise.
3127 (vectorizable_call): Likewise.
3128 (vectorizable_conversion): Likewise.
3129 (vectorizable_operation): Likewise.
3130 (vectorizable_load): Likewise.
3131 (vect_analyze_stmt): Likewise.
3132 (vect_is_simple_use): Likewise.
3133 (vect_get_vector_types_for_stmt): Likewise.
3134 (vect_get_mask_type_for_stmt): Likewise.
3135 * tree-vectorizer.c (increase_alignment): Likewise.
3137 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
3139 * doc/rtl.texi: Adjust vec_select description.
3140 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3141 non-constant selectors.
3143 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
3145 * config/aarch64/aarch64-protos.h
3146 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3147 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3148 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3149 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3150 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3151 (AARCH64_FL_PROFILE): Move index so flags are ordered.
3152 (AARCH64_ISA_RCPC8_4): New flag.
3153 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
3154 to aarch64_offset_9bit_signed_unscaled_p.
3155 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
3157 * config/aarch64/constraints.md (Ust): New constraint.
3158 * config/aarch64/predicates.md.
3159 (aarch64_9bit_offset_memory_operand): New predicate.
3160 (aarch64_rcpc_memory_operand): New predicate.
3162 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
3164 PR rtl-optimization/87361
3165 * rtlanal.c (nonzero_bits1): Revert accidental change.
3167 2018-09-19 Richard Biener <rguenther@suse.de>
3169 PR tree-optimization/87349
3170 PR tree-optimization/87342
3171 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3173 2018-09-18 Marek Polacek <polacek@redhat.com>
3175 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3176 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3178 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
3180 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3182 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
3184 PR rtl-optimization/86882
3185 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3187 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
3189 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3190 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3192 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
3195 * doc/invoke.texi (Link Options): Fix formatting and grammar.
3197 2018-09-18 Richard Biener <rguenther@suse.de>
3200 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3201 (compute_samebase_partition_bases): Likewise.
3202 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3203 (gimple_can_coalesce_p): Simplify.
3205 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
3207 Handle a library implementation of ffs calling __builtin_ffs.
3208 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3209 (mmix_init_libfuncs): New function: make __builtin_ffs expand
3212 2018-09-17 David Malcolm <dmalcolm@redhat.com>
3214 * diagnostic-show-locus.c (class layout_range): Add field
3216 (layout_range::layout_range): Add "original_idx" param and use it
3217 to initialize new field.
3218 (make_range): Use 0 for original_idx.
3219 (layout::layout): Pass in index to calls to
3220 maybe_add_location_range.
3221 (layout::maybe_add_location_range): Add param "original_idx" and
3222 pass it on to layout_range.
3223 (layout::print_any_labels): Pass on range->m_original_idx to
3225 (gcc_rich_location::add_location_if_nearby): Use 0 for
3227 * gcc-rich-location.h (text_range_label::get_text): Update for new
3229 (range_label_for_type_mismatch::get_text): Likewise.
3231 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
3233 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
3235 2018-09-17 David Malcolm <dmalcolm@redhat.com>
3237 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
3238 format_string_diagnostic_t.
3239 (fmtwarn_n): Likewise.
3240 * substring-locations.c
3241 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
3242 (format_warning_n_va): Convert to...
3243 (format_string_diagnostic_t::emit_warning_n_va): ...this.
3244 (format_warning_va): Convert to...
3245 (format_string_diagnostic_t::emit_warning_va): ...this.
3246 (format_warning_at_substring): Convert to...
3247 (format_string_diagnostic_t::emit_warning): ...this.
3248 (format_warning_at_substring_n): Convert to...
3249 (format_string_diagnostic_t::emit_warning_n): ...this.
3250 * substring-locations.h (class format_string_diagnostic_t): New
3252 (format_warning_va): Convert to
3253 format_string_diagnostic_t::emit_warning_va.
3254 (format_warning_n_va): Convert to
3255 format_string_diagnostic_t::emit_warning_n_va.
3256 (format_warning_at_substring): Convert to
3257 format_string_diagnostic_t::emit_warning.
3258 (format_warning_at_substring_n): Convert to
3259 format_string_diagnostic_t::emit_warning_n.
3261 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
3262 Bernd Schmidt <bernds_cb1@t-online.de>
3264 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
3267 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
3269 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
3270 operand 0 predicate to nonimmediate operand.
3271 (rint<mode>2_frndint): Remove insn pattern.
3272 (rint<mode>2): Change operand 1 predicate to general_operand.
3273 Extend operand 1 to XFmode and generate rintxf2 insn.
3274 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
3275 Do not use X87MODEF mode macro.
3276 (frndintxf2_<rounding>_i387): Rename from
3277 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
3278 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
3279 to XFmode and generate significandxf3 insn.
3281 2018-09-17 Richard Biener <rguenther@suse.de>
3283 PR tree-optimization/87328
3284 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
3285 visiting unexecutable backedges when not iterating.
3286 (do_rpo_vn): Mark all edges not executable even when not
3289 2018-09-17 Martin Jambor <mjambor@suse.cz>
3292 * doc/invoke.texi (Warning Options): Likewise.
3294 2018-09-17 Richard Biener <rguenther@suse.de>
3296 PR tree-optimization/87301
3297 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
3298 clean EH info from leftover copy assignments.
3300 2018-09-17 Martin Liska <mliska@suse.cz>
3302 PR gcov-profile/85871
3303 * gcov.c (output_intermediate_file): Fix out of bounds
3306 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
3308 * config/arc/arc.c: Object attributes for core4 not reflected
3310 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
3313 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
3315 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
3317 2018-09-17 Martin Liska <mliska@suse.cz>
3319 * doc/gcov.texi: Document new option --use-hotness-colors.
3320 * gcov.c (struct source_info): Declare new field.
3321 (source_info::source_info): Set default for maximum_count.
3322 (print_usage): Add new -q option.
3323 (process_args): Process it.
3324 (accumulate_line_info): Save src->maximum_count.
3325 (output_line_beginning): Make color line number if
3326 flag_use_hotness_colors is set.
3327 (output_line_details): Pass default argument value.
3328 (output_lines): Pass src->maximum_count.
3330 2018-09-17 Martin Liska <mliska@suse.cz>
3332 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
3333 Use processor_names table.
3334 * config/i386/i386.c (ix86_default_align): Use
3335 processor_cost_table for alignment values.
3336 (ix86_option_override_internal): Use processor_names.
3337 (ix86_function_specific_print): Likewise.
3338 * config/i386/i386.h (struct processor_costs):
3339 Add alignment values.
3340 (struct ptt): Remove and replace with const char *.
3341 * config/i386/x86-tune-costs.h (struct processor_costs):
3342 Declare default alignments for all costs.
3344 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
3346 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
3347 symbolics or VR_VARYING ranges for ABS_EXPR.
3348 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
3349 when range will wrap.
3351 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
3354 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
3355 before and after a JUMP_TABLE_DATA.
3357 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
3360 * dojump.c (do_compare_and_jump): Canonicalize function pointers
3361 when one operand is a function pointer. Use POINTER_TYPE_P and
3362 FUNC_OR_METHOD_TYPE_P.
3363 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3364 * fold-const.c (build_range_check): Likewise.
3365 * match.pd (simple_comparison): Likewise.
3367 2018-09-14 David Malcolm <dmalcolm@redhat.com>
3370 * spellcheck.c (get_edit_distance_cutoff): New function.
3371 (selftest::test_edit_distance_unit_test_oneway): Rename to...
3372 (selftest::test_get_edit_distance_one_way): ...this.
3373 (selftest::test_get_edit_distance_unit): Rename to...
3374 (selftest::test_get_edit_distance_both_ways): ...this.
3375 (selftest::test_edit_distances): Move tests to this new function,
3376 and test some more pairs of strings. Update for above renaming.
3377 (selftest::get_old_cutoff): New function.
3378 (selftest::test_get_edit_distance_cutoff): New function.
3379 (selftest::assert_suggested_for): New function.
3380 (ASSERT_SUGGESTED_FOR): New macro.
3381 (selftest::assert_not_suggested_for): New function.
3382 (ASSERT_NOT_SUGGESTED_FOR): New macro.
3383 (selftest::test_suggestions): New function.
3384 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
3385 tests to selftest::test_edit_distances and call it. Add calls to
3386 selftest::test_get_edit_distance_cutoff and
3387 selftest::test_suggestions.
3388 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
3389 (best_match::consider): Replace hard-coded cutoff calculation with
3391 (best_match::get_cutoff): New declaration.
3392 (best_match::get_best_meaningful_candidate): Likewise.
3394 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
3396 * builtins.c (fold_builtin_strlen): Remove TODO comment.
3398 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
3401 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3403 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3404 terminated string literal.
3406 2018-09-14 Martin Sebor <msebor@redhat.com>
3408 * builtins.c (unterminated_array): Handle ARRAY_REF.
3409 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
3410 * builtins.h (unterminated_array): Declare extern.
3411 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
3413 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
3416 2018-09-14 Martin Sebor <msebor@redhat.com>
3417 Jeff Law <law@redhat.com>
3419 * builtins.c (unterminated_array): New.
3420 (expand_builtin_strcpy): Adjust.
3421 (expand_builtin_strcpy_args): Detect unterminated arrays.
3422 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
3423 unterminated arrays.
3424 * gimple-fold.h (get_maxval_strlen): Add argument.
3425 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
3427 * gimple-fold.c (get_range_strlen): Add argument.
3428 (get_maxval_strlen): Adjust.
3429 * gimple-fold.h (get_range_strlen): Add argument.
3431 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
3433 * config/i386/movdirintrin.h: Fix copyright year.
3435 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
3437 * reg-stack.c: Include regs.h.
3438 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
3439 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
3440 FIRST_STACK_REG, not DFmode.
3441 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
3442 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
3443 (change stack): Default register mode to the reg_raw_mode of
3444 FIRST_STACK_REG, not DFmode.
3445 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
3446 (*swapxf): Rename from swapxf.
3448 2018-09-14 Carl Love <cel@us.ibm.com>
3450 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
3451 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
3453 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
3456 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
3459 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
3462 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
3464 2018-09-14 Jason Merrill <jason@redhat.com>
3466 Fix --enable-gather-detailed-mem-stats.
3467 * hash-table.c (hash_table_usage): Change from variable to function.
3468 * hash-table.h: Adjust.
3469 * Makefile.in: Add missing dependencies on hash-table.h.
3471 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3473 PR tree-optimization/87259
3475 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
3476 execute_cse_reciprocals_1 has tried transforming.
3478 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
3480 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
3481 VR_VARYING for PLUS/MINUS_EXPR.
3483 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
3485 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
3488 2018-09-14 Richard Biener <rguenther@suse.de>
3491 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
3492 bits for the merged partition.
3494 2018-09-13 Martin Sebor <msebor@redhat.com>
3495 Bernd Edlinger <bernd.edlinger@hotmail.de>
3497 * builtins.h (c_srlen): Add argument.
3498 * builtins.c (warn_string_no_nul): New function.
3499 (c_strlen): Add argument and use it. Update recursive calls.
3500 Pass DECL argument to string_constant to get info on non
3501 terminated strings. Update *NONSTR as needed.
3502 (fold_builtin_strlen): Add argument to calls to c_strlen.
3503 Warn for unterminated arrays.
3504 (warn_string_no_null): Add prototype.
3505 * expr.c (string_constant): Update arguments. Update recursive
3506 calls appropriately. Detect missing NUL terminator and outermost
3507 declaration its missing in.
3508 Improve checks for arrays with nonzero lower bound or elements
3509 that are not a single byte. Simplify offset computation.
3510 Simplify checks for non-NUL terminated strings.
3511 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
3512 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
3514 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
3516 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
3518 * fold-const.c (c_getstr): Fix function comment. Remove unused third
3519 argument. Fix range checks.
3520 * fold-const.h (c_getstr): Adjust protoype.
3521 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
3522 string is constant but contains no NUL byte.
3524 * expr.c (string_constant): Adjust function comment.
3525 Remove bogus check for zero termination.
3527 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
3529 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
3530 (get_constant_size): Don't make STRING_CSTs larger than they are.
3531 (check_string_literal): New check function for STRING_CSTs.
3532 (output_constant): Use it.
3534 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
3537 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
3539 2018-09-13 Richard Biener <rguenther@suse.de>
3541 PR tree-optimization/87263
3542 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
3543 (struct unwind_state): Add max_rpo field.
3544 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
3545 Compute max_rpo, the max RPO number a block can be backwards reached
3546 from. Re-write non-iterating mode to a RPO ordered worklist approach,
3547 separating it from the iterating mode.
3549 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
3551 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
3552 (rfs_decision): New scheduling decision.
3554 2018-09-13 Richard Biener <rguenther@suse.de>
3557 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
3558 (vn_nary_op_insert_pieces_predicated): Do not write useless
3559 valid_dominated_by_p entry outside of the allocated storage.
3561 2018-09-13 Omar Sandoval <osandov@osandov.com>
3562 Tom de Vries <tdevries@suse.de>
3565 * dwarf2out.c (is_c): New function.
3566 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
3568 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
3571 * config/aarch64/aarch64.md (*aarch64_bfxil):
3573 * config/aarch64/constraints.md (Ulc): Define.
3574 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
3576 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
3579 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
3581 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
3582 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
3583 aarch64_layout_frame call.
3584 (aarch64_expand_epilogue): Likewise.
3585 (aarch64_initial_elimination_offset): Likewise.
3586 (aarch64_get_separate_components): Likewise.
3587 (aarch64_use_return_insn_p): Likewise.
3588 (aarch64_layout_frame): Remove unneeded check.
3590 2018-09-13 Jakub Jelinek <jakub@redhat.com>
3592 * configure.ac: Only append
3593 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
3594 gcc_config_arguments if it was never reconfigured or last reconfigure
3595 was with different arguments.
3596 * configure: Regenerated.
3598 2018-09-13 Jakub Jelinek <jakub@redhat.com>
3599 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3602 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
3603 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
3605 2018-09-13 Jakub Jelinek <jakub@redhat.com>
3607 PR tree-optimization/87287
3608 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
3609 X % C == 0 to X % (unsigned) C == 0 optimization to ...
3610 * match.pd (X % C == 0): ... here. New optimization.
3612 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3615 * expr.h (maybe_optimize_mod_cmp): Declare.
3616 * expr.c (mod_inv): New function.
3617 (maybe_optimize_mod_cmp): New function.
3618 (do_store_flag): Use it.
3619 * cfgexpand.c (expand_gimple_cond): Likewise.
3621 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
3622 Julian Brown <julian@codesourcery.com>
3625 * gimplify.c (gimplify_scan_omp_clauses): Set
3626 target_firstprivatize_array_bases in OpenACC parallel and kernels
3627 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
3628 OpenACC data regions.
3630 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
3632 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
3633 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
3635 2018-09-12 Richard Biener <rguenther@suse.de>
3637 PR tree-optimization/87280
3638 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
3639 edge but unreachable target.
3640 (do_rpo_vn): For conservatively handling a PHI only mark
3641 the backedge executable but not the block reachable.
3643 2018-09-12 Richard Biener <rguenther@suse.de>
3645 PR tree-optimization/87266
3646 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
3649 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
3651 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
3653 ("trunc<BFP:mode><DFP_ALL:mode>2")
3654 ("trunc<DFP_ALL:mode><BFP:mode>2")
3655 ("extend<BFP:mode><DFP_ALL:mode>2")
3656 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
3657 according to the target operand type.
3659 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3660 Andreas Krebbel <krebbel@linux.ibm.com>
3662 PR tree-optimization/86844
3663 * gimple-ssa-store-merging.c
3664 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
3665 there are any overlapping stores in between them, make sure they are
3666 also coalesced or we give up completely.
3668 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3671 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
3672 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
3674 2018-09-12 Tom de Vries <tdevries@suse.de>
3676 * common.opt (gdescribe-dies): Add option.
3677 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
3678 attribute for artifical and nameless decls.
3679 (dwarf2out_register_external_die): Add description attribute to
3680 external reference die.
3681 (add_desc_attribute): New functions.
3682 (gen_subprogram_die): Add description attribute to
3683 DW_TAG_call_site_parameter.
3684 * tree-pretty-print.c (print_generic_expr_to_str): New function.
3685 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
3686 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
3688 (@item -gdescribe-dies): Add.
3690 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
3692 * tree-vrp.c (vrp_shift_undefined_p): Remove.
3693 (extract_range_from_binary_expr_1: Call
3694 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
3695 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
3698 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
3700 * gimple-ssa-warn-alloca.c
3701 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
3702 field for ALLOCA_BOUND_*_LARGE.
3704 2018-09-11 Nathan Sidwell <nathan@acm.org>
3706 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
3708 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
3710 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
3711 for clobbers. Remove obsolete comment.
3713 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
3715 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
3716 mpxchk, mpxld and mpxst types.
3717 (define_attr length_immediate): Remove all processing of mpx types.
3718 (define_attr prefix_0f): Ditto.
3719 (define_attr memory): Ditto.
3721 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
3723 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
3724 (log<mode>2): Change operand 1 predicate to general_operand.
3725 Extend operand 1 to XFmode and generate logxf3 insn.
3726 (log10<mode>2): Change operand 1 predicate to general_operand.
3727 Extend operand 1 to XFmode and generate log10xf3 insn.
3728 (log2<mode>2): Change operand 1 predicate to general_operand.
3729 Extend operand 1 to XFmode and generate log2xf3 insn.
3730 (fyl2xp1_extend<mode>xf3_i387): Remove.
3731 (log1p<mode>2): Change operand 1 predicate to general_operand.
3732 Extend operand 1 to XFmode and generate log1pxf3 insn.
3733 (fxtract_extend<mode>xf3_i387): Remove.
3734 (logb<mode>2): Change operand 1 predicate to general_operand.
3735 Extend operand 1 to XFmode and generate logbxf3 insn.
3736 (ilogb<mode>2): Change operand 1 predicate to general_operand.
3737 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
3738 (significand<mode>2): Change operand 1 predicate to general_operand.
3739 Extend operand 1 to XFmode and generate significandxf3 insn.
3741 2018-09-11 Nathan Sidwell <nathan@acm.org>
3743 * gcc.c (perror_with_name, pfatal_with_name): Delete.
3744 (load_specs): Use fatal_error.
3745 (DELETE_IF_ORDINARY, process_command): Use error.
3746 (execute, run_attempt): Use fatal_error.
3748 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
3750 * diagnostic-core.h (sorry_at): New prototype.
3751 * diagnostic.c (sorry_at): New function.
3753 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
3755 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
3756 by zero as VR_UNDEFINED.
3758 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
3760 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
3761 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
3762 (<sincos>mode2): New expander.
3763 (sincos_extend<mode>xf3_i387): Remove insn pattern.
3764 (sincos -> sin, cos splitters): Remove splitter patterns.
3765 (sincos<mode>3): Change operand 2 predicate to general_operand.
3766 Extend operand 2 to XFmode and generate sincosxf3 insn.
3767 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
3768 Change operand 3 predicate to const1_operand.
3769 (fptan_extend<mode>xf4_i387): Remove insn pattern.
3770 (tanxf2): Update operands in the call to fptanxf4_i387.
3771 (tan<mode>2): Change operand 1 predicate to general_operand.
3772 Extend operand 1 to XFmode and generate tanxf3 insn.
3773 (atan2xf3): Rename from *fpatanxf3_i387.
3774 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
3775 (atan2xf3): Remove expander.
3776 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
3777 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
3778 (atan<mode>2): Change operand 1 predicate to general_operand.
3779 Extend operand 1 to XFmode and generate atanxf3 insn.
3781 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
3783 * config/i386/i386.md (x87/SSE constant load splitter): Use
3784 memory_operand instead of nonimmediate_operand for input operand
3787 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
3789 * config/i386/i386.md (float partial SSE register stall splitter): Move
3790 splitter near its instruction pattern.
3791 (float_extend partial SSE register stall splitter): Ditto.
3792 (float_truncate partial SSE register stall splitter): Ditto.
3794 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
3797 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3798 to speculation_safe_value_not_needed.
3801 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
3802 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
3803 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
3804 leaf_function_p, instead use has_hard_reg_initial_val.
3806 2018-09-09 Nathan Sidwell <nathan@acm.org>
3808 * gcc.h (pfatal_with_name): Don't declare here.
3809 * gcc.c (pfatal_with_name): Make static.
3811 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
3813 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
3816 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
3818 PR rtl-optimization/85458
3819 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
3820 priority hook to reduce the priority of EXPR.
3822 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
3824 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
3825 DImode for x87 on 32bit targets. Conditionally disable x87 modes
3826 with X87_ENABLE_FLOAT. Remove preparation code.
3827 (*float<SWI48:mode><MODEF:mode>2): Rename from
3828 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
3829 math using "enabled" attribute.
3830 (*floatdi<MODEF:mode>2_i387): Rename from
3831 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
3832 enable for 32bit targets only.
3833 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
3835 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
3836 as operand 1 predicate. Rewrite as define_insn_and_split.
3837 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
3839 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
3841 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
3842 to fallthru to FLOAT case.
3844 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3847 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
3848 around folding of vec_sl to handle out of range shift values.
3850 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
3852 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
3853 Update callers to gen_fix_trunc<mode>_i387_fisttp
3854 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
3855 nonimmediate_operand.
3856 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
3857 and corresponding splitters.
3858 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
3859 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
3860 (fix_truncdi_i387_with_temp): Remove insn pattern
3861 and corresponding splitters.
3862 (fix_trunc<mode>_i387): Change operand 0 predicate to
3863 nonimmediate_operand.
3864 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
3865 and corresponding splitters.
3866 (*fistdi2_1): Remove.
3868 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
3869 (lrintxfdi2): Remove expander. Reimplement as define_insn.
3870 (*fist<mode>2_1): Remove.
3871 (fist<mode>2): Ditto.
3872 (fist<mode>2_with_temp): Remove insn pattern and corresponding
3874 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
3875 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
3876 (fistdi2_<rounding>): Change operand 0 predicate to
3877 nonimmediate_operand.
3878 (fistdi2_<rounding>_with_temp): Remove insn pattern
3879 and corresponding splitters.
3880 (fist<mode>2_<rounding>): Change operand 0 predicate to
3881 nonimmediate_operand.
3882 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
3883 and corresponding splitters.
3885 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
3887 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
3889 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
3892 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3894 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
3895 early gimple folding of vec_splat().
3896 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
3897 * gimple-fold.h: Add an extern define for tree_vec_extract().
3899 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3901 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
3902 wrappers around TREE_TYPE comparisons.
3904 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3907 * config/s390/predicates.md: Add nonsym_memory_operand.
3908 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
3909 contains a SYMBOL_REF, load it into an intermediate pseudo.
3910 (s390_emit_compare_and_swap): Legitimize operand.
3911 * config/s390/s390.md: Use the new nonsym_memory_operand
3912 with UNSPECV_CAS patterns.
3914 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3917 * config/s390/s390-passes.def: New file.
3918 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3920 (make_pass_s390_early_mach): Add declaration.
3921 * config/s390/s390.c (make_pass_s390_early_mach):
3922 (s390_option_override): Remove dynamic registration.
3923 * config/s390/t-s390: Add s390-passes.def.
3925 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3927 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3928 Remove UNSPEC_LTREL_BASE check.
3929 (annotate_constant_pool_refs): Likewise.
3930 (find_constant_pool_ref): Likewise.
3931 (find_ltrel_base): Removed.
3932 (replace_ltrel_base): Removed.
3933 (s390_mainpool_finish): Remove replace_ltrel_base call.
3934 (s390_chunkify_start): Remove pending LTREL_BASE logic.
3935 (s390_chunkify_finish): Remove replace_ltrel_base call.
3936 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3938 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
3941 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3942 to speculation_safe_value_not_needed.
3944 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
3945 Bernd Schmidt <bernds_cb1@t-online.de>
3947 * config/nvptx/nvptx-opts.h: New file.
3948 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3949 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3951 (TARGET_SM35): New macro.
3952 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3954 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3956 (misa=): New option.
3957 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3959 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
3961 * config/i386/i386.md (truncdfsf2): Remove expander.
3962 (truncdfsf2_with_temp): Ditto.
3963 (truncxf<mode>2): Ditto.
3964 (*truncdfsf_fast_mixed): Remove insn pattern.
3965 (*truncdfsf_fast_i387): Ditto.
3966 (*truncdfsf_mixed): Ditto.
3967 (*truncdfsf_i387): Ditto.
3968 (*truncdfsf2_i387_1): Ditto.
3969 (*truncxfsf2_mixed): Ditto.
3970 (*truncxfdf2_mixed): Ditto.
3971 (*truncxf<mode>2_i387_noop): Ditto. Update callers
3972 to call gen_truncxf<mode>2 instead.
3973 (*truncxf<mode>2_i387): Remove.
3974 (reg->reg splitters): Remove splitter pattern.
3975 (reg->mem splitters): Ditto.
3977 (truncdfsf2): New insn pattern.
3978 (truncxf<mode>2): Ditto.
3980 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3982 * tree-ssa-math-opts.c (is_mult_by): New function.
3983 (is_square_of): Use the above.
3984 (optimize_recip_sqrt): New function.
3985 (pass_cse_reciprocals::execute): Use the above.
3987 2018-09-05 Richard Biener <rguenther@suse.de>
3990 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3991 to zero-init the emplaced vec.
3993 2018-09-05 Martin Liska <mliska@suse.cz>
3995 PR tree-optimization/87205
3996 * tree-switch-conversion.c (pass_lower_switch::execute):
3997 Group cases for switch statements.
3999 2018-09-05 Richard Biener <rguenther@suse.de>
4001 PR tree-optimization/87217
4002 * tree-ssa-sccvn.c (vuse_valueize): New.
4003 (vn_reference_lookup_pieces): Use it.
4004 (vn_reference_lookup): Likewise.
4006 2018-09-05 Nathan Sidwell <nathan@acm.org>
4009 * stor-layout.c (place_field): Scan forwards to check last
4010 bitfield when ms_bitfield_placement is in effect.
4012 2018-09-05 Richard Biener <rguenther@suse.de>
4015 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
4018 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
4019 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4021 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
4022 * config.gcc (extra_objs): Build it.
4023 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
4025 * config/aarch64/aarch64-passes.def
4026 (pass_tag_collision_avoidance): New pass.
4027 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
4028 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
4029 (aarch64_classify_address): Remove static qualifier.
4030 (aarch64_address_info, aarch64_address_type): Move to...
4031 * config/aarch64/aarch64-protos.h: ... here.
4032 (make_pass_tag_collision_avoidance): New function.
4033 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
4036 2018-09-05 Martin Liska <mliska@suse.cz>
4038 * doc/gcov.texi: Update documentation of humar
4040 * gcov.c (format_count): Print one decimal place, it provides
4041 more fine number of situations like '1G' vs. '1.4G'.
4043 2018-09-05 Martin Liska <mliska@suse.cz>
4046 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
4047 * optc-gen.awk: Allow 'Var' for Deprecated options in order
4048 to generate a MASK value.
4050 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
4053 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
4054 if frame pointer isn't used.
4055 (compute_frame_pointer_to_fb_displacement): Likewise.
4057 2018-09-04 Jakub Jelinek <jakub@redhat.com>
4060 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
4061 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
4062 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
4063 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
4064 and OPTION_MASK_ISA_XSAVEC_UNSET.
4066 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
4068 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
4069 XOR operations in NAND case.
4071 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
4073 * wide-int-range.cc (wide_int_range_convert): New.
4074 * wide-int-range.h (wide_int_range_convert): New.
4075 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
4076 code into wide_int_range_convert.
4077 (extract_range_into_wide_ints): Do not munge anti range constants
4078 into the entire domain. Just return the range back.
4080 2018-09-04 Martin Liska <mliska@suse.cz>
4082 * genmatch.c (output_line_directive): Add new argument
4084 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
4086 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
4088 * doc/invoke.texi (Option Summary): Add whitespace.
4090 * doc/invoke.texi (Option Summary): Add -Waligned-new.
4092 2018-09-04 Richard Biener <rguenther@suse.de>
4094 PR tree-optimization/87211
4095 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
4096 backedge value we're supposed to treat as VARYING also number
4097 the PHI to VARYING in case it got a different value-number already.
4099 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
4101 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
4102 (extract_range_from_binary_expr_1): Do not call
4103 vrp_can_optimize_bit_op.
4104 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
4106 (wide_int_range_get_mask_and_bounds): New.
4107 (wide_int_range_optimize_bit_op): New.
4108 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4109 (wide_int_range_bit_and): Same.
4110 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4111 (wide_int_range_optimize_bit_op): New.
4112 (wide_int_range_get_mask_and_bounds): New.
4114 2018-09-04 Richard Biener <rguenther@suse.de>
4116 PR tree-optimization/87176
4117 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4118 variable. When value-numbering a virtual PHI node make sure
4119 to not value-number to the backedge value.
4121 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
4123 * doc/extend.texi (Long Long, Hex Floats): Document support for
4124 long long and hex floats in more recent versions of ISO C++.
4126 2018-09-03 Richard Biener <rguenther@suse.de>
4128 PR tree-optimization/87177
4129 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4132 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
4134 * bb-reorder.c (edge_order): Convert to C-qsort-style
4135 tri-state comparator.
4136 (reorder_basic_blocks_simple): Change std::stable_sort to
4139 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
4141 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4142 tri-state comparator.
4143 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4145 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
4147 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
4148 (mergesort): ... here as maximum count for using netsort.
4149 (gcc_qsort): Set nlim to 3 if stable sort is requested.
4150 (gcc_stablesort): New.
4151 * system.h (gcc_stablesort): Declare.
4153 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
4155 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
4156 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
4157 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
4159 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4161 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
4162 lxsdx and stxsdx alternatives.
4163 (*mov<mode>_hardfloat64): Ditto.
4164 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4166 2018-09-03 Richard Biener <rguenther@suse.de>
4168 PR tree-optimization/87200
4169 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4172 2018-09-03 Martin Liska <mliska@suse.cz>
4174 PR tree-optimization/87201
4175 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4176 Fix parenthesis in an expression.
4178 2018-09-03 Richard Biener <rguenther@suse.de>
4180 PR tree-optimization/87197
4181 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4182 visited. CSE the VN_INFO hashtable lookup.
4184 PR tree-optimization/87169
4185 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4186 iterating make sure there's no extra backedges from irreducible
4187 regions feeding the header. Mark the destination block
4190 2018-09-03 Martin Liska <mliska@suse.cz>
4193 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4194 * common/common-targhooks.c (default_get_valid_option_values):
4196 * common/common-targhooks.h (default_get_valid_option_values):
4198 * common/config/i386/i386-common.c: Move processor_target_table
4200 (ix86_get_valid_option_values): New function.
4201 (TARGET_GET_VALID_OPTION_VALUES): New macro.
4202 * config/i386/i386.c (struct ptt): Move to i386-common.c.
4203 (PTA_*): Move all defined masks into i386-common.c.
4204 (ix86_function_specific_restore): Use new processor_cost_table.
4205 * config/i386/i386.h (struct ptt): Moved from i386.c.
4206 (struct pta): Likewise.
4207 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4208 * doc/tm.texi.in: Likewise.
4209 * opt-suggestions.c (option_proposer::suggest_option):
4210 Pass prefix to build_option_suggestions.
4211 (option_proposer::get_completions): Likewise.
4212 (option_proposer::build_option_suggestions): Use the new target
4214 * opts.c (struct option_help_tuple): New struct.
4215 (print_filtered_help): Use the new target hook.
4217 2018-09-03 Martin Liska <mliska@suse.cz>
4220 * predict.c (set_even_probabilities): Add likely_edges
4221 argument and handle cases where we have precisely one
4223 (combine_predictions_for_bb): Catch also likely_edges.
4224 (tree_predict_by_opcode): Handle gswitch statements.
4225 * tree-cfg.h (find_case_label_for_value): New declaration.
4226 (find_taken_edge_switch_expr): Likewise.
4227 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4228 Find pivot in decision tree based on probabily, not by number of
4231 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
4233 * doc/standards.texi (Standards): Update Objective-C reference.
4235 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
4237 * doc/install.texi (Prerequisites): Update link for MPC.
4239 2018-09-01 Michael Matz <matz@suse.de>
4241 PR tree-optimization/87074
4242 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
4243 PHIs for outer-loop uses.
4245 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
4247 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
4248 * doc/invoke.texi (C Dialect Options): Ditto.
4250 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
4252 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
4254 2018-08-31 Richard Biener <rguenther@suse.de>
4256 PR tree-optimization/87168
4257 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
4258 (rpo_elim::eliminate_avail): When OP was not visited it must
4261 2018-08-31 David Malcolm <dmalcolm@redhat.com>
4263 * tree-vrp.c (copy_value_range): Convert param "from" from
4264 "value_range *" to "const value_range *".
4265 (range_is_null): Likewise for param "vr".
4266 (range_int_cst_p): Likewise.
4267 (range_int_cst_singleton_p): Likewise.
4268 (symbolic_range_p): Likewise.
4269 (value_ranges_intersect_p): Likewise for both params.
4270 (value_range_nonnegative_p): Likewise for param "vr".
4271 (value_range_constant_singleton): Likewise.
4272 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
4273 (extract_range_into_wide_ints): Likewise for param "vr".
4274 (extract_range_from_multiplicative_op): Likewise for params "vr0"
4276 (vrp_can_optimize_bit_op): Likewise.
4277 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
4279 (extract_range_from_unary_expr): Likewise.
4280 (debug_value_range): Likewise for param "vr".
4281 (value_range::dump): Add "const" qualifier.
4282 (vrp_prop::check_array_ref): Convert local "vr" from
4283 "value_range *" to "const value_range *".
4284 (vrp_prop::check_mem_ref): Likewise.
4285 (vrp_prop::visit_stmt): Likewise for local "old_vr".
4286 (vrp_intersect_ranges_1): Likewise for param "vr_1".
4287 (vrp_intersect_ranges): Likewise.
4288 (simplify_stmt_for_jump_threading): Likewise for local "vr".
4289 (vrp_prop::vrp_finalize): Likewise.
4290 * tree-vrp.h (value_range::dump): Add "const" qualifier.
4291 (vrp_intersect_ranges): Add "const" qualifier to params as above.
4292 (extract_range_from_unary_expr): Likewise.
4293 (value_range_constant_singleton): Likewise.
4294 (symbolic_range_p): Likewise.
4295 (copy_value_range): Likewise.
4296 (extract_range_from_binary_expr_1): Likewise.
4297 (range_int_cst_p): Likewise.
4298 (vrp_set_zero_nonzero_bits): Likewise.
4299 (range_int_cst_singleton_p): Likewise.
4301 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
4303 * config/aarch64/arm_neon.h (vabsd_s64): New.
4304 (vnegd_s64): Likewise.
4306 2018-08-31 Martin Jambor <mjambor@suse.cz>
4308 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
4310 2018-08-31 Martin Liska <mliska@suse.cz>
4312 * ipa-icf.c (sem_item::add_type): Use
4313 sem_item::m_type_hash_cache.
4314 * ipa-icf.h: Move the cache from sem_item_optimizer
4317 2018-08-31 Nathan Sidwell <nathan@acm.org>
4319 * doc/extend.texi (Backwards Compatibility): Remove implicit
4320 extern C leeway of () being (...).
4322 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4324 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
4326 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
4330 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
4332 2018-08-31 Jakub Jelinek <jakub@redhat.com>
4335 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
4336 gen_int_mode. Formatting fixes.
4338 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
4340 * target.def (custom_function_descriptors): Improve documentation.
4341 * doc/tm.texi.in (Trampolines): Expand discussion of function
4342 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
4343 beginning of the section.
4344 * doc/tm.texi: Regenerated.
4346 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
4348 * cfg.h (class auto_edge_flag): Spell out the template-id of the
4349 base class in the initializer list. This is a workaround for
4350 building with older GCC.
4351 (class auto_bb_flag): Likewise.
4353 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
4355 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
4356 (altivec_vcmpequ<VI_char>_p): Remove star.
4357 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
4359 (expand_strncmp_vec_sequence): New function.
4360 (emit_final_str_compare_vec): New function.
4361 (expand_strn_compare): Add support for vector strncmp.
4362 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
4363 length specification to bytes.
4364 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
4365 (vcmpnezb_p): New pattern.
4366 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
4367 for option -mstring-compare-inline-limit.
4369 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
4371 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
4372 (PTA_SKYLAKE): Add PTA_AES.
4373 (PTA_GOLDMONT): Likewise.
4375 2018-08-29 Jan Hubicka <jh@suse.cz>
4378 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
4379 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
4381 2018-08-29 Jan Hubicka <jh@suse.cz>
4383 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
4385 (hash_tree): Do not visit TYPE_STUB_DECL.
4386 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
4387 stream TYPE_STUB_DECL.
4388 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
4389 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
4390 after free_lang_data.
4391 (type_in_anonymous_namespace_p): Likewise.
4393 2018-08-29 Jan Hubicka <jh@suse.cz>
4395 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4396 comment that it has to be even number.
4397 (class sreal): Change m_sig type to int32_t.
4398 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4399 int64_t for temporary calculations.
4400 (sreal_verify_basics): Drop one bit from minimum and maximum.
4402 2018-08-30 Richard Biener <rguenther@suse.de>
4404 PR tree-optimization/87147
4405 * tree-ssa-sccvn.c (SSA_VISITED): New function.
4406 (visit_phi): When the degenerate result is from the backedge and
4407 we didn't visit its definition yet drop to VARYING.
4408 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
4410 2018-08-29 Jan Hubicka <jh@suse.cz>
4412 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
4414 (hash_tree): Likewise.
4416 2018-08-29 Jan Hubicka <jh@suse.cz>
4418 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
4419 and TYPE_NEXT_REF_TO.
4421 2018-08-29 Jan Hubicka <jh@suse.cz>
4423 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4424 comment that it has to be even number.
4425 (class sreal): Change m_sig type to int32_t.
4426 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4427 int64_t for temporary calculations.
4428 (sreal_verify_basics): Drop one bit from minimum and maximum.
4430 2018-08-30 Tamar Christina <tamar.christina@arm.com>
4432 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
4434 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
4437 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
4438 if to_add is negative.
4440 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
4443 * builtins.c (c_strlen): Improve range checks.
4445 2018-08-29 Martin Sebor <msebor@redhat.com>
4446 Jeff Law <law@redhat.com>
4448 PR tree-optimization/86714
4449 PR tree-optimization/86711
4450 * builtins.c (c_strlen): Add arguments to call to string_constant.
4451 * expr.c (string_constant): Add argument. Detect missing nul
4452 terminator and outermost declaration it's missing in.
4453 * expr.h (string_constant): Add argument.
4454 * fold-const.c (read_from_constant_string): Add arguments to call to
4456 (c_getstr): Likewise.
4457 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4459 * tree-ssa-strlen.c (get_stridx): Likewise.
4461 2018-08-29 Jan Hubicka <jh@suse.cz>
4463 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
4464 Do not stream DECL_VINDEX.
4465 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
4466 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
4467 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
4469 2018-08-29 Richard Biener <rguenther@suse.de>
4471 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
4472 virtual operands that are not default defs to honor region
4474 (rpo_vn_valueize): Remove ineffective code here.
4476 2018-08-29 Richard Biener <rguenther@suse.de>
4478 PR tree-optimization/87132
4479 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
4480 when skipping defs reachable over backedges.
4482 2018-08-29 Richard Biener <rguenther@suse.de>
4484 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
4485 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
4486 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
4487 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
4488 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
4489 (vn_reference_lookup_3): Remove use of const_parms.
4490 (free_rpo_vn): Do not free const_parms.
4491 (do_rpo_vn): Do not call init_const_parms.
4492 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
4493 SSA_NAME_POINTS_TO_READONLY_MEMORY.
4494 (call_may_clobber_ref_p_1): Likewise.
4496 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
4499 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
4501 (-ftree-scev-cprop): Document.
4503 2018-08-29 Jan Hubicka <jh@suse.cz>
4505 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
4507 (sreal constructor): Update.
4508 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
4509 sreal:operator/): Update.
4511 2018-08-29 Martin Liska <mliska@suse.cz>
4513 * tree-switch-conversion.c (switch_conversion::expand):
4514 Strenghten assumption about gswitch statements.
4516 2018-08-29 Richard Biener <rguenther@suse.de>
4518 PR tree-optimization/87117
4519 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
4520 re-value-number released SSA VDEFs.
4522 2018-08-29 Richard Biener <rguenther@suse.de>
4524 PR tree-optimization/87126
4525 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
4527 2018-08-28 Jim Wilson <jimw@sifive.com>
4529 * config/riscv/pic.md: Rewrite.
4530 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
4532 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
4533 (SOFTF, default_load, softload, softstore): New.
4535 2018-08-28 Jeff Law <law@redhat.com>
4537 * fold-const.c (fold_binary_loc): Remove recently added assert.
4539 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
4541 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
4542 to OP parmeter of generated function.
4544 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
4546 PR tree-optimization/87009
4547 * match.pd: Add boolean optimizations.
4549 2018-08-28 Martin Sebor <msebor@redhat.com>
4552 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
4553 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
4554 (pass_walloca::gate): Use it.
4555 (alloca_call_type): Same.
4556 (pass_walloca::execute): Same.
4557 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
4559 2018-08-28 David Malcolm <dmalcolm@redhat.com>
4561 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
4562 GCC_VERSION for usage of "__gcc_dump_printf__" format from
4565 2018-08-28 Richard Biener <rguenther@suse.de>
4567 PR tree-optimization/87124
4568 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
4569 constants before looking up avail.
4571 2018-08-28 Jakub Jelinek <jakub@redhat.com>
4574 * calls.c (maybe_warn_nonstring_arg): Punt early if
4575 warn_stringop_overflow is zero. Don't call get_range_strlen
4576 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
4577 Swap comparison operands to have constants on rhs. Only use
4578 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
4579 increment lenrng[0].
4581 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
4583 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
4584 use of tree_to_shwi. Remove duplicated test for the size being
4585 a whole number of bytes.
4587 2018-08-28 Richard Biener <rguenther@suse.de>
4589 PR tree-optimization/87117
4590 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
4591 Handle removed stmt without LHS (GIMPLE_NOP).
4593 2018-08-28 Richard Biener <rguenther@suse.de>
4595 PR tree-optimization/87117
4596 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
4597 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
4599 2018-08-28 Richard Biener <rguenther@suse.de>
4601 PR tree-optimization/87117
4602 * tree-ssa-pre.c (compute_avail): Do not make expressions
4603 with predicated values available.
4604 (get_expr_value_id): Assert we do not run into predicated value
4607 2018-08-28 Richard Biener <rguenther@suse.de>
4609 PR tree-optimization/87117
4610 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
4611 get virtual operands.
4612 (get_expr_operands): Handle STRING_CST like other decls.
4614 2018-08-28 Martin Liska <mliska@suse.cz>
4616 * tree.h: Update documentation of fndecl_built_in_p
4620 2018-08-27 Jeff Law <law@redhat.com>
4621 PR tree-optimization/87110
4622 * tree-ssa-dse.c (compute_trims): Handle non-constant
4625 2018-08-27 Martin Sebor <msebor@redhat.com>
4627 PR tree-optimization/86914
4628 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
4630 2018-08-27 Martin Sebor <msebor@redhat.com>
4632 PR tree-optimization/87112
4633 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
4634 the type of the bound argument.
4636 2018-08-27 Jeff Law <law@redhat.com>
4638 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
4639 type does not have a TYPE_SIZE_UNIT.
4641 2018-08-27 Steve Ellcey <sellcey@cavium.com>
4643 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
4644 with include of backend.h.
4646 2018-08-27 Richard Biener <rguenther@suse.de>
4648 PR tree-optimization/86927
4649 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
4650 use const cond reduction code.
4652 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
4654 PR tree-optimization/85758
4655 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
4657 2018-08-27 David Malcolm <dmalcolm@redhat.com>
4660 * diagnostic-show-locus.c (class layout_range): Update for
4661 conversion of show_caret_p to a tri-state.
4662 (layout_range::layout_range): Likewise.
4663 (make_range): Likewise.
4664 (layout::maybe_add_location_range): Likewise.
4665 (layout::should_print_annotation_line_p): Don't show annotation
4666 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
4667 (layout::get_state_at_point): Update for conversion of
4668 show_caret_p to a tri-state. Bail out early for
4669 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
4670 underlining or source colorization.
4671 (gcc_rich_location::add_location_if_nearby): Update for conversion
4672 of show_caret_p to a tri-state.
4673 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
4674 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
4676 (selftest::test_one_liner_labels): Likewise.
4677 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
4678 conversion of show_caret_p to a tri-state.
4679 * pretty-print.c (text_info::set_location): Likewise.
4680 * pretty-print.h (text_info::set_location): Likewise.
4681 * substring-locations.c (format_warning_n_va): Likewise.
4682 * tree-diagnostic.c (default_tree_printer): Likewise.
4683 * tree-pretty-print.c (newline_and_indent): Likewise.
4685 2018-08-27 David Malcolm <dmalcolm@redhat.com>
4688 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
4689 line above for line-insertion fix-it hints.
4690 (selftest::test_fixit_insert_containing_newline): Update the
4691 expected results, and add a test with line-numbering enabled.
4693 2018-08-27 Martin Liska <mliska@suse.cz>
4696 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
4697 params with DECL_HAS_VALUE_EXPR_P.
4699 2018-08-27 Martin Liska <mliska@suse.cz>
4701 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
4702 selected expansion strategy.
4704 2018-08-27 Martin Liska <mliska@suse.cz>
4706 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
4707 * builtins.c (is_builtin_fn): Likewise.
4708 * attribs.c (diag_attr_exclusions): Use new function
4709 fndecl_built_in_p and remove check for FUNCTION_DECL if
4711 (builtin_mathfn_code): Likewise.
4712 (fold_builtin_expect): Likewise.
4713 (fold_call_expr): Likewise.
4714 (fold_builtin_call_array): Likewise.
4715 (fold_call_stmt): Likewise.
4716 (set_builtin_user_assembler_name): Likewise.
4717 (is_simple_builtin): Likewise.
4718 * calls.c (gimple_alloca_call_p): Likewise.
4719 (maybe_warn_nonstring_arg): Likewise.
4720 * cfgexpand.c (expand_call_stmt): Likewise.
4721 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
4722 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
4723 (cgraph_node::verify_node): Likewise.
4724 * cgraphclones.c (build_function_decl_skip_args): Likewise.
4725 (cgraph_node::create_clone): Likewise.
4726 * config/arm/arm.c (arm_insert_attributes): Likewise.
4727 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
4728 * dse.c (scan_insn): Likewise.
4729 * expr.c (expand_expr_real_1): Likewise.
4730 * fold-const.c (operand_equal_p): Likewise.
4731 (fold_binary_loc): Likewise.
4732 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
4733 * gimple-low.c (lower_stmt): Likewise.
4734 * gimple-pretty-print.c (dump_gimple_call): Likewise.
4735 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
4737 * gimple.c (gimple_build_call_from_tree): Likewise.
4738 (gimple_call_builtin_p): Likewise.
4739 (gimple_call_combined_fn): Likewise.
4740 * gimplify.c (gimplify_call_expr): Likewise.
4741 (gimple_boolify): Likewise.
4742 (gimplify_modify_expr): Likewise.
4743 (gimplify_addr_expr): Likewise.
4744 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
4745 * ipa-cp.c (determine_versionability): Likewise.
4746 * ipa-fnsummary.c (compute_fn_summary): Likewise.
4747 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
4748 * ipa-split.c (visit_bb): Likewise.
4749 (split_function): Likewise.
4750 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4751 * lto-cgraph.c (input_node): Likewise.
4752 * lto-streamer-out.c (write_symbol): Likewise.
4753 * omp-low.c (setjmp_or_longjmp_p): Likewise.
4754 (lower_omp_1): Likewise.
4755 * predict.c (strip_predict_hints): Likewise.
4756 * print-tree.c (print_node): Likewise.
4757 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
4758 * trans-mem.c (is_tm_irrevocable): Likewise.
4759 (is_tm_load): Likewise.
4760 (is_tm_simple_load): Likewise.
4761 (is_tm_store): Likewise.
4762 (is_tm_simple_store): Likewise.
4763 (is_tm_abort): Likewise.
4764 (tm_region_init_1): Likewise.
4765 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
4766 * tree-cfg.c (verify_gimple_call): Likewise.
4767 (move_stmt_r): Likewise.
4768 (stmt_can_terminate_bb_p): Likewise.
4769 * tree-eh.c (lower_eh_constructs_2): Likewise.
4770 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
4771 * tree-inline.c (remap_gimple_stmt): Likewise.
4772 (copy_bb): Likewise.
4773 (estimate_num_insns): Likewise.
4774 (fold_marked_statements): Likewise.
4775 * tree-sra.c (scan_function): Likewise.
4776 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
4777 (optimize_stack_restore): Likewise.
4778 (pass_fold_builtins::execute): Likewise.
4779 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
4780 (mark_all_reaching_defs_necessary_1): Likewise.
4781 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
4782 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4783 (pass_forwprop::execute): Likewise.
4784 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4785 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4786 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
4787 * tree-ssa-strlen.c (get_string_length): Likewise.
4788 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
4789 (find_func_aliases_for_call): Likewise.
4790 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
4791 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
4792 * tree-tailcall.c (find_tail_calls): Likewise.
4793 * tree.c (need_assembler_name_p): Likewise.
4794 (free_lang_data_in_decl): Likewise.
4795 (get_call_combined_fn): Likewise.
4796 * ubsan.c (is_ubsan_builtin_p): Likewise.
4797 * varasm.c (incorporeal_function_p): Likewise.
4798 * tree.h (DECL_BUILT_IN): Remove and replace with
4800 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
4801 (fndecl_built_in_p): New.
4803 2018-08-27 Martin Liska <mliska@suse.cz>
4805 PR tree-optimization/86847
4806 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
4807 Dump also subtree probability.
4808 (switch_decision_tree::do_jump_if_equal): New function.
4809 (switch_decision_tree::emit_case_nodes): Handle special
4810 situations in balanced tree that can be emitted much simpler.
4811 Fix calculation of probabilities that happen in tree expansion.
4812 * tree-switch-conversion.h (struct cluster): Add
4814 (struct simple_cluster): Likewise.
4815 (struct case_tree_node): Add new function has_child.
4816 (do_jump_if_equal): New.
4818 2018-08-27 Martin Liska <mliska@suse.cz>
4820 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
4821 Add new argument to bit_test_cluster constructor.
4822 (bit_test_cluster::emit): Set bits really number of values
4824 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
4825 probability argument.
4826 * tree-switch-conversion.h (struct bit_test_cluster):
4827 Add m_handles_entire_switch.
4829 2018-08-27 Martin Liska <mliska@suse.cz>
4831 PR tree-optimization/86702
4832 * tree-switch-conversion.c (jump_table_cluster::emit):
4833 Make probabilities even for values in jump table
4834 according to number of cases handled.
4835 (switch_decision_tree::compute_cases_per_edge): Pass
4836 argument to reset_out_edges_aux function.
4837 (switch_decision_tree::analyze_switch_statement): Likewise.
4838 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
4841 2018-08-27 Martin Liska <mliska@suse.cz>
4843 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
4844 cfun argument explicitly.
4845 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
4846 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
4847 function gimple_switch_default_bb.
4848 (convert_switch_statements):
4850 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
4851 * stmt.c (label_to_block_fn): Use label_to_block and pass
4852 cfun argument explicitly and use gimple_switch_label_bb.
4853 (expand_case): Likewise.
4854 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
4855 cfun argument explicitly. Likewise.
4856 (make_edges_bb): Likewise.
4857 (make_cond_expr_edges): Likewise.
4858 (get_cases_for_edge): Likewise.
4859 (make_gimple_switch_edges): Likewise.
4860 (label_to_block_fn): Likewise.
4861 (label_to_block): Likewise.
4862 (make_goto_expr_edges): Likewise.
4863 (make_gimple_asm_edges): Likewise.
4864 (main_block_label): Likewise.
4865 (group_case_labels_stmt): Likewise.
4866 (find_taken_edge_computed_goto): Likewise.
4867 (find_taken_edge_switch_expr): Likewise.
4868 (gimple_verify_flow_info): Likewise.
4869 (gimple_redirect_edge_and_branch): Likewise.
4870 (gimple_switch_label_bb): New function.
4871 (gimple_switch_default_bb): Likewise.
4872 (gimple_switch_edge): Likewise.
4873 (gimple_switch_default_edge): Likewise.
4874 * tree-cfg.h (label_to_block_fn): Remove and replace ...
4875 (label_to_block): ... with this.
4876 (gimple_switch_label_bb): New.
4877 (gimple_switch_default_bb): Likewise.
4878 (gimple_switch_edge): Likewise.
4879 (gimple_switch_default_edge): Likewise.
4880 * tree-cfgcleanup.c (convert_single_case_switch): Use
4881 new gimple functions and pass new argument to label_to_block.
4882 (cleanup_control_flow_bb):
4883 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
4884 cfun argument explicitly.
4885 (make_eh_edges): Likewise.
4886 (redirect_eh_dispatch_edge): Likewise.
4887 (lower_resx): Likewise.
4888 (lower_eh_dispatch): Likewise.
4889 (maybe_remove_unreachable_handlers): Likewise.
4890 (unsplit_eh): Likewise.
4891 (cleanup_empty_eh): Likewise.
4892 (verify_eh_edges): Likewise.
4893 (verify_eh_dispatch_edge): Likewise.
4894 * tree-ssa-dom.c (record_edge_info): Likewise.
4895 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
4896 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
4897 (thread_through_normal_block): Likewise.
4898 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
4899 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
4900 * tree-switch-conversion.c (switch_conversion::collect): Use new
4902 (switch_conversion::check_final_bb): Likewise.
4903 (switch_conversion::gather_default_values): Pass new argument
4905 (switch_conversion::build_constructors): Likewise.
4906 (switch_decision_tree::compute_cases_per_edge): Use new
4907 gimple_switch_edge function.
4908 (switch_decision_tree::analyze_switch_statement): Pass new argument
4910 (switch_decision_tree::try_switch_expansion): Use
4911 gimple_switch_default_edge.
4912 * tree-vrp.c (find_switch_asserts): Pass new argument
4914 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4915 (vr_values::simplify_switch_using_ranges): Likewise.
4917 2018-08-27 Richard Biener <rguenther@suse.de>
4919 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4920 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4922 * tree-ssa-sccvn.h (struct vn_pval): New structure.
4923 (struct vn_nary_op_s): Add unwind_to member. Add
4924 predicated_values flag and put result into a union together
4925 with a linked list of vn_pval.
4926 (struct vn_ssa_aux): Add name member to make maintaining
4927 a map of SSA name to vn_ssa_aux possible. Remove no longer
4928 needed info, dfsnum, low, visited, on_sccstack, use_processed
4929 and range_info_anti_range_p members.
4930 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4931 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4933 (vn_valueize): New global.
4934 (vn_context_bb): Likewise.
4935 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4936 VN_INFO_PTR_INFO): Remove.
4937 * tree-ssa-sccvn.c: ... (rewrite)
4938 (pass_fre::execute): For -O2+ initialize loops and run
4939 RPO VN in optimistic mode (iterating). For -O1 and -Og
4940 run RPO VN in non-optimistic mode.
4941 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4942 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4943 * doc/invoke.texi (sccvn-max-scc-size): Remove.
4944 (rpo-vn-max-loop-depth): Document.
4945 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4946 when valuezing the VUSE signals we walked out of the region.
4947 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4948 (phi_translate): Set VN context block to use for availability
4950 (compute_avail): Likewise.
4951 (pre_valueize): New function.
4952 (pass_pre::execute): Adjust to the RPO VN API.
4954 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4955 (propagate_constants_for_unrolling): Remove.
4956 (tree_unroll_loops_completely): Perform value-numbering
4957 on the unrolled bodies loop parent.
4959 2018-08-27 Richard Biener <rguenther@suse.de>
4961 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4962 for partial antic compute.
4964 2018-08-27 Jakub Jelinek <jakub@redhat.com>
4966 PR rtl-optimization/87065
4967 * combine.c (simplify_if_then_else): Formatting fix.
4968 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4970 (known_cond): Don't return const_true_rtx for vector modes. Use
4971 CONST0_RTX instead of const0_rtx. Formatting fixes.
4973 2018-08-27 Martin Liska <mliska@suse.cz>
4975 PR gcov-profile/87069
4976 * gcov.c (process_file): Record files already processed
4977 and warn about a file being processed multiple times.
4979 2018-08-27 Martin Liska <mliska@suse.cz>
4982 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4983 Set default values for x_aarch64_*_string strings.
4984 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4985 prefix. For -mabi do not print '=ABI' in help and use
4986 <option_value> format for -msve-vector-bits and -moverride
4989 2018-08-26 Jeff Law <law@redhat.com>
4991 * config/mips/frame-header-opt.c: Include "backend.h" rather than
4994 2018-08-26 Marek Polacek <polacek@redhat.com>
4996 PR c++/87029, Implement -Wredundant-move.
4997 * doc/invoke.texi: Document -Wredundant-move.
4999 2018-08-25 Martin Sebor <msebor@redhat.com>
5001 PR tree-optimization/87059
5002 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
5003 to the same type as the other.
5004 * fold-const.c (fold_binary_loc): Assert expectation.
5006 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
5008 * config/darwin.c (machopic_legitimize_pic_address): Clean up
5009 extraneous parentheses, dead code section and formatting.
5011 2018-08-24 David Malcolm <dmalcolm@redhat.com>
5014 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
5015 wide enough for jumps in the line-numbering to be visible.
5016 (layout::print_gap_in_line_numbering): New member function.
5017 (layout::calculate_line_spans): When using line numbering, merge
5018 line spans that are only 1 line apart.
5019 (diagnostic_show_locus): When printing line numbers, show gaps in
5020 line numbering directly, rather than printing headers.
5021 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
5022 line-numbering with multiple line spans.
5023 (selftest::test_fixit_insert_containing_newline_2): Add test of
5024 line-numbering, in which the spans are close enough to be merged.
5026 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
5028 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
5029 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
5030 * tree-vrp.c (range_is_nonnull): Remove.
5031 (range_includes_zero_p): Accept value_range instead of min/max.
5032 (extract_range_from_binary_expr_1): Do not early bail on
5034 Use range_includes_zero_p instead of range_is_nonnull.
5035 (extract_range_from_unary_expr): Use range_includes_zero_p instead
5036 of range_is_nonnull.
5037 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
5038 special case VR_ANTI_RANGE.
5039 (vrp_finalize): Same.
5040 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
5042 (range_is_nonnull): Remove.
5043 * vr-values.c (vrp_stmt_computes_nonzero): Use
5044 range_includes_zero_p instead of range_is_nonnull.
5045 (extract_range_basic): Pass value_range to range_includes_zero_p
5046 instead of range_is_nonnull.
5048 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
5050 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
5051 * emit-rtl.h (rtl_data): Remove return_bnd.
5052 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
5053 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
5054 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
5055 (POINTER_BOUNDS_MODE): Remove definition.
5056 (make_pointer_bounds_mode): Remove.
5057 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
5058 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
5059 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
5060 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
5061 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
5062 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
5063 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
5065 * config/i386/i386-modes.def (BND32, BND64): Remove.
5066 * config/i386/i386.c (dbx_register_map): Remove bound registers.
5067 (dbx64_register_map): Ditto.
5068 (svr4_dbx_register_map): Ditto.
5069 (indirect_thunk_bnd_needed): Remove.
5070 (indirect_thunks_bnd_used): Ditto.
5071 (indirect_return_bnd_needed): Ditto.
5072 (indirect_return_via_cx_bnd): Ditto.
5073 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
5074 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
5075 (output_indirect_thunk): Ditto. Remove need_prefix argument.
5076 (output_indirect_thunk_function): Remove handling of
5077 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
5078 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
5079 (ix86_save_reg): Remove handling of crtl->return_bnd.
5080 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
5081 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
5082 and UNSPEC_BNDLX_ADDR.
5083 (ix86_output_indirect_branch_via_reg): Remove handling of
5084 indirect_thunk_prefix_bnd.
5085 (ix86_output_indirect_branch_via_push): Ditto.
5086 (ix86_output_function_return): Ditto.
5087 (ix86_output_indirect_function_return): Ditto.
5088 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
5089 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
5090 (CALL_USED_REGISTERS): Ditto.
5091 (REG_ALLOC_ORDER): Update for removal of bound registers.
5092 (HI_REGISTER_NAMES): Ditto.
5093 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
5094 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
5095 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
5096 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
5097 (FIRST_PSEUDO_REG): Update.
5098 (BND): Remove mode iterator.
5099 * config/i386/predicates.md (bnd_mem_operator): Remove.
5101 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
5103 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
5106 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
5108 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5109 the case in which the permute needs only a single element and
5110 repeats for every vector of the result. Extend that case to
5111 handle variable-length vectors.
5112 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5114 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
5117 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5118 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5120 2018-08-24 Richard Biener <rguenther@suse.de>
5122 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5123 bb_flags_allocated members.
5124 (auto_flag): New RAII class for allocating flags.
5125 (auto_edge_flag): New RAII class for allocating edge flags.
5126 (auto_bb_flag): New RAII class for allocating bb flags.
5127 * cfgloop.c (verify_loop_structure): Allocate temporary edge
5129 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5130 in favor of temporarily allocated BB flag.
5131 * hsa-brig.c: Re-order includes.
5132 * hsa-dump.c: Likewise.
5133 * hsa-regalloc.c: Likewise.
5134 * print-rtl.c: Likewise.
5135 * profile-count.c: Likewise.
5137 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
5140 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5143 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
5146 * wide-int-range.cc (wide_int_range_div): Do not ignore result
5147 from wide_int_range_multiplicative_op.
5149 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5151 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5154 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
5156 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
5159 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
5161 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
5162 full GENERIC expression used for replacement.
5164 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
5166 * tree-vrp.c (abs_extent_range): Remove.
5167 (extract_range_into_wide_ints): Pass wide ints by reference.
5168 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5169 Pass wide ints by reference in all calls to
5170 extract_range_into_wide_ints.
5171 * wide-int-range.cc (wide_int_range_div): New.
5172 * wide-int-range.h (wide_int_range_div): New.
5173 (wide_int_range_includes_zero_p): New.
5174 (wide_int_range_zero_p): New.
5176 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
5178 * config/aarch64/aarch64.md (arches): New enum.
5179 (arch): New enum attr.
5180 (arch_enabled): New attr.
5181 (enabled): Now uses arch_enabled only.
5182 (simd, sve, fp16): Removed attribute.
5183 (fp): Attr now defined in terms of 'arch'.
5184 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5185 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5186 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5187 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5188 attributes into 'arch'.
5189 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5190 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5191 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5192 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5193 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5194 'simd' attribute into 'arch'.
5195 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5196 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5197 Convert use of 'fp' attribute to 'arch'.
5198 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5199 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5201 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5202 (different modes) Merge 'fp' and 'simd' into 'arch'.
5203 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5206 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
5208 PR rtl-optimization/87026
5209 * expmed.c (canonicalize_comparison): If we can no longer create
5210 pseudoregisters, don't.
5212 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
5215 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
5217 * config/arm/arm.c (speculation_barrier_libfunc): New static
5219 (arm_init_libfuncs): Initialize it.
5220 (arm_emit_speculation_barrier): New function.
5221 * config/arm/arm.md (speculation_barrier): Call
5222 arm_emit_speculation_barrier for architectures that do not have
5224 (speculation_barrier_insn): Only match on Armv7 or later.
5226 2018-08-23 Richard Biener <rguenther@suse.de>
5229 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
5232 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
5234 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
5235 of single-vector TBLs.
5236 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
5239 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
5242 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
5243 aarch64_evpc_tbl guard.
5245 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
5247 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
5250 2018-08-22 Martin Sebor <msebor@redhat.com>
5253 * tree-pretty-print.c (pretty_print_string): Add argument.
5254 (dump_generic_node): Call to pretty_print_string with string size.
5256 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
5258 PR rtl-optimization/86771
5259 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
5260 of two SETs into those two SETs, one to be placed at i2, if that SETs
5261 destination is modified between i2 and i3.
5263 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5265 PR tree-optimization/86725
5266 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
5268 (vect_analyze_scalar_cycles_1): Check it.
5270 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5272 PR tree-optimization/86725
5273 * tree-vect-loop.c (vect_is_simple_reduction): When treating
5274 an outer loop phi as a double reduction, make sure that the
5275 single user of the phi result is an inner loop phi.
5277 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5279 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
5280 grouped stores with gaps to a strided group.
5282 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5284 * tree-vect-stmts.c (get_group_load_store_type)
5285 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
5286 first statement in a group.
5288 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5290 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
5291 the sequence used in gcc/gcc.c.
5293 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5296 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
5299 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5301 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
5302 Darwin10-specific unwinder-shim.
5303 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
5304 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
5305 New to cater for Darwin10 Rosetta.
5307 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5309 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
5310 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
5312 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5317 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
5318 (dwarf2out_switch_text_section): Generate a local label for the
5319 second function sub-section and apply it as the second FDE start
5321 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
5322 second sub-section start.
5324 2018-08-22 Richard Biener <rguenther@suse.de>
5326 PR tree-optimization/86988
5327 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
5329 2018-08-22 Richard Biener <rguenther@suse.de>
5331 PR tree-optimization/86945
5332 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
5334 2018-08-22 Alexandre Oliva <oliva@adacore.com>
5336 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
5337 a comment about how uses of r2 for .sdata2 come about.
5339 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
5341 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
5343 2018-08-21 Marek Polacek <polacek@redhat.com>
5345 PR c++/86981, Implement -Wpessimizing-move.
5346 * doc/invoke.texi: Document -Wpessimizing-move.
5348 2018-08-21 Jan Hubicka <jh@suse.cz>
5350 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
5351 * tree.h (is_redundant_typedef): Remove.
5352 * dwarf2out.c (is_redundant_typedef): Turn into static function.
5354 2018-08-21 Jan Hubicka <jh@suse.cz>
5356 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
5359 2018-08-21 Tamar Christina <tamar.christina@arm.com>
5361 * expmed.c (extract_low_bits): Reject invalid subregs early.
5363 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
5366 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
5369 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5371 * config/vxworks.h: Guard vxworks_asm_out_constructor and
5372 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
5373 * config/vxworks.c: Likewise.
5375 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5377 * config/vxworks.c: Set targetm.have_ctors_dtors
5378 if HAVE_INITFINI_ARRAY_SUPPORT.
5379 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
5380 if HAVE_INITFINI_ARRAY_SUPPORT.
5382 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5384 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
5385 default search path for VxWorks < 7.
5387 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
5389 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
5390 (get_format_string): Refer to c_getstr.
5392 2018-08-21 Tom de Vries <tdevries@suse.de>
5394 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5395 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5396 (debuginfo_early_stop): Declare.
5397 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5398 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5399 (debuginfo_early_stop): New function.
5400 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
5401 and debuginfo_early_stop.
5402 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
5403 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
5404 (general_init): Call debuginfo_early_init.
5405 (finalize): Call debuginfo_fini.
5406 (do_compile): Call debuginfo_init.
5407 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
5409 (@item -fdump-debug, @item -fdump-earlydebug): Add.
5411 2018-08-21 Tom de Vries <tdevries@suse.de>
5413 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
5414 flag_dump_noaddr and flag_dump_unnumbered.
5416 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
5418 * wide-int-range.cc (wide_int_range_abs): New.
5419 (wide_int_range_order_set): Rename from wide_int_range_min_max.
5420 * wide-int-range.h (wide_int_range_abs): New.
5421 (wide_int_range_min_max): New.
5422 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
5423 case to call wide_int_range_abs.
5424 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
5425 (extract_range_from_abs_expr): Delete.
5427 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
5430 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
5431 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
5434 2018-08-20 Nathan Sidwell <nathan@acm.org>
5435 Jeff Law <law@redhat.com>
5437 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
5438 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
5440 2018-08-20 David Malcolm <dmalcolm@redhat.com>
5443 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
5444 (decl_attributes): Likewise.
5445 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
5447 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5448 * diagnostic-core.h (class auto_diagnostic_group): New class.
5449 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
5450 (diagnostic_report_diagnostic): Handle the first diagnostics within
5452 (emit_diagnostic): Add auto_diagnostic_group instance.
5454 (inform_n): Likewise.
5455 (warning): Likewise.
5456 (warning_at): Likewise.
5457 (warning_n): Likewise.
5458 (pedwarn): Likewise.
5459 (permerror): Likewise.
5461 (error_n): Likewise.
5462 (error_at): Likewise.
5464 (fatal_error): Likewise.
5465 (internal_error): Likewise.
5466 (internal_error_no_backtrace): Likewise.
5467 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
5468 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
5469 * diagnostic.h (struct diagnostic_context): Add fields
5470 "diagnostic_group_nesting_depth",
5471 "diagnostic_group_emission_count", "begin_group_cb",
5473 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
5474 Add auto_diagnostic_group instance(s).
5475 (find_explicit_erroneous_behavior): Likewise.
5476 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
5477 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
5478 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
5479 (gimplify_va_arg_expr): Likewise.
5480 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
5481 (HSA_SORRY_AT): Likewise.
5482 * ipa-devirt.c (compare_virtual_tables): Likewise.
5483 (warn_odr): Likewise.
5484 * multiple_target.c (expand_target_clones): Likewise.
5485 * opts-common.c (cmdline_handle_error): Likewise.
5486 * reginfo.c (globalize_reg): Likewise.
5487 * substring-locations.c (format_warning_n_va): Likewise.
5488 * tree-inline.c (expand_call_inline): Likewise.
5489 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
5490 * tree-ssa-loop-niter.c
5491 (do_warn_aggressive_loop_optimizations): Likewise.
5492 * tree-ssa-uninit.c (warn_uninit): Likewise.
5493 * tree.c (warn_deprecated_use): Likewise.
5495 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
5498 * config/i386/i386.md (eh_return): Always update EH return
5499 address in word_mode.
5501 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
5503 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
5504 TARGET_SPLIT_COMPLEX_ARG is defined.
5506 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
5508 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
5510 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
5513 * expr.c (expand_assignment): Assert that bitpos is positive.
5514 (store_field): Likewise
5515 (expand_expr_real_1): Make sure that bitpos is positive.
5516 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
5519 2018-08-20 Nathan Sidwell <nathan@acm.org>
5521 * Makefile.in (CPP_ID_DATA_H): Delete.
5522 (CPP_INTERNAL_H): Don't add it.
5523 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
5524 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
5526 2018-08-20 Richard Biener <rguenther@suse.de>
5528 PR tree-optimization/78655
5529 * tree-vrp.c (extract_range_from_binary_expr_1): Make
5530 pointer + offset nonnull if either operand is nonnull work.
5532 2018-08-20 Tom de Vries <tdevries@suse.de>
5534 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
5535 unless the referenced die describes the added property using
5536 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
5537 Otherwise, add a DW_AT_location to the referenced die.
5539 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
5542 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
5543 register_operand when calling ix86_set_reg_reg_cost.
5544 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
5545 Set *total to 0 for operands that satisfy x86_64_immediate_operand
5546 predicate and to 1 otherwise.
5548 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
5550 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
5551 emit a diagnostic that it is not supported and reset the option.
5552 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
5553 supported and consume the option. (ASM_FINAL_SPEC): New.
5555 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
5557 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
5560 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
5562 C-SKY port: Documentation
5564 * doc/extend.texi (C-SKY Function Attributes): New section.
5565 * doc/invoke.texi (Option Summary): Add C-SKY options.
5566 (C-SKY Options): New section.
5567 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
5569 2018-08-17 Jojo <jijie_rong@c-sky.com>
5570 Huibin Wang <huibin_wang@c-sky.com>
5571 Sandra Loosemore <sandra@codesourcery.com>
5572 Chung-Lin Tang <cltang@codesourcery.com>
5574 C-SKY port: Backend implementation
5576 * config/csky/*: New.
5577 * common/config/csky/*: New.
5579 2018-08-17 Jojo <jijie_rong@c-sky.com>
5580 Huibin Wang <huibin_wang@c-sky.com>
5581 Sandra Loosemore <sandra@codesourcery.com>
5582 Chung-Lin Tang <cltang@codesourcery.com>
5583 Andrew Jenner <andrew@codesourcery.com>
5585 C-SKY port: Configury
5587 * config.gcc (csky-*-*): New.
5588 * configure.ac: Add csky to targets for dwarf2 debug_line support.
5589 * configure: Regenerated.
5591 2018-08-17 David Malcolm <dmalcolm@redhat.com>
5593 * dump-context.h: Include "dumpfile.h".
5594 (dump_context::dump_printf_va): Convert final param from va_list
5595 to va_list *. Convert from ATTRIBUTE_PRINTF to
5596 ATTRIBUTE_GCC_DUMP_PRINTF.
5597 (dump_context::dump_printf_loc_va): Likewise.
5598 * dumpfile.c: Include "stringpool.h".
5599 (make_item_for_dump_printf_va): Delete.
5600 (make_item_for_dump_printf): Delete.
5601 (class dump_pretty_printer): New class.
5602 (dump_pretty_printer::dump_pretty_printer): New ctor.
5603 (dump_pretty_printer::emit_items): New member function.
5604 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
5606 (dump_pretty_printer::emit_item): New member function.
5607 (dump_pretty_printer::stash_item): New member function.
5608 (dump_pretty_printer::format_decoder_cb): New member function.
5609 (dump_pretty_printer::decode_format): New member function.
5610 (dump_context::dump_printf_va): Reimplement in terms of
5611 dump_pretty_printer.
5612 (dump_context::dump_printf_loc_va): Convert final param from va_list
5614 (dump_context::begin_scope): Reimplement call to
5615 make_item_for_dump_printf.
5616 (dump_printf): Update for change to dump_printf_va.
5617 (dump_printf_loc): Likewise.
5618 (selftest::test_capture_of_dump_calls): Convert "stmt" from
5619 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
5620 with %T, %E, and %G.
5621 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
5622 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
5623 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
5624 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
5625 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
5626 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
5627 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
5628 within a dump_printf_loc call to "%wu".
5629 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
5630 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
5631 missing space after "=".
5632 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
5633 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
5634 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
5635 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
5636 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
5637 duplicate "vectorized" from message.
5639 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
5641 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
5642 polyNxK_t element's TYPE_STRING_FLAG.
5644 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
5646 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
5647 (they were unnamed before). Fix comments.
5649 2018-08-17 Nathan Sidwell <nathan@acm.org>
5651 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
5653 2018-08-17 Richard Biener <rguenther@suse.de>
5655 PR tree-optimization/86841
5656 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
5658 2018-08-17 Martin Liska <mliska@suse.cz>
5660 * common.opt: Remove Warn, Init and Report for options with
5661 Ignore/Deprecated flag. Warning is done automatically for
5663 * config/i386/i386.opt: Likewise.
5664 * config/ia64/ia64.opt: Likewise.
5665 * config/rs6000/rs6000.opt: Likewise.
5666 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5667 Remove usage of flag_check_pointer_bounds.
5668 * lto-wrapper.c (merge_and_complain): Do not handle
5669 OPT_fcheck_pointer_bounds.
5670 (append_compiler_options): Likewise.
5671 * opt-functions.awk: Do not handle Deprecated.
5672 * optc-gen.awk: Check that Var, Report and Init are not
5673 used for an option with Ignore/Deprecated flag.
5674 * opts-common.c (decode_cmdline_option): Do not report
5676 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
5678 * opts.h (struct cl_option): Remove cl_deprecated flag.
5679 (CL_ERR_DEPRECATED): Remove error enum value.
5681 2018-08-17 Richard Biener <rguenther@suse.de>
5684 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
5685 across a va-arg-pack using call adjust its return value accordingly.
5687 2018-08-16 Martin Sebor <msebor@redhat.com>
5689 PR tree-optimization/86853
5690 * gimple-ssa-sprintf.c (struct format_result): Rename member.
5691 (struct fmtresult): Add member and initialize it in ctors.
5692 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
5693 (format_string): Handle %S the same as %ls. Set MAYFAIL.
5694 (format_directive): Set POSUNDER4K when MAYFAIL is set.
5695 (parse_directive): Handle %C same as %c.
5696 (sprintf_dom_walker::compute_format_length): Adjust.
5697 (is_call_safe): Adjust.
5699 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5701 * builtins.c (c_strlen): Add new parameter eltsize. Use it
5702 for determining how to count the elements.
5703 * builtins.h (c_strlen): Adjust prototype.
5704 * expr.c (string_constant): Add new parameter mem_size.
5705 Set *mem_size appropriately.
5706 * expr.h (string_constant): Adjust protoype.
5707 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
5708 * gimple-fold.h (get_range_strlen): Adjust prototype.
5709 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
5710 (format_string): Call get_string_length with eltsize.
5712 2018-08-16 David Malcolm <dmalcolm@redhat.com>
5714 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
5715 to emit the span, rather than setting it as the prefix.
5717 2018-08-16 David Malcolm <dmalcolm@redhat.com>
5719 * diagnostic-show-locus.c (layout::start_annotation_line): Add
5720 "margin_char" parameter, defaulting to space. Use it in place
5721 of pp_space for the initial part of the margin.
5722 (layout::print_leading_fixits): Use '+' when filling the margin
5723 of line-insertion fix-it hints.
5725 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5727 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
5730 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5732 * config/rs6000/altivec.md: Don't set length attribute to the default
5734 * config/rs6000/darwin.md: Ditto.
5735 * config/rs6000/dfp.md: Ditto.
5736 * config/rs6000/htm.md: Ditto.
5737 * config/rs6000/rs6000.md: Ditto.
5738 * config/rs6000/sync.md: Ditto.
5739 * config/rs6000/vsx.md: Ditto.
5741 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5743 * config/rs6000/altivec.md: Don't set length attribute to the default
5744 value, for branch instructions.
5745 * config/rs6000/darwin.md: Ditto.
5746 * config/rs6000/rs6000.md: Ditto.
5748 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5750 * config/rs6000/rs6000.md (length): Always define as const_int 4.
5751 (unnamed conditional branch define_insn): Set length to 4 or 8
5752 depending on offset.
5753 (<bd>_<mode>): Similar, for alternative 0.
5754 (<bd>tf_<mode>): Ditto.
5756 2018-08-16 Tamar Christina <tamar.christina@arm.com>
5758 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
5760 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
5762 * doc/rtl.texi: Replace old RTX class names with new names.
5765 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
5767 * expmed.h (canonicalize_comparison): New declaration.
5768 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
5769 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
5770 * optabs.c (prepare_cmp_insn): Likewise.
5771 * rtl.h (unsigned_condition_p): New function which checks if a
5772 comparison operator is unsigned.
5774 2018-08-16 Nathan Sidwell <nathan@acm.org>
5776 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
5777 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
5779 2018-08-16 Tamar Christina <tamar.christina@arm.com>
5782 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
5783 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
5784 (mov<mov>): ..this and enable unconditionally.
5786 2018-08-16 Tamar Christina <tamar.christina@arm.com>
5788 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
5790 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
5792 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
5793 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
5794 "Common" with "Target".
5796 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
5798 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
5799 * doc/invoke.texi (mmitigate-rop): Remove.
5800 * config/i386/i386.c: Do not include "regrename.h".
5801 (ix86_rop_should_change_byte_p, reg_encoded_number)
5802 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
5804 (ix86_reorg): Remove call to ix86_mitigate_rop.
5805 * config/i386/i386.md (attr "modrm_class"): Remove.
5806 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
5807 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
5808 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
5810 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
5812 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
5813 allow folding of mergeh() and mergel() for the float and double types.
5814 (fold_mergehl_helper): Rework to handle building a permute tree
5817 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
5819 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
5822 2018-08-15 David Malcolm <dmalcolm@redhat.com>
5824 * common.opt (fdiagnostics-show-labels): New option.
5825 * diagnostic-show-locus.c (class layout_range): Add field
5827 (class layout): Add field "m_show_labels_p".
5828 (layout_range::layout_range): Add param "label" and use it to
5830 (make_range): Pass in NULL for new "label" param of layout_range's
5832 (layout::layout): Initialize m_show_labels_p.
5833 (layout::maybe_add_location_range): Pass in loc_range->m_label
5834 when constructing layout_range instances.
5835 (struct line_label): New struct.
5836 (layout::print_any_labels): New member function.
5837 (layout::print_line): Call it if label-printing is enabled.
5838 (selftest::test_one_liner_labels): New test.
5839 (selftest::test_diagnostic_show_locus_one_liner): Call it.
5840 * diagnostic.c (diagnostic_initialize): Initialize
5841 context->show_labels_p.
5842 * diagnostic.h (struct diagnostic_context): Add field
5844 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5845 -fno-diagnostics-show-labels.
5846 * dwarf2out.c (gen_producer_string): Add
5847 OPT_fdiagnostics_show_labels to the ignored options.
5848 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
5850 (gcc_rich_location::maybe_add_expr): Likewise.
5851 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
5852 label" param, defaulting to NULL.
5853 (gcc_rich_location::add_expr): Add "label" param.
5854 (gcc_rich_location::maybe_add_expr): Likewise.
5855 (class text_range_label): New class.
5856 (class range_label_for_type_mismatch): New class.
5857 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
5858 of format_warning_va.
5859 (fmtwarn_n): Likewise for new params of format_warning_n_va.
5860 * lto-wrapper.c (merge_and_complain): Add
5861 OPT_fdiagnostics_show_labels to the "pick one setting" options.
5862 (append_compiler_options): Likewise to the dropped options.
5863 (append_diag_options): Likewise to the passed-on options.
5864 * opts.c (common_handle_option): Handle the new option.
5865 * selftest-diagnostic.c
5866 (test_diagnostic_context::test_diagnostic_context): Enable
5868 * substring-locations.c: Include "gcc-rich-location.h".
5869 (format_warning_n_va): Add "fmt_label" and "param_label" params
5870 and use them as appropriate.
5871 (format_warning_va): Add "fmt_label" and "param_label" params,
5872 passing them on to format_warning_n_va.
5873 (format_warning_at_substring): Likewise.
5874 (format_warning_at_substring_n): Likewise.
5875 * substring-locations.h (format_warning_va): Add "fmt_label" and
5876 "param_label" params.
5877 (format_warning_n_va): Likewise.
5878 (format_warning_at_substring): Likewise.
5879 (format_warning_at_substring_n): Likewise.
5880 * toplev.c (general_init): Initialize global_dc->show_labels_p.
5882 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
5885 * builtins.c (expand_builtin_memcmp): Do not expand the call
5886 when overflow is detected.
5888 2018-08-15 Martin Sebor <msebor@redhat.com>
5890 PR tree-optimization/71625
5891 * config/aarch64/aarch64-builtins.c
5892 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
5894 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
5896 * config/s390/s390.c (s390_reorg): Remove loop.
5898 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
5901 (darwin_function_switched_text_sections): Delete.
5902 * gcc/config/darwin.h
5903 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
5905 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
5908 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
5909 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
5910 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
5912 2018-08-15 Martin Liska <mliska@suse.cz>
5914 PR tree-optimization/86925
5915 * predict.c (expr_expected_value_1): When taking
5916 later predictor, assign also probability.
5917 Use fold_build2_initializer_loc in order to fold
5918 the expression in -frounding-math.
5920 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
5922 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5924 (expand_vec_perm_1): Try the new method.
5926 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
5929 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5930 Check whether lra_live_max_point is 0 before dividing.
5932 2018-08-14 Martin Sebor <msebor@redhat.com>
5934 PR tree-optimization/86650
5935 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5936 (vrp_prop::check_mem_ref): Same.
5938 2018-08-13 Liu Hao <lh_mouse@126.com>
5940 * pretty-print.c (eat_esc_sequence): Swap the foreground and
5941 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5942 and clear it thereafter, as it only works for DBCS.
5944 2018-08-13 Liu Hao <lh_mouse@126.com>
5946 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5947 handle returned by _get_osf_handle().
5949 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5951 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5952 for folding vec_perm.
5954 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5956 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5957 Add support for gimple-folding of vec_pack() and vec_unpack()
5960 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5962 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5963 vec_xst variants to the list.
5964 (rs6000_gimple_fold_builtin): Add support for folding unaligned
5965 vector loads and stores.
5967 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
5969 * config.gcc (rs6000-ibm-aix4.x): Delete.
5970 (rs6000-ibm-aix5.1): Delete.
5971 (rs6000-ibm-aix5.2): Delete.
5972 (rs6000-ibm-aix5.3): Delete.
5973 * config/rs6000/aix43.h: Delete.
5974 * config/rs6000/aix51.h: Delete.
5975 * config/rs6000/aix52.h: Delete.
5976 * config/rs6000/t-aix43: Delete.
5978 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
5980 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5982 (s390_decompose_address): Factor out constant pool ref
5985 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5987 * config/nds32/nds32-predicates.c
5988 (nds32_can_use_bclr_p): Change return type as bool.
5989 (nds32_can_use_bset_p): Ditto.
5990 (nds32_can_use_btgl_p): Ditto.
5991 (nds32_can_use_bitci_p): Ditto.
5992 * config/nds32/nds32-protos.h
5993 (nds32_can_use_bclr_p): Change declaration.
5994 (nds32_can_use_bset_p): Ditto.
5995 (nds32_can_use_btgl_p): Ditto.
5996 (nds32_can_use_bitci_p): Ditto.
5998 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
6000 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
6001 Support -msched-prolog-epilog option.
6002 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
6004 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
6006 * common/config/nds32/nds32-common.c
6007 (nds32_option_optimization_table): Enalbe -malways-align.
6009 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
6011 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
6013 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
6014 OPT_misr_secure_ case.
6015 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
6016 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
6017 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
6019 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
6020 (nds32_isr_info): New field security_level.
6021 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
6022 * config/nds32/nds32.md (return_internal): Consider critical attribute.
6023 * config/nds32/nds32.opt (misr-secure): New option.
6024 * config/nds32/nds32_init.inc: New file.
6025 * config/nds32/nds32_isr.h: New file.
6027 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
6029 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
6030 Update comment for atomic instructions.
6031 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
6032 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
6034 (atomic_loaddi): Revise fence expansion to only emit fence prior to
6035 load for __ATOMIC_SEQ_CST model.
6036 (atomic_loaddi_1): Remove float register target.
6037 (atomic_storedi): Handle CONST_INT values.
6038 (atomic_storedi_1): Remove float register source. Add special case
6040 (memory_barrier): New expander and insn.
6042 2018-08-11 Jakub Jelinek <jakub@redhat.com>
6044 PR tree-optimization/86835
6045 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
6046 new_stmt after def_gsi, make sure to insert new_square_stmt after
6047 that stmt, not 2 stmts before it.
6049 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
6052 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
6055 2018-08-10 Martin Liska <mliska@suse.cz>
6058 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
6060 * builtins.c (expand_builtin_expect_with_probability):
6062 (expand_builtin_expect_with_probability): New function.
6063 (build_builtin_expect_predicate): Add new argumnet probability
6064 for BUILT_IN_EXPECT_WITH_PROBABILITY.
6065 (fold_builtin_expect):
6068 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
6069 * builtins.h (fold_builtin_expect): Set new argument.
6070 * doc/extend.texi: Document __builtin_expect_with_probability.
6071 * doc/invoke.texi: Likewise.
6072 * gimple-fold.c (gimple_fold_call): Pass new argument.
6073 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
6074 also BUILT_IN_EXPECT_WITH_PROBABILITY.
6075 * predict.c (get_predictor_value): New function.
6076 (expr_expected_value): Add new argument probability. Assume
6077 that predictor and probability are always non-null.
6078 (expr_expected_value_1): Likewise. For __builtin_expect and
6079 __builtin_expect_with_probability set probability. Handle
6080 combination in binary expressions.
6081 (tree_predict_by_opcode): Simplify code by simply calling
6082 get_predictor_value.
6083 (pass_strip_predict_hints::execute): Add handling of
6084 BUILT_IN_EXPECT_WITH_PROBABILITY.
6085 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
6087 * tree.h (DECL_BUILT_IN_P): New function.
6089 2018-08-10 Martin Liska <mliska@suse.cz>
6091 PR tree-optimization/85799
6092 * passes.def: Add argument for pass_strip_predict_hints.
6093 * predict.c (class pass_strip_predict_hints): Add new argument
6095 (strip_predictor_early): New function.
6096 (pass_strip_predict_hints::execute): Call the function to
6098 (strip_predict_hints): New function.
6099 * predict.def: Fix comment.
6101 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
6103 * Makefile.in: Clarify which tm.texi to copy over to assert the
6104 right to grant a GFDL license for all.
6106 2018-08-09 Jeff Law <law@redhat.com>
6108 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6111 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
6113 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6116 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
6118 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6119 reductions for variable-length vectors.
6121 2018-08-09 David Malcolm <dmalcolm@redhat.com>
6124 * common.opt (fdiagnostics-show-line-numbers): New option.
6125 * diagnostic-show-locus.c (class layout): Add fields
6126 "m_show_line_numbers_p" and "m_linenum_width";
6127 (num_digits): New function.
6128 (test_num_digits): New function.
6129 (layout::layout): Initialize new fields. Update m_x_offset
6130 logic to handle any left margin.
6131 (layout::print_source_line): Print line number when requested.
6132 (layout::start_annotation_line): New member function.
6133 (layout::print_annotation_line): Call it.
6134 (layout::print_leading_fixits): Likewise.
6135 (layout::print_trailing_fixits): Likewise. Update calls to
6136 move_to_column for new parameter.
6137 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6138 it to potentially call start_annotation_line.
6139 (layout::show_ruler): Call start_annotation_line.
6140 (selftest::test_line_numbers_multiline_range): New selftest.
6141 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6142 and selftest::test_line_numbers_multiline_range.
6143 * diagnostic.c (diagnostic_initialize): Initialize
6144 show_line_numbers_p.
6145 * diagnostic.h (struct diagnostic_context): Add field
6146 "show_line_numbers_p".
6147 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6148 -fno-diagnostics-show-line-numbers.
6149 * dwarf2out.c (gen_producer_string): Add
6150 OPT_fdiagnostics_show_line_numbers to the ignored options.
6151 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6152 one setting" options.
6153 (append_compiler_options): Likewise to the dropped options.
6154 (append_diag_options): Likewise to the passed-on options.
6155 * opts.c (common_handle_option): Handle the new option.
6156 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
6158 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
6160 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
6161 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
6162 third argument of type "const signed char" to descriptions of
6163 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
6164 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
6165 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6166 __builtin_bcdsub_ov functions.
6168 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
6170 PR tree-optimization/86858
6171 * tree-vect-loop.c (vect_is_simple_reduction): Restore
6172 flow_bb_inside_loop_p calls.
6174 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
6176 PR tree-optimization/86871
6177 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6178 instead of gimple_assign_lhs.
6180 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
6183 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6184 register constraint to operand 0.
6185 (add<mode>3_carryinC): Likewise.
6186 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6188 2018-08-09 Martin Liska <mliska@suse.cz>
6191 * common.opt: Remove extra line.
6193 2018-08-09 Martin Liska <mliska@suse.cz>
6195 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6196 at the end of a line, make first letter capital and end up
6197 a sentence with a dot.
6198 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6199 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6200 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6201 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6202 (PARAM_MAX_ISL_OPERATIONS): Likewise.
6203 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6204 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6205 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6206 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6207 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6208 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6209 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6210 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
6211 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
6212 (PARAM_TREE_REASSOC_WIDTH): Likewise.
6213 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
6214 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
6215 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
6217 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
6220 * config/s390/s390.c (s390_option_override_internal): Reduce the
6221 stack-clash-protection-probe-interval param if it would be too big
6224 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
6227 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
6228 * config/m68k/m68k.c (handle_move_double): Don't call
6229 m68k_final_prescan_insn.
6230 (m68k_adjust_decorated_operand): Renamed from
6231 m68k_final_prescan_insn, remove first and third operand and
6233 (print_operand): Call it.
6234 (print_operand_address): Call it.
6236 2018-08-08 Nathan Sidwell <nathan@acm.org>
6238 * diagnostic.c (diagnostic_report_current_module): Use
6239 linemap_included_from & linemap_included_from_linemap.
6241 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
6243 * config/aarch64/aarch64-cores.def: Add phecda core.
6244 * config/aarch64/aarch64-tune.md: Regenerate.
6245 * doc/invoke.texi: Add phecda core.
6247 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
6250 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
6252 * config/s390/s390.md ("movti"): Add more alternatives for
6253 constant to GPR copies.
6255 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
6257 * config/s390/s390.c: Fix whitespace damage throughout the file.
6258 * config/s390/s390.h: Likewise.
6259 * config/s390/tpf.h: Likewise.
6261 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
6263 * config/s390/s390.c (s390_loadrelative_operand_p):
6264 Remove TARGET_CPU_ZARCH usages.
6265 (s390_rtx_costs): Likewise.
6266 (s390_legitimate_constant_p): Likewise.
6267 (s390_cannot_force_const_mem): Likewise.
6268 (legitimate_reload_constant_p): Likewise.
6269 (s390_preferred_reload_class): Likewise.
6270 (legitimize_pic_address): Likewise.
6271 (legitimize_tls_address): Likewise.
6272 (s390_split_branches): Removed.
6273 (s390_add_execute): Removed.
6274 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
6275 (s390_mainpool_start): Likewise.
6276 (s390_mainpool_finish): Likewise.
6277 (s390_mainpool_cancel): Removed.
6278 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
6279 (s390_chunkify_cancel): Likewise.
6280 (s390_return_addr_rtx): Likewise.
6281 (s390_register_info): Remove split_branches_pending_p uages.
6282 (s390_optimize_register_info): Likewise.
6283 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
6284 split_branches_pending_p usages.
6285 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
6286 (s390_load_got): Likewise.
6287 (s390_expand_split_stack_prologue): Likewise.
6288 (output_asm_nops): Likewise.
6289 (s390_function_profiler): Likewise.
6290 (s390_emit_call): Likewise.
6291 (s390_conditional_register_usage): Likewise.
6292 (s390_optimize_prologue): Likewise.
6293 (s390_reorg): Remove TARGET_CPU_ZARCH and
6294 split_branches_pending_p usages.
6295 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
6297 (s390_output_indirect_thunk_function): Likewise.
6298 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
6299 (TARGET_CPU_ZARCH_P): Removed.
6300 (struct machine_function): Remove split_branches_pending_p.
6301 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
6303 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
6305 * common/config/s390/s390-common.c (processor_flags_table):
6307 * config.gcc: Remove with_arch/with_tune support.
6308 * config/s390/2064.md: Remove cpu attribute comparisons.
6309 * config/s390/driver-native.c (s390_host_detect_local_cpu):
6311 * config/s390/linux.h (ASM_SPEC):
6312 Remove -march support.
6313 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
6314 Use a table to get an arch level.
6315 * config/s390/s390-opts.h (enum processor_type):
6317 * config/s390/s390.c
6318 (processor_table): Remove entries, add arch_level values.
6319 (s390_issue_rate): Remove cases.
6320 (s390_option_override): Adjust
6321 s390_option_override_internal() call.
6322 (s390_option_override_internal): Remove deprecation warning.
6323 (s390_valid_target_attribute_tree): Adjust
6324 s390_option_override_internal() call.
6325 * config/s390/s390.h (struct s390_processor):
6326 Share with s390-c.c, add arch_level field.
6327 * config/s390/s390.md:
6328 Remove occurrences in cpu attribute.
6329 * config/s390/s390.opt: Remove -march/-mtune support.
6330 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
6331 * doc/invoke.texi: Remove deprecation warning.
6333 2018-08-08 Luis Machado <luis.machado@linaro.org>
6335 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
6337 (qdf24xx_tunings): Set vector cost structure to
6338 qdf24xx_vector_cost.
6340 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
6341 <register_sextend>: Set to 3.
6343 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
6346 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
6347 * config/aarch64/aarch64-simd.md
6348 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
6349 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
6350 (aarch64_frecpx<mode>): ...this new pattern.
6351 * config/aarch64/aarch64-simd-builtins.def: Remove comment
6352 about aarch64_frecp<FRECP:frecp_suffix><mode>.
6354 2018-08-07 Martin Liska <mliska@suse.cz>
6357 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
6358 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
6359 * predict.def (PRED_MALLOC_NONNULL): New predictor.
6360 * doc/extend.texi: Document that malloc attribute adds
6363 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
6366 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6367 Define to speculation_safe_value_not_needed.
6369 2018-08-06 Jeff Law <law@redhat.com>
6371 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
6372 the vr_values instance to cprop_into_stmt.
6373 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
6374 (cprop_operand): Also query EVRP to determine if OP is a constant.
6376 2018-08-06 Nathan Sidwell <nathan@acm.org>
6378 * diagnostic.c (diagnostic_report_current_module): Reroll
6379 included-at loop. Translate text.
6381 2018-08-06 David Malcolm <dmalcolm@redhat.com>
6383 * function-tests.c (selftest::test_expansion_to_rtl): Call
6384 free_after_compilation.
6386 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6388 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
6390 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
6392 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
6393 loops with memory block operations from getting unrolled.
6395 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
6398 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6399 Define to speculation_safe_value_not_needed.
6401 2018-08-06 Jeff Law <law@redhat.com>
6403 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
6406 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6409 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
6410 with all enabled __intN types.
6412 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
6414 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6416 * alias.c (record_set): Check for clobber high.
6417 * cfgexpand.c (expand_gimple_stmt): Likewise.
6418 * combine-stack-adj.c (single_set_for_csa): Likewise.
6419 * combine.c (find_single_use_1): Likewise.
6420 (set_nonzero_bits_and_sign_copies): Likewise.
6421 (get_combine_src_dest): Likewise.
6422 (is_parallel_of_n_reg_sets): Likewise.
6423 (try_combine): Likewise.
6424 (record_dead_and_set_regs_1): Likewise.
6425 (reg_dead_at_p_1): Likewise.
6426 (reg_dead_at_p): Likewise.
6427 * dce.c (deletable_insn_p): Likewise.
6428 (mark_nonreg_stores_1): Likewise.
6429 (mark_nonreg_stores_2): Likewise.
6430 * df-scan.c (df_find_hard_reg_defs): Likewise.
6431 (df_uses_record): Likewise.
6432 (df_get_call_refs): Likewise.
6433 * dwarf2out.c (mem_loc_descriptor): Likewise.
6434 * haifa-sched.c (haifa_classify_rtx): Likewise.
6435 * ira-build.c (create_insn_allocnos): Likewise.
6436 * ira-costs.c (scan_one_insn): Likewise.
6437 * ira.c (equiv_init_movable_p): Likewise.
6438 (rtx_moveable_p): Likewise.
6439 (interesting_dest_for_shprep): Likewise.
6440 * jump.c (mark_jump_label_1): Likewise.
6441 * postreload-gcse.c (record_opr_changes): Likewise.
6442 * postreload.c (reload_cse_simplify): Likewise.
6443 (struct reg_use): Add source expr.
6444 (reload_combine): Check for clobber high.
6445 (reload_combine_note_use): Likewise.
6446 (reload_cse_move2add): Likewise.
6447 (move2add_note_store): Likewise.
6448 * print-rtl.c (print_pattern): Likewise.
6449 * recog.c (decode_asm_operands): Likewise.
6450 (store_data_bypass_p): Likewise.
6451 (if_test_bypass_p): Likewise.
6452 * regcprop.c (kill_clobbered_value): Likewise.
6453 (kill_set_value): Likewise.
6454 * reginfo.c (reg_scan_mark_refs): Likewise.
6455 * reload1.c (maybe_fix_stack_asms): Likewise.
6456 (eliminate_regs_1): Likewise.
6457 (elimination_effects): Likewise.
6458 (mark_not_eliminable): Likewise.
6459 (scan_paradoxical_subregs): Likewise.
6460 (forget_old_reloads_1): Likewise.
6461 * reorg.c (find_end_label): Likewise.
6462 (try_merge_delay_insns): Likewise.
6463 (redundant_insn): Likewise.
6464 (own_thread_p): Likewise.
6465 (fill_simple_delay_slots): Likewise.
6466 (fill_slots_from_thread): Likewise.
6467 (dbr_schedule): Likewise.
6468 * resource.c (update_live_status): Likewise.
6469 (mark_referenced_resources): Likewise.
6470 (mark_set_resources): Likewise.
6471 * rtl.c (copy_rtx): Likewise.
6472 * rtlanal.c (reg_referenced_p): Likewise.
6473 (single_set_2): Likewise.
6474 (noop_move_p): Likewise.
6475 (note_stores): Likewise.
6476 * sched-deps.c (sched_analyze_reg): Likewise.
6477 (sched_analyze_insn): Likewise.
6479 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6481 * cse.c (invalidate_reg): New function extracted from...
6482 (invalidate): ...here.
6483 (canonicalize_insn): Check for clobber high.
6484 (invalidate_from_clobbers): invalidate clobber highs.
6485 (invalidate_from_sets_and_clobbers): Likewise.
6486 (count_reg_usage): Check for clobber high.
6487 (insn_live_p): Likewise.
6488 * cselib.c (cselib_expand_value_rtx_1):Likewise.
6489 (cselib_invalidate_regno): Check for clobber in setter.
6490 (cselib_invalidate_rtx): Pass through setter.
6491 (cselib_invalidate_rtx_note_stores):
6492 (cselib_process_insn): Check for clobber high.
6493 * cselib.h (cselib_invalidate_rtx): Add operand.
6495 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6497 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
6498 (mark_not_eliminable): Likewise.
6499 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
6500 * lra-lives.c (process_bb_lives): Check for clobber high.
6501 * lra.c (new_insn_reg): Remember clobber highs.
6502 (collect_non_operand_hard_regs): Check for clobber high.
6503 (lra_set_insn_recog_data): Likewise.
6504 (add_regs_to_insn_regno_info): Likewise.
6505 (lra_update_insn_regno_info): Likewise.
6507 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6509 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
6510 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
6512 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6514 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
6515 (copy_insn_1): Likewise.
6516 (gen_hard_reg_clobber_high): New gen function.
6517 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
6518 * genemit.c (gen_exp): Likewise.
6519 (gen_emit_seq): Pass through info.
6520 (gen_insn): Check for CLOBBER_HIGH.
6521 (gen_expand): Pass through info.
6522 (gen_split): Likewise.
6523 (output_add_clobbers): Likewise.
6524 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
6525 (remove_clobbers): Likewise.
6526 * rtl.h (gen_hard_reg_clobber_high): New declaration.
6528 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6530 * doc/rtl.texi (clobber_high): Add.
6531 (parallel): Add in clobber high
6532 * rtl.c (rtl_check_failed_code3): Add function.
6533 * rtl.def (CLOBBER_HIGH): Add expression.
6534 * rtl.h (RTL_CHECKC3): Add macro.
6535 (rtl_check_failed_code3): Add declaration.
6536 (XC3EXP): Add macro.
6538 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
6541 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
6542 cfun->machine->max_used_stack_alignment if needed.
6544 2018-08-04 Martin Sebor <msebor@redhat.com>
6546 PR tree-optimization/86571
6547 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
6550 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
6553 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6556 2018-08-03 Jeff Law <law@redhat.com>
6559 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6560 Define to speculation_safe_value_not_needed.
6562 2018-08-03 David Malcolm <dmalcolm@redhat.com>
6564 * doc/gcov.texi (-x): Remove duplicate "to".
6565 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
6566 (-Wif-not-aligned): Remove duplicate "is".
6567 (-flto): Remove duplicate "the".
6568 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
6569 duplicate "v5.00.b".
6570 (MSP430 Options): Remove duplicate "and" from the description
6571 of "-mgprel-sec=regexp".
6572 (x86 Options): Remove duplicate copies of "vmldLog102" and
6573 vmlsLog104 from description of "-mveclibabi=type".
6575 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
6577 * internal-fn.h (first_commutative_argument): Declare.
6578 * internal-fn.c (first_commutative_argument): New function.
6579 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
6580 restrictions for pattern statements. Use first_commutative_argument
6581 to look for commutative operands in calls to internal functions.
6583 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
6585 * Makefile.in (wide-int-range.o): New.
6586 * tree-vrp.c: Move all the wide_int_* functions to...
6587 * wide-int-range.cc: ...here.
6588 * tree-vrp.h: Move all the wide_int_* prototypes to...
6589 * wide-int-range.h: ...here.
6591 2018-08-03 Tom de Vries <tdevries@suse.de>
6593 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
6595 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
6596 * except.c (output_function_exception_table): Do early exit if
6597 targetm_common.except_unwind_info (&global_options) == UI_NONE.
6599 2018-08-03 Martin Liska <mliska@suse.cz>
6601 * predict.c (dump_prediction): Change to 2 digits
6604 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
6606 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
6608 (falkor_am_1_gtov_gtov): ... a new insn reservation.
6610 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
6612 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
6613 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
6614 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
6615 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
6616 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
6618 2018-08-02 David Malcolm <dmalcolm@redhat.com>
6620 * diagnostic-show-locus.c (diagnostic_show_locus): Use
6621 pp_take_prefix when saving the existing prefix.
6622 * diagnostic.c (diagnostic_append_note): Likewise.
6623 * langhooks.c (lhd_print_error_function): Likewise.
6624 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
6625 param's type. Free the existing prefix.
6626 (pp_take_prefix): New function.
6627 (pretty_printer::pretty_printer): Drop the prefix parameter.
6628 Rename the length parameter to match the comment.
6629 (pretty_printer::~pretty_printer): Free the prefix.
6630 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
6632 (struct pretty_printer): Drop the "const" from "prefix" field's
6633 type and clarify memory management.
6634 (pp_set_prefix): Drop the "const" from the 2nd param.
6635 (pp_take_prefix): New decl.
6637 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
6639 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
6640 for word_mode_ok here instead of passing as argument.
6641 (expand_block_compare): Change select_block_compare_mode() call.
6642 (expand_strncmp_gpr_sequence): New function.
6643 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
6645 2018-08-02 Jeff Law <law@redhat.com>
6648 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6649 Define to speculation_safe_value_not_needed.
6652 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6653 Define to speculation_safe_value_not_needed.
6655 2018-08-02 Tom de Vries <tdevries@suse.de>
6658 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
6659 function. Return UI_TARGET unconditionally.
6660 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
6661 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
6663 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
6665 * genemit.c (print_overload_test): Fix typo.
6667 2018-08-02 Richard Biener <rguenther@suse.de>
6669 PR tree-optimization/86816
6670 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
6671 which checks for value availability before querying it.
6672 (gvn_uses_equal): Use it.
6673 (same_succ_hash): Likewise.
6674 (gimple_equal_p): Likewise.
6676 2018-08-02 Nick Clifton <nickc@redhat.com>
6679 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6680 Define to speculation_safe_value_not_needed.
6683 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6684 Define to speculation_safe_value_not_needed.
6687 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6688 Define to speculation_safe_value_not_needed.
6691 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6692 Define to speculation_safe_value_not_needed.
6695 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6696 Define to speculation_safe_value_not_needed.
6699 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6700 Define to speculation_safe_value_not_needed.
6703 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6704 Define to speculation_safe_value_not_needed.
6707 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6708 Define to speculation_safe_value_not_needed.
6711 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6712 speculation_safe_value_not_needed.
6715 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
6716 to speculation_safe_value_not_needed.
6718 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
6720 * doc/md.texi: Expand the documentation of instruction names
6721 to mention port-local uses. Document '@' in pattern names.
6722 * read-md.h (overloaded_instance, overloaded_name): New structs.
6724 (md_reader::handle_overloaded_name): New member function.
6725 (md_reader::get_overloads): Likewise.
6726 (md_reader::m_first_overload): New member variable.
6727 (md_reader::m_next_overload_ptr): Likewise.
6728 (md_reader::m_overloads_htab): Likewise.
6729 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
6730 m_next_overload_ptr and m_overloads_htab.
6731 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
6732 (get_mode_token, get_code_token, get_int_token): New functions.
6733 (map_attr_string): Add an optional argument that passes back
6734 the associated iterator.
6735 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
6736 (md_reader::handle_overloaded_name, add_overload_instance): New
6738 (apply_iterators): Handle '@' names. Report an error if '@'
6739 is used without iterators.
6740 (initialize_iterators): Initialize the new iterator_group fields.
6741 * genopinit.c (handle_overloaded_code_for)
6742 (handle_overloaded_gen): New functions.
6743 (main): Use them to print declarations of maybe_code_for_* and
6744 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
6745 * genemit.c (print_overload_arguments, print_overload_test)
6746 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
6747 (main): Use it to print definitions of maybe_code_for_* and
6748 maybe_gen_* functions.
6749 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
6750 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
6751 instead of explicit mode checks.
6752 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
6753 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
6754 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
6755 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
6756 (aarch64_expand_compare_and_swap): Likewise
6757 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
6758 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
6759 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
6760 (aarch64_constant_pool_reload_icode): Delete.
6761 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
6762 instead of aarch64_constant_pool_reload_icode. Use
6763 code_for_aarch64_reload_mov instead of explicit mode checks.
6764 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
6765 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
6766 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
6767 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
6768 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
6769 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
6770 (aarch64_atomic_load_op_code): Delete.
6771 (aarch64_emit_atomic_load_op): Likewise.
6772 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
6773 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
6774 instead of aarch64_emit_atomic_load_op.
6775 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
6776 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
6777 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
6778 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
6779 character before the pattern name.
6780 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
6781 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
6782 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
6783 (aarch64_frecps<mode>): Likewise.
6784 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
6785 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
6786 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
6787 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
6788 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
6790 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
6792 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
6793 Allow HFmode constants if TARGET_FP_F16INST.
6795 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
6798 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
6799 No longer check last store for clobber of address register.
6801 2018-08-02 Martin Liska <mliska@suse.cz>
6803 PR gcov-profile/86817
6804 * gcov.c (process_all_functions): New function.
6806 (process_file): Move functions processing to
6807 process_all_functions.
6809 2018-08-02 David Malcolm <dmalcolm@redhat.com>
6811 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
6812 "const" to the "gimple *" and "rtx_insn *" parameters.
6813 * dumpfile.h (dump_user_location_t::dump_user_location_t):
6815 (dump_location_t::dump_location_t): Likewise.
6817 2018-08-01 Martin Sebor <msebor@redhat.com>
6819 PR tree-optimization/86650
6820 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
6821 rather than a "gcall *". Directly pass the data of interest
6822 to percent_K_format, rather than building a temporary CALL_EXPR
6824 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
6825 (gimple_fold_builtin_strncat): Adjust.
6826 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
6827 gcall* argument with gimple*.
6828 * gimple-ssa-warn-restrict.c (check_call): Same.
6829 (wrestrict_dom_walker::before_dom_children): Same.
6830 (builtin_access::builtin_access): Same.
6831 (check_bounds_or_overlap): Same
6832 (maybe_diag_overlap): Same.
6833 (maybe_diag_offset_bounds): Same.
6834 * tree-diagnostic.c (default_tree_printer): Move usage of
6835 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
6837 * tree-pretty-print.c (percent_K_format): Add argument.
6838 * tree-pretty-print.h: Add argument.
6839 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
6840 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
6841 (maybe_diag_stxncpy_trunc): Same.
6842 (handle_builtin_stxncpy): Same.
6843 (handle_builtin_strcat): Same.
6845 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6847 * match.pd: Optimise pointer range checks.
6849 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6851 PR tree-optimization/86758
6852 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
6853 to remove pattern statements.
6855 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6857 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
6858 result of dfs_enumerate_from when constructing stmt_vec_infos,
6859 instead of additionally calling get_loop_body.
6861 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6863 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
6865 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
6866 When creating an iv, assert that the step is not known to be zero.
6867 (vect_setup_realignment): Update call accordingly.
6868 * tree-vect-stmts.c (vectorizable_store): Likewise.
6869 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
6871 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6873 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
6874 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
6875 (vectorizable_reduction): Likewise.
6876 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6877 (vect_detect_hybrid_slp_stmts): Likewise.
6878 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6880 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
6882 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
6883 (wide_int_set_zero_nonzero_bits): ...this.
6884 (zero_nonzero_bits_from_vr): Rename to...
6885 (vrp_set_zero_nonzero_bits): ...this.
6886 (extract_range_from_multiplicative_op_1): Abstract wide int
6888 (wide_int_range_multiplicative_op): ...here.
6889 (extract_range_from_binary_expr_1): Extract wide int binary
6890 operations into their own functions.
6891 (wide_int_range_lshift): New.
6892 (wide_int_range_can_optimize_bit_op): New.
6893 (wide_int_range_shift_undefined_p): New.
6894 (wide_int_range_bit_xor): New.
6895 (wide_int_range_bit_ior): New.
6896 (wide_int_range_bit_and): New.
6897 (wide_int_range_trunc_mod): New.
6898 (extract_range_into_wide_ints): New.
6899 (vrp_shift_undefined_p): New.
6900 (extract_range_from_multiplicative_op): New.
6901 (vrp_can_optimize_bit_op): New.
6902 * tree-vrp.h (value_range::dump): New.
6903 (wide_int_range_multiplicative_op): New.
6904 (wide_int_range_lshift):New.
6905 (wide_int_range_shift_undefined_p): New.
6906 (wide_int_range_bit_xor): New.
6907 (wide_int_range_bit_ior): New.
6908 (wide_int_range_bit_and): New.
6909 (wide_int_range_trunc_mod): New.
6910 (zero_nonzero_bits_from_bounds): Rename to...
6911 (wide_int_set_zero_nonzero_bits): ...this.
6912 (zero_nonzero_bits_from_vr): Rename to...
6913 (vrp_set_zero_nonzero_bits): ...this.
6914 (range_easy_mask_min_max): Rename to...
6915 (wide_int_range_can_optimize_bit_op): this.
6916 * vr-values.c (simplify_bit_ops_using_ranges): Rename
6917 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6919 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6921 * tree-vectorizer.h (vect_orig_stmt): New function.
6922 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6923 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6924 (vect_create_epilog_for_reduction): Likewise.
6925 (vectorizable_live_operation): Likewise.
6926 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6927 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6928 * tree-vect-stmts.c (vectorizable_call): Likewise.
6929 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6931 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6933 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6935 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6936 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6937 (vect_transform_loop): Likewise.
6938 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6940 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6942 * tree-vectorizer.h (vect_schedule_slp): Return void.
6943 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6944 (vect_schedule_slp): Likewise.
6946 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6948 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6950 (vect_transform_loop): Update calls accordingly. Schedule SLP
6951 instances before the main loop, if any exist.
6953 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6955 PR tree-optimization/86749
6956 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6957 If the lhs is used in a COND_EXPR, check that it is being used
6958 as the "then" or "else" value.
6960 2018-08-01 Tom de Vries <tdevries@suse.de>
6963 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6964 speculation_safe_value_not_needed.
6966 2018-08-01 Richard Biener <rguenther@suse.de>
6968 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6971 2018-08-01 Martin Liska <mliska@suse.cz>
6973 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6974 format how successful transformation is dumped.
6975 (gimple_mod_pow2_value_transform): Likewise.
6976 (gimple_mod_subtract_transform): Likewise.
6977 (gimple_stringops_transform): Likewise.
6979 2018-08-01 Martin Liska <mliska@suse.cz>
6982 * value-prof.c (interesting_stringop_to_profile_p):
6983 Simplify the code and add BUILT_IN_MEMMOVE.
6984 (gimple_stringops_transform): Likewise.
6986 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
6988 * config/aarch64/aarch64-simd.md
6989 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6990 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6991 use GPI iterator instead of SI mode.
6993 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6995 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6996 rs6000_speculation_barrier.
6997 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6998 new barrier pattern name.
7000 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
7002 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
7003 (speculation_barrier): New insn.
7005 2018-08-01 Richard Biener <rguenther@suse.de>
7008 * graphite.h: Include isl/id.h and isl/space.h to allow build
7011 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
7014 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7015 mode for COFF targets.
7016 * defaults.h (TARGET_COFF): Define.
7017 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
7018 TARGET_COFF): Define.
7019 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
7020 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
7021 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
7023 2018-07-31 Alexandre Oliva <oliva@adacore.com>
7025 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
7026 * gimple-streamer-out.c (output_bb): Save it.
7027 * lto-streamer-in.c (input_struct_function_base): Restore
7028 instance discriminator if available. Create map on demand.
7029 * lto-streamer-out.c (output_struct_function_base): Save it if
7031 * final.c (decl_to_instance_map): Document LTO strategy.
7033 2018-07-31 Alexandre Oliva <oliva@adacore.com>
7034 Olivier Hainque <hainque@adacore.com>
7036 * debug.h (decl_to_instance_map_t): New type.
7037 (decl_to_instance_map): Declare.
7038 (maybe_create_decl_to_instance_map): New inline function.
7039 * final.c (bb_discriminator, last_bb_discriminator): New statics,
7040 to track basic block discriminators.
7041 (final_start_function_1): Initialize them.
7042 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
7044 (decl_to_instance_map): New variable.
7045 (map_decl_to_instance, maybe_set_discriminator): New functions.
7046 (notice_source_line): Set discriminator.
7048 2018-07-31 Ian Lance Taylor <iant@golang.org>
7050 * targhooks.c (default_have_speculation_safe_value): Add
7053 2018-07-31 David Malcolm <dmalcolm@redhat.com>
7055 * dump-context.h: Include "pretty-print.h".
7056 (dump_context::refresh_dumps_are_enabled): New decl.
7057 (dump_context::emit_item): New decl.
7058 (class dump_context): Add fields "m_test_pp" and
7060 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7061 (temp_dump_context::get_dumped_text): New decl.
7062 (class temp_dump_context): Add field "m_pp".
7063 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
7064 (dump_context::refresh_dumps_are_enabled): ...and add a test for
7066 (set_dump_file): Update for above change.
7067 (set_alt_dump_file): Likewise.
7068 (dump_loc): New overload, taking a pretty_printer *.
7069 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
7070 to any test pretty-printer.
7071 (make_item_for_dump_gimple_stmt): New function, adapted from
7072 optinfo::add_gimple_stmt.
7073 (dump_context::dump_gimple_stmt): Call it, and use the result,
7074 eliminating the direct usage of dump_file and alt_dump_file in
7075 favor of indirectly using them via emit_item.
7076 (make_item_for_dump_gimple_expr): New function, adapted from
7077 optinfo::add_gimple_expr.
7078 (dump_context::dump_gimple_expr): Call it, and use the result,
7079 eliminating the direct usage of dump_file and alt_dump_file in
7080 favor of indirectly using them via emit_item.
7081 (make_item_for_dump_generic_expr): New function, adapted from
7083 (dump_context::dump_generic_expr): Call it, and use the result,
7084 eliminating the direct usage of dump_file and alt_dump_file in
7085 favor of indirectly using them via emit_item.
7086 (make_item_for_dump_printf_va): New function, adapted from
7087 optinfo::add_printf_va.
7088 (make_item_for_dump_printf): New function.
7089 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
7090 and use the result, eliminating the direct usage of dump_file and
7091 alt_dump_file in favor of indirectly using them via emit_item.
7092 (make_item_for_dump_dec): New function.
7093 (dump_context::dump_dec): Call it, and use the result,
7094 eliminating the direct usage of dump_file and alt_dump_file in
7095 favor of indirectly using them via emit_item.
7096 (make_item_for_dump_symtab_node): New function, adapted from
7097 optinfo::add_symtab_node.
7098 (dump_context::dump_symtab_node): Call it, and use the result,
7099 eliminating the direct usage of dump_file and alt_dump_file in
7100 favor of indirectly using them via emit_item.
7101 (dump_context::begin_scope): Reimplement, avoiding direct usage
7102 of dump_file and alt_dump_file in favor of indirectly using them
7104 (dump_context::emit_item): New member function.
7105 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7106 Set up test pretty-printer on the underlying context. Call
7107 refresh_dumps_are_enabled.
7108 (temp_dump_context::~temp_dump_context): Call
7109 refresh_dumps_are_enabled.
7110 (temp_dump_context::get_dumped_text): New member function.
7111 (selftest::verify_dumped_text): New function.
7112 (ASSERT_DUMPED_TEXT_EQ): New macro.
7113 (selftest::test_capture_of_dump_calls): Run all tests twice, with
7114 and then without optinfo enabled. Add uses of
7115 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
7116 * dumpfile.h: Update comment for the dump_* API.
7117 * optinfo-emit-json.cc
7118 (selftest::test_building_json_from_dump_calls): Update for new
7119 param for temp_dump_context ctor.
7120 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7121 and "m_owned" field.
7122 (optinfo_item::~optinfo_item): Likewise.
7123 (optinfo::add_item): New member function.
7124 (optinfo::emit): Update comment.
7125 (optinfo::add_string): Delete.
7126 (optinfo::add_printf): Delete.
7127 (optinfo::add_printf_va): Delete.
7128 (optinfo::add_gimple_stmt): Delete.
7129 (optinfo::add_gimple_expr): Delete.
7130 (optinfo::add_tree): Delete.
7131 (optinfo::add_symtab_node): Delete.
7132 (optinfo::add_dec): Delete.
7133 * optinfo.h (class dump_context): New forward decl.
7134 (optinfo::add_item): New decl.
7135 (optinfo::add_string): Delete.
7136 (optinfo::add_printf): Delete.
7137 (optinfo::add_printf_va): Delete.
7138 (optinfo::add_gimple_stmt): Delete.
7139 (optinfo::add_gimple_expr): Delete.
7140 (optinfo::add_tree): Delete.
7141 (optinfo::add_symtab_node): Delete.
7142 (optinfo::add_dec): Delete.
7143 (optinfo::add_poly_int): Delete.
7144 (optinfo_item::optinfo_item): Remove "owned" param.
7145 (class optinfo_item): Remove field "m_owned".
7147 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7150 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7151 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7152 requested variable alignment.
7153 (expand_one_ssa_partition): Likewise.
7154 (expand_one_var): Likewise.
7156 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7158 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7159 to speculation_safe_value_not_needed.
7161 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7163 * targhooks.h (speculation_safe_value_not_needed): New prototype.
7164 * targhooks.c (speculation_safe_value_not_needed): New function.
7165 * target.def (have_speculation_safe_value): Update documentation.
7166 * doc/tm.texi: Regenerated.
7168 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7170 * config/aarch64/iterators.md (ALLI_TI): New iterator.
7171 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7173 (despeculate_copy<ALLI:mode>_insn): New insn.
7174 (despeculate_copyti_insn): New insn.
7175 (despeculate_simple<ALLI:mode>): New insn
7176 (despeculate_simpleti): New insn.
7177 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7179 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7180 aarch64_speculation_safe_value.
7181 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7183 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7185 * config/aarch64/aarch64-speculation.cc: New file.
7186 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7187 before pass_reorder_blocks.
7188 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7190 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7191 X14 and X15 when tracking speculation.
7192 * config/aarch64/aarch64.md (register name constants): Add
7193 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7194 (unspec): Add UNSPEC_SPECULATION_TRACKER.
7195 (speculation_barrier): New insn attribute.
7196 (cmp<mode>): Allow SP in comparisons.
7197 (speculation_tracker): New insn.
7198 (speculation_barrier): Add speculation_barrier attribute.
7199 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7200 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7201 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7203 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7205 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7206 aarch64_track_speculation is true.
7207 (tb<optab><mode>1): Likewise.
7208 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7209 generate CB[N]Z when tracking speculation.
7210 (aarch64_split_compare_and_swap): Likewise.
7211 (aarch64_split_atomic_op): Likewise.
7213 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7215 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
7217 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7219 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
7220 (speculation_barrier): New insn.
7222 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7224 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
7225 * config/arm/arm.md (speculation_barrier): New expand.
7226 (speculation_barrier_insn): New pattern.
7228 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7230 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
7231 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
7232 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
7233 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
7235 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
7236 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
7237 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
7238 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
7239 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
7240 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
7241 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
7242 * builtins.c (expand_speculation_safe_value): New function.
7243 (expand_builtin): Call it.
7244 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
7245 * doc/extend.texi: Document __builtin_speculation_safe_value.
7246 * doc/md.texi: Document "speculation_barrier" pattern.
7247 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
7248 TARGET_HAVE_SPECULATION_SAFE_VALUE.
7249 * doc/tm.texi: Regenerated.
7250 * target.def (have_speculation_safe_value, speculation_safe_value): New
7252 * targhooks.c (default_have_speculation_safe_value): New function.
7253 (default_speculation_safe_value): New function.
7254 * targhooks.h (default_have_speculation_safe_value): Add prototype.
7255 (default_speculation_safe_value): Add prototype.
7257 2018-07-31 David Malcolm <dmalcolm@redhat.com>
7259 * dump-context.h (dump_context::dump_loc): New decl.
7260 * dumpfile.c (dump_context::dump_loc): New member function.
7261 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
7262 and dump_gimple_stmt.
7263 (dump_context::dump_gimple_expr_loc): Likewise, using
7265 (dump_context::dump_generic_expr_loc): Likewise, using
7267 (dump_context::dump_printf_loc_va): Likewise, using
7269 (dump_context::begin_scope): Explicitly using the global function
7270 "dump_loc", rather than the member function.
7272 2018-07-31 Martin Sebor <msebor@redhat.com>
7274 PR tree-optimization/86741
7275 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
7277 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
7279 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
7280 depend on whether prefetch instructions will be emitted or not.
7281 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
7282 will be emitted or not.
7283 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
7284 (TARGET_SETMEM_PFD): New macros.
7286 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7288 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
7289 (NULL_STMT_VEC_INFO): Delete.
7290 (stmt_vec_info::operator*): Likewise.
7291 (stmt_vec_info::operator gimple *): Likewise.
7292 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
7293 of NULL_STMT_VEC_INFO.
7294 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7295 (vect_reassociating_reduction_p): Likewise.
7296 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7297 (vectorizable_store): Likewise.
7298 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
7299 (vec_info::free_stmt_vec_infos): Likewise.
7301 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7303 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
7304 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
7306 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7308 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
7309 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7310 (vec_info::free_stmt_vec_info): New private member functions.
7311 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
7312 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
7313 * tree-parloops.c (gather_scalar_reductions): Remove calls to
7314 set_stmt_vec_info_vec and free_stmt_vec_infos.
7315 * tree-vect-loop.c (_loop_vec_info): Remove call to
7316 set_stmt_vec_info_vec.
7317 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
7318 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
7319 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
7320 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7321 (vec_info::free_stmt_vec_info): ...these new functions. Remove
7322 assignments in {vec_info::,}new_stmt_vec_info that are redundant
7323 with the clearing in the xcalloc.
7324 (stmt_vec_info_vec): Delete.
7325 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
7326 (vectorize_loops): Likewise.
7327 (vec_info::~vec_info): Remove argument from call to
7328 free_stmt_vec_infos.
7329 (vec_info::add_stmt): Remove vinfo argument from call to
7332 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7334 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
7335 rather than a gimple stmt.
7336 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
7337 information for pattern statements when passed the original
7338 statement; instead wait to be passed the pattern statement itself.
7339 Don't call set_vinfo_for_stmt here.
7340 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
7341 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
7342 stmt_vec_infos here.
7343 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
7344 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
7345 stmt_vec_infos entry.
7347 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7349 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
7350 * tree-vectorizer.c (vec_info::replace_stmt): New function.
7351 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
7352 * tree-vect-stmts.c (vectorizable_call): Likewise.
7353 (vectorizable_simd_clone_call): Likewise.
7355 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7357 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
7358 * tree-vectorizer.c (vec_info::remove_stmt): New function.
7359 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
7360 * tree-vect-loop.c (vect_transform_loop): Likewise.
7361 * tree-vect-slp.c (vect_schedule_slp): Likewise.
7362 * tree-vect-stmts.c (vect_remove_stores): Likewise.
7364 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7366 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
7367 (vect_dr_stmt): Delete.
7368 * tree-vectorizer.c (vec_info::lookup_dr): New function.
7369 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
7371 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7372 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
7373 (vect_verify_datarefs_alignment, vect_peeling_supportable)
7374 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7375 (vect_analyze_data_refs): Likewise.
7376 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
7378 (vect_find_same_alignment_drs): Likewise.
7379 (vect_slp_analyze_node_dependences): Update calls accordingly.
7380 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
7381 instead of DR_VECT_AUX.
7382 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
7383 of a vector data references. Use vec_info::lookup_dr instead of
7385 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
7386 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
7387 instead of DR_VECT_AUX.
7389 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7391 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
7393 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7395 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7396 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7397 (vect_gen_prolog_loop_niters): Likewise.
7399 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7401 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
7402 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
7403 (vect_known_alignment_in_bytes, vect_dr_behavior)
7404 (vect_get_scalar_dr_size): Take references as dr_vec_infos
7405 instead of data_references. Update calls to other routines for
7406 which the same change has been made.
7407 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
7408 dr_vec_infos instead of stmt_vec_infos.
7409 (vect_analyze_data_ref_dependence): Update call accordingly.
7410 (vect_slp_analyze_data_ref_dependence)
7411 (vect_record_base_alignments): Use DR_VECT_AUX.
7412 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
7413 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
7414 (vector_alignment_reachable_p, vect_get_data_access_cost)
7415 (vect_peeling_supportable, vect_analyze_group_access_1)
7416 (vect_analyze_group_access, vect_analyze_data_ref_access)
7417 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
7418 (vect_compile_time_alias, vect_small_gap_p)
7419 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
7420 (vect_supportable_dr_alignment): Take references as dr_vec_infos
7421 instead of data_references. Update calls to other routines for
7422 which the same change has been made.
7423 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
7424 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7425 (vect_slp_analyze_and_verify_node_alignment)
7426 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7427 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7428 (vect_setup_realignment): Use dr_vec_infos. Update calls after
7430 (_vect_peel_info::dr): Replace with...
7431 (_vect_peel_info::dr_info): ...this new field.
7432 (vect_peeling_hash_get_most_frequent)
7433 (vect_peeling_hash_choose_best_peeling): Update accordingly.
7434 (vect_peeling_hash_get_lowest_cost):
7435 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
7436 routines for which the same change has been made.
7437 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
7439 * tree-vect-loop-manip.c (get_misalign_in_elems)
7440 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
7442 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7443 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
7444 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
7445 (get_group_load_store_type, get_negative_load_store_type)
7446 (vect_get_data_ptr_increment, vectorizable_store)
7447 (vectorizable_load): Likewise.
7448 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
7449 Update calls to other routines for which the same change has been made.
7451 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7453 * tree-vectorizer.h (vec_info::move_dr): New member function.
7454 (dataref_aux): Rename to...
7455 (dr_vec_info): ...this and add "dr" and "stmt" fields.
7456 (_stmt_vec_info::dr_aux): Update accordingly.
7457 (_stmt_vec_info::data_ref_info): Delete.
7458 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
7459 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
7460 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
7461 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
7463 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
7464 (STMT_VINFO_DR_INFO): New macro.
7465 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
7466 (set_dr_misalignment): Update after rename of dataref_aux.
7467 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
7468 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
7469 initialization of STMT_VINFO_DATA_REF.
7470 * tree-vectorizer.c (vec_info::move_dr): New function.
7471 * tree-vect-patterns.c (vect_recog_bool_pattern)
7472 (vect_recog_mask_conversion_pattern)
7473 (vect_recog_gather_scatter_pattern): Use it.
7474 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
7475 the "dr" and "stmt" fields of dr_vec_info instead of
7476 STMT_VINFO_DATA_REF.
7478 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7480 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
7481 (is_pattern_stmt_p): Use it.
7482 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
7483 on pattern statements.
7485 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7487 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
7488 original stmt as a stmt_vec_info rather than a gimple stmt.
7489 (vect_pattern_recog_1): Take the statement directly as a
7490 stmt_vec_info, rather than via a gimple_stmt_iterator.
7491 Update call to vect_mark_pattern_stmts.
7492 (vect_pattern_recog): Update calls accordingly.
7494 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7496 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
7497 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
7498 a vect_def_type for the first argument.
7499 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
7500 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
7501 operand if it isn't defined by a vectorized statement.
7502 (vect_build_gather_load_calls): Remove the mask_dt argument and
7503 update calls to vect_get_vec_def_for_stmt_copy.
7504 (vectorizable_bswap): Likewise the dt argument.
7505 (vectorizable_call): Update calls to vectorizable_bswap and
7506 vect_get_vec_def_for_stmt_copy.
7507 (vectorizable_simd_clone_call, vectorizable_assignment)
7508 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
7509 (vectorizable_comparison): Update calls to
7510 vect_get_vec_def_for_stmt_copy.
7511 (vectorizable_store): Likewise. Remove now-unnecessary calls to
7513 (vect_get_loop_based_defs): Remove dt argument and update call
7514 to vect_get_vec_def_for_stmt_copy.
7515 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
7516 and vect_get_vec_def_for_stmt_copy.
7517 (vectorizable_load): Update calls to vect_build_gather_load_calls
7518 and vect_get_vec_def_for_stmt_copy.
7519 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7520 (vectorizable_reduction, vectorizable_live_operation): Update calls
7521 to vect_get_vec_def_for_stmt_copy.
7523 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7525 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
7526 and gimple stmt arguments with a stmt_vec_info.
7527 (vect_record_base_alignments): Update calls accordingly.
7528 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
7529 and gimple stmt arguments with a stmt_vec_info.
7530 (vect_build_slp_tree_1): Remove vinfo argument and update call
7531 to vect_record_max_nunits.
7532 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
7533 and vect_record_max_nunits.
7535 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7537 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
7538 file and take a stmt_vec_info instead of a gimple stmt.
7539 (supportable_widening_operation, vect_finish_replace_stmt)
7540 (vect_finish_stmt_generation, vect_get_store_rhs)
7541 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
7542 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
7543 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
7544 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
7545 (vect_create_data_ref_ptr, bump_vector_ptr)
7546 (vect_permute_store_chain, vect_setup_realignment)
7547 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
7548 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
7549 (vectorizable_reduction, vectorizable_induction)
7550 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
7551 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
7552 than gimple stmts as arguments.
7553 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
7554 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
7555 (can_group_stmts_p, vect_check_gather_scatter)
7556 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7557 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
7558 (vect_permute_load_chain, vect_shift_permute_load_chain)
7559 (vect_transform_grouped_load)
7560 (vect_record_grouped_load_vectors): Likewise.
7561 * tree-vect-loop.c (vect_fixup_reduc_chain)
7562 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
7563 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
7564 (vectorizable_reduction, vectorizable_induction)
7565 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
7566 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
7567 (vect_get_load_store_mask): Likewise.
7568 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
7569 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
7570 * tree-vect-stmts.c (vect_mark_relevant)
7571 (is_simple_and_all_uses_invariant)
7572 (exist_non_indexing_operands_for_use_p, process_use)
7573 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
7574 (vect_get_vec_def_for_operand, vect_get_vec_defs)
7575 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
7576 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
7577 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
7578 (get_negative_load_store_type, get_load_store_type)
7579 (vect_check_load_store_mask, vect_check_store_rhs)
7580 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
7581 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7582 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7583 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7584 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
7585 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
7586 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
7587 (supportable_widening_operation): Likewise.
7589 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7591 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
7592 a stmt_vec_info instead of a gcall.
7593 (vect_check_gather_scatter): Update call accordingly.
7594 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
7596 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
7597 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
7598 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
7599 instead of a gimple stmt.
7600 (vect_transform_loop): Update calls accordingly.
7601 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
7602 stmt_vec_infos instead of gimple stmts.
7603 (vect_analyze_slp_instance): Update use accordingly.
7604 * tree-vect-stmts.c (read_vector_array, write_vector_array)
7605 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
7606 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
7607 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
7608 (vect_gen_widened_results_half, vect_get_loop_based_defs)
7609 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
7610 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
7611 down to subroutines.
7613 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7615 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
7616 of the worklist from a vector of gimple stmts to a vector of
7618 * tree-vect-stmts.c (vect_mark_relevant, process_use)
7619 (vect_mark_stmts_to_be_vectorized): Likewise
7621 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7623 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
7624 statement before passing it to vect_analyze_stmt.
7625 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
7626 the chain of phi vector definitions. Track the exit phi via its
7628 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
7629 STMT_VINFO_REDUC_DEF.
7630 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
7631 stmt_vec_infos to handle the statement chains.
7632 (vect_get_slp_defs): Record the first statement in the node
7633 using a stmt_vec_info.
7634 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
7635 statements here and pass their stmt_vec_info down to subroutines.
7636 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
7637 down to vect_finish_stmt_generation.
7638 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
7639 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
7640 stmt_vec_infos to subroutines.
7641 (vect_remove_stores): Use stmt_vec_infos to handle the statement
7644 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7646 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
7647 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
7648 (vect_permute_store_chain, vect_setup_realignment)
7649 (vect_permute_load_chain, vect_shift_permute_load_chain)
7650 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
7651 stmts internally, and when passing values to other vectorizer routines.
7652 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7653 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
7654 (vect_analyze_loop_operations, get_initial_def_for_reduction)
7655 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
7656 (vectorizable_reduction, vectorizable_induction)
7657 (vectorizable_live_operation, vect_transform_loop_stmt)
7658 (vect_transform_loop): Likewise.
7659 * tree-vect-patterns.c (vect_reassociating_reduction_p)
7660 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
7661 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
7662 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7663 (vect_slp_analyze_node_operations_1): Likewise.
7664 * tree-vect-stmts.c (vect_mark_relevant, process_use)
7665 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
7666 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
7667 (vect_finish_stmt_generation_1, get_group_load_store_type)
7668 (get_load_store_type, vect_build_gather_load_calls)
7669 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7670 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7671 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7672 (vectorizable_store, vectorizable_load, vectorizable_condition)
7673 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
7674 (supportable_widening_operation): Likewise.
7675 (vect_get_vector_types_for_stmt): Likewise.
7676 * tree-vectorizer.h (vect_dr_behavior): Likewise.
7678 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7680 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7681 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
7682 (vect_permute_store_chain, vect_permute_load_chain)
7683 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
7684 repeated stmt_vec_info lookups.
7685 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7686 (vect_update_ivs_after_vectorizer): Likewise.
7687 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
7688 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7689 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
7690 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7691 (vect_bb_slp_scalar_cost): Likewise.
7692 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
7694 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7696 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
7697 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
7698 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
7699 to get gassigns and gcalls, rather than operating on generc gimple
7701 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
7702 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
7703 (vectorizable_load, vect_analyze_stmt): Likewise.
7704 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
7706 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7708 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
7709 return stmt_vec_infos rather than gimple stmts. Do not accept
7711 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
7713 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7714 Update use of get_later_stmt.
7715 (vect_get_constant_vectors): Update call accordingly.
7716 (vect_schedule_slp_instance): Likewise
7717 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
7718 (vect_slp_analyze_instance_dependence): Likewise.
7719 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
7721 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7723 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
7724 (stmt_info_for_cost::stmt_info): ...this new field.
7725 (add_stmt_costs): Update accordingly.
7726 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7727 (vect_get_known_peeling_cost): Likewise.
7728 (vect_estimate_min_profitable_iters): Likewise.
7729 * tree-vect-stmts.c (record_stmt_cost): Likewise.
7731 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7733 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
7734 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7735 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7737 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
7739 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7741 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
7742 a gimple stmt to a stmt_vec_info.
7743 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7745 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7747 * tree-vectorizer.h (vec_info::grouped_stores): Change from
7748 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7749 (_loop_vec_info::reduction_chains): Likewise.
7750 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
7752 * tree-vect-slp.c (vect_analyze_slp): Likewise.
7754 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7756 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
7757 a gimple stmt to a stmt_vec_info.
7758 (_stmt_vec_info::next_element): Likewise.
7759 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
7760 (vect_slp_analyze_and_verify_node_alignment)
7761 (vect_analyze_group_access_1, vect_analyze_group_access)
7762 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
7763 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
7764 (vect_supportable_dr_alignment): Update accordingly.
7765 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
7766 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
7767 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
7768 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7769 * tree-vect-slp.c (vect_build_slp_tree_1)
7770 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
7771 (vect_split_slp_store_group, vect_analyze_slp_instance)
7772 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
7773 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
7774 (get_group_load_store_type, get_load_store_type)
7775 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
7776 (vect_transform_stmt, vect_remove_stores): Likewise.
7778 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7780 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
7782 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7783 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
7784 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
7785 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
7786 (vector_alignment_reachable_p, vect_get_data_access_cost)
7787 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
7788 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
7789 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7790 (vect_analyze_group_access_1, vect_analyze_group_access)
7791 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
7792 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
7793 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
7794 result of vect_dr_stmt and use the stmt_vec_info instead of
7795 the associated gimple stmt.
7796 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7797 (vect_gen_prolog_loop_niters): Likewise.
7798 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7800 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7802 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
7803 to a vec<stmt_vec_info>.
7804 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
7805 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
7807 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
7808 to a vec<stmt_vec_info>.
7809 (bst_traits::value_type, bst_traits::value_type): Likewise.
7810 (bst_traits::hash): Update accordingly.
7811 (vect_get_and_check_slp_defs): Change the stmts parameter from
7812 a vec<gimple *> to a vec<stmt_vec_info>.
7813 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
7814 (vect_build_slp_tree): Likewise.
7815 (vect_build_slp_tree_2): Likewise. Update uses of
7816 SLP_TREE_SCALAR_STMTS.
7817 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
7818 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
7819 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
7820 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
7821 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
7822 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
7823 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
7824 (vect_get_constant_vectors, vect_get_slp_defs)
7825 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
7826 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
7827 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
7828 instead of gimple stmts.
7829 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
7830 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
7831 (vect_slp_analyze_instance_dependence): Update uses of
7832 SLP_TREE_SCALAR_STMTS.
7833 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7834 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
7835 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
7836 (get_initial_defs_for_reduction): Likewise.
7837 (vect_create_epilog_for_reduction): Likewise.
7838 (vectorize_fold_left_reduction): Likewise.
7839 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
7840 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
7841 (can_vectorize_live_stmts): Likewise.
7843 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7845 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
7846 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7847 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
7849 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
7851 (gather_scalar_reductions): Update after above interface changes.
7852 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
7853 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
7855 (vect_force_simple_reduction): Likewise.
7856 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
7857 LOOP_VINFO_REDUCTIONS.
7858 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7860 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7862 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
7863 a gimple stmt to a stmt_vec_info.
7864 * tree-vect-loop.c (vect_active_double_reduction_p)
7865 (vect_force_simple_reduction, vectorizable_reduction): Update
7868 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7870 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
7871 vec<gimple *> to a vec<stmt_vec_info>.
7872 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
7873 the reduction_phis argument from a vec<gimple *> to a
7875 (vectorizable_reduction): Likewise the phis local variable that
7876 is passed to vect_create_epilog_for_reduction. Update for new type
7877 of SLP_TREE_VEC_STMTS.
7878 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
7879 (vectorizable_live_operation): Likewise.
7880 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
7881 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
7883 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7885 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
7886 a gimple stmt to a stmt_vec_info.
7887 (vectorizable_condition, vectorizable_live_operation)
7888 (vectorizable_reduction, vectorizable_induction): Pass back the
7889 vectorized statement as a stmt_vec_info.
7890 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
7891 use of STMT_VINFO_VEC_STMT.
7892 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
7893 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
7894 as stmt_vec_infos rather than gimple stmts.
7895 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
7897 (vectorizable_live_operation): Likewise.
7898 (vectorizable_reduction, vectorizable_induction): Likewise,
7899 updating use of STMT_VINFO_VEC_STMT.
7900 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
7901 of STMT_VINFO_VEC_STMT.
7902 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
7903 (vectorizable_simd_clone_call, vectorizable_conversion)
7904 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7905 (vectorizable_store, vectorizable_load, vectorizable_condition)
7906 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
7907 from a gimple stmt to a stmt_vec_info.
7908 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
7909 pointer to a stmt_vec_info to the vectorizable_* routines.
7911 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7913 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7914 a gimple stmt to a stmt_vec_info.
7915 (is_pattern_stmt_p): Update accordingly.
7916 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7917 (vect_record_grouped_load_vectors): Likewise.
7918 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7919 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7920 (vect_model_reduction_cost): Likewise.
7921 (vect_create_epilog_for_reduction): Likewise.
7922 (vectorizable_reduction, vectorizable_induction): Likewise.
7923 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7924 Return the stmt_vec_info for the pattern statement.
7925 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7926 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7927 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7928 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7929 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7930 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7931 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7932 (free_stmt_vec_info, vect_is_simple_use): Likewise.
7934 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7936 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7937 (vect_finish_stmt_generation): Likewise.
7938 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7939 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7940 (vect_build_gather_load_calls): Use the return value of the above
7941 functions instead of a separate call to vinfo_for_stmt. Use narrow
7942 scopes for the input gimple stmt and wider scopes for the associated
7943 stmt_vec_info. Use vec_info::lookup_def when setting these
7944 stmt_vec_infos from an SSA_NAME definition.
7945 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7946 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7947 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7948 (vectorizable_store, vectorizable_load, vectorizable_condition)
7949 (vectorizable_comparison): Likewise.
7950 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7951 (vectorizable_reduction): Likewise.
7953 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7955 * tree-vectorizer.h (vect_is_simple_use): Add an optional
7956 stmt_vec_info * parameter before the optional gimple **.
7957 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7958 (process_use, vect_get_vec_def_for_operand_1): Update callers.
7959 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7960 * tree-vect-loop.c (vectorizable_reduction): Likewise.
7961 (vectorizable_live_operation): Likewise.
7962 * tree-vect-patterns.c (type_conversion_p): Likewise.
7963 (vect_look_through_possible_promotion): Likewise.
7964 (vect_recog_rotate_pattern): Likewise.
7965 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7967 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7969 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7970 a typedef to a wrapper class.
7971 (NULL_STMT_VEC_INFO): New macro.
7972 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7973 (stmt_vec_info::operator*): New function.
7974 (stmt_vec_info::operator gimple *): Likewise.
7975 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7976 (add_stmt_costs): Likewise.
7977 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7978 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7979 (vect_get_known_peeling_cost): Likewise.
7980 (vect_estimate_min_profitable_iters): Likewise.
7981 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7982 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7983 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7984 (vectorizable_store, free_stmt_vec_infos): Likewise.
7985 (new_stmt_vec_info): Change return type of xcalloc to
7988 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7990 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7991 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7992 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7993 a single_imm_use-based sequence.
7994 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7996 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7998 * tree-vectorizer.h (vec_info::lookup_def): Declare.
7999 * tree-vectorizer.c (vec_info::lookup_def): New function.
8000 * tree-vect-patterns.c (vect_get_internal_def): Use it.
8001 (vect_widened_op_tree): Likewise.
8002 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8003 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
8004 (vectorizable_reduction): Likewise.
8005 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
8007 (vect_is_slp_reduction): Update calls accordingly. Use
8008 vec_info::lookup_def.
8009 (vect_is_simple_reduction): Likewise
8010 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
8012 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8014 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
8015 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
8016 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
8018 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
8019 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
8020 (vect_update_vf_for_slp, vect_analyze_loop_operations)
8021 (vect_is_slp_reduction, vectorizable_induction)
8022 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
8023 * tree-vect-patterns.c (vect_init_pattern_stmt):
8024 (vect_determine_min_output_precision_1, vect_determine_precisions)
8025 (vect_pattern_recog): Likewise.
8026 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
8027 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
8028 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
8029 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8030 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
8031 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
8032 info field from a loop to a loop_vec_info.
8034 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8036 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
8037 (vec_info::add_stmt): Declare.
8038 * tree-vectorizer.c (vec_info::add_stmt): New function.
8039 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
8040 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
8041 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8042 (vectorizable_induction): Likewise.
8043 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
8044 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8045 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
8046 (vectorizable_load): Likewise.
8047 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8048 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
8049 (vect_recog_gather_scatter_pattern): Likewise.
8050 (append_pattern_def_seq): Likewise. Remove a check that is
8051 performed by add_stmt itself.
8053 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8055 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
8056 which make_ssa_name was called with new_stmt before new_stmt
8059 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8061 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
8063 (vect_is_slp_reduction): ...here...
8064 (vect_is_simple_reduction): ...and here. Remove repetition of tests
8065 that are already known to be false.
8067 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8069 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
8070 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
8071 STMT_VINFO_NUM_SLP_USES when it's true.
8072 (vect_free_slp_instance): Add a final_p parameter and pass it to
8074 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
8075 (vect_analyze_slp_instance): Likewise.
8076 (vect_slp_analyze_operations): Likewise.
8077 (vect_slp_analyze_bb_1): Likewise.
8078 * tree-vectorizer.c (vec_info): Likewise.
8079 * tree-vect-loop.c (vect_transform_loop): Likewise.
8081 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8083 * tree-vect-loop.c (vectorizable_reduction): Assert that the
8084 function is not called for second and subsequent members of
8087 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8089 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
8090 cases for nested loops from here to ...
8091 (vect_create_epilog_for_reduction): ...here. Only call
8092 vect_is_simple_use for inner-loop reductions.
8094 2018-07-31 Martin Liska <mliska@suse.cz>
8096 PR gcov-profile/85338
8097 PR gcov-profile/85350
8098 PR gcov-profile/85372
8099 * profile.c (struct location_triplet): New.
8100 (struct location_triplet_hash): Likewise.
8101 (output_location): Do not output a BB that
8102 is already recorded for a line.
8103 (branch_prob): Use streamed_locations.
8105 2018-07-31 Martin Liska <mliska@suse.cz>
8107 PR gcov-profile/85370
8108 * coverage.c (coverage_begin_function): Do not mark target
8109 clones as artificial functions.
8111 2018-07-31 Martin Liska <mliska@suse.cz>
8113 PR gcov-profile/83813
8114 PR gcov-profile/84758
8115 PR gcov-profile/85217
8116 PR gcov-profile/85332
8117 * profile.c (branch_prob): Do not record GOTO expressions
8118 for GIMPLE statements which locations are already streamed.
8120 2018-07-31 Olivier Hainque <hainque@adacore.com>
8122 * gcc.c (handle_spec_function): Accept a soft_matched_part
8123 argument, as do_spec_1. Pass it down to ...
8124 (eval_spec_function): Accept a soft_matched_part argument,
8125 and pass it down to ...
8126 (do_spec_2): Accept a soft_matched_part argument, and pass
8127 it down to do_spec_1.
8128 (do_spec_1): Pass soft_matched_part to handle_spec_function.
8129 (handle_braces): Update call to handle_spec_function.
8130 (driver::set_up_specs): Update calls to do_spec_2.
8131 (compare_debug_dump_opt_spec_function): Likewise.
8132 (compare_debug_self_opt_spec_function): Likewise.
8134 2018-07-31 Olivier Hainque <hainque@adacore.com>
8136 * common.opt (nolibc): New option.
8137 * doc/invoke.texi (Link Options): Document it.
8138 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8139 * config/alpha/linux.h: Likewise.
8140 * config/arc/elf.h: Likewise.
8141 * config/arm/uclinux-elf.h: Likewise.
8142 * config/arm/unknown-elf.h: Likewise.
8143 * config/avr/avrlibc.h: Likewise.
8144 * config/bfin/bfin.h: Likewise.
8145 * config/bfin/linux.h: Likewise.
8146 * config/bfin/uclinux.h: Likewise.
8147 * config/darwin.h: Likewise.
8148 * config/darwin10.h: Likewise.
8149 * config/darwin12.h: Likewise.
8150 * config/gnu-user.h: Likewise.
8151 * config/lm32/uclinux-elf.h: Likewise.
8152 * config/pa/pa-hpux11.h: Likewise.
8153 * config/pa/pa64-hpux.h: Likewise.
8154 * config/sparc/sparc.h: Likewise.
8156 2018-07-31 Olivier Hainque <hainque@adacore.com>
8158 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
8159 undefined variables.
8161 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
8164 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8167 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8169 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8170 terminated string literal.
8172 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
8174 PR rtl-optimization/85160
8175 * combine.c (is_just_move): New function.
8176 (try_combine): Allow combining two instructions into two if neither of
8177 the original instructions was a move.
8179 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
8182 * doc/extend.texi (Global Register Variables): Discourage use of type
8184 (Local Register Variables): Likewise.
8186 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
8188 PR tree-optimization/86506
8189 * hwint.h (ceil_log2): Resync with hwint.c implementation.
8191 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
8194 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8195 hard_regno, make sure no insn between `from` and `to` clobbers it.
8197 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
8198 Tom de Vries <tdevries@suse.de>
8200 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8201 (PTX_DEFAULT_RUNTIME_DIM): ... this.
8202 (nvptx_goacc_validate_dims): Set default worker and gang dims to
8203 PTX_DEFAULT_RUNTIME_DIM.
8204 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8206 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
8208 * config/pa/pa.c (pa_output_addr_vec): Align address table.
8209 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8210 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
8212 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
8214 * config/rs6000/constraints.md (wG constraint): Delete, no longer
8216 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
8217 predicate to reflect toc fusion has been deleted.
8218 (toc_fusion_mem_raw): Delete, no longer used.
8219 (toc_fusion_mem_wrapped): Likewise.
8220 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
8222 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
8223 Delete, no longer used.
8224 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
8225 meant to be used for toc fusion.
8226 (rs6000_debug_print_mode): Delete toc fusion debugging.
8227 (rs6000_debug_reg_global): Likewise.
8228 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
8229 fusion and secondary reload support that were never used.
8230 (rs6000_option_override_internal): Delete TOC fusion, that was only
8231 partially defined, and it did not work unless you also used the
8233 (rs6000_legitimate_address_p): Delete TOC fusion support.
8234 (rs6000_opt_masks): Likewise.
8235 (fusion_wrap_memory_address): Delete function, no longer used.
8236 (fusion_split_address); Delete TOC fusion support.
8237 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
8238 longer used with toc fusion being deleted.
8239 (TARGET_TOC_FUSION_FP): Likewise.
8240 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
8242 (toc fusion spliter): Delete TOC fusion support.
8243 (toc_fusionload_<mode>): Likewise.
8244 (toc_fusionload_di): Likewise.
8245 (fusion_gpr_load_<mode>): Delete generator function, this insn no
8246 longer needs to be named. Rename predicate to delete TOC fusion.
8247 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8248 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8249 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8250 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8251 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
8253 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
8255 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
8256 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
8257 __int128 in built-in function prototypes.
8258 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
8259 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
8261 2018-07-27 Martin Sebor <msebor@redhat.com>
8263 PR tree-optimization/86696
8264 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
8265 types, including enums.
8266 (handle_char_store): Be prepared for the above function to fail.
8268 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
8270 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
8271 when optimization level is lower than 2 or optimize for size.
8273 2018-07-26 Martin Sebor <msebor@redhat.com>
8275 PR tree-optimization/86043
8276 PR tree-optimization/86042
8277 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
8278 (get_string_cst_length): Rename...
8279 (get_min_string_length): ...to this. Add argument.
8280 (handle_char_store): Extend to handle multi-character stores by
8282 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
8283 * tree.h (initializer_zerop): Add argument.
8285 2018-07-26 Jakub Jelinek <jakub@redhat.com>
8288 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
8289 declare target to variables if they have always,{to,from,tofrom} map
8292 2018-07-26 Martin Liska <mliska@suse.cz>
8295 * lto-wrapper.c: Add linker_output as prefix
8296 for ltrans_output_file.
8298 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
8300 PR rtl-optimization/85805
8301 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
8302 value for hard registers if that was written in the same mode.
8304 2018-07-26 Martin Liska <mliska@suse.cz>
8306 PR gcov-profile/86536
8307 * gcov.c (format_gcov): Use printf format %.*f directly
8308 and do not handle special values.
8310 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8312 * common/config/arc/arc-common.c (arc_option_optimization_table):
8313 Update default optimizations for size.
8315 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8317 * config/arc/arc.md (movsf_insn): Add short instruction selection.
8318 * config/arc/constraints.md (CfZ): New constraint.
8319 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
8320 (subsf3_fpu): Likewise.
8321 (cmpsf_fpu): Likewise.
8322 (cmpsf_fpu_uneq): Likewise.
8324 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8326 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
8328 (arc_is_uncached_mem_p): uncached applies to both the variable and
8331 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8333 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
8336 2018-07-25 David Malcolm <dmalcolm@redhat.com>
8338 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
8339 field "m_scopes" from vec to auto_vec.
8341 2018-07-25 Martin Liska <mliska@suse.cz>
8343 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
8346 2018-07-25 Richard Biener <rguenther@suse.de>
8349 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
8350 special wrt context_die late.
8351 (gen_subprogram_die): Re-use DIEs in local scope.
8353 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
8355 PR tree-optimization/86644
8356 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
8358 2018-07-25 Martin Liska <mliska@suse.cz>
8361 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
8362 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
8364 2018-07-25 Martin Liska <mliska@suse.cz>
8367 * params.def: Explain ASan abbreviation and provide
8368 a documentation link.
8370 2018-07-24 Martin Sebor <msebor@redhat.com>
8372 PR tree-optimization/86622
8373 PR tree-optimization/86532
8374 * builtins.h (string_length): Declare.
8375 * builtins.c (c_strlen): Correct handling of non-constant offsets.
8376 (check_access): Be prepared for non-constant length ranges.
8377 (string_length): Make extern.
8378 * expr.c (string_constant): Only handle the minor non-constant
8379 array index. Use string_constant to compute the length of
8380 a generic string constant.
8382 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
8384 PR tree-optimization/86618
8385 * tree-vect-stmts.c (vectorizable_call): Don't take the address
8386 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
8388 2018-07-24 David Malcolm <dmalcolm@redhat.com>
8390 PR tree-optimization/86636
8391 * json.cc (json::object::set): Fix comment. Add assertions.
8392 (json::array::append): Move here from json.h. Add comment and an
8394 (json::string::string): Likewise.
8395 * json.h (json::array::append): Move to json.cc.
8396 (json::string::string): Likewise.
8397 * optinfo-emit-json.cc
8398 (optrecord_json_writer::impl_location_to_json): Assert that we
8399 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
8400 wrapper around it. Expand the location once, rather than three
8402 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
8403 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
8405 (optrecord_json_writer::optinfo_to_json): Likewise, in four
8406 places. Fix some overlong lines.
8408 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
8410 * config/aarch64/aarch64-simd.md
8411 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
8412 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
8413 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
8414 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
8415 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
8416 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
8417 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
8418 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
8419 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
8421 2018-07-24 Jakub Jelinek <jakub@redhat.com>
8424 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
8425 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
8426 and abs_d == d, do the power of two handling if profitable.
8428 2018-07-24 Richard Biener <rguenther@suse.de>
8430 * match.pd: Add BIT_FIELD_REF canonicalizations.
8432 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
8435 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
8437 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
8439 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
8440 terminated STRING_CST object.
8442 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
8444 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
8446 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
8448 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
8449 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
8450 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
8451 the elements into a register.
8452 (rs6000_split_v4si_init_di_reg): Delete.
8453 (rs6000_split_v4si_init): Delete.
8454 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
8455 (vsx_init_v4si): Rewrite as a define_expand.
8457 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
8459 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
8460 zero_extend argument from memory): New.
8462 2018-07-22 Martin Sebor <msebor@redhat.com>
8465 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
8466 diagnosing calls with unknown arguments unless -Walloca-larger-than
8467 is restricted to less than PTRDIFF_MAX bytes.
8469 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
8471 * doc/gcov.texi (Invoking Gcov): Editorial changes.
8473 2018-07-20 David Malcolm <dmalcolm@redhat.com>
8475 * pretty-print.c (text_info::set_location): Remove redundant
8476 "line_table" parameter from call to rich_location::set_range.
8478 2018-07-20 Martin Sebor <msebor@redhat.com>
8481 * builtins.c (expand_builtin_alloca): Adjust.
8482 * calls.c (alloc_max_size): Simplify.
8483 * cgraphunit.c (cgraph_node::expand): Adjust.
8484 * common.opt (larger_than_size, warn_frame_larger_than): Remove
8486 (frame_larger_than_size): Same.
8487 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
8488 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
8490 * doc/invoke.texi (GCC Command Options): Document option arguments.
8491 Explain byte-size arguments and suffixes.
8492 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
8493 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
8494 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
8495 * doc/options.texi (UInteger): Expand.
8496 (Host_Wide_Int, ByteSize): Document new properties.
8497 * final.c (final_start_function_1): Include sizes in an error message.
8498 * function.c (frame_offset_overflow): Same.
8499 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
8500 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
8501 Diagnose unbounded alloca calls only for limits of less than
8503 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
8504 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
8506 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
8507 only for limits of less than PTRDIFF_MAX.
8508 * langhooks-def.h (lhd_handle_option): Change function argument
8510 * langhooks.c (lhd_handle_option): Same.
8511 * langhooks.h (handle_option): Same.
8512 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
8514 (var_type, var_type_struct): Same.
8515 (var_set): Handle ByteSize flag.
8516 * optc-gen.awk: Add comments to output to ease debugging. Make
8517 use of HOST_WIDE_INT where appropriate.
8518 * opts-gen-save.awk: Use %lx to format unsigned long.
8519 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
8520 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
8521 arguments. Parse bytes-size suffixes.
8522 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
8523 (enum_value_to_arg): Same.
8524 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
8525 (handle_option): Adjust.
8526 (generate_option): Change function argument to HOST_WIDE_INT.
8527 (cmdline_handle_error): Adjust.
8528 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
8529 (set_option): Change function argument to HOST_WIDE_INT.
8530 (option_enabled): Handle cl_host_wide_int.
8531 (get_option_state): Handle CLVC_SIZE.
8532 (control_warning_option): Same.
8533 * opts.c (common_handle_option): Change function argument to
8534 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
8535 OPT_Wvla_larger_than_.
8536 * opts.h (enum cl_var_type): Add an enumerator.
8537 * stor-layout.c (layout_decl): Print a more meaningful warning.
8538 * toplev.c (output_stack_usage): Adjust.
8540 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
8542 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
8543 call to inline_expand_builtin_string_cmp.
8544 (expand_builtin_strcmp): Likewise.
8545 (expand_builtin_strncmp): Likewise.
8546 (inline_string_cmp): Delete the last parameter, change char_type_node
8547 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
8549 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
8550 the inlining expansion on target where the type of the call has same or
8551 narrower precision than unsigned char.
8553 2018-07-20 David Malcolm <dmalcolm@redhat.com>
8555 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
8556 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
8557 * common.opt (fsave-optimization-record): New option.
8558 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
8559 * doc/invoke.texi (-fsave-optimization-record): New option.
8560 * dumpfile.c: Include "optinfo-emit-json.h".
8561 (struct kv_pair): Move to coretypes.h.
8562 (optgroup_options): Make non-static.
8563 (dump_context::end_scope): Call
8564 optimization_records_maybe_pop_dump_scope.
8565 * dumpfile.h (optgroup_options): New decl.
8566 * json.cc: New file.
8568 * optinfo-emit-json.cc: New file.
8569 * optinfo-emit-json.h: New file.
8570 * optinfo.cc: Include "optinfo-emit-json.h".
8571 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
8572 (optinfo_enabled_p): Check optimization_records_enabled_p.
8573 (optinfo_wants_inlining_info_p): Likewise.
8574 * optinfo.h: Update comment.
8575 * profile-count.c (profile_quality_as_string): New function.
8576 * profile-count.h (profile_quality_as_string): New decl.
8577 (profile_count::quality): New accessor.
8578 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
8579 and optinfo_emit_json_cc_tests.
8580 * selftest.h (selftest::json_cc_tests): New decl.
8581 (selftest::optinfo_emit_json_cc_tests): New decl.
8582 * toplev.c: Include "optinfo-emit-json.h".
8583 (compile_file): Call optimization_records_finish.
8584 (do_compile): Call optimization_records_start.
8585 * tree-ssa-live.c: Include optinfo.h.
8586 (remove_unused_scope_block_p): Retain inlining information if
8587 optinfo_wants_inlining_info_p returns true.
8589 2018-07-20 Richard Biener <rguenther@suse.de>
8592 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
8593 to cover -flto-partition=none.
8595 2018-07-20 Martin Liska <mliska@suse.cz>
8597 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
8598 (get_decl_source_range): Remove unused function.
8600 2018-07-20 Richard Biener <rguenther@suse.de>
8602 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
8603 (struct vn_phi_s): Likewise.
8604 (struct vn_reference_s): Likewise.
8605 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
8606 for searching the slot of an entry known to be in the hash itself.
8607 (vn_phi_hasher::equal): Likewise.
8608 (vn_reference_hasher::equal): Likewise.
8609 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
8611 (optimistic_info, current_info): Remove, keeping only valid_info.
8612 (vn_reference_lookup_1): Remove fallback lookup.
8613 (vn_reference_lookup_2): Likewise.
8614 (vn_nary_op_lookup_1): Likewise.
8615 (vn_phi_lookup): Likewise.
8616 (vn_nary_build_or_lookup_1): Make sure to not chain the built
8618 (vn_reference_insert): Adjust, chain the inserted hash element
8619 at last_inserted_ref.
8620 (vn_reference_insert_pieces): Likewise.
8621 (visit_reference_op_call): Likewise.
8622 (vn_nary_op_insert_into): Chain the inserted hash element at
8624 (vn_nary_op_insert_pieces): Adjust.
8625 (vn_nary_op_insert): Likewise.
8626 (vn_nary_op_insert_stmt): Likewise.
8627 (vn_phi_insert): Adjust, chain the inserted hash element at
8629 (process_scc): Remove clearing and copying the optimistic
8630 table. Instead remove elements inserted during an optimistic
8631 iteration from the single table we maintain.
8632 (init_scc_vn): Adjust.
8633 (free_scc_vn): Likewise.
8634 (sccvn_dom_walker::record_cond): Likewise.
8635 (sccvn_dom_walker::after_dom_children): Likewise.
8637 2018-07-19 Martin Sebor <msebor@redhat.com>
8639 PR tree-optimization/84047
8640 PR tree-optimization/83776
8641 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
8642 (check_array_bounds): Call it.
8644 2018-07-19 Martin Sebor <msebor@redhat.com>
8646 * align.h (align_flags): Use member initialization.
8648 2018-07-19 David Malcolm <dmalcolm@redhat.com>
8650 * Makefile.in (OBJS): Add optinfo.o.
8651 * coretypes.h (class symtab_node): New forward decl.
8652 (struct cgraph_node): New forward decl.
8653 (class varpool_node): New forward decl.
8654 * dump-context.h: New file.
8655 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
8657 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
8658 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
8659 (set_alt_dump_file): Likewise.
8660 (dump_context::~dump_context): New dtor.
8661 (dump_gimple_stmt): Move implementation to...
8662 (dump_context::dump_gimple_stmt): ...this new member function.
8663 Add the stmt to any pending optinfo, creating one if need be.
8664 (dump_gimple_stmt_loc): Move implementation to...
8665 (dump_context::dump_gimple_stmt_loc): ...this new member function.
8666 Start a new optinfo and add the stmt to it.
8667 (dump_gimple_expr): Move implementation to...
8668 (dump_context::dump_gimple_expr): ...this new member function.
8669 Add the stmt to any pending optinfo, creating one if need be.
8670 (dump_gimple_expr_loc): Move implementation to...
8671 (dump_context::dump_gimple_expr_loc): ...this new member function.
8672 Start a new optinfo and add the stmt to it.
8673 (dump_generic_expr): Move implementation to...
8674 (dump_context::dump_generic_expr): ...this new member function.
8675 Add the tree to any pending optinfo, creating one if need be.
8676 (dump_generic_expr_loc): Move implementation to...
8677 (dump_context::dump_generic_expr_loc): ...this new member
8678 function. Add the tree to any pending optinfo, creating one if
8680 (dump_printf): Move implementation to...
8681 (dump_context::dump_printf_va): ...this new member function. Add
8682 the text to any pending optinfo, creating one if need be.
8683 (dump_printf_loc): Move implementation to...
8684 (dump_context::dump_printf_loc_va): ...this new member function.
8685 Start a new optinfo and add the stmt to it.
8686 (dump_dec): Move implementation to...
8687 (dump_context::dump_dec): ...this new member function. Add the
8688 value to any pending optinfo, creating one if need be.
8689 (dump_context::dump_symtab_node): New member function.
8690 (dump_context::get_scope_depth): New member function.
8691 (dump_context::begin_scope): New member function.
8692 (dump_context::end_scope): New member function.
8693 (dump_context::ensure_pending_optinfo): New member function.
8694 (dump_context::begin_next_optinfo): New member function.
8695 (dump_context::end_any_optinfo): New member function.
8696 (dump_context::s_current): New global.
8697 (dump_context::s_default): New global.
8698 (dump_scope_depth): Delete global.
8699 (dumpfile_ensure_any_optinfo_are_flushed): New function.
8700 (dump_symtab_node): New function.
8701 (get_dump_scope_depth): Reimplement in terms of dump_context.
8702 (dump_begin_scope): Likewise.
8703 (dump_end_scope): Likewise.
8704 (selftest::temp_dump_context::temp_dump_context): New ctor.
8705 (selftest::temp_dump_context::~temp_dump_context): New dtor.
8706 (selftest::verify_item): New function.
8707 (ASSERT_IS_TEXT): New macro.
8708 (ASSERT_IS_TREE): New macro.
8709 (ASSERT_IS_GIMPLE): New macro.
8710 (selftest::test_capture_of_dump_calls): New test.
8711 (selftest::dumpfile_c_tests): Call it.
8712 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
8713 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
8714 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
8715 descriptive comment.
8716 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
8717 (dump_node, dump_bb): Move these unrelated decls.
8718 (class dump_manager): Add leading comment.
8719 * optinfo.cc: New file.
8720 * optinfo.h: New file.
8722 2018-07-19 Michael Collison <michael.collison@arm.com>
8723 Richard Henderson <rth@redhat.com>
8725 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
8726 (subti): Handle op1 zero.
8727 (subvti4, usub4ti4): New.
8728 (*sub<GPI>3_compare1_imm): New.
8729 (sub<GPI>3_carryinCV): New.
8730 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
8731 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
8733 2018-07-19 Michael Collison <michael.collison@arm.com>
8734 Richard Henderson <rth@redhat.com>
8736 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
8737 (addti3): Create simpler code if low part is already known to be 0.
8738 (addvti4, uaddvti4): New.
8739 (*add<GPI>3_compareC_cconly_imm): New.
8740 (*add<GPI>3_compareC_cconly): New.
8741 (*add<GPI>3_compareC_imm): New.
8742 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
8743 handle constants within this pattern..
8744 (*add<GPI>3_compareV_cconly_imm): New.
8745 (*add<GPI>3_compareV_cconly): New.
8746 (*add<GPI>3_compareV_imm): New.
8747 (add<GPI>3_compareV): New.
8748 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
8749 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
8750 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
8751 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
8752 with 'comparison' operator.
8753 (*add<GPI>3_compareV_cconly_imm): Ditto.
8754 (*add<GPI>3_compareV_cconly): Ditto.
8755 (*add<GPI>3_compareV_imm): Ditto.
8756 (add<GPI>3_compareV): Ditto.
8757 (add<mode>3_carryinC): Ditto.
8758 (*add<mode>3_carryinC_zero): Ditto.
8759 (*add<mode>3_carryinC): Ditto.
8760 (add<mode>3_carryinV): Ditto.
8761 (*add<mode>3_carryinV_zero): Ditto.
8762 (*add<mode>3_carryinV): Ditto.
8764 2018-07-19 Michael Collison <michael.collison@arm.com>
8765 Richard Henderson <rth@redhat.com>
8767 * config/aarch64/aarch64-modes.def (CC_V): New.
8768 * config/aarch64/aarch64-protos.h
8769 (aarch64_addti_scratch_regs): Declare
8770 (aarch64_subvti_scratch_regs): Declare.
8771 (aarch64_expand_subvti): Declare.
8772 (aarch64_gen_unlikely_cbranch): Declare
8773 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
8774 for signed overflow using CC_Vmode.
8775 (aarch64_get_condition_code_1): Handle CC_Vmode.
8776 (aarch64_gen_unlikely_cbranch): New function.
8777 (aarch64_addti_scratch_regs): New function.
8778 (aarch64_subvti_scratch_regs): New function.
8779 (aarch64_expand_subvti): New function.
8781 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
8783 * config/aarch64/aarch64-option-extensions.def: New entry for profile
8785 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
8786 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
8789 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
8792 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8793 address check not strict.
8795 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
8797 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
8799 (store_pair_lanes<mode>): Likewise.
8800 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
8801 enum value 'ADDR_QUERY_LDP_STP_N'.
8802 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
8803 (aarch64_print_address_internal): Add declaration.
8804 (aarch64_print_ldpstp_address): Remove.
8805 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
8806 (aarch64_print_operand): Change printing of 'y'.
8807 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
8808 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
8809 'true' rather than '1'.
8810 * config/aarch64/constraints.md (Uml): Likewise.
8811 (Uml): Rename to Umn.
8814 2018-07-19 Richard Biener <rguenther@suse.de>
8816 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
8818 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
8819 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
8820 (vn_reference_hasher): Likewise.
8821 (struct vn_tables_s): Remove obstack and alloc-pool members.
8822 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
8823 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
8824 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
8825 (vn_reference_insert_pieces): Likewise.
8826 (alloc_vn_nary_op_noinit): Adjust.
8827 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
8828 (vn_phi_eq): Adjust.
8829 (shared_lookup_phiargs): Remove.
8830 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
8831 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
8832 (visit_reference_op_call): Likewise.
8833 (copy_nary, copy_phi, copy_reference): Remove.
8834 (process_scc): Rewind the obstack when iterating. Do not
8835 copy the elements to valid_info but just move them from one
8836 hashtable to the other.
8837 (allocate_vn_table): Adjust.
8838 (free_vn_table): Likewise.
8839 (init_scc_vn): Likewise.
8840 (free_scc_vn): Likewise.
8842 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
8845 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
8846 indirect_return as function type attribute.
8847 (ix86_attribute_table): Change indirect_return to function
8849 * doc/extend.texi: Update indirect_return attribute.
8851 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
8853 * wide-int.h (widest2_int): New.
8854 * gimple-fold.c (arith_overflowed_p): Use it.
8855 * tree.h (widest2_int_cst): New.
8856 * tree-vrp.c (wide_int_binop_overflow): Rename from
8857 vrp_int_const_binop.
8858 Rewrite to work on trees.
8859 (extract_range_from_multiplicative_op_1): Abstract code to...
8860 (wide_int_range_min_max): ...here.
8861 (wide_int_range_cross_product): ...and here.
8862 (extract_range_from_binary_expr_1): Abstract overflow code to...
8863 (wide_int_range_mult_wrapping): ...here.
8864 * tree-vrp.h (wide_int_range_cross_product): New.
8865 (wide_int_range_mult_wrapping): New.
8867 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
8868 Julia Koval <julia.koval@intel.com>
8870 * config/i386/x86-tune-costs.h (skylake_memcpy,
8871 skylake_memset): Replace rep_prefix with unrolling for size 512.
8873 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
8876 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
8877 comparision with EQ_EXPR in last stmt.
8879 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
8881 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
8882 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
8883 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
8884 previously known as "PowerPC AltiVec Built-in Functions". Move
8885 some material to new subsubsections "PowerPC AltiVec Built-in
8886 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
8888 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
8889 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
8890 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
8891 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
8893 2018-07-18 Richard Biener <rguenther@suse.de>
8895 PR tree-optimization/86557
8896 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
8899 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
8901 * config/s390/s390.c (s390_function_profiler): Generate CFI.
8903 2018-07-17 Jeff Law <law@redhat.com>
8905 * config/arm/arm.c (get_label_padding): Update for recent
8906 changes to label_to_alignment.
8908 PR tree-optimization/86010
8909 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
8911 * config/mips/mips.c (vr4130_align_insns): Update for recent
8912 changes to label_to_alignment.
8914 * config/frv/frv.c (frv_label_align): Update for recent changes
8915 to label_to_alignment.
8917 * config/nios2/nios2.c (nios2_label_align): Update for recent
8918 changes which dropped ALIGN_LABELS_LOG.
8920 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
8922 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8923 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8925 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
8927 * config/arc/arc.c (arc_label_align): Use align_labels instead of
8928 deprecated align_labels_log.
8930 2018-07-17 Richard Biener <rguenther@suse.de>
8933 * dwarf2out.c (init_sections_and_labels): Always generate
8934 a debug_line_str_section for early LTO debug.
8935 (dwarf2out_finish): Reset debug_line_str_hash output early.
8936 Bump counter for extra dwarf5 .debug_loc labels to not conflict
8938 (dwarf2out_early_finish): Output debug_line_str.
8940 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8942 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8943 index register on z196 or later.
8945 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8947 * config/s390/s390.c (s390_default_align): Set default function
8949 (s390_override_options_after_change): Call s390_default align.
8950 (s390_option_override_internal): Call s390_default_align.
8951 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8953 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8956 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8957 to _looptemp_ clauses, other than the first two.
8959 2018-07-17 Martin Liska <mliska@suse.cz>
8961 * opts.c: Do not enable OPT_falign_* for -Os.
8963 2018-07-17 Martin Liska <mliska@suse.cz>
8965 * align.h (MAX_CODE_ALIGN): New.
8966 (MAX_CODE_ALIGN_VALUE): New.
8967 * common/config/i386/i386-common.c (ix86_handle_option):
8968 (MAX_CODE_ALIGN): Moved to align.h.
8969 * final.c (MAX_CODE_ALIGN): Likewise.
8970 * opts.c (parse_and_check_align_values):
8971 (MAX_CODE_ALIGN): Likewise.
8972 (MAX_CODE_ALIGN_VALUE): Likewise.
8974 2018-07-17 Martin Liska <mliska@suse.cz>
8976 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8977 in order to fulfil coding style.
8978 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8979 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8980 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8981 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8982 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8983 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8984 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8985 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8986 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8988 2018-07-17 Martin Liska <mliska@suse.cz>
8990 * align.h: New file.
8991 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8993 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8994 align_flags of label_to_alignment.
8995 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8997 * config/m68k/m68k.c: Do not use removed align_labels_value and
8999 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
9000 (LOOP_ALIGN): Likewise.
9001 (LABEL_ALIGN): Likewise.
9002 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9003 Remove not used macro.
9004 (rs6000_loop_align): Change return type to align_flags.
9005 (rs6000_loop_align_max_skip): Remove.
9006 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
9007 Change return type to align_flags.
9008 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9009 Remove not used macro.
9010 (rs6000_loop_align): Change return type to align_flags.
9011 (rs6000_loop_align_max_skip): Remove.
9012 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
9013 * config/rx/rx-protos.h (rx_align_for_label): Make it
9015 * config/rx/rx.c (rx_align_for_label): Change return type
9017 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
9019 into align_flags class.
9020 (LABEL_ALIGN): Likewise.
9021 (LOOP_ALIGN): Likewise.
9022 * config/s390/s390.c (s390_label_align): Use align_flags
9024 (s390_asm_output_function_label): Likewise.
9025 * config/sh/sh.c (sh_override_options_after_change):
9026 Use align_flags class directly without macros.
9027 (find_barrier): Likewise.
9028 (barrier_align): Likewise.
9029 (sh_loop_align): Likewise.
9030 * config/spu/spu.c (spu_option_override):
9031 Use align_flags_tuple::get_value instead of removed macros.
9032 (spu_sched_init): Likewise.
9033 * config/spu/spu.h (GTY): Likewise.
9034 * config/visium/visium.c (visium_option_override):
9035 Set "8" as default secondary alignment.
9036 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
9037 in order to guarantee secondary alignment of 8.
9038 * coretypes.h: Include align.h header file.
9039 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
9040 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
9041 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
9042 * doc/tm.texi.in: Likewise.
9043 * final.c (struct label_alignment): Remove not used structure.
9044 (LABEL_ALIGN): Change type to align_flags.
9045 (LOOP_ALIGN): Likewise.
9046 (JUMP_ALIGN): Likewise.
9047 (default_loop_align_max_skip): Remove.
9048 (default_label_align_max_skip): Likewise.
9049 (default_jump_align_max_skip): Likewise.
9050 (default_label_align_after_barrier_max_skip):
9051 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
9052 (LABEL_TO_MAX_SKIP): Remove.
9053 (label_to_alignment): Return align_flags type instead of integer.
9054 (label_to_max_skip): Remove.
9055 (align_fuzz): Use align_flags type.
9056 (compute_alignments): Use align_flags type and use align_flags::max
9057 to combine multiple alignments.
9058 (grow_label_align): Grow vec instead of C array.
9059 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
9060 (shorten_branches): Use align_flags type and use align_flags::max
9061 to combine multiple alignments.
9062 (final_scan_insn_1): Remove usage of secondary alignment that comes
9063 from label alignment, but instead use proper secondary alignment
9064 which is computed in grow_label_align.
9065 * flags.h (struct align_flags_tuple): Move to align.h.
9066 (struct align_flags): Likewise.
9067 (state_align_loops): Rename to align_loops.
9068 (state_align_jumps): Rename to align_jumps.
9069 (state_align_labels): Rename to align_labels.
9070 (state_align_functions): Rename to align_functions.
9071 (align_loops_log): Remove.
9072 (align_jumps_log): Remove.
9073 (align_labels_log): Remove.
9074 (align_functions_log): Remove.
9075 (align_loops_max_skip): Remove.
9076 (align_jumps_max_skip): Remove.
9077 (align_labels_max_skip): Remove.
9078 (align_functions_max_skip): Remove.
9079 (align_loops_value): Remove.
9080 (align_jumps_value): Remove.
9081 (align_labels_value): Remove.
9082 (align_functions_value): Remove.
9083 * output.h (label_to_alignment): Change return type to align_flags.
9084 (label_to_max_skip): Remove.
9085 * target.def: Remove loop_align_max_skip, label_align_max_skip,
9086 jump_align_max_skip macros.
9087 * targhooks.h (default_loop_align_max_skip): Remove.
9088 (default_label_align_max_skip): Likewise.
9089 (default_jump_align_max_skip): Likewise.
9090 (default_label_align_after_barrier_max_skip): Remove.
9091 * toplev.c (read_log_maxskip): Use ::normalize function.
9092 (parse_N_M): Remove not used argument and also call ::normalize.
9093 (parse_alignment_opts): Do not pass unused arguments.
9094 * varasm.c (assemble_start_function): Use directly align_functions
9095 instead of removed macros.
9096 * system.h: Do not poison removed macros.
9098 2018-07-17 Jakub Jelinek <jakub@redhat.com>
9101 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
9102 and cond temporaries don't have reference type if iterator has
9103 pointer type. For init use &for_pre_body instead of pre_p if
9104 for_pre_body is non-empty.
9106 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
9108 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9109 double-double modes to SFmode directly directly.
9110 (trunc<mode>sf2_fprs): Delete.
9112 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
9114 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9115 for conversions between IFmode and the decimal floating point modes.
9116 (init_float128_ieee): Use the correct names for conversions between
9117 KFmode and the decimal floating point modes.
9119 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
9121 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9122 for the conversions between TDmode and IFmode.
9123 (init_float128_ieee): Use more correct names for the conversions
9124 between TDmode and KFmode.
9126 2018-07-16 Jakub Jelinek <jakub@redhat.com>
9128 PR tree-optimization/86526
9129 * builtins.c (expand_builtin_memcmp): Formatting fixes.
9130 (inline_expand_builtin_string_cmp): Likewise.
9131 (inline_string_cmp): Likewise. Use c_readstr instead of
9132 builtin_memcpy_read_str. Add unit_mode temporary.
9134 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
9137 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9138 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9140 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9142 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9143 Alphabetize prototypes of built-in functions, separating out
9144 built-in functions that are listed in this section but should be
9145 described elsewhere.
9147 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
9150 * expmed.c (emit_store_flag): Do not emit setcc followed by a
9151 conditional move when trapping comparison was split to a
9152 non-trapping one (and vice versa).
9154 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9156 * config/s390/s390.c (s390_function_profiler): Generate nops
9157 instead of profiler call sequences.
9158 * config/s390/s390.opt: Add the new option.
9160 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9162 * config/s390/s390.c (s390_function_profiler): Generate
9163 __mcount_loc section.
9164 * config/s390/s390.opt: Add the new option.
9166 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9168 * common.opt: Add the new warning.
9169 * config/s390/s390.c (s390_function_profiler): Emit "brasl
9170 %r0,__fentry__" when -mfentry is specified.
9171 (s390_option_override_internal): Disallow -mfentry for 31-bit
9173 * config/s390/s390.opt: Add the new option.
9175 2018-07-16 Richard Biener <rguenther@suse.de>
9178 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9179 for function-local FUNCTION_DECL and RESULT_DECL immediately.
9181 2018-07-16 Martin Liska <mliska@suse.cz>
9184 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9187 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
9189 * config/arc/arcHS.md: Update ARCHS scheduling rules.
9191 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
9193 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9195 * config/arc/arc-cpus.def (hs4x): New cpu.
9197 * config/arc/arc-tables.opt: Regenerate.
9198 * config/arc/arc.c (arc_sched_issue_rate): New function.
9199 (TARGET_SCHED_ISSUE_RATE): Define.
9200 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9201 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9203 (attr tune): Add ARCHS4x tune values.
9204 (attr tune_dspmpy): Define.
9205 (*tst): Correct instruction type.
9206 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9207 * config/arc/arcHS4x.md: New file.
9208 * config/arc/fpu.md: Update instruction type attributes.
9209 * config/arc/t-multilib: Regenerate.
9211 2018-07-16 Tom de Vries <tdevries@suse.de>
9214 * var-tracking.c (vt_initialize): Fix pre_dec handling.
9216 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9218 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
9221 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
9223 PR tree-optimization/86514
9224 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
9225 conversion to a boolean type from a type with greater precision.
9227 2018-07-16 Tom de Vries <tdevries@suse.de>
9229 * var-tracking.c (vt_initialize): Print adjusted insn slim if
9230 dump_flags request TDF_SLIM.
9232 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
9234 * fold-const.c (int_const_binop_1): Abstract...
9235 (wide_int_binop): ...wide int code here.
9236 (poly_int_binop): ...poly int code here.
9237 Abstract the rest of int_const_binop_1 into int_const_binop.
9238 * fold-const.h (wide_int_binop): New.
9239 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
9240 Remove useless PLUS/MINUS_EXPR case.
9241 (zero_nonzero_bits_from_vr): Move wide int code...
9242 (zero_nonzero_bits_from_bounds): ...here.
9243 (extract_range_from_binary_expr_1): Move mask optimization code...
9244 (range_easy_mask_min_max): ...here.
9245 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
9246 (range_easy_mask_min_max): New.
9248 2018-07-15 Jeff Law <law@redhat.com>
9251 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
9254 2018-07-14 Jim Wilson <jimw@sifive.com>
9256 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
9258 2018-07-14 Paul Koning <ni1d@arrl.net>
9260 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
9262 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
9264 * lto-streamer-out.c (copy_function_or_variable): Dump info about
9267 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
9268 Steve Munroe <munroesj52@gmail.com>
9270 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
9271 (_mm_andnot_si128): Likewise.
9272 (_mm_or_si128): Likewise.
9273 (_mm_xor_si128): Likewise.
9275 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
9278 * builtins.c (expand_builtin_memcmp): Inline the calls first
9279 when result_eq is false.
9280 (expand_builtin_strcmp): Inline the calls first.
9281 (expand_builtin_strncmp): Likewise.
9282 (inline_string_cmp): New routine. Expand a string compare
9283 call by using a sequence of char comparison.
9284 (inline_expand_builtin_string_cmp): New routine. Inline expansion
9285 a call to str(n)cmp/memcmp.
9286 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
9288 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
9290 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
9292 * config/arm/driver-arm.c: Include arm-native.h.
9293 (host_detect_local_cpu): Use auto-generated data tables.
9294 (vendors, arm_cpu_table): Delete. Move part information to ...
9295 * config/arm/arm-cpus.in: ... here.
9296 * config/arm/parsecpu.awk (gen_native): New function.
9297 (vendor, part): New CPU fields.
9298 (END): Add support for building the native CPU detection tables.
9299 * config/arm/t-arm (arm-native.h): Add build rule.
9300 (driver-arm.o): Add dependency on arm-native.h.
9302 2018-07-13 Richard Biener <rguenther@suse.de>
9305 * match.pd (addr1 - addr2): Allow either of the operand to
9308 2018-07-13 Tom de Vries <tdevries@suse.de>
9310 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
9313 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
9315 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
9316 arrays instead of numbered variables.
9318 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
9320 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
9321 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
9323 2018-07-13 Richard Biener <rguenther@suse.de>
9326 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
9327 instead of get_context_die.
9329 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
9330 Richard Biener <rguenther@suse.de>
9333 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
9334 that the loop latch destination where phi is defined.
9336 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
9338 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
9339 (riscv_expand_epilogue): Add assertion to check interrupt mode.
9340 (riscv_set_current_function): Extract getting interrupt type to new
9342 (riscv_get_interrupt_type): New function.
9343 (riscv_merge_decl_attributes): New function, checking interrupt type is
9345 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
9347 2018-07-12 Paul Koning <ni1d@arrl.net>
9349 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
9352 2018-07-12 Paul Koning <ni1d@arrl.net>
9354 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
9355 zero reference, add doloop_end instead.
9356 * doc/md.texi (decrement_and_branch_until_zero): Remove.
9357 (Looping patterns): Remove decrement_and_branch_until_zero. Add
9358 detail for doloop_end.
9360 2018-07-12 Martin Sebor <msebor@redhat.com>
9363 * attribs.c (decl_attributes): Reject conflicting attributes before
9364 calling attribute handlers.
9366 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
9368 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
9370 (gcc::dump_manager::get_dump_file_name): likewise.
9371 (dump_begin): Likewise.
9372 * dumpfile.h (dump_begin): Update prototype.
9373 (gcc::dump_manager::get_dump_file_name,
9374 gcc::dump_manager::get_dump_file_name): Update prototype.
9376 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9378 * internal-fn.h (vectorizable_internal_fn_p): New function.
9379 * tree-vect-slp.c (compatible_calls_p): Likewise.
9380 (vect_build_slp_tree_1): Remove nops argument. Handle calls
9381 to internal functions.
9382 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
9384 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9386 * fold-const.h (inverse_conditions_p): Declare.
9387 * fold-const.c (inverse_conditions_p): New function.
9388 * match.pd: Use inverse_conditions_p. Add folds of view_converts
9389 that test the inverse condition of a conditional internal function.
9390 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
9391 * internal-fn.c (internal_fn_mask_index): Handle conditional
9393 (vectorized_internal_fn_supported_p): New function.
9394 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
9395 (any_pred_load_store): Replace with...
9396 (need_to_predicate): ...this new variable.
9397 (redundant_ssa_names): New variable.
9398 (ifcvt_can_use_mask_load_store): Move initial checks to...
9399 (ifcvt_can_predicate): ...this new function. Handle tree codes
9400 for which a conditional internal function exists.
9401 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
9402 instead of ifcvt_can_use_mask_load_store. Update after variable
9404 (predicate_load_or_store): New function, split out from
9405 predicate_mem_writes.
9406 (check_redundant_cond_expr): New function.
9407 (value_available_p): Likewise.
9408 (predicate_rhs_code): Likewise.
9409 (predicate_mem_writes): Rename to...
9410 (predicate_statements): ...this. Use predicate_load_or_store
9411 and predicate_rhs_code.
9412 (combine_blocks, tree_if_conversion): Update after above name changes.
9413 (ifcvt_local_dce): Handle redundant_ssa_names.
9414 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
9415 general conditional functions.
9416 * tree-vect-stmts.c (vectorizable_call): Likewise.
9418 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9419 Alan Hayward <alan.hayward@arm.com>
9420 David Sherwood <david.sherwood@arm.com>
9422 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
9423 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
9424 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
9425 plus and minus and convert them into IFN_COND_FMA-based sequences.
9426 (convert_mult_to_fma): Handle conditional plus and minus.
9428 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9430 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
9431 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
9432 (cond_fnms_optab): New optabs.
9433 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
9435 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
9436 * internal-fn.h (get_conditional_internal_fn): Declare.
9437 (get_unconditional_internal_fn): Likewise.
9438 * internal-fn.c (cond_ternary_direct): New macro.
9439 (expand_cond_ternary_optab_fn): Likewise.
9440 (direct_cond_ternary_optab_supported_p): Likewise.
9441 (FOR_EACH_COND_FN_PAIR): Likewise.
9442 (get_conditional_internal_fn): New function.
9443 (get_unconditional_internal_fn): Likewise.
9444 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
9445 (gimple_match_op::gimple_match_op): Add a new overload for 5
9447 (gimple_match_op::set_op): Likewise.
9448 (gimple_resimplify5): Declare.
9449 * genmatch.c (decision_tree::gen): Generate simplifications for
9451 * gimple-match-head.c (gimple_simplify): Define an overload for
9452 5 operands. Handle calls with 5 arguments in the top-level overload.
9453 (convert_conditional_op): Handle conversions from unconditional
9454 internal functions to conditional ones.
9455 (gimple_resimplify5): New function.
9456 (build_call_internal): Pass a fifth operand.
9457 (maybe_push_res_to_seq): Likewise.
9458 (try_conditional_simplification): Try converting conditional
9459 internal functions to unconditional internal functions.
9460 Handle 3-operand unconditional forms.
9461 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
9462 Define ternary equivalents of the current rules for binary conditional
9464 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
9466 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
9467 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
9468 (optab): Handle them.
9469 (SVE_COND_FP_TERNARY): New int iterator.
9470 (sve_fmla_op, sve_fmad_op): New int attributes.
9471 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
9472 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
9473 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
9475 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9477 * target.def (preferred_else_value): New target hook.
9478 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
9479 * doc/tm.texi: Regenerate.
9480 * targhooks.h (default_preferred_else_value): Declare.
9481 * targhooks.c (default_preferred_else_value): New function.
9482 * internal-fn.h (conditional_internal_fn_code): Declare.
9483 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
9484 (get_conditional_internal_fn): Use it.
9485 (conditional_internal_fn_code): New function.
9486 * gimple-match.h (gimple_match_cond): New struct.
9487 (gimple_match_op): Add a cond member function.
9488 (gimple_match_op::gimple_match_op): Update all forms to take a
9490 * genmatch.c (expr::gen_transform): Use the same condition as res_op
9491 for the suboperation, but don't specify a particular else_value.
9492 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
9493 (visit_nary_op, visit_reference_op_load): Pass
9494 gimple_match_cond::UNCOND to the gimple_match_op constructor.
9495 * gimple-match-head.c: Include tree-eh.h
9496 (convert_conditional_op): New function.
9497 (maybe_resimplify_conditional_op): Likewise.
9498 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
9499 (gimple_resimplify2): Likewise.
9500 (gimple_resimplify3): Likewise.
9501 (gimple_resimplify4): Likewise.
9502 (maybe_push_res_to_seq): Return null for conditional operations.
9503 (try_conditional_simplification): New function.
9504 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
9506 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
9508 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
9510 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
9512 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
9514 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9516 (hash_tree): Do not hash DECL_FCONTEXT
9517 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
9518 Do not stream DECL_FCONTEXT.
9519 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
9520 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
9522 2018-07-12 Richard Biener <rguenther@suse.de>
9525 * dwarf2out.c (gen_block_die): Only output blocks when they have
9526 at least one !DECL_IGNORED_P variable.
9528 2018-07-12 Richard Biener <rguenther@suse.de>
9531 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
9532 Remove -mieee-fp handling.
9534 2018-07-12 Richard Biener <rguenther@suse.de>
9536 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
9537 left-over from last patch.
9539 2018-07-12 Jakub Jelinek <jakub@redhat.com>
9541 PR tree-optimization/86492
9542 * gimple-ssa-store-merging.c
9543 (imm_store_chain_info::coalesce_immediate_stores): Call
9544 check_no_overlap even for the merge_overlapping case. Formatting fix.
9546 2018-07-12 Richard Biener <rguenther@suse.de>
9549 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
9550 move possibly trapping operations into the conditional.
9552 2018-07-12 Richard Biener <rguenther@suse.de>
9554 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
9555 (vn_lookup_simplify_result): Remove recursion limit applied
9557 (vn_nary_build_or_lookup_1): Adjust.
9558 (try_to_simplify): Likewise.
9559 * gimple-match-head.c (gimple_resimplify1): Instead apply one
9561 (gimple_resimplify2): Likewise.
9562 (gimple_resimplify3): Likewise.
9563 (gimple_resimplify4): Likewise.
9565 2018-07-11 Jakub Jelinek <jakub@redhat.com>
9567 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
9568 Use __mmask64 type instead of __mmask8 for __M argument.
9569 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
9570 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
9572 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
9573 __mmask16 for __M argument.
9574 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
9575 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
9576 to __mmask16 instead of __mmask8.
9577 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
9578 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
9579 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
9580 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
9581 instead of __mmask16 for __U argument.
9582 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
9583 __mmask16 instead of __mmask8 for __U argument.
9584 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
9586 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
9588 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
9590 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
9592 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
9594 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
9596 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
9598 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
9600 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
9602 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
9604 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
9605 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
9606 return type as well as __M argument type and all casts from __mmask8
9608 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
9609 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
9610 return type as well as __M argument type and all casts from __mmask8
9612 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
9613 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
9614 return type as well as __M argument type and all casts from __mmask8
9616 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
9617 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
9618 return type as well as __M argument type and all casts from __mmask8
9620 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
9621 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
9624 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
9626 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
9627 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
9630 2018-07-11 Paul Koning <ni1d@arrl.net>
9632 * doc/md.texi (define_subst): Document how multiple occurrences of
9633 the same argument in the replacement pattern are handled.
9635 2018-07-11 Paul Koning <ni1d@arrl.net>
9637 * doc/extend.texi (Common Variable Attributes): Move "mode" into
9639 (Common Type Attributes): Add "mode" attribute.
9641 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
9643 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
9644 stream DECL_ORIGINAL_TYPE.
9645 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
9646 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
9647 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
9648 Do not walk original type.
9649 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
9651 (write_ts_decl_non_common_tree_pointers): Do not stream
9653 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
9654 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
9656 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
9658 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
9659 thread twice from the same starting edge.
9661 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
9663 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
9664 * gimple.c (gimple_call_nonnull_result_p): ...here...
9665 (gimple_call_nonnull_arg): ...and here.
9666 * gimple.h (gimple_call_nonnull_result_p): New.
9667 (gimple_call_nonnull_arg): New.
9669 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
9671 * config/arm/arm-cpus.in: Move information from fpu field of each
9672 cpu definition to the isa field.
9673 * config/arm/parsecpu.awk (fpu): Delete match rule.
9674 (gen_comm_data): Don't add bits from the CPU's FPU entry.
9676 2018-07-11 Richard Biener <rguenther@suse.de>
9679 * dwarf2out.c (init_sections_and_labels): Use
9680 output_asm_line_debug_info consistently.
9681 (dwarf2out_early_finish): Likewise.
9682 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
9685 2018-07-11 Richard Biener <rguenther@suse.de>
9687 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
9688 Rework father_bb setting in a way to avoid propagating constants
9689 multiple times on a loop body.
9691 2018-07-10 Mark Wielaard <mark@klomp.org>
9694 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
9696 2018-07-10 Richard Biener <rguenther@suse.de>
9698 * hash-map.h (hash_map::iterator::operator*): Return
9699 references to key and value.
9701 2018-07-10 Jakub Jelinek <jakub@redhat.com>
9704 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
9705 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
9706 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
9707 (gimplify_omp_for): For composite loops, move outer
9708 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
9709 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
9710 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
9711 TREE_LIST for both the original class iterator and the "last" helper
9712 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
9713 loop, remember has_decl_expr from outer composite loops for the
9714 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
9716 2018-07-09 Martin Sebor <msebor@redhat.com>
9720 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
9721 accessing implicitly initialized array elements.
9722 * expr.c (string_constant): Handle string initializers of
9723 character arrays within aggregates.
9724 * gimple-fold.c (fold_array_ctor_reference): Add argument.
9725 Store element offset. As a special case, handle zero size.
9726 (fold_nonarray_ctor_reference): Same.
9727 (fold_ctor_reference): Add argument. Store subobject offset.
9728 * gimple-fold.h (fold_ctor_reference): Add argument.
9730 2018-07-09 Paul Koning <ni1d@arrl.net>
9732 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
9733 (pdp11_insn_cost): New function.
9734 (pdp11_md_asm_adjust): New function.
9735 (TARGET_INVALID_WITHIN_DOLOOP): Define.
9736 (pdp11_rtx_costs): Update to match machine better.
9737 (output_addr_const_pdp11): Correct format mismatch warnings.
9738 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
9739 * config/pdp11/pdp11.md: General change to add base_cost and/or
9740 length attributes for use by new pdp11_insn_cost function.
9741 (MIN_BRANCH): Correct definition.
9743 (doloop_end): Use standard pattern name for looping pattern.
9744 (doloop_end_nocc): New.
9745 (movsf): Add another constraint alternative.
9746 (zero_extendqihi2): Add constraint alternatives for not in place
9748 (zero_extendhisi2): Remove.
9749 (shift patterns): Add CC handling variants.
9753 (define_peephole2): New peephole to recognize mov that sets CC for
9756 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9758 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
9759 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
9762 2018-07-09 Jakub Jelinek <jakub@redhat.com>
9765 * real.c (real_nextafter): Return true if result is denormal.
9767 2018-07-09 Martin Liska <mliska@suse.cz>
9769 * common.opt: Add back wrongly removed attribute.
9771 2018-07-09 Richard Biener <rguenther@suse.de>
9774 * dwarf2out.c (gen_block_die): For an early generated DIE
9775 always output high/low PC attributes.
9777 2018-07-09 Tom de Vries <tdevries@suse.de>
9779 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
9780 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
9781 onto VAR_DECL with abstract origin.
9783 2018-07-07 Jim Wilson <jimw@sifive.com>
9785 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
9787 2018-07-07 Tom de Vries <tdevries@suse.de>
9789 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
9791 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
9793 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
9795 (combine_bound): Use wide-int overflow calculation instead of
9797 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
9799 * fold-const.c (int_const_binop_2): Same.
9800 (extract_muldiv_1): Same.
9801 (fold_div_compare): Same.
9802 (fold_abs_const): Same.
9804 * poly-int.h (add): Same.
9808 * predict.c (predict_iv_comparison): Same.
9809 * profile-count.c (slow_safe_scale_64bit): Same.
9810 * simplify-rtx.c (simplify_const_binary_operation): Same.
9811 * tree-chrec.c (tree_fold_binomial): Same.
9812 * tree-data-ref.c (split_constant_offset_1): Same.
9813 * tree-if-conv.c (idx_within_array_bound): Same.
9814 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
9815 * tree-ssa-phiopt.c (minmax_replacement): Same.
9816 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
9817 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
9818 * vr-values.c (vr_values::adjust_range_with_scev): Same.
9819 * wide-int.cc (wi::add_large): Same.
9820 (wi::mul_internal): Same.
9821 (wi::sub_large): Same.
9822 (wi::divmod_internal): Same.
9823 * wide-int.h: Change overflow type to overflow_type for neg, add,
9824 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
9825 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
9826 mul_internal, divmod_internal.
9827 (overflow_type): New enum.
9828 (accumulate_overflow): New.
9830 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
9832 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
9833 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
9835 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
9837 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
9838 argument is checked for zero before entering loop, avoid checking again.
9840 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
9842 * gimplify.h (generic_expr_could_trap_p): Set as global function.
9843 * gimplify.h (generic_expr_could_trap_p): Likwise.
9844 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
9846 2018-07-06 Jakub Jelinek <jakub@redhat.com>
9848 PR tree-optimization/86401
9849 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
9850 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
9851 (fold_bit_and_mask): ... here. New helper function for match.pd.
9852 * fold-const.h (fold_bit_and_mask): Declare.
9853 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
9855 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
9858 * target.def (translate_mode_attribute): New hook.
9859 * targhooks.h (default_translate_mode_attribute): Declare.
9860 * targhooks.c (default_translate_mode_attribute): New function.
9861 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
9862 * doc/tm.texi: Regenerate.
9863 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
9864 (rs6000_translate_mode_attribute): New function.
9866 2018-07-06 Paul Koning <ni1d@arrl.net>
9868 * doc/md.texi (define_split): Document DONE and FAIL.
9869 (define_peephole2): Ditto.
9871 2018-07-05 Jeff Law <law@redhat.com>
9873 PR tree-optimization/86010
9874 * tree-ssa-dse.c (compute_trims): More aggressively trim at
9875 both the head and tail of mem* and str* calls.
9877 2018-07-05 Jim Wilson <jimw@sifive.com>
9879 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
9881 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
9883 * config/aarch64/aarch64-simd.md: correct flags text for
9884 MIN_EXPR replacement.
9886 2018-07-05 James Clarke <jrtc27@jrtc27.com>
9888 * configure: Regenerated.
9890 2018-07-05 Carl Love <cel@us.ibm.com>
9892 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
9893 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9894 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
9895 VSX_BUILTIN_DOUBLEL_V4SF.
9897 2018-07-05 Martin Sebor <msebor@redhat.com>
9900 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
9901 than its domain to compute its the upper bound of a char array.
9903 2018-07-05 Nathan Sidwell <nathan@acm.org>
9905 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
9906 * doc/cpp.texi: Update comment.
9907 * doc/tm.texi: Rebuilt.
9908 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
9909 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
9910 * doc/extend.texi (Backwards Compatibility): Clarify it is system
9911 headers affected by extern "C".
9912 * system.h: Poison NO_IMPLICIT_EXTERN_C.
9913 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9914 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9915 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9916 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9917 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9918 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9919 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9920 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9921 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9922 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9923 config/sparc/sp64-elf.h, config/spu/spu.h,
9924 config/stormy16/stormy16.h, config/v850/v850.h,
9925 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9926 define NO_IMPLICIT_EXTERN_C.
9927 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9929 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9932 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9933 instead of GET_MODE_SIZE when comparing Units.
9935 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9938 * rtlanal.c (set_noop_p): Constrain on mode change,
9939 include hard-reg-set.h
9941 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9943 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9945 2018-07-05 Jakub Jelinek <jakub@redhat.com>
9948 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9951 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9953 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9955 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9958 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9960 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9962 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
9965 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9967 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
9969 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9970 neg_*_op* variables.
9972 2018-07-04 Martin Liska <mliska@suse.cz>
9974 * tree-switch-conversion.c: Define
9975 max_ratio_for_speed and max_ratio_for_size constants.
9977 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9978 Martin Liska <mliska@suse.cz>
9983 * common.opt: Rename align options with 'str_' prefix.
9984 * common/config/i386/i386-common.c (set_malign_value): New
9986 (ix86_handle_option): Use it to set -falign-* options/
9987 * config/aarch64/aarch64-protos.h (struct tune_params): Change
9988 type from int to string.
9989 * config/aarch64/aarch64.c: Update default values from int
9991 * config/alpha/alpha.c (alpha_override_options_after_change):
9993 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9994 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9995 max skip conditionally.
9996 * config/i386/freebsd.h (SUBALIGN_LOG): New.
9997 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9998 max skip conditionally.
9999 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10000 max skip conditionally.
10001 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
10002 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10003 max skip conditionally.
10004 * config/i386/i386.c (struct ptt): Change type from int to
10006 (ix86_default_align): Set default values.
10007 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
10008 max skip conditionally.
10009 * config/i386/iamcu.h (SUBALIGN_LOG): New.
10010 (ASM_OUTPUT_MAX_SKIP_ALIGN):
10011 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
10012 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10013 max skip conditionally.
10014 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
10015 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
10016 * config/i386/x86-64.h (SUBALIGN_LOG): New.
10017 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10018 max skip conditionally.
10019 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10020 * config/ia64/ia64.c (ia64_option_override): Set default values
10021 for alignment options.
10022 * config/m68k/m68k.c: Handle new str_align_* options.
10023 * config/mips/mips.c (mips_set_compression_mode): Change
10025 (mips_option_override): Set default values for options.
10026 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
10028 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10030 * config/rx/rx.c (rx_option_override): Likewise.
10031 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
10032 (LABEL_ALIGN): Use align_labels_log.
10033 (LOOP_ALIGN): Use align_loops_align.
10034 * config/s390/s390.c (s390_asm_output_function_label): Use new
10036 * config/sh/sh.c (sh_override_options_after_change):
10037 Change type of constants.
10038 * config/spu/spu.c (spu_sched_init): Likewise.
10039 * config/sparc/sparc.c (sparc_option_override): Set default
10040 values for options.
10041 * config/visium/visium.c (visium_option_override): Likewise.
10042 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
10043 emit p2align format with last argument if it's not needed.
10044 * doc/invoke.texi: Document extended format of -falign-*.
10045 * final.c: Use align_labels alignment.
10046 * flags.h (struct target_flag_state): Change type to use
10048 (struct align_flags_tuple): New.
10049 (struct align_flags): Likewise.
10050 (align_loops_log): Redefine macro to use new types.
10051 (align_loops_max_skip): Redefine macro to use new types.
10052 (align_jumps_log): Redefine macro to use new types.
10053 (align_jumps_max_skip): Redefine macro to use new types.
10054 (align_labels_log): Redefine macro to use new types.
10055 (align_labels_max_skip): Redefine macro to use new types.
10056 (align_functions_log): Redefine macro to use new types.
10057 (align_loops): Redefine macro to use new types.
10058 (align_jumps): Redefine macro to use new types.
10059 (align_labels): Redefine macro to use new types.
10060 (align_functions): Redefine macro to use new types.
10061 (align_functions_max_skip): Redefine macro to use new types.
10062 (align_loops_value): New macro.
10063 (align_jumps_value): New macro.
10064 (align_labels_value): New macro.
10065 (align_functions_value): New macro.
10066 * function.c (invoke_set_current_function_hook): Propagate
10067 alignment values from flags to global variables default in
10069 * ipa-icf.c (sem_function::equals_wpa): Use
10070 cl_optimization_option_eq instead of memcmp.
10071 * lto-streamer.h (cl_optimization_stream_out): Support streaming
10073 (cl_optimization_stream_in): Likewise.
10074 * optc-save-gen.awk: Support strings in cl_optimization.
10075 * opth-gen.awk: Likewise.
10076 * opts.c (finish_options): Remove error checking of invalid
10078 (MAX_CODE_ALIGN): Remove.
10079 (MAX_CODE_ALIGN_VALUE): Likewise.
10080 (parse_and_check_align_values): New function.
10081 (check_alignment_argument): Likewise.
10082 (common_handle_option): Use check_alignment_argument.
10083 * opts.h (parse_and_check_align_values): Declare.
10084 * toplev.c (init_alignments): Remove.
10085 (read_log_maxskip): New.
10086 (parse_N_M): Likewise.
10087 (parse_alignment_opts): Likewise.
10088 (backend_init_target): Remove usage of init_alignments.
10089 * toplev.h (parse_alignment_opts): Declare.
10090 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
10092 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
10093 * tree.c (cl_option_hasher::equal): New.
10094 * varasm.c: Use new global macros.
10096 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
10098 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
10099 Use a simpler align directive also if MAXSKIP = ALIGN-1.
10100 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10101 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10102 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10103 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10104 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10105 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
10106 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10107 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10108 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10109 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10110 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10112 2018-07-04 Martin Liska <mliska@suse.cz>
10113 Jonathan Wakely <jwakely@redhat.com>
10115 * coverage.c: Use correct type.
10116 * doc/invoke.texi: Language correction.
10118 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
10121 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10122 ENDBRANCH for non-tail call which may return via indirect branch.
10123 * doc/extend.texi: Document indirect_return attribute.
10125 2018-07-03 Martin Sebor <msebor@redhat.com>
10127 PR tree-optimization/86274
10128 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10130 (format_floating): Correct handling of infinities and NaNs.
10132 2018-07-03 Martin Sebor <msebor@redhat.com>
10134 * print-tree.c (print_real_cst): New function.
10135 (print_node_brief): Call it.
10136 (print_node): Ditto.
10138 2018-07-03 Jeff Law <law@redhat.com>
10140 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10141 into a single pattern.
10143 * config/h8300/h8300.md (ors code_iterator): New.
10144 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10145 a single pattern and single splitter.
10146 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10147 (iorqi3_1, xorqi3_1): Likewise.
10148 (iorqi3, xorqi3 expanders): Similarly.
10150 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10151 (movmd_internal) into a single pattern using the P mode iterator.
10152 (movmd splitters): Similarly.
10153 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
10154 (movsd splitters): Similarly.
10156 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
10157 ADDB, ADDW and ADDL into a single ADD attribute which selects the
10158 right table based on the size of the operand.
10159 * config/h8300/h8300.md (length_table): Corresponding changes. All
10160 references to "addb", "addw" and "addl" changed to "add".
10161 (btst patterns): Merge two variants into a single pattern.
10162 (tstqi, tsthi): Likewise.
10163 (addhi3_incdec, addsi3_incdec): Likewise.
10164 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
10165 (mulhi3, mulsi3): Likewise.
10166 (udivhi3, udivsi3): Likewise.
10167 (divhi3, divsi3): Likewise.
10168 (andorqi3, andorhi3, andorsi3): Likewise.
10170 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
10173 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10174 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10176 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10178 PR tree-optimization/85694
10179 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10180 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10182 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10183 (<u>avg<mode>3_ceil): New patterns.
10185 2018-07-03 David Malcolm <dmalcolm@redhat.com>
10187 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10188 scan-tree-dump directive.
10189 * gcc.dg/vect/slp-perm-2.c: Likewise.
10190 * gcc.dg/vect/slp-perm-3.c: Likewise.
10191 * gcc.dg/vect/slp-perm-5.c: Likewise.
10192 * gcc.dg/vect/slp-perm-6.c: Likewise.
10193 * gcc.dg/vect/slp-perm-7.c: Likewise.
10194 * gcc.dg/vect/slp-perm-8.c: Likewise.
10196 2018-07-03 Marek Polacek <polacek@redhat.com>
10198 PR middle-end/86202
10199 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10201 2018-07-03 Richard Biener <rguenther@suse.de>
10204 * tree-ssa-structalias.c (find_func_clobbers): Properly
10205 handle indirect calls.
10207 2018-07-03 Jeff Law <law@redhat.com>
10209 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10210 (shifts): New code iterator.
10211 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
10212 expander. Fix HImode handling on H8/SX.
10213 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
10214 (subqi3, subhi3, subsi3 expanders): Likewise.
10215 (andqi3, andhi3, andsi3 expanders): Likewise.
10216 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
10217 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
10218 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
10219 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
10220 (zero_extendqihi2, zero_extendqisi2): Likewise.
10221 (extendqihi2, extendqisi2): Likewise.
10222 (rotlqi3, rotlhi3, rotlsi3): Likewise.
10223 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
10224 (rotlqi3_1, rotlhi3_1): Likewise.
10225 (logicalhi3_sn, logicalsi3_sn): Likewise.
10226 (logicalhi3, logicalsi3): Likewise.
10228 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10230 * tree-vect-patterns.c (vect_recog_rotate_pattern)
10231 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10232 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
10233 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
10234 type to append_pattern_def_seq instead of creating a stmt_vec_info
10236 (build_mask_conversion): Likewise. Remove vinfo argument.
10237 (vect_add_conversion_to_patterm): Likewise, renaming to...
10238 (vect_add_conversion_to_pattern): ...this.
10239 (vect_recog_mask_conversion_pattern): Update call to
10240 build_mask_conversion. Pass the vector type to
10241 append_pattern_def_seq here too.
10242 (vect_recog_gather_scatter_pattern): Update call to
10243 vect_add_conversion_to_pattern.
10245 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10247 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
10248 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10249 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
10250 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
10251 STMT_VINFO_PATTERN_DEF_SEQ to null here.
10252 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
10253 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10254 append_pattern_def_seq instead of new_pattern_def_seq.
10255 (vect_recog_divmod_pattern): Do both of the above.
10256 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
10259 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10261 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10262 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
10263 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
10264 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
10265 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
10266 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10267 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
10268 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10269 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10270 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
10271 parameter with a single stmt_vec_info.
10272 (vect_recog_func_ptr): Likewise.
10273 (vect_recog_gather_scatter_pattern): Likewise, folding in...
10274 (vect_try_gather_scatter_pattern): ...this.
10275 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
10276 the stmt_vec_info of the statement to be matched. Don't clear
10277 STMT_VINFO_RELATED_STMT.
10278 (vect_pattern_recog): Update call accordingly.
10280 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10282 PR tree-optimization/85694
10283 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
10284 (uavgM3_ceil): Document new optabs.
10285 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
10286 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
10288 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
10289 (savg_ceil_optab): New optabs.
10290 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
10291 (vect_vect_recog_func_ptrs): Add it.
10292 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
10293 constant directly from the associated lhs.
10295 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10297 * tree-vect-patterns.c (vect_split_statement): New function.
10298 (vect_convert_input): Use it to try to split an existing cast.
10300 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10302 * poly-int.h (print_hex): New function.
10303 * dumpfile.h (dump_dec, dump_hex): Declare.
10304 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
10305 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
10306 min_input_precision, operation_precision and operation_sign.
10307 * tree-vect-patterns.c (vect_get_range_info): New function.
10308 (vect_same_loop_or_bb_p, vect_single_imm_use)
10309 (vect_operation_fits_smaller_type): Delete.
10310 (vect_look_through_possible_promotion): Add an optional
10311 single_use_p parameter.
10312 (vect_recog_over_widening_pattern): Rewrite to use new
10313 stmt_vec_info infomration. Handle one operation at a time.
10314 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
10315 (vect_truncatable_operation_p, vect_set_operation_type)
10316 (vect_set_min_input_precision): New functions.
10317 (vect_determine_min_output_precision_1): Likewise.
10318 (vect_determine_min_output_precision): Likewise.
10319 (vect_determine_precisions_from_range): Likewise.
10320 (vect_determine_precisions_from_users): Likewise.
10321 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
10322 (vect_vect_recog_func_ptrs): Put over_widening first.
10324 (vect_pattern_recog): Call vect_determine_precisions.
10326 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10328 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
10329 statements that have been replaced by further pattern statements.
10330 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
10332 2018-07-03 Richard Biener <rguenther@suse.de>
10334 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10335 always set *dt. Dump vectype in vectype overload.
10336 * dumpfile.h (dump_gimple_expr): New function.
10337 (dump_gimple_expr_loc): Likewise.
10338 * dumpfile.c (dump_gimple_expr): New function.
10339 (dump_gimple_expr_loc): Likewise.
10341 2018-07-02 Jeff Law <law@redhat.com>
10343 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
10344 the H8/300, H8/300H and H8/S variants into a single pattern.
10345 (movhi_h8300, movqi_h8300hs): Similarly.
10346 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
10347 (QHI mode iterator): New.
10349 * config/h8300/h8300.md: Remove trailing whitespace.
10351 2018-07-02 Jim Wilson <jimw@sifive.com>
10353 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
10354 instead of emit_insn for interrupt returns.
10355 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
10356 (riscv_sret, riscv_uret): Likewise.
10358 2018-07-02 David Malcolm <dmalcolm@redhat.com>
10360 * pretty-print.c (selftest::test_pp_format): Move save and restore
10361 of quotes to class auto_fix_quotes, and add an instance.
10362 * selftest.c: Include "intl.h".
10363 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
10364 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
10365 * selftest.h (selftest::auto_fix_quotes): New class.
10367 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
10369 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
10370 (aarch64_sve_prepare_conditional_op): Remove.
10371 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
10372 Allow aarch64_simd_reg_or_zero as select operand; remove
10373 the aarch64_sve_prepare_conditional_op call.
10374 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
10375 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10376 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
10377 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
10378 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
10379 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
10380 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
10381 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
10382 and a splitters to match all of the *_any patterns.
10383 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
10385 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
10386 (SVE_COND_FP_BINARY_REV): Remove.
10387 (sve_int_op_rev, sve_fp_op_rev): New.
10388 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
10389 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
10390 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
10391 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
10392 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
10393 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
10394 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
10395 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
10396 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
10398 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
10399 Remove match_dup 1 from the inner unspec.
10400 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10402 * config/aarch64/aarch64.md (movprfx): New attr.
10403 (length): Default movprfx to 8.
10404 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
10405 (*madd<SVE_I>, *msub<SVE_I): Likewise.
10406 (*<su>mul<SVE_I>3_highpart): Likewise.
10407 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
10408 (*v<ASHIFT><SVE_I>3): Likewise.
10409 (*<su><MAXMIN><SVE_I>3): Likewise.
10410 (*<su><MAXMIN><SVE_F>3): Likewise.
10411 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
10412 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
10413 (*div<SVE_F>4): Likewise.
10415 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
10417 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
10420 2018-07-02 Richard Biener <rguenther@suse.de>
10422 PR tree-optimization/86363
10423 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
10424 memset argument refers to a non-variable address.
10426 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
10428 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
10429 {PLUS,MINUS}_EXPR code to...
10430 (adjust_symbolic_bound): ...here,
10431 (combine_bound): ...here,
10432 (set_value_range_with_overflow): ...and here.
10434 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
10436 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
10438 (extract_range_from_abs_expr): ...here.
10440 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
10442 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
10443 -fno-omit-frame-pointer when not optimizing.
10445 2018-07-02 Martin Liska <mliska@suse.cz>
10448 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
10449 (propagate_nothrow): Likewise.
10451 2018-07-02 Martin Liska <mliska@suse.cz>
10454 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
10456 2018-07-02 David Malcolm <dmalcolm@redhat.com>
10458 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
10459 function in r262149, changing "loc" param from source_location to
10460 const dump_location_t &.
10461 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
10462 declaration, as above.
10464 2018-07-01 Paul Koning <ni1d@arrl.net>
10466 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10467 -munit-asm, -mgnu-asm, -mdec-asm.
10468 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
10469 (pdp11_output_labelref): New.
10470 (pdp11_output_def): New.
10471 (pdp11_output_addr_vec_elt): New.
10472 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
10473 %# and %@ format codes.
10474 (pdp11_option_override): New.
10475 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
10476 (pdp11_output_ident): New.
10477 (pdp11_asm_named_section): New.
10478 (pdp11_asm_init_sections): New.
10479 (pdp11_file_start): New.
10480 (pdp11_file_end): New.
10481 (output_ascii): Use .ascii/.asciz for -mdec-asm.
10482 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
10483 %o, like %c but octal.
10484 (pdp11_option_override): New.
10485 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
10487 (DATA_SECTION_ASM_OP): Ditto.
10488 (READONLY_DATA_SECTION_ASM_OP): New.
10489 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
10490 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
10491 (ASM_OUTPUT_LABELREF): Ditto.
10492 (ASM_OUTPUT_DEF): Ditto.
10493 (ASM_OUTPUT_EXTERNAL): New.
10494 (ASM_OUTPUT_SOURCE_FILENAME): New.
10495 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
10496 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
10497 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
10498 %# and %@ format codes.
10499 * config/pdp11/pdp11.opt (mgnu-asm): New.
10500 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
10501 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
10502 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
10504 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
10506 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
10507 dereferencing path[] beyond its length.
10509 (debug_all_paths): New.
10510 (rewire_first_differing_edge): New.
10511 (adjust_paths_after_duplication): New.
10512 (duplicate_thread_path): Call adjust_paths_after_duplication.
10514 (thread_through_all_blocks): Add new argument to
10515 duplicate_thread_path.
10517 2018-06-30 Jim Wilson <jimw@sifive.com>
10519 * config/riscv/predicates.md (p2m1_shift_operand): New.
10520 (high_mask_shift_operand): New.
10521 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
10522 pattern using p2m1_shift_operand.
10523 (lshsi3_zero_extend_3+2): New combiner pattern using
10524 high_mask_shift_operand.
10526 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10528 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
10530 (vect_recog_rotate_pattern): ...here.
10531 (vect_convert_input): Try to insert casts of invariants in the
10533 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
10534 preheader to be empty.
10536 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10538 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
10539 vector type. If given, install it in the new statement's
10540 STMT_VINFO_VECTYPE.
10541 (vect_element_precision): New function.
10542 (vect_unpromoted_value): New struct.
10543 (vect_unpromoted_value::vect_unpromoted_value): New function.
10544 (vect_unpromoted_value::set_op): Likewise.
10545 (vect_look_through_possible_promotion): Likewise.
10546 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
10547 (vect_widened_op_tree, vect_convert_input): Likewise.
10548 (vect_convert_inputs, vect_convert_output): Likewise.
10549 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
10550 to handle the optional cast of the multiplication result and
10551 vect_widened_op_tree to detect the widened multiplication itself.
10552 Do not require the input and output of promotion casts to have
10553 the same sign, but base the signedness of the operation on the
10554 input rather than the result. If the pattern includes two
10555 promotions, check that those promotions have the same sign.
10556 Do not restrict the MULT_EXPR handling to a double-width result;
10557 handle quadruple-width results and wider. Use vect_convert_inputs
10558 to convert the inputs to the common type.
10559 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
10560 to handle the optional cast of the ABS result. Also allow a sign
10561 change or a sign extension between the ABS and MINUS.
10562 Use vect_widened_op_tree to detect the widened subtraction and use
10563 vect_convert_inputs to convert the inputs to the common type.
10564 (vect_handle_widen_op_by_const): Delete.
10565 (vect_recog_widen_op_pattern): New function.
10566 (vect_recog_widen_mult_pattern): Use it.
10567 (vect_recog_widen_shift_pattern): Likewise.
10568 (vect_recog_widen_sum_pattern): Use
10569 vect_look_through_possible_promotion to handle the promoted
10572 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10574 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
10575 the containing gimple_seq *.
10576 * gimple-iterator.h (gsi_for_stmt): Declare it.
10577 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
10578 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
10579 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10580 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10581 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
10583 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
10585 (vect_mark_pattern_stmts): ...here. Handle cases in which the
10586 statement being replaced is part of an existing pattern
10587 definition sequence, inserting the new pattern statements before
10589 (vect_pattern_recog_1): Don't return a bool. If the statement
10590 is already part of a pattern, instead apply pattern matching
10591 to the pattern definition statements. Don't clear the
10592 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
10593 (vect_pattern_recog): Don't break after the first match;
10594 continue processing the pattern definition statements instead.
10595 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
10597 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10599 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10600 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10601 (vect_recog_widen_sum_pattern): Use it.
10603 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10605 * tree-vect-loop.c (vectorizable_reduction): Assert that the
10606 phi is not a pattern statement and has not been replaced by
10607 a pattern statement.
10608 * tree-vect-patterns.c (type_conversion_p): Don't check
10609 STMT_VINFO_IN_PATTERN_P.
10610 (vect_recog_vector_vector_shift_pattern): Likewise.
10611 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
10612 the pattern statement rather than the original statement; check
10613 directly for a WIDEN_MULT_EXPR here.
10614 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
10615 vect_is_simple_use to return the pattern statement rather
10616 than the original statement; use is_pattern_stmt_p to check
10617 for such a pattern statement.
10618 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
10619 to return the pattern statement rather than the original statement;
10620 don't do the same transformation here.
10621 (vect_is_simple_use): If the defining statement has been replaced
10622 by a pattern statement, return the pattern statement instead.
10623 Remove the corresponding (local) transformation from the vectype
10626 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10628 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
10629 end and default to null.
10630 * tree-vect-loop.c (vect_create_epilog_for_reduction)
10631 (vectorizable_reduction): Update calls accordingly, dropping the
10632 gimple ** argument if the passed-back statement isn't needed.
10633 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
10634 (vect_recog_rotate_pattern): Likewise.
10635 (vect_recog_mask_conversion_pattern): Likewise.
10636 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10637 (vect_mask_constant_operand_p): Likewise.
10638 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
10639 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
10640 (get_group_load_store_type, get_load_store_type): Likewise.
10641 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
10642 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
10643 (vectorizable_conversion, vectorizable_assignment): Likewise.
10644 (vectorizable_shift, vectorizable_operation): Likewise.
10645 (vectorizable_store, vect_is_simple_cond): Likewise.
10646 (vectorizable_condition, vectorizable_comparison): Likewise.
10647 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
10648 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
10649 and move it to the end. Cope with null def_stmt_outs.
10651 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
10653 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
10655 2018-06-29 Jeff Law <law@redhat.com>
10657 * config/v850/v850.c (v850_legitimate_address_p): Handle large
10658 displacements for TARGET_V850E2V3 and newer.
10659 (TARGET_LRA_P): Remove. Defaults to LRA now.
10660 * config/v850/v850.md (sign23byte_load): Remove.
10661 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
10662 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
10664 2018-06-29 Martin Liska <mliska@suse.cz>
10667 * coverage.c (coverage_init): Mangle full path name.
10668 * doc/invoke.texi: Document the change.
10669 * gcov-io.c (mangle_path): New.
10670 * gcov-io.h (mangle_path): Likewise.
10671 * gcov.c (mangle_name): Use mangle_path for path mangling.
10673 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10675 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
10676 if starting source register is not even.
10678 2018-06-29 Martin Liska <mliska@suse.cz>
10680 PR tree-optimization/86263
10681 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
10682 Make edge redirection.
10684 2018-06-29 David Malcolm <dmalcolm@redhat.com>
10686 * dumpfile.c (dump_loc): Add indentation based on scope depth.
10687 (dump_scope_depth): New variable.
10688 (get_dump_scope_depth): New function.
10689 (dump_begin_scope): New function.
10690 (dump_end_scope): New function.
10691 * dumpfile.h (get_dump_scope_depth): New declaration.
10692 (dump_begin_scope): New declaration.
10693 (dump_end_scope): New declaration.
10694 (class auto_dump_scope): New class.
10695 (AUTO_DUMP_SCOPE): New macro.
10696 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
10699 2018-06-29 Richard Biener <rguenther@suse.de>
10701 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
10702 compute_all_dependences succeeds.
10703 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
10704 exceed --param loop-max-datarefs-for-datadeps.
10706 2018-06-29 Jakub Jelinek <jakub@redhat.com>
10708 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
10710 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
10713 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
10714 alternative 0 in preferred_for_speed attribute.
10716 2018-06-28 Paul Koning <ni1d@arrl.net>
10718 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
10719 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
10720 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
10721 * config/pdp11/pdp11.md: Correct "length" attribute calculation
10722 for shift insn patterns.
10724 2018-06-28 David Malcolm <dmalcolm@redhat.com>
10726 * cgraph.c (cgraph_node::get_body): Replace assignments to
10727 "dump_file" with calls to set_dump_file.
10728 * dumpfile.c (alt_dump_file): Make static, and group with...
10729 (alt_flags): ...this definition.
10730 (dumps_are_enabled): New variable.
10731 (refresh_dumps_are_enabled): New function.
10732 (set_dump_file): New function.
10733 (set_alt_dump_file): New function.
10734 (gcc::dump_manager::dump_start): Replace assignments to
10735 "dump_file" and "alt_dump_file" with calls to set_dump_file and
10737 (gcc::dump_manager::dump_finish): Likewise.
10738 * dumpfile.h (alt_dump_file): Delete decl.
10739 (dumps_are_enabled): New variable decl.
10740 (set_dump_file): New function decl.
10741 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
10743 * tree-nested.c (lower_nested_functions): Replace assignments to
10744 "dump_file" with calls to set_dump_file.
10746 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
10748 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
10749 goto_locus of each outgoing edge of each basic block.
10751 2018-06-28 Richard Biener <rguenther@suse.de>
10753 * dwarf2out.c (decl_scope_table): Remove.
10754 (push_decl_scope): Likewise.
10755 (pop_decl_scope): Likewise.
10756 (gen_type_die_for_member): Do not call push/pop_decl_scope.
10757 (gen_struct_or_union_type_die): Likewise.
10758 (gen_tagged_type_die): Likewise.
10759 (dwarf2out_init): Do not initialize decl_scope_table.
10760 (dwarf2out_c_finalize): Do not free it.
10762 2018-06-28 Richard Biener <rguenther@suse.de>
10764 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
10765 deciding whether to not re-use a DIE.
10767 2018-06-28 Richard Biener <rguenther@suse.de>
10769 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
10770 DW_AT_abstract_origin attribute.
10772 2018-06-28 Martin Liska <mliska@suse.cz>
10774 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10775 Use newly introduced constants.
10776 * tree-switch-conversion.h (struct jump_table_cluster):
10777 Define max_ratio_for_size and max_ratio_for_speed.
10779 2018-06-28 Martin Liska <mliska@suse.cz>
10781 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10782 Add new checking assert to catch invalid state.
10783 (jump_table_cluster::can_be_handled): Handle single case
10785 (jump_table_cluster::is_beneficial): Bail out for such case.
10786 (bit_test_cluster::find_bit_tests):
10787 Add new checking assert to catch invalid state.
10788 (bit_test_cluster::can_be_handled): Handle single case
10790 (bit_test_cluster::is_beneficial): Bail out for such case.
10791 (switch_decision_tree::analyze_switch_statement):
10794 2018-06-28 Martin Liska <mliska@suse.cz>
10796 * common.opt: Introduce -completion option.
10797 * gcc.c (driver_handle_option): Handle it.
10798 (driver::main): Print completions if completion
10800 * opt-suggestions.c (option_proposer::get_completions):
10802 (option_proposer::suggest_completion): Likewise.
10803 (option_proposer::find_param_completions): Likewise.
10804 (verify_autocompletions): Likewise.
10805 (test_completion_valid_options): Likewise.
10806 (test_completion_valid_params): Likewise.
10807 (in_completion_p): Likewise.
10808 (empty_completion_p): Likewise.
10809 (test_completion_partial_match): Likewise.
10810 (test_completion_garbage): Likewise.
10811 (opt_proposer_c_tests): Likewise.
10812 * opt-suggestions.h: Declare new functions.
10813 * opts.c (common_handle_option): Handle OPT__completion_.
10814 * selftest-run-tests.c (selftest::run_tests): Add
10815 opt_proposer_c_tests.
10816 * selftest.c (assert_str_startswith): New.
10817 * selftest.h (assert_str_startswith): Likewise.
10818 (opt_proposer_c_tests): New.
10819 (ASSERT_STR_STARTSWITH): Likewise.
10821 2018-06-28 Martin Liska <mliska@suse.cz>
10823 * Makefile.in: Add opt-suggestions.o.
10824 * gcc-main.c: Include opt-suggestions.h.
10825 * gcc.c (driver::driver): Likewise.
10826 (driver::~driver): Remove m_option_suggestions.
10827 (driver::build_option_suggestions): Moved to option_proposer.
10828 (driver::suggest_option): Likewise.
10829 (driver::handle_unrecognized_options): Use option_proposer.
10830 * gcc.h (class driver): Add new memver m_option_proposer.
10831 * opt-suggestions.c: New file.
10832 * opt-suggestions.h: New file.
10834 2018-06-28 Martin Liska <mliska@suse.cz>
10836 * vec.h (class auto_string_vec): New (moved from auto_argvec).
10837 (auto_string_vec::~auto_string_vec): Likewise.
10839 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
10841 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
10842 prevent_decl_creation_for_types fields up and add reset_location field.
10843 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
10844 statement if id->reset_location is true.
10845 (copy_edges_for_bb): Do not set goto_locus on the new edges if
10846 id->reset_location is true.
10847 (copy_phis_for_bb): Force input_location on the arguments if
10848 id->reset_location is true.
10849 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10850 is set on the function to be inlined.
10852 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
10854 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
10856 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
10858 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
10859 registers for Pmode.
10860 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
10861 hard registers for the clobbered pseudo.
10863 2018-06-27 Paul Koning <ni1d@arrl.net>
10865 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10866 mutually exclusive options.
10867 * config/pdp11/constraints.md (h): New constraint.
10868 (O): Update definition to match shift code generation.
10869 (D): New constraint.
10870 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
10872 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
10874 (output_jump): Change arguments.
10875 (pdp11_fixed_cc_regs): New function.
10876 (pdp11_cc_mode): Ditto.
10877 (pdp11_expand_shift): Ditto.
10878 (pdp11_assemble_shift): Ditto.
10879 (pdp11_small_shift): Ditto.
10880 (pdp11_branch_cost): Remove.
10881 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
10883 (pdp11_register_move_cost): Update for CC registers.
10884 (pdp11_rtx_costs): Add case for LSHIFTRT.
10885 (pdp11_output_jump): Add CCNZ mode conditional branches.
10886 (notice_update_cc_on_set): Remove.
10887 (pdp11_cc_mode): New function.
10888 (simple_memory_operand): Correct pre/post decrement case.
10889 (no_side_effect_operand): New function.
10890 (pdp11_regno_reg_class): Add CC_REGS class.
10891 (pdp11_fixed_cc_regs): New function.
10892 (pdp11_small_shift): New function.
10893 (pdp11_expand_shift): New function to expand shift insns.
10894 (pdp11_assemble_shift): New function to output shifts.
10895 (pdp11_branch_cost): Remove.
10896 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
10897 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
10898 (WCHAR_TYPE): Ditto.
10899 (PTRDIFF_TYPE): Ditto.
10900 (ADJUST_INSN_LENGTH): New macro.
10901 (FIXED_REGISTERS): Add CC registers.
10902 (CALL_USED_REGISTERS): Ditto.
10903 (reg_class): Ditto.
10904 (REG_CLASS_NAMES): Ditto.
10905 (REG_CLASS_CONTENTS): Ditto.
10906 (SELECT_CC_MODE): Use new function.
10907 (TARGET_FLAGS_REGNUM): New macro.
10908 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
10909 (cc0_reg_rtx): Remove.
10910 (CC_STATUS_MDEP): Remove.
10911 (CC_STATUS_MDEFP_INIT): Remove.
10912 (CC_IN_FPU): Remove.
10913 (NOTICE_UPDATE_CC): Remove.
10914 (REGISTER_NAMES): Add CC registers.
10915 (BRANCH_COST): Change to constant 1.
10916 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10918 * config/pdp11/pdp11.opt (mbcopy): Remove.
10919 (mbcopy-builtin): Remove.
10920 (mbranch-cheap): Remove.
10921 (mbranch-expensive): Remove.
10922 * config/pdp11/predicates.md (expand_shift_operand): Update to
10923 match shift code generation.
10924 (ccnz_operator): New predicate.
10925 * doc/invoke.texi (PDP-11 Options): Remove deleted options
10926 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10927 Remove non-existent option -mabshi, -mno-abshi. Document mutually
10929 * doc/md.texi (PDP-11): Document new D and h constraints. Update
10930 description of O constraint.
10932 2018-06-27 Jeff Law <law@redhat.com>
10933 Austin Law <austinklaw@gmail.com>
10935 * config/v850/v850.md (addsi3_set_flags): New pattern.
10936 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10937 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10938 (zero_extendhisi2_v850_set_flags): Likewise.
10939 (zero_extendqisi2_v850_set_flags): Likewise.
10940 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10941 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10942 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10944 * config/v850/v850-protos.h (notice_update_cc): Remove.
10945 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10946 (v850_print_operand): Handle 'D' and "d".
10947 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10948 Add handling of arithmetic/logical operations compared against zero.
10949 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10950 Do not look at v850_compare_op, instead get mode from last argument.
10951 (v850_gen_compare): Remove
10952 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10953 after reload for prologue insns.
10954 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10956 (construct_save_jarl): Likewise.
10957 (TARGET_FLAGS_REGNUM): Define.
10958 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10959 (NOTICE_UPDATE_CC): Remove.
10960 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10961 than cc0. Conditionalize on reload_completed.
10962 (cmpsi_insn, setfcc_insn): Likewise.
10963 (tst1 splitter): Turn into define_and_split which sets the flags
10965 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10966 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10967 (cstoresf4, cstoredf4): Clobber the flags.
10968 (cmpsi, cmpsf, cmpdf): Remove expanders.
10969 (setf_insn): Remove pattern.
10970 (addsi3): Turn into define_and_split which clobbers the flags after
10971 reload and a suitable pattern (addsi3_clobber_flags) for use after
10973 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10974 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10975 (ashrsi3, ashrsi3_v850e2): Likewise.
10976 (bins): Clobber the flags.
10977 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10978 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10979 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10980 (call_value_internal_short, call_value_internal_long): Likewise.
10981 (callt_save_interrupt, callt_return_interrupt): Likewise.
10982 (save_interrupt, return_interrupt): Likewise.
10983 (callt_save_all_interrupt, save_all_interrupt): Likewise.
10984 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10985 (restore_all_interrupt, _restore_all_interrupt): Likewise.
10986 (All FP comparisons): Only allow after reload has completed.
10988 (divh, divhu): Tweak output template.
10989 (branch_z_normal, branch_z_invert): Remove
10990 (branch_nz_normal, branch_nz_invert): Likewise.
10991 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10993 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10994 * config/v850/v850.c (notice_update_cc): Remove.
10995 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
10996 (CC_NO_CARRY): Likewise.
10997 (NOTICE_UPDATE_CC): Define to nothing.
10998 * config/v850/v850.md: Remove block comment on cc0 handling
10999 Remove "cc" attribute from all patterns. Remove cc_status handling
11000 from all patterns. Minor formatting fixes.
11002 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11004 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
11005 (cortex-a76.cortex-a55): Likewise.
11006 * config/aarch64/aarch64-tune.md: Regenerate.
11007 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
11008 cortex-a76.cortex-a55.
11010 2018-06-27 Jeff Law <law@redhat.com>
11012 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
11013 (MULTILIB_DIRNAMES): Similarly.
11015 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
11017 * gimple.h (gimple_return_retbnd): Delete.
11018 (gimple_return_set_retbnd): Likewise.
11019 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11020 gimple_return_set_retbnd.
11021 * gimple-pretty-print.c (dump_gimple_return): Remove call to
11022 gimple_return_retbnd and adjust.
11023 * tree-inline.h (struct copy_body_data): Remove retbnd field.
11024 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
11025 Explicitly return NULL in a couple more cases. Move assertion
11026 on debug statements and remove unreachable code.
11027 (reset_debug_binding): Do not test id->retbnd.
11028 (expand_call_inline): Do not set it.
11030 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
11032 * configure.ac: Add --disable-gcov option.
11033 * configure: Regenerate.
11034 * Makefile.in: Honour @enable_gcov@.
11035 * doc/install.texi: Document --disable-gcov.
11037 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11039 * config/arm/arm-cpus.in (cortex-a76): New entry.
11040 (cortex-a76.cortex-a55): Likewise.
11041 * config/arm/arm-tables.opt: Regenerate.
11042 * config/arm/arm-tune.md: Likewise.
11043 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
11044 * doc/invoke.texi (ARM Options): Document cortex-a76 and
11045 cortex-a76.cortex-a55.
11047 2018-06-27 Tamar Christina <tamar.christina@arm.com>
11050 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
11052 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
11054 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
11056 (EPILOGUE_USES): Likewise.
11058 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
11060 * tree-inline.c (remap_location): New function extracted from...
11061 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
11062 (copy_phis_for_bb): ...here. Call remap_location.
11063 (copy_cfg_body): Adjust call to copy_edges_for_bb.
11065 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
11067 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
11068 unaligned vsx for 16B memset.
11070 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
11073 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
11074 ieee128_float_type_node to long_double_type_node unless
11075 TARGET_LONG_DOUBLE_128 is set.
11077 2018-06-26 David Malcolm <dmalcolm@redhat.com>
11079 * cfgloop.c (get_loop_location): Convert return type from
11080 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
11081 by implicit construction from rtx_insn *, and using
11082 dump_user_location_t::from_function_decl for the fallback case.
11083 * cfgloop.h (get_loop_location): Convert return type from
11084 location_t to dump_user_location_t.
11085 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
11086 dump_printf_loc to pass in a dump_location_t rather than a
11087 location_t, via the gimple stmt.
11088 * coverage.c (get_coverage_counts): Update calls to
11089 dump_printf_loc to pass in dump_location_t rather than a
11091 * doc/optinfo.texi (Dump types): Convert example of
11092 dump_printf_loc from taking "locus" to taking "insn". Update
11093 description of the "_loc" calls to cover dump_location_t.
11094 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
11096 (dump_user_location_t::dump_user_location_t): New constructors,
11097 from gimple *stmt and rtx_insn *.
11098 (dump_user_location_t::from_function_decl): New function.
11099 (dump_loc): Make static.
11100 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
11101 const dump_location_t &.
11102 (dump_generic_expr_loc): Delete.
11103 (dump_printf_loc): Convert param "loc" from location_t to
11104 const dump_location_t &.
11105 (selftest::test_impl_location): New function.
11106 (selftest::dumpfile_c_tests): New function.
11107 * dumpfile.h: Include "profile-count.h".
11108 (class dump_user_location_t): New class.
11109 (struct dump_impl_location_t): New struct.
11110 (class dump_location_t): New class.
11111 (dump_printf_loc): Convert 2nd param from source_location to
11112 const dump_location_t &.
11113 (dump_generic_expr_loc): Delete.
11114 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11115 const dump_location_t &.
11116 * gimple-fold.c (fold_gimple_assign): Update call to
11117 dump_printf_loc to pass in a dump_location_t rather than a
11118 location_t, via the gimple stmt.
11119 (gimple_fold_call): Likewise.
11120 * gimple-loop-interchange.cc
11121 (loop_cand::analyze_iloop_reduction_var): Update for change to
11122 check_reduction_path.
11123 (tree_loop_interchange::interchange): Update for change to
11124 find_loop_location.
11125 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11126 change in return-type of find_loop_location.
11127 (graphite_regenerate_ast_isl): Likewise.
11128 * graphite-optimize-isl.c (optimize_isl): Likewise.
11129 * graphite.c (graphite_transform_loops): Likewise.
11130 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11131 pass in a dump_location_t rather than a location_t, via the
11133 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11134 * ipa.c (walk_polymorphic_call_targets): Likewise.
11135 * loop-unroll.c (report_unroll): Convert "locus" param from
11136 location_t to dump_location_t.
11137 (decide_unrolling): Update for change to get_loop_location's
11139 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11140 location_t to dump_user_location_t.
11141 (grid_find_single_omp_among_assignments_1): Updates calls to
11142 dump_printf_loc to pass in a dump_location_t rather than a
11143 location_t, via the gimple stmt.
11144 (grid_parallel_clauses_gridifiable): Convert "tloc" from
11145 location_t to dump_location_t. Updates calls to dump_printf_loc
11146 to pass in a dump_location_t rather than a location_t, via the
11148 (grid_inner_loop_gridifiable_p): Likewise.
11149 (grid_dist_follows_simple_pattern): Likewise.
11150 (grid_gfor_follows_tiling_pattern): Likewise.
11151 (grid_target_follows_gridifiable_pattern): Likewise.
11152 (grid_attempt_target_gridification): Convert initialization
11153 of local "grid" from memset to zero-initialization; FIXME: does
11154 this require C++11? Update call to dump_printf_loc to pass in a
11155 optinfo_location rather than a location_t, via the gimple stmt.
11156 * profile.c (read_profile_edge_counts): Updates call to
11157 dump_printf_loc to pass in a dump_location_t rather than a
11159 (compute_branch_probabilities): Likewise.
11160 * selftest-run-tests.c (selftest::run_tests): Call
11162 * selftest.h (dumpfile_c_tests): New decl.
11163 * tree-loop-distribution.c (pass_loop_distribution::execute):
11164 Update for change in return type of find_loop_location.
11165 * tree-parloops.c (parallelize_loops): Likewise.
11166 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11167 "locus" from location_t to dump_user_location_t.
11168 (canonicalize_loop_induction_variables): Likewise.
11169 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11170 for change in return type of find_loop_location.
11171 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11172 to dump_printf_loc to pass in a dump_location_t rather than a
11173 location_t, via the stmt.
11174 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11176 * tree-vect-loop-manip.c (find_loop_location): Convert return
11177 type from source_location to dump_user_location_t.
11178 (vect_do_peeling): Update for above change.
11179 (vect_loop_versioning): Update for change in type of
11181 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11182 from location_t to dump_user_location_t.
11183 (vect_estimate_min_profitable_iters): Update for change in type
11185 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11186 location_t to dump_location_t.
11187 (vect_slp_bb): Update for change in type of vect_location.
11188 * tree-vectorizer.c (vect_location): Convert from source_location
11189 to dump_user_location_t.
11190 (try_vectorize_loop_1): Update for change in vect_location's type.
11191 (vectorize_loops): Likewise.
11192 (increase_alignment): Likewise.
11193 * tree-vectorizer.h (vect_location): Convert from source_location
11194 to dump_user_location_t.
11195 (find_loop_location): Convert return type from source_location to
11196 dump_user_location_t.
11197 (check_reduction_path): Convert 1st param from location_t to
11198 dump_user_location_t.
11199 * value-prof.c (check_counter): Update call to dump_printf_loc to
11200 pass in a dump_user_location_t rather than a location_t; update
11201 call to error_at for change in type of "locus".
11202 (check_ic_target): Update call to dump_printf_loc to
11203 pass in a dump_user_location_t rather than a location_t, via the
11206 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
11208 * config/s390/s390.h (enum processor_flags): Do not use
11209 default tune parameter when -march was specified.
11211 2018-06-26 Jakub Jelinek <jakub@redhat.com>
11214 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
11215 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
11218 2018-06-26 Richard Biener <rguenther@suse.de>
11220 PR tree-optimization/86287
11222 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
11223 (vect_analyze_loop): Initialize n_stmts.
11225 2018-06-26 Richard Biener <rguenther@suse.de>
11227 PR middle-end/86271
11228 * fold-const.c (fold_convertible_p): Pointer extension
11231 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
11234 * dwarf2out.c (loc_list_has_views): Adjust comments.
11235 (dw_loc_list): Split single cross-partition range with
11238 2018-06-25 Jeff Law <law@redhat.com>
11240 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
11241 on -mbig-switch by default.
11243 * config/v850/predicates.md (const_float_1_operand): Fix match_code
11245 (const_float_0_operand): Remove unused predicate.
11246 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
11247 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
11248 (recipsf2): New expander. Original pattern now called
11250 (recipdf2, recipdf2_insn): Similarly.
11251 (rsqrtsf2, rsqrtsf2_insn): Similarly
11252 (rsqrtdf2, rsqrtdf2_insn): Similarly
11254 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
11256 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
11257 Simplify logic for FreeBSD (twice).
11259 2018-06-25 Martin Sebor <msebor@redhat.com>
11261 PR tree-optimization/86204
11262 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
11263 a strnlen result if it's less than the length of the string.
11265 2018-06-25 Martin Sebor <msebor@redhat.com>
11267 PR tree-optimization/85700
11268 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
11269 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
11270 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
11272 2018-06-25 Martin Sebor <msebor@redhat.com>
11274 * doc/extend.texi (Zero-length arrays): Update and clarify.
11276 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
11278 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
11279 added IEEE/IBM long double multilib support on PowerPC little
11280 endian Linux systems.
11281 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11282 (MULTILIB_DEFAULTS): Likewise.
11283 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11285 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
11286 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
11287 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
11289 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
11291 PR middle-end/86311
11292 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
11293 (REORDER_45): Likewise.
11295 2018-06-25 Jeff Law <law@redhat.com>
11297 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
11298 dividend to 32 bits. Adjust length.
11299 (udivmodhi4): Cleanup output template. Fix length.
11301 2018-06-25 Carl Love <cel@us.ibm.com>
11303 * config/rs6000/vsx.md: Change word selector to prefered location.
11305 2018-06-25 Richard Biener <rguenther@suse.de>
11307 PR tree-optimization/86304
11308 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
11309 epilogue-if-converted loops as well.
11311 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
11313 * lto-section-out.c (lto_begin_section): Do not print section
11314 name for noaddr and unnumbered dumps.
11316 2018-06-25 Richard Biener <rguenther@suse.de>
11318 * tree-vectorizer.h (struct vec_info_shared): New structure
11319 with parts split out from struct vec_info and loop_nest from
11320 struct _loop_vec_info.
11321 (struct vec_info): Adjust accordingly.
11322 (struct _loop_vec_info): Likewise.
11323 (LOOP_VINFO_LOOP_NEST): Adjust.
11324 (LOOP_VINFO_DATAREFS): Likewise.
11325 (LOOP_VINFO_DDRS): Likewise.
11326 (struct _bb_vec_info): Likewise.
11327 (BB_VINFO_DATAREFS): Likewise.
11328 (BB_VINFO_DDRS): Likewise.
11329 (struct _stmt_vec_info): Add dr_aux member.
11330 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
11331 (DR_MISALIGNMENT_UNINITIALIZED): New.
11332 (set_dr_misalignment): Adjust.
11333 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
11334 (vect_analyze_loop): Adjust prototype.
11335 (vect_analyze_loop_form): Likewise.
11336 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11337 Compute dependences lazily.
11338 (vect_record_base_alignments): Use shared datarefs/ddrs.
11339 (vect_verify_datarefs_alignment): Likewise.
11340 (vect_analyze_data_refs_alignment): Likewise.
11341 (vect_analyze_data_ref_accesses): Likewise.
11342 (vect_analyze_data_refs): Likewise.
11343 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
11344 constructor parameter for shared part.
11345 (vect_analyze_loop_form): Pass in shared part and adjust.
11346 (vect_analyze_loop_2): Pass in storage for the number of
11347 stmts. Move loop nest finding to the caller. Compute
11349 (vect_analyze_loop): Pass in shared part.
11350 (vect_transform_loop): Verify shared datarefs are unchanged.
11351 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
11352 constructor parameter for shared part.
11353 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
11354 (vect_slp_bb): Verify shared datarefs are unchanged before
11356 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
11358 (new_stmt_vec_info): Initialize DR_AUX misalignment to
11359 DR_MISALIGNMENT_UNINITIALIZED.
11360 * tree-vectorizer.c (vec_info::vec_info): Add constructor
11361 parameter for shared part.
11362 (vec_info::~vec_info): Adjust.
11363 (vec_info_shared::vec_info_shared): New.
11364 (vec_info_shared::~vec_info_shared): Likewise.
11365 (vec_info_shared::save_datarefs): Likewise.
11366 (vec_info_shared::check_datarefs): Likewise.
11367 (try_vectorize_loop_1): Construct shared part live for analyses
11368 of a single loop for multiple vector sizes.
11369 * tree-parloops.c (gather_scalar_reductions): Adjust.
11371 2018-06-25 Richard Biener <rguenther@suse.de>
11373 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
11374 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
11375 (vect_analyze_data_refs): Remove similar code from here and
11376 simplify accordingly.
11378 2018-06-25 Richard Biener <rguenther@suse.de>
11380 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
11381 for reverse storage order accesses rather than asserting
11382 they cannot happen here.
11384 2018-06-25 Tom de Vries <tdevries@suse.de>
11387 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
11388 Use data16 instead of .byte for insn prefix.
11390 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
11393 * parser.c (make_char_string_pack): Pass this literal chars
11394 through cpp_interpret_string.
11395 (cp_parser_userdef_numeric_literal): Check the result of
11396 make_char_string_pack.
11398 2018-06-24 Maya Rashish <coypu@sdf.org>
11400 * ginclude/stddef.h: Simplify conditions around avoiding
11401 re-definition of __size_t.
11403 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
11405 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
11406 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
11408 2018-06-22 Maya Rashish <coypu@sdf.org>
11410 * doc/invoke.texi (mno-fancy-math-387): Update for changes
11411 made to OpenBSD and NetBSD through the years.
11413 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11415 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11416 behavior of vec_pack (vector double, vector double) to match
11417 behavior of vec_float2 (vector double, vector double).
11419 2018-06-22 Olivier Hainque <hainque@adacore.com>
11421 * gimplify.c (gimplify_function_tree): Prevent creation
11422 of a trampoline for the address of the current function
11423 passed to entry/exit instrumentation hooks.
11425 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
11428 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
11431 2018-06-22 Martin Liska <mliska@suse.cz>
11433 PR tree-optimization/86263
11434 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11435 Bail out if is_enabled is false.
11436 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
11438 (jump_table_cluster::is_enabled): New function.
11440 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
11442 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11443 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
11444 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
11445 (lto_input_ts_binfo_tree_pointers): Likewise.
11446 * tree-streamer-out.c (streamer_write_tree_bitfields,
11447 write_ts_binfo_tree_pointers): Likewise.
11448 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
11450 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
11452 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
11454 2018-06-22 Martin Liska <mliska@suse.cz>
11456 * symbol-summary.h (get): Make it pure and inline move
11457 functionality from ::get function.
11458 (get): Remove and inline into ::get and ::get_create.
11459 (get_create): Move code from ::get function.
11461 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11464 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
11465 -x assembler-with-cpp.
11467 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11469 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
11470 _FILE_OFFSET_BITS=64 for C++.
11472 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
11474 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
11475 conversion insn that shows up when pr85657-3.c is compiled using
11476 IEEE 128-bit long double.
11477 (neg<mode>2_internal): Use the correct mode to check whether the
11478 mode is IBM extended.
11479 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
11480 multiply and divide external functions from being created more
11483 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11485 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
11487 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
11488 the edge can be forwarded.
11489 (cfg_layout_merge_blocks): Likewise.
11491 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11493 * except.c (finish_eh_generation): Commit edge insertions only after
11494 the EH edges have been redirected from post-landing to landing pads.
11496 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11498 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
11499 create_tmp_var_for to create the FRAME decl.
11500 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
11502 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11504 * tree-inline.c (copy_edges_for_bb): Minor tweak.
11505 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
11506 debug statement when resetting its value.
11507 (expand_call_inline): Copy the locus of the call onto the assignment
11508 of the return value, if any. Use local variable in more cases.
11510 2018-06-21 Martin Liska <mliska@suse.cz>
11512 * ipa-pure-const.c (propagate_nothrow): Use
11513 funct_state_summaries->get.
11514 (dump_malloc_lattice): Likewise.
11515 (propagate_malloc): Likewise.
11517 2018-06-21 Richard Biener <rguenther@suse.de>
11519 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
11520 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
11521 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
11523 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
11524 BLOCK_ABSTRACT_ORIGIN unconditionally.
11526 2018-06-21 David Malcolm <dmalcolm@redhat.com>
11528 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
11530 * ipa-reference.c (ipa_reference_c_finalize): Delete
11531 ipa_ref_opt_sum_summaries and set it to NULL.
11533 2018-06-21 Tom de Vries <tdevries@suse.de>
11535 PR tree-optimization/85859
11536 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
11537 test with comment from bb_no_side_effects_p.
11539 2018-06-21 Richard Biener <rguenther@suse.de>
11541 PR tree-optimization/86232
11542 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
11543 max for constant niter.
11545 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
11547 * config/aarch64/aarch64-simd.md
11548 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
11550 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
11552 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
11553 Make opernads of the unspec commutative.
11555 2018-06-21 Richard Biener <rguenther@suse.de>
11557 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
11558 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
11559 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
11560 (vect_analyze_data_ref_dependence): Re-order checks to deal with
11562 (vect_record_base_alignments): Do not record base alignment
11563 for gathers or scatters.
11564 (vect_compute_data_ref_alignment): Drop return value that is always
11565 true. Bail out early for gathers or scatters.
11566 (vect_enhance_data_refs_alignment): Bail out early for gathers
11568 (vect_find_same_alignment_drs): Likewise.
11569 (vect_analyze_data_refs_alignment): Remove dead code.
11570 (vect_slp_analyze_and_verify_node_alignment): Likewise.
11571 (vect_analyze_data_refs): For possible gathers or scatters do
11572 not create an alternate DR, just check their possible validity
11573 and mark them. Adjust DECL_NONALIASED handling to not rely
11574 on DR_BASE_ADDRESS.
11575 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
11576 update inits of gathers or scatters.
11577 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
11578 Also copy gather/scatter flag to pattern vinfo.
11580 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
11582 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11583 behavior of vec_packsu (vector unsigned long long, vector unsigned
11584 long long) to match behavior of vec_packs with same signature.
11586 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
11587 Thomas Schwinge <thomas@codesourcery.com>
11588 Cesar Philippidis <cesar@codesourcery.com>
11590 * gimplify.c (gimplify_scan_omp_clauses): Add support for
11591 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11592 (gimplify_adjust_omp_clauses): Likewise.
11593 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
11594 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
11595 (gimplify_omp_target_update): Update handling of acc update and
11597 * omp-low.c (install_var_field): Remove unused parameter
11598 base_pointers_restrict.
11599 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
11600 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
11602 (omp_target_base_pointers_restrict_p): Delete.
11603 (scan_omp_target): Update call to scan_sharing_clauses.
11604 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
11606 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
11607 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11608 (convert_local_omp_clauses): Likewise.
11609 * tree-pretty-print.c (dump_omp_clause): Likewise.
11610 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
11612 (omp_clause_code_name): Likewise.
11614 2018-06-20 Jakub Jelinek <jakub@redhat.com>
11617 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
11620 PR tree-optimization/86231
11621 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
11622 anti-range don't overwrite *vr0min before using it to compute *vr0max.
11624 2018-06-20 Tom de Vries <tdevries@suse.de>
11626 PR tree-optimization/86097
11627 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
11628 iv type if signedness of iv type is not the same as that of *nit.
11630 2018-06-20 Jakub Jelinek <jakub@redhat.com>
11632 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
11633 EDGE_EH edges, verify they are all EDGE_EH.
11635 2018-06-20 Maya Rashish <coypu@sdf.org>
11637 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
11639 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11641 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
11642 * config/aarch64/aarch64.c (xgene1_tunings): Add
11643 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
11644 (aarch64_mode_valid_for_sched_fusion_p):
11645 Allow 16-byte modes.
11646 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
11647 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
11649 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
11651 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
11652 * config/aarch64/iterators.md (VQ2): New mode iterator.
11654 2018-06-20 Martin Liska <mliska@suse.cz>
11656 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11657 Change default ratio from 10 to 8.
11659 2018-06-20 Martin Liska <mliska@suse.cz>
11661 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11663 (bit_test_cluster::find_bit_tests): Likewise.
11664 (switch_decision_tree::analyze_switch_statement): Find clusters.
11665 * tree-switch-conversion.h (struct jump_table_cluster): Document
11668 2018-06-20 Martin Liska <mliska@suse.cz>
11670 * tree-switch-conversion.c (switch_conversion::collect):
11671 Record m_uniq property.
11672 (switch_conversion::expand): Bail out for special conditions.
11673 (group_cluster::~group_cluster): New.
11674 (group_cluster::group_cluster): Likewise.
11675 (group_cluster::dump): Likewise.
11676 (jump_table_cluster::emit): New.
11677 (switch_decision_tree::fix_phi_operands_for_edges): New.
11678 (struct case_node): Remove struct.
11679 (jump_table_cluster::can_be_handled): New.
11680 (case_values_threshold): Moved to header.
11681 (reset_out_edges_aux): Likewise.
11682 (jump_table_cluster::is_beneficial): New.
11683 (bit_test_cluster::can_be_handled): Likewise.
11684 (add_case_node): Remove.
11685 (bit_test_cluster::is_beneficial): New.
11686 (case_bit_test::cmp): New.
11687 (bit_test_cluster::emit): New.
11688 (expand_switch_as_decision_tree_p): Remove.
11689 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
11690 (fix_phi_operands_for_edge): Likewise.
11691 (switch_decision_tree::analyze_switch_statement): New.
11692 (compute_cases_per_edge): Move ...
11693 (switch_decision_tree::compute_cases_per_edge): ... here.
11694 (try_switch_expansion): Likewise.
11695 (switch_decision_tree::try_switch_expansion): Likewise.
11696 (record_phi_operand_mapping): Likewise.
11697 (switch_decision_tree::record_phi_operand_mapping): Likewise.
11698 (emit_case_decision_tree): Likewise.
11699 (switch_decision_tree::emit): Likewise.
11700 (balance_case_nodes): Likewise.
11701 (switch_decision_tree::balance_case_nodes): Likewise.
11702 (dump_case_nodes): Likewise.
11703 (switch_decision_tree::dump_case_nodes): Likewise.
11704 (emit_jump): Likewise.
11705 (switch_decision_tree::emit_jump): Likewise.
11706 (emit_cmp_and_jump_insns): Likewise.
11707 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
11708 (emit_case_nodes): Likewise.
11709 (switch_decision_tree::emit_case_nodes): Likewise.
11710 (conditional_probability): Remove.
11711 * tree-switch-conversion.h (enum cluster_type): New.
11713 (struct cluster): Likewise.
11714 (cluster::cluster): Likewise.
11715 (struct simple_cluster): Likewise.
11716 (simple_cluster::simple_cluster): Likewise.
11717 (struct group_cluster): Likewise.
11718 (struct jump_table_cluster): Likewise.
11719 (struct bit_test_cluster): Likewise.
11720 (struct min_cluster_item): Likewise.
11721 (struct case_tree_node): Likewise.
11722 (case_tree_node::case_tree_node): Likewise.
11723 (jump_table_cluster::case_values_threshold): Likewise.
11724 (struct case_bit_test): Likewise.
11725 (struct switch_decision_tree): Likewise.
11726 (struct switch_conversion): Likewise.
11727 (switch_decision_tree::reset_out_edges_aux): Likewise.
11729 2018-06-20 Martin Liska <mliska@suse.cz>
11731 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
11732 (hoist_edge_and_branch_if_true): Likewise.
11733 (expand_switch_using_bit_tests_p): Likewise.
11734 (struct case_bit_test): Likewise.
11735 (case_bit_test_cmp): Likewise.
11736 (emit_case_bit_tests): Likewise.
11737 (switch_conversion::switch_conversion): New class.
11738 (struct switch_conv_info): Remove old struct.
11739 (collect_switch_conv_info): More to ...
11740 (switch_conversion::collect): ... this.
11741 (check_range): Likewise.
11742 (switch_conversion::check_range): Likewise.
11743 (check_all_empty_except_final): Likewise.
11744 (switch_conversion::check_all_empty_except_final): Likewise.
11745 (check_final_bb): Likewise.
11746 (switch_conversion::check_final_bb): Likewise.
11747 (create_temp_arrays): Likewise.
11748 (switch_conversion::create_temp_arrays): Likewise.
11749 (free_temp_arrays): Likewise.
11750 (gather_default_values): Likewise.
11751 (switch_conversion::gather_default_values): Likewise.
11752 (build_constructors): Likewise.
11753 (switch_conversion::build_constructors): Likewise.
11754 (constructor_contains_same_values_p): Likewise.
11755 (switch_conversion::contains_same_values_p): Likewise.
11756 (array_value_type): Likewise.
11757 (switch_conversion::array_value_type): Likewise.
11758 (build_one_array): Likewise.
11759 (switch_conversion::build_one_array): Likewise.
11760 (build_arrays): Likewise.
11761 (switch_conversion::build_arrays): Likewise.
11762 (gen_def_assigns): Likewise.
11763 (switch_conversion::gen_def_assigns): Likewise.
11764 (prune_bbs): Likewise.
11765 (switch_conversion::prune_bbs): Likewise.
11766 (fix_phi_nodes): Likewise.
11767 (switch_conversion::fix_phi_nodes): Likewise.
11768 (gen_inbound_check): Likewise.
11769 (switch_conversion::gen_inbound_check): Likewise.
11770 (process_switch): Use the newly created class.
11771 (switch_conversion::expand): New.
11772 (switch_conversion::~switch_conversion): New.
11773 * tree-switch-conversion.h: New file.
11775 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11777 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
11778 tree-vect-patterns.c.
11779 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
11780 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
11781 (vect_recog_sad_pattern): Likewise.
11782 (vect_recog_widen_sum_pattern): Likewise.
11783 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
11784 (vect_recog_widen_shift_pattern): Remove the type_in argument.
11785 (vect_recog_rotate_pattern): Likewise.
11786 (vect_recog_mult_pattern): Likewise.
11787 (vect_recog_vector_vector_shift_pattern): Likewise.
11788 (vect_recog_divmod_pattern): Likewise.
11789 (vect_recog_mixed_size_cond_pattern): Likewise.
11790 (vect_recog_bool_pattern): Likewise.
11791 (vect_recog_mask_conversion_pattern): Likewise.
11792 (vect_try_gather_scatter_pattern): Likewise.
11793 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
11794 (vect_recog_over_widening_pattern): Likewise.
11795 (vect_recog_gather_scatter_pattern): Likewise.
11796 (vect_recog_func_ptr): Move from tree-vectorizer.h
11797 (vect_vect_recog_func_ptrs): Move further down the file.
11798 (vect_recog_func): Likewise. Remove the third argument.
11799 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
11800 (vect_pattern_recog_1): Expect the pattern function to do any
11801 necessary target tests. Also expect it to provide a vector type.
11802 Remove the type_in handling.
11804 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11806 * tree-vect-patterns.c (vect_pattern_detected): New function.
11807 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
11808 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
11809 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
11810 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
11811 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11812 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11813 (vect_recog_mask_conversion_pattern)
11814 (vect_try_gather_scatter_pattern): Likewise.
11816 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11818 * tree-vect-patterns.c (vect_get_internal_def): New function.
11819 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11820 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
11821 (search_type_for_mask_1): Use it.
11823 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11825 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11826 redundant WIDEN_SUM_EXPR handling.
11827 (vect_recog_sad_pattern): Likewise.
11829 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11831 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11832 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
11833 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
11834 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
11835 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
11837 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11839 * tree-vect-stmts.c (vectorizable_call): Make sure that we
11840 use the stmt_vec_info of the original bb statement for the
11841 new zero assignment, even if the call is part of a pattern.
11843 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11845 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
11846 that the sequence is attached to the original statement rather
11847 than the pattern statement.
11848 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
11849 PATTERN_DEF_SEQ from the original statement rather than
11850 the main pattern statement.
11851 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
11852 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
11853 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
11855 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11857 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
11858 definition statements before the early exit for statements that aren't
11860 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
11862 (vect_transform_loop): ...here. Process pattern definition
11863 statements without first checking whether the main pattern
11864 statement is live or relevant.
11866 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11868 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
11869 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
11871 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
11873 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
11874 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
11875 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
11876 (expand_block_compare): Change select_block_compare_mode call.
11877 (expand_strncmp_align_check): Use new functions, fix comment.
11878 (emit_final_str_compare_gpr): New function.
11879 (expand_strn_compare): Refactor and clean up code.
11880 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
11882 2018-06-19 Tony Reix <tony.reix@atos.com>
11883 Damien Bergamini <damien.bergamini@atos.com>
11884 David Edelsohn <dje.gcc@gmail.com>
11886 * collect2.c (static_obj): New variable.
11887 (static_libs): New variable.
11888 (is_in_list): Uncomment declaration.
11889 (main): Track AIX libraries linked statically.
11890 (is_in_list): Uncomment definition.
11891 (scan_prog_file): Don't add AIX shared libraries initializer
11892 to constructor list if linking statically.
11894 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
11896 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
11898 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
11900 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
11902 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
11905 2018-06-19 Martin Liska <mliska@suse.cz>
11907 * config/i386/i386.c (ix86_can_inline_p): Do not use
11908 ipa_fn_summaries::get_create.
11909 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
11911 (devirtualization_time_bonus): Likewise.
11912 (ipcp_propagate_stage): Likewise.
11913 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11914 (edge_set_predicate): Likewise.
11915 (evaluate_conditions_for_known_args): Likewise.
11916 (evaluate_properties_for_edge): Likewise.
11917 (ipa_call_summary::reset): Tranform to ...
11918 (ipa_call_summary::~ipa_call_summary): ... this.
11919 (ipa_fn_summary::reset): Transform to ...
11920 (ipa_fn_summary::~ipa_fn_summary): ... this.
11921 (ipa_fn_summary_t::remove): Rename to ...
11922 (ipa_fn_summary_t::remove_callees): ... this.
11923 (ipa_fn_summary_t::duplicate): Use placement new
11924 instead of memory copy.
11925 (ipa_call_summary_t::duplicate): Likewise.
11926 (ipa_call_summary_t::remove): Remove.
11927 (dump_ipa_call_summary): Change get_create to get.
11928 (ipa_dump_fn_summary): Dump only when summary exists.
11929 (analyze_function_body): Use symbol_summary::get instead
11931 (compute_fn_summary): Likewise.
11932 (estimate_edge_devirt_benefit): Likewise.
11933 (estimate_edge_size_and_time): Likewise.
11934 (inline_update_callee_summaries): Likewise.
11935 (remap_edge_change_prob): Likewise.
11936 (remap_edge_summaries): Likewise.
11937 (ipa_merge_fn_summary_after_inlining): Likewise.
11938 (write_ipa_call_summary): Likewise.
11939 (ipa_fn_summary_write): Likewise.
11940 (ipa_free_fn_summary): Likewise.
11941 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11942 (struct ipa_call_summary): Likewise.
11943 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11945 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11946 (estimate_size_after_inlining): Likewise.
11947 (estimate_growth): Likewise.
11948 (growth_likely_positive): Likewise.
11949 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11950 (inline_call): Likewise.
11951 * ipa-inline.c (caller_growth_limits): Likewise.
11952 (can_inline_edge_p): Likewise.
11953 (can_inline_edge_by_limits_p): Likewise.
11954 (compute_uninlined_call_time): Likewise.
11955 (compute_inlined_call_time): Likewise.
11956 (want_inline_small_function_p): Likewise.
11957 (edge_badness): Likewise.
11958 (update_caller_keys): Likewise.
11959 (update_callee_keys): Likewise.
11960 (inline_small_functions): Likewise.
11961 (inline_to_all_callers_1): Likewise.
11962 (dump_overall_stats): Likewise.
11963 (early_inline_small_functions): Likewise.
11964 (early_inliner): Likewise.
11965 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11966 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11967 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11968 * ipa-split.c (execute_split_functions): Likewise.
11969 * symbol-summary.h: Likewise.
11970 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11972 2018-06-19 Richard Biener <rguenther@suse.de>
11974 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11975 (vectorize_loops): ... here. Fix dbgcnt handling.
11976 (try_vectorize_loop): Wrap try_vectorize_loop_1.
11978 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
11981 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11982 ieee128 argument takes up only one (vector) register, not two (floating
11985 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11987 * gimplify.c (gimplify_init_constructor): Really never clear for an
11988 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11990 2018-06-19 Richard Biener <rguenther@suse.de>
11992 PR tree-optimization/86179
11993 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11994 after failed recognition.
11996 2018-06-18 Martin Sebor <msebor@redhat.com>
11998 PR middle-end/85602
11999 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
12000 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
12001 Handle integer subtraction.
12002 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
12003 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
12005 2018-06-18 David Malcolm <dmalcolm@redhat.com>
12007 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
12008 param from rtx to rtx_insn *.
12009 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
12011 (frv_ifcvt_modify_insn): Likwise.
12012 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
12013 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
12014 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
12015 as_a <rtx_insn *> cast to local "unprotected_region" once
12016 it's been established that it's not NULL or pc_rtx.
12017 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
12018 param "sethi" from rtx to rtx_insn *.
12019 (nds32_group_float_insns): Likewise for param "insn".
12020 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
12022 (vax_output_int_subtract): Likewise.
12023 * config/vax/vax.c (vax_output_int_add): Likewise for param
12025 (vax_output_int_subtract): Likewise.
12026 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
12027 (emit_pattern_after): Likewise for param "after".
12028 (emit_insn_after): Likewise.
12029 (emit_jump_insn_after): Likewise.
12030 (emit_call_insn_after): Likewise.
12031 (emit_debug_insn_after): Likewise.
12032 (emit_pattern_before): Likewise for param "before".
12033 (emit_insn_before): Likewise.
12034 (emit_jump_insn_before): Likewise.
12035 * final.c (get_insn_template): Likewise for param "insn", removing
12037 * output.h (get_insn_template): Likewise for 2nd param.
12038 * rtl.h (emit_insn_before): Likewise.
12039 (emit_jump_insn_before): Likewise.
12040 (emit_debug_insn_before_noloc): Likewise.
12041 (emit_insn_after): Likewise.
12042 (emit_jump_insn_after): Likewise.
12043 (emit_call_insn_after): Likewise.
12044 (emit_debug_insn_after): Likewise.
12045 (set_insn_deleted): Likewise for param.
12047 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
12050 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
12051 floating point modes, so that IFmode is numerically greater than
12052 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
12053 to declare the ordering. This prevents IFmode from being
12054 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
12055 machine. Include rs6000-modes.h to share the fractional values
12056 between genmodes* and the rest of the compiler.
12057 (IFmode): Likewise.
12058 (KFmode): Likewise.
12059 (TFmode): Likewise.
12060 * config/rs6000/rs6000-modes.h: New file.
12061 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
12062 meaning of rs6000_long_double_size so that 126..128 selects an
12063 appropriate 128-bit floating point type.
12064 (rs6000_option_override_internal): Likewise.
12065 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
12066 (TARGET_LONG_DOUBLE_128): Change the meaning of
12067 rs6000_long_double_size so that 126..128 selects an appropriate
12068 128-bit floating point type.
12069 (LONG_DOUBLE_TYPE_SIZE): Update comment.
12070 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
12071 source and destination to match the standard usage.
12072 (truncifkf2): Likewise.
12073 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
12074 ISA 2.07 to use an explicit clobber, instead of passing in a
12076 (copysign<mode>3_soft): Likewise.
12078 2018-06-18 David Malcolm <dmalcolm@redhat.com>
12080 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12081 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
12082 (vect_slp_analyze_instance_dependence): Likewise.
12083 (vect_enhance_data_refs_alignment): Likewise.
12084 (vect_analyze_data_refs_alignment): Likewise.
12085 (vect_slp_analyze_and_verify_instance_alignment
12086 (vect_analyze_data_ref_accesses): Likewise.
12087 (vect_prune_runtime_alias_test_list): Likewise.
12088 (vect_analyze_data_refs): Likewise.
12089 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
12090 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12091 (vect_analyze_scalar_cycles_1): Likewise.
12092 (vect_get_loop_niters): Likewise.
12093 (vect_analyze_loop_form_1): Likewise.
12094 (vect_update_vf_for_slp): Likewise.
12095 (vect_analyze_loop_operations): Likewise.
12096 (vect_analyze_loop): Likewise.
12097 (vectorizable_induction): Likewise.
12098 (vect_transform_loop): Likewise.
12099 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
12100 * tree-vect-slp.c (vect_analyze_slp): Likewise.
12101 (vect_make_slp_decision): Likewise.
12102 (vect_detect_hybrid_slp): Likewise.
12103 (vect_slp_analyze_operations): Likewise.
12104 (vect_slp_bb): Likewise.
12105 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
12106 (vectorizable_bswap): Likewise.
12107 (vectorizable_call): Likewise.
12108 (vectorizable_simd_clone_call): Likewise.
12109 (vectorizable_conversion): Likewise.
12110 (vectorizable_assignment): Likewise.
12111 (vectorizable_shift): Likewise.
12112 (vectorizable_operation): Likewise.
12113 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12115 2018-06-18 Martin Sebor <msebor@redhat.com>
12117 PR tree-optimization/81384
12118 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12119 * builtins.c (expand_builtin_strnlen): New function.
12120 (expand_builtin): Call it.
12121 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12122 * builtins.def (BUILT_IN_STRNLEN): New.
12123 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12124 Warn for bounds in excess of maximum object size.
12125 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12126 single-value ranges. Handle strnlen.
12127 (handle_builtin_strlen): Handle strnlen.
12128 (strlen_check_and_optimize_stmt): Same.
12129 * doc/extend.texi (Other Builtins): Document strnlen.
12131 2018-06-18 Maya Rashish <coypu@sdf.org>
12133 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12134 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12135 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12137 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12139 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12141 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12143 * tree.c (escaped_string::escape): Replace cast to char * by
12144 const_cast<char *> (unescaped).
12146 2018-06-18 Nick Clifton <nickc@redhat.com>
12149 * tree.c (escaped_string): New class. Converts an unescaped
12150 string into its escaped equivalent.
12151 (warn_deprecated_use): Use the new class to convert the
12152 deprecation message, if present.
12153 (test_escaped_strings): New self test.
12154 (test_c_tests): Add test_escaped_strings.
12155 * doc/extend.texi (deprecated): Add a note that the
12156 deprecation message is affected by the -fmessage-length
12157 option, and that control characters will be escaped.
12158 (#pragma GCC error): Document this pragma.
12159 (#pragma GCC warning): Likewise.
12160 * doc/invoke.texi (-fmessage-length): Document this option's
12161 effect on the #warning and #error preprocessor directives and
12162 the deprecated attribute.
12164 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
12166 * tree.c (decl_value_expr_lookup): Revert latest change.
12167 (decl_value_expr_insert): Likewise.
12169 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
12171 * gimplify.c (nonlocal_vlas): Delete.
12172 (nonlocal_vla_vars): Likewise.
12173 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12175 (gimplify_body): Do not create and destroy nonlocal_vlas.
12176 * tree-nested.c: Include diagnostic.h.
12177 (use_pointer_in_frame): Tweak.
12178 (lookup_field_for_decl): Add assertion and declare the transformation.
12179 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12180 internal error when the reference is in a wrong context. Do not
12181 create a debug decl by default.
12182 (note_nonlocal_block_vlas): Delete.
12183 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12184 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
12185 create a debug decl by default.
12186 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12187 call is in a wrong context.
12188 (fixup_vla_decls): New function.
12189 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
12190 debug variables were created.
12191 * tree.c (decl_value_expr_lookup): Add checking assertion.
12192 (decl_value_expr_insert): Likewise.
12194 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
12196 PR middle-end/82479
12197 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12198 * tree-scalar-evolution.c (interpret_expr): Likewise.
12199 (expression_expensive_p): Likewise.
12200 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12201 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12202 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12203 (ssa_defined_by_minus_one_stmt_p): New.
12205 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
12207 PR middle-end/64946
12208 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12209 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12210 * dojump.c (do_jump): Likewise.
12211 * expr.c (expand_expr_real_2): Check operand type's sign.
12212 * fold-const.c (const_unop): Handle ABSU_EXPR.
12213 (fold_abs_const): Likewise.
12214 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
12215 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12216 (strip_sign_op_1): Likesise.
12217 * match.pd: Add new pattern to generate ABSU_EXPR.
12218 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
12219 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12220 * tree-eh.c (operation_could_trap_helper_p): Likewise.
12221 * tree-inline.c (estimate_operator_cost): Likewise.
12222 * tree-pretty-print.c (dump_generic_node): Likewise.
12223 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
12224 * tree.def (ABSU_EXPR): New.
12226 2018-06-16 Jakub Jelinek <jakub@redhat.com>
12228 PR middle-end/86095
12229 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
12230 documented as preserved for backward compatibility only.
12231 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
12233 PR rtl-optimization/86108
12234 * bb-reorder.c (create_forwarder_block): Renamed to ...
12235 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
12236 jump from new landing pad to the second part.
12237 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
12240 2018-06-15 Jakub Jelinek <jakub@redhat.com>
12242 PR middle-end/85878
12243 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
12244 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
12245 Only call store_expr for halves if the mode is the same.
12247 PR middle-end/86123
12248 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
12249 Fix up comment formatting.
12251 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
12253 * typed-splay-tree.h (typed_splay_tree::remove): New function.
12254 (typed_splay_tree::closure,
12255 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
12256 (typed_splay_tree::typed_splay_tree,
12257 typed_splay_tree::operator =): Declared private.
12258 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
12259 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
12260 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
12261 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
12262 typed_splay_tree::splay_tree_splay,
12263 typed_splay_tree::splay_tree_foreach_helper,
12264 typed_splay_tree::splay_tree_insert,
12265 typed_splay_tree::splay_tree_remove,
12266 typed_splay_tree::splay_tree_lookup,
12267 typed_splay_tree::splay_tree_predecessor,
12268 typed_splay_tree::splay_tree_successor,
12269 typed_splay_tree::splay_tree_min,
12270 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
12271 (typed_splay_tree::root, typed_splay_tree::comp,
12272 typed_splay_tree::delete_key,
12273 typed_splay_tree::delete_value): New data members.
12274 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
12275 typed_splay_tree::remove.
12277 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
12279 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
12280 -mginv and -mno-ginv to the assembler.
12281 * config/mips/mips.opt (-mcrc): New option.
12282 (-mginv): Likewise.
12283 * doc/invoke.text (-mcrc): Document.
12284 (-mginv): Likewise.
12286 2018-06-15 Nick Clifton <nickc@redhat.com>
12289 * tree.c (escaped_string): New class. Converts an unescaped
12290 string into its escaped equivalent.
12291 (warn_deprecated_use): Use the new class to convert the
12292 deprecation message, if present.
12293 (test_escaped_strings): New self test.
12294 (test_c_tests): Add test_escaped_strings.
12295 * doc/extend.texi (deprecated): Add a note that the
12296 deprecation message is affected by the -fmessage-length
12297 option, and that control characters will be escaped.
12298 (#pragma GCC error): Document this pragma.
12299 (#pragma GCC warning): Likewise.
12300 * doc/invoke.texi (-fmessage-length): Document this option's
12301 effect on the #warning and #error preprocessor directives and
12302 the deprecated attribute.
12304 2018-06-15 Richard Biener <rguenther@suse.de>
12306 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
12307 here, also noting vector size used.
12308 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
12309 size used in MSG_OPTIMIZED_LOCATIONS dump.
12310 (pass_slp_vectorize::execute): Adjust.
12312 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
12315 * config/arc/arc.c (arc_return_address_register): Fix
12318 2018-06-15 Richard Biener <rguenther@suse.de>
12320 PR middle-end/86159
12321 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
12322 leave useless conversion stripping to force_gimple_operand_gsi.
12323 (gimplify_build2): Likewise.
12324 (gimplify_build1): Likewise.
12326 2018-06-15 Richard Biener <rguenther@suse.de>
12328 PR middle-end/86076
12329 * tree-cfg.c (move_stmt_op): unshare invariant addresses
12330 before adjusting their block.
12332 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
12334 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
12335 multilibs for *-*-rtems*.
12336 * config/riscv/t-rtems: New file.
12338 2018-06-14 Jakub Jelinek <jakub@redhat.com>
12340 PR middle-end/86122
12341 * match.pd ((A +- CST1) +- CST2): Punt if last resort
12342 unsigned_type_for returns NULL.
12345 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
12346 subregs of multi-word pseudos unless the float mode has word size.
12348 2018-06-14 Richard Biener <rguenther@suse.de>
12350 PR middle-end/86139
12351 * tree-vect-generic.c (build_word_mode_vector_type): Remove
12352 duplicate and harmful type_hash_canon.
12353 * tree.c (type_hash_canon): Assert we didn't find ourselves.
12355 2018-06-14 Richard Biener <rguenther@suse.de>
12358 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
12361 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
12363 * config/rtems.h (STDINT_LONG32): Define.
12365 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
12366 Prachi Godbole <prachi.godbole@imgtec.com>
12368 * config/mips/mips-cpus.def: Define P6600.
12369 * config/mips/mips-tables.opt: Regenerate.
12370 * config/mips/mips.c (mips_ucbranch_type): New enum.
12371 (mips_rtx_cost_data): Add support for P6600.
12372 (mips_issue_rate): Likewise.
12373 (mips_multipass_dfa_lookahead): Likewise.
12374 (mips_avoid_hazard): Likewise.
12375 (mips_reorg_process_insns): Likewise.
12376 (mips_classify_branch_p6600): New function.
12377 * config/mips/mips.h (TUNE_P6600): New define.
12378 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
12379 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
12380 * config/mips/mips.md: Include p6600.md.
12381 (processor): Add p6600.
12382 * config/mips/p6600.md: New file.
12383 * doc/invoke.texi: Add p6600 to supported architectures.
12385 2018-06-13 Martin Sebor <msebor@redhat.com>
12387 PR tree-optimization/86114
12388 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
12390 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
12392 2018-06-13 Richard Biener <rguenther@suse.de>
12394 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
12395 Properly set vector type of the intermediate stmt.
12396 * tree-vect-stmts.c (vectorizable_operation): The destination
12397 var always has vectype_out type.
12399 2018-06-13 Jeff Law <law@redhat.com>
12401 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
12402 integer 0 for argument to print_rtl_with_bb.
12403 (rl78_reorg): Likewise.
12405 2018-06-13 David Malcolm <dmalcolm@redhat.com>
12407 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
12408 from rtx to rtx_insn *.
12409 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
12411 (add_sched_insns_for_speculation): Likewise for local "target",
12412 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12413 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
12414 from rtx_insn ** to rtx_code_label **.
12415 (reorg_emit_nops): Likewise.
12416 (c6x_reorg): Likewise for local "call_labels".
12417 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
12419 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
12420 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
12421 the loops over LABEL_REFS.
12422 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
12424 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12425 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
12426 (split_branches): Strengthen local "olabel" from rtx to
12427 rtx_insn *, adding a safe_as_a cast.
12428 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
12430 (add_insn_after): Likewise for first two params.
12431 (add_insn_before): Likewise.
12432 (remove_insn): Likewise for param.
12433 (emit_pattern_before_noloc): Likewise for second and third params.
12434 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
12435 (emit_call_insn_before_noloc): Likewise.
12436 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
12438 (emit_barrier_before): Likewise.
12439 (emit_label_before): Strengthen "label" param from "rtx" to
12440 "rtx_code_label *". Strengthen "before" param from "rtx" to
12442 (emit_insn_after_1): Strengthen "after" param from "rtx" to
12444 (emit_pattern_after_noloc): Likewise.
12445 (emit_insn_after_noloc): Likewise.
12446 (emit_jump_insn_after_noloc): Likewise.
12447 (emit_call_insn_after_noloc): Likewise.
12448 (emit_debug_insn_after_noloc): Likewise.
12449 (emit_barrier_after): Likewise.
12450 (emit_label_after): Likewise for both params.
12451 (emit_pattern_after_setloc): Likewise for "after" param. Convert
12452 "loc" param from "int" to "location_t".
12453 (emit_insn_after_setloc): Likewise.
12454 (emit_jump_insn_after_setloc): Likewise.
12455 (emit_call_insn_after_setloc): Likewise.
12456 (emit_debug_insn_after_setloc): Likewise.
12457 (emit_pattern_before_setloc): Likewise for "before" param. Convert
12458 "loc" param from "int" to "location_t".
12459 (emit_pattern_before): Convert NULL_RTX to NULL.
12460 (emit_insn_before_setloc): Convert "loc" param from "int" to
12462 (emit_jump_insn_before_setloc): Likewise.
12463 (emit_call_insn_before_setloc): Likewise.
12464 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
12465 rtx_insn *. Convert "loc" param from "int" to "location_t".
12466 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
12467 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
12468 Convert 3rd param from "int" to "location_t".
12469 (emit_barrier_before, emit_barrier_after, next_real_insn):
12470 Strengthen param from rtx to rtx_insn *.
12471 (emit_label_before): Strengthen 1st param from "rtx" to
12472 "rtx_code_label *". Strengthen 2nd param from "rtx" to
12474 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
12475 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
12476 Strengthen 2nd param from "rtx" to "rtx_insn *".
12477 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
12478 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
12479 Likewise. Convert 3rd param from "int" to "location_t".
12480 (emit_label_after): Strengthen 1st param from "rtx" to
12481 "rtx_code_label *".
12482 (next_real_insn, remove_insn): Strengthen param from "rtx" to
12484 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
12485 from "rtx" to "rtx_insn *".
12487 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
12489 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
12490 bodies streamed in with -Q.
12491 * dumpfile.c (dump_files): Add lto-stream-out dump file.
12492 * dumpfile.h (tree_dump_index): Add lto_stream_out.
12493 * gimple-streamer-out.c: Include gimple-pretty-print.h
12494 (output_bb): Dump stmts streamed.
12495 * lto-section-out.c: Include print-tree.h
12496 (lto_begin_section): Dump sections created.
12497 (lto_output_decl_index): Dump decl encoded.
12498 * lto-streamer-out.c: Include print-tree.h
12499 (create_output_block): Dump output block created.
12500 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
12501 (output_function): Dump function output.
12502 (output_constructor): Dump constructor streamed.
12503 (write_global_stream): Output indexes encoded.
12504 (produce_asm_for_decls): Dump streams encoded.
12505 * lto-streamer.c (streamer_dump_file): New global var.
12506 * lto-streamer.h (streamer_dump_file): Declare.
12507 * passes.c (ipa_write_summaries): Initialize streamer dump.
12508 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
12511 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
12514 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
12515 offsets for register save directives. Emit a second batch of save
12516 directives, if need be, when the function accesses prior frames.
12518 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12520 * config/arc/fpu.md (fmasf4): Force operand to register.
12521 (fnmasf4): Likewise.
12523 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12525 * config/arc/arc-protos.h (arc_pad_return): Remove.
12526 * config/arc/arc.c (machine_function): Remove force_short_suffix
12528 (arc_print_operand): Adjust printing of '&'.
12529 (arc_verify_short): Remove conditional printing of short suffix.
12530 (arc_final_prescan_insn): Remove reference to size_reason.
12531 (pad_return): New function.
12532 (arc_reorg): Call pad_return.
12533 (arc_pad_return): Remove.
12534 (arc_init_machine_status): Remove reference to force_short_suffix.
12535 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
12536 (attr length): When attribute iscompact is true force to 2
12537 regardless; in the case of maybe check if we want to force the
12538 instruction to have 4 bytes length.
12539 (nopv): Change it to generate 4 byte long nop as well.
12540 (blockage): New pattern.
12541 (simple_return): Remove call to arc_pad_return.
12542 (p_return_i): Likewise.
12544 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12546 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
12548 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12550 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
12553 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12555 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
12556 for ARC700 and ARCv2.
12558 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
12561 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
12562 operands[2] instead of operands[1].
12565 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
12567 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
12568 case, check whether the outer register overlaps an unallocatable
12569 register, not just whether it fits the required class.
12571 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
12573 * poly-int.h (can_div_trunc_p): Add new overload in which all values
12575 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
12576 (memrefs_conflict_p): Likewise.
12577 (init_alias_analysis): Likewise.
12578 * cfgexpand.c (expand_debug_expr): Likewise.
12579 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
12580 * cse.c (fold_rtx): Likewise.
12581 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
12582 * expr.c (emit_block_move_hints): Likewise.
12583 (clear_storage_hints, push_block, emit_push_insn): Likewise.
12584 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
12585 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
12586 (emit_group_store): Likewise.
12587 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
12588 to read the PRE/POST_MODIFY increment.
12589 * calls.c (store_one_arg): Use strip_offset.
12590 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
12592 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
12594 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
12595 (simplify_binary_operation_1): Extend CONST_INT handling to
12597 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
12598 than a HOST_WIDE_INT.
12599 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
12601 (adjust_mems, add_stores): Update accodingly.
12602 (vt_canonicalize_addr): Track polynomial offsets.
12603 (emit_note_insn_var_location): Likewise.
12604 (vt_add_function_parameter): Likewise.
12605 (vt_initialize): Likewise.
12607 2018-06-12 Jeff Law <law@redhat.com>
12609 * config.gcc (alpha*-*-freebsd*): Remove.
12610 * config/alpha/freebsd.h: Remove.
12612 2018-06-12 David Malcolm <dmalcolm@redhat.com>
12615 * spellcheck-tree.c (levenshtein_distance): Rename to...
12616 (get_edit_distance): ...this, and update for underlying renaming.
12617 * spellcheck-tree.h (levenshtein_distance): Rename to...
12618 (get_edit_distance): ...this.
12619 * spellcheck.c (levenshtein_distance): Rename to...
12620 (get_edit_distance): ...this. Convert from Levenshtein distance
12621 to Damerau-Levenshtein distance by supporting transpositions of
12622 adjacent characters. Rename "v1" to "v_next" and "v0" to
12624 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
12625 (selftest::test_edit_distance_unit_test_oneway): ...this, and
12626 update for underlying renaming.
12627 (selftest::levenshtein_distance_unit_test): Rename to...
12628 (selftest::test_get_edit_distance_unit): ...this, and update for
12629 underlying renaming.
12630 (selftest::test_find_closest_string): Add example from PR 69968
12631 where transposition helps
12632 (selftest::test_metric_conditions): Update for renaming.
12633 (selftest::test_metric_conditions): Likewise.
12634 (selftest::spellcheck_c_tests): Likewise.
12635 * spellcheck.h (levenshtein_distance): Rename both overloads to...
12636 (get_edit_distance): ...this.
12637 (best_match::consider): Update for renaming.
12639 2018-06-12 Martin Sebor <msebor@redhat.com>
12641 PR tree-optimization/85259
12642 * builtins.c (compute_objsize): Handle constant offsets.
12643 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
12644 true iff a warning has been issued.
12645 * gimple.h (gimple_nonartificial_location): New function.
12646 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
12647 gimple_nonartificial_location and handle -Wno-system-headers.
12648 (handle_builtin_stxncpy): Same.
12650 2018-06-12 Martin Sebor <msebor@redhat.com>
12653 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
12655 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
12657 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
12658 BUILTIN_VEC_XST entries for pointer to double and long long.
12660 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
12663 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12665 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12668 2018-06-12 Martin Liska <mliska@suse.cz>
12670 * doc/options.texi: Document IntegerRange.
12672 2018-06-12 Martin Liska <mliska@suse.cz>
12674 * config/i386/i386.opt: Make MPX-related options as Deprecated.
12675 * opt-functions.awk: Handle Deprecated flag.
12676 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
12678 (read_cmdline_option): Report warning for a deprecated option.
12679 * opts.h (struct cl_option): Add new field cl_deprecated.
12680 (CL_ERR_DEPRECATED): New.
12682 2018-06-12 Martin Liska <mliska@suse.cz>
12684 * doc/options.texi: Document Deprecated option flag.
12686 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12688 * config/arc/arc-arch.h (arc_extras): New enum.
12689 (arc_cpu_t):Add field extra.
12690 (arc_cpu_types): Consider the extras.
12691 * config/arc/arc-cpus.def: Add extras info.
12692 * config/arc/arc-opts.h (processor_type): Consider extra field.
12693 * config/arc/arc.c (arc_override_options): Handle extra field.
12695 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12697 * config/arc/arc-arch.h: Update ARC_OPTX macro.
12698 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
12700 * config/arc/arc.c (arc_init): Update pic warning.
12701 (irq_range): Update irq range parsing warnings.
12702 (arc_override_options): Update various warning messages.
12703 (arc_handle_aux_attribute): Likewise.
12705 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
12707 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
12709 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12711 * doc/sourcebuild.texi: Document usage of line number 0 in verify
12712 compiler messages directives.
12714 2018-06-12 Matthew Fortune <mfortune@gmail.com>
12716 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
12717 * config/mips/mips-tables.opt: Regenerate.
12718 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
12720 * doc/invoke.texi: Document -march=i6500.
12722 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
12724 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
12725 (i6400_gpmul): Add cpu_unit.
12726 (i6400_gpdiv): Likewise.
12727 (i6400_msa_add_d): Update reservations.
12728 (i6400_msa_int_add) Likewise.
12729 (i6400_msa_short_logic3) Likewise.
12730 (i6400_msa_short_logic2) Likewise.
12731 (i6400_msa_short_logic) Likewise.
12732 (i6400_msa_move) Likewise.
12733 (i6400_msa_cmp) Likewise.
12734 (i6400_msa_short_float2) Likewise.
12735 (i6400_msa_div_d) Likewise.
12736 (i6400_msa_long_logic1) Likewise.
12737 (i6400_msa_long_logic2) Likewise.
12738 (i6400_msa_mult) Likewise.
12739 (i6400_msa_long_float2) Likewise.
12740 (i6400_msa_long_float4) Likewise.
12741 (i6400_msa_long_float5) Likewise.
12742 (i6400_msa_long_float8) Likewise.
12743 (i6400_fpu_fadd): Include frint type.
12744 (i6400_fpu_store): New define_insn_reservation.
12745 (i6400_fpu_load): Likewise.
12746 (i6400_fpu_move): Likewise.
12747 (i6400_fpu_fcmp): Likewise.
12748 (i6400_fpu_fmadd): Likewise.
12749 (i6400_int_mult): Include imul3nc type and update reservation.
12750 (i6400_int_div): Include idiv3 type and update reservation.
12751 (i6400_int_load): Update to check type not move_type.
12752 (i6400_int_store): Likewise.
12753 (i6400_int_prefetch): Set zero latency.
12755 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
12757 * gcc.c: Document new %@{...} sequence.
12758 (LINK_COMMAND_SPEC): Use it for the -L switches.
12759 (cpp_unique_options): Use it for the -I switches.
12760 (at_file_argbuf): New global variable.
12761 (in_at_file): Likewise.
12762 (alloc_args): Create at_file_argbuf.
12763 (clear_args): Truncate at_file_argbuf.
12764 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
12765 (open_at_file): New function.
12766 (close_at_file): Likewise.
12767 (create_at_file): Delete.
12768 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
12771 (validate_switches_from_spec): Deal with %@{...} sequence.
12772 (validate_switches): Likewise.
12773 (driver::finalize): Call clear_args.
12775 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
12777 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
12779 2018-06-11 Martin Sebor <msebor@redhat.com>
12781 * doc/invoke.texi (-Wall): List -Wc++17-compat.
12782 (Wno-class-memaccess): Add @opindex.
12783 (Wno-templates, Wno-multiple-inheritance): Same.
12784 (Wno-virtual-inheritance, Wno-namespaces): Same.
12785 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
12786 (Wno-format-overflow, Wno-format-truncation): Same.
12787 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
12788 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
12789 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
12790 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
12791 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
12792 (Wno-misspelled-isr): Same.
12794 2018-06-11 Martin Sebor <msebor@redhat.com>
12796 * PR tree-optimization/86083
12797 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
12799 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
12801 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
12803 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
12806 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
12807 on the correct operand.
12808 (*movdi_internal64): Ditto.
12810 2018-06-11 Martin Liska <mliska@suse.cz>
12812 PR tree-optimization/86089
12813 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
12815 2018-06-11 Julia Koval <julia.koval@intel.com>
12817 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
12818 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
12819 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
12821 2018-06-11 Olivier Hainque <hainque@adacore.com>
12823 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
12824 for Ada with strict dwarf2.
12826 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
12829 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
12832 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
12834 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
12836 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
12838 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
12841 2018-06-08 Martin Liska <mliska@suse.cz>
12843 * tree-cfg.h (debug_function): Fix argument type to match
12846 2018-06-08 Martin Liska <mliska@suse.cz>
12848 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
12849 Remove usage of MPX-related (and removed) fields.
12850 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
12852 2018-06-08 David Malcolm <dmalcolm@redhat.com>
12854 * cfg.c (debug): Use TDF_NONE rather than 0.
12855 * cfghooks.c (debug): Likewise.
12856 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
12857 (struct dump_option_value_info): Convert to...
12858 (struct kv_pair): ...this template type.
12859 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
12861 (optinfo_verbosity_options): Likewise.
12862 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
12864 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
12865 than int for "optgroup_flags" param.
12866 (dump_generic_expr_loc): Use dump_flags_t rather than int for
12868 (dump_dec): Likewise.
12869 (dump_finish): Use TDF_NONE rather than 0.
12870 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
12871 rather than int for "optgroup_flags" param. Use TDF_NONE rather
12872 than 0. Update for change to option_ptr.
12873 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
12874 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
12875 0. Update for changes to optinfo_verbosity_options and
12877 (opt_info_switch_p): Convert optgroup_flags from int to
12879 (dump_basic_block): Use dump_flags_t rather than int
12880 for "dump_kind" param.
12881 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
12882 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
12883 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
12884 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
12885 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
12886 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
12887 TDF_NONE): Convert from macros to...
12888 (enum dump_flag): ...this new enum.
12889 (dump_flags_t): Update to use enum.
12890 (operator|, operator&, operator~, operator|=, operator&=):
12891 Implement for dump_flags_t.
12892 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
12893 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
12894 Convert from macros to...
12895 (enum optgroup_flag): ...this new enum.
12896 (optgroup_flags_t): New typedef.
12897 (operator|, operator|=): Implement for optgroup_flags_t.
12898 (struct dump_file_info): Convert field "alt_flags" to
12899 dump_flags_t. Convert field "optgroup_flags" to
12901 (dump_basic_block): Use dump_flags_t rather than int for param.
12902 (dump_generic_expr_loc): Likewise.
12903 (dump_dec): Likewise.
12904 (dump_register): Convert param "optgroup_flags" to
12906 (opt_info_enable_passes): Likewise.
12907 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
12909 * gimple-pretty-print.c (debug): Likewise.
12910 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
12911 (merged_store_group::apply_stores): Likewise.
12912 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12913 * gimple.c (verify_gimple_pp): Likewise.
12914 * graphite-poly.c (print_pbb_body): Likewise.
12915 * passes.c (pass_manager::register_one_dump_file): Convert
12916 local "optgroup_flags" to optgroup_flags_t.
12917 * print-tree.c (print_node): Use TDF_NONE rather than 0.
12919 (debug_body): Likewise.
12920 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12921 to optgroup_flags_t.
12922 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12924 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12925 (convert_mult_to_fma): Likewise.
12926 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12927 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12928 * tree-vect-data-refs.c (dump_lower_bound): Convert param
12929 "dump_kind" to dump_flags_t.
12931 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12933 * config/rs6000/rs6000.c (min, max): Delete.
12935 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12937 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12938 -mabi=spe and -mabi=no-spe.
12940 2018-06-08 Martin Liska <mliska@suse.cz>
12942 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12943 where we expect an existing summary.
12945 2018-06-08 Martin Liska <mliska@suse.cz>
12947 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12948 * ipa-inline.h (estimate_edge_growth): Likewise.
12950 2018-06-08 Martin Liska <mliska@suse.cz>
12952 * cgraph.c (function_version_hasher::hash): Use
12953 cgraph_node::get_uid ().
12954 (function_version_hasher::equal):
12955 * cgraph.h (cgraph_node::get_uid): New method.
12956 * ipa-inline.c (update_caller_keys): Use
12957 cgraph_node::get_uid ().
12958 (update_callee_keys): Likewise.
12959 * ipa-utils.c (searchc): Likewise.
12960 (ipa_reduced_postorder): Likewise.
12961 * lto-cgraph.c (input_node): Likewise.
12962 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12963 * symbol-summary.h (symtab_insertion): Likewise.
12964 (symtab_removal): Likewise.
12965 (symtab_duplication): Likewise.
12966 * tree-pretty-print.c (dump_function_header): Likewise.
12967 * tree-sra.c (convert_callers_for_node): Likewise.
12969 2018-06-08 Martin Liska <mliska@suse.cz>
12971 * cgraph.c (symbol_table::create_edge): Always assign a new
12973 (symbol_table::free_edge): Do not recycle numbers.
12974 * cgraph.h (cgraph_edge::get): New method.
12975 * symbol-summary.h (symtab_removal): Use it.
12976 (symtab_duplication): Likewise.
12977 (call_summary::hashable_uid): Remove.
12979 2018-06-08 Martin Liska <mliska@suse.cz>
12981 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12982 (initialize_growth_caches): Remove.
12983 (free_growth_caches): Likewise.
12984 (do_estimate_edge_time): Use edge_growth_cache.
12985 (do_estimate_edge_size): Likewise.
12986 (do_estimate_edge_hints): Likewise.
12987 * ipa-inline.c (reset_edge_caches): Likewise.
12988 (recursive_inlining): Likewise.
12989 (inline_small_functions): Likewise.
12990 * ipa-inline.h (initialize_growth_caches): Remove.
12991 (estimate_edge_size): Likewise.
12992 (estimate_edge_time): Likewise.
12993 (estimate_edge_hints): Likewise.
12994 (reset_edge_growth_cache): Likewise.
12995 * symbol-summary.h (call_summary::remove): New method.
12997 2018-06-08 Martin Liska <mliska@suse.cz>
12999 * ipa-cp.c (class edge_clone_summary): New summary.
13000 (grow_edge_clone_vectors): Remove.
13001 (ipcp_edge_duplication_hook): Remove.
13002 (class edge_clone_summary_t): New call_summary class.
13003 (ipcp_edge_removal_hook): Remove.
13004 (edge_clone_summary_t::duplicate): New function.
13005 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
13006 (create_specialized_node): Likewise.
13007 (ipcp_driver): Initialize edge_clone_summaries and do not
13010 2018-06-08 Martin Liska <mliska@suse.cz>
13012 * symbol-summary.h (get): New function.
13013 (call_summary::m_initialize_when_cloning): New class member.
13015 2018-06-08 Martin Liska <mliska@suse.cz>
13017 * cgraph.c (cgraph_node::remove): Do not recycle uid.
13018 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
13019 (symbol_table::allocate_cgraph_symbol): Do not set uid.
13020 * passes.c (uid_hash_t): Record removed_nodes by their uids.
13021 (remove_cgraph_node_from_order): Use the removed_nodes set.
13022 (do_per_function_toporder): Likwise.
13023 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
13024 instead of summary_uid.
13025 (symtab_removal): Likewise.
13026 (symtab_duplication): Likewise.
13028 2018-06-08 Martin Liska <mliska@suse.cz>
13030 * ipa-cp.c (ipcp_store_bits_results): Use
13031 ipcp_transformation_sum.
13032 (ipcp_store_vr_results): Likewise.
13033 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
13035 (ipcp_transformation_initialize): ... this.
13036 (ipa_set_node_agg_value_chain):
13037 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
13038 (write_ipcp_transformation_info): Likewise.
13039 (read_ipcp_transformation_info): Likewise.
13040 (ipcp_update_bits): Likewise.
13041 (ipcp_update_vr): Likewise.
13042 (ipcp_transform_function): Likewise.
13043 * ipa-prop.h: Rename ipcp_transformation_summary to
13044 ipcp_transformation.
13045 (class ipcp_transformation_t): New function summary.
13046 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
13047 (ipa_get_agg_replacements_for_node): Likewise.
13049 2018-06-08 Martin Liska <mliska@suse.cz>
13051 * ipa-pure-const.c (struct funct_state_d): Do it class instead
13053 (class funct_state_summary_t): New function_summary class.
13054 (has_function_state): Remove.
13055 (get_function_state): Likewise.
13056 (set_function_state): Likewise.
13057 (add_new_function): Likewise.
13058 (funct_state_summary_t::insert): New function.
13059 (duplicate_node_data): Remove.
13060 (remove_node_data): Remove.
13061 (funct_state_summary_t::duplicate): New function.
13062 (register_hooks): Create new funct_state_summaries.
13063 (pure_const_generate_summary): Use it.
13064 (pure_const_write_summary): Likewise.
13065 (pure_const_read_summary): Likewise.
13066 (propagate_pure_const): Likewise.
13067 (propagate_nothrow): Likewise.
13068 (dump_malloc_lattice): Likewise.
13069 (propagate_malloc): Likewise.
13070 (execute): Do not register hooks, just remove summary
13072 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
13075 2018-06-08 Martin Liska <mliska@suse.cz>
13077 * ipa-reference.c (remove_node_data): Remove.
13078 (duplicate_node_data): Likewise.
13079 (class ipa_ref_var_info_summary_t): New class.
13080 (class ipa_ref_opt_summary_t): Likewise.
13081 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
13082 (get_reference_optimization_summary): Use
13083 ipa_ref_opt_sum_summaries.
13084 (set_reference_vars_info): Remove.
13085 (set_reference_optimization_summary): Likewise.
13086 (ipa_init): Create summaries.
13087 (init_function_info): Use function summary.
13088 (ipa_ref_opt_summary_t::duplicate): New function.
13089 (ipa_ref_opt_summary_t::remove): New function.
13090 (get_read_write_all_from_node): Fix GNU coding style.
13091 (propagate): Use function summary.
13092 (write_node_summary_p): Fix GNU coding style.
13093 (stream_out_bitmap): Likewise.
13094 (ipa_reference_read_optimization_summary): Use function summary.
13095 (ipa_reference_c_finalize): Do not release hooks.
13097 2018-06-08 Martin Liska <mliska@suse.cz>
13099 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
13100 (analyze_function_body): Extract multiple calls of get_create.
13101 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13102 * ipa-inline.c (recursive_inlining): Use ::get method.
13103 * ipa-inline.h (estimate_edge_growth): Likewise.
13105 2018-06-08 Martin Liska <mliska@suse.cz>
13107 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13109 (hsa_function_summary::hsa_function_summary): Use the new enum
13111 (hsa_gpu_implementation_p): Use hsa_summaries::get.
13112 * hsa-gen.c (hsa_get_host_function): Likewise.
13113 (get_brig_function_name): Likewise.
13114 * ipa-hsa.c (process_hsa_functions): Likewise.
13115 (ipa_hsa_write_summary): Likewise.
13116 * symbol-summary.h (symtab_duplication): Use ::get function/
13117 (get): New function.
13119 2018-06-08 Martin Liska <mliska@suse.cz>
13121 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13123 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13124 (hsa_register_kernel): Likewise.
13125 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13126 * hsa-gen.c (hsa_get_host_function): Likewise.
13127 (get_brig_function_name): Likewise.
13128 (generate_hsa): Likewise.
13129 (pass_gen_hsail::execute): Likewise.
13130 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13131 (devirtualization_time_bonus): Likewise.
13132 (ipcp_propagate_stage): Likewise.
13133 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13134 (edge_set_predicate): Likewise.
13135 (evaluate_conditions_for_known_args): Likewise.
13136 (evaluate_properties_for_edge): Likewise.
13137 (ipa_fn_summary::reset): Likewise.
13138 (ipa_fn_summary_t::duplicate): Likewise.
13139 (dump_ipa_call_summary): Likewise.
13140 (ipa_dump_fn_summary): Likewise.
13141 (analyze_function_body): Likewise.
13142 (compute_fn_summary): Likewise.
13143 (estimate_edge_devirt_benefit): Likewise.
13144 (estimate_edge_size_and_time): Likewise.
13145 (estimate_calls_size_and_time): Likewise.
13146 (estimate_node_size_and_time): Likewise.
13147 (inline_update_callee_summaries): Likewise.
13148 (remap_edge_change_prob): Likewise.
13149 (remap_edge_summaries): Likewise.
13150 (ipa_merge_fn_summary_after_inlining): Likewise.
13151 (ipa_update_overall_fn_summary): Likewise.
13152 (read_ipa_call_summary): Likewise.
13153 (inline_read_section): Likewise.
13154 (write_ipa_call_summary): Likewise.
13155 (ipa_fn_summary_write): Likewise.
13156 (ipa_free_fn_summary): Likewise.
13157 * ipa-hsa.c (process_hsa_functions): Likewise.
13158 (ipa_hsa_write_summary): Likewise.
13159 (ipa_hsa_read_section): Likewise.
13160 * ipa-icf.c (sem_function::merge): Likewise.
13161 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13162 (do_estimate_edge_time): Likewise.
13163 (estimate_size_after_inlining): Likewise.
13164 (estimate_growth): Likewise.
13165 (growth_likely_positive): Likewise.
13166 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13167 (inline_call): Likewise.
13168 * ipa-inline.c (caller_growth_limits): Likewise.
13169 (can_inline_edge_p): Likewise.
13170 (can_inline_edge_by_limits_p): Likewise.
13171 (compute_uninlined_call_time): Likewise.
13172 (compute_inlined_call_time): Likewise.
13173 (want_inline_small_function_p): Likewise.
13174 (edge_badness): Likewise.
13175 (update_caller_keys): Likewise.
13176 (update_callee_keys): Likewise.
13177 (recursive_inlining): Likewise.
13178 (inline_small_functions): Likewise.
13179 (inline_to_all_callers_1): Likewise.
13180 (dump_overall_stats): Likewise.
13181 (early_inline_small_functions): Likewise.
13182 (early_inliner): Likewise.
13183 * ipa-inline.h (estimate_edge_growth): Likewise.
13184 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13185 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13186 * ipa-prop.h (IPA_NODE_REF): Likewise.
13187 (IPA_EDGE_REF): Likewise.
13188 * ipa-pure-const.c (malloc_candidate_p): Likewise.
13189 (propagate_malloc): Likewise.
13190 * ipa-split.c (execute_split_functions): Likewise.
13191 * symbol-summary.h: Rename get to get_create.
13193 (get_create): Likewise.
13194 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13196 2018-06-08 Martin Liska <mliska@suse.cz>
13198 * symbol-summary.h (release): Move definition out of class
13200 (symtab_removal): Likewise.
13201 (symtab_duplication): Likewise.
13203 2018-06-08 Martin Liska <mliska@suse.cz>
13205 * symbol-summary.h (function_summary): Move constructor
13206 implementation out of class declaration.
13207 (release): Likewise.
13208 (symtab_insertion): Likewise.
13209 (symtab_removal): Likewise.
13210 (symtab_duplication): Likewise.
13213 2018-06-08 Martin Liska <mliska@suse.cz>
13215 * Makefile.in: Remove support for MPX (macros, related functions,
13216 fields in cgraph_node, ...).
13217 * builtin-types.def (BT_BND): Likewise.
13218 (BT_FN_BND_CONST_PTR): Likewise.
13219 (BT_FN_CONST_PTR_BND): Likewise.
13220 (BT_FN_VOID_PTR_BND): Likewise.
13221 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
13222 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
13223 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
13224 (expand_builtin_mempcpy_with_bounds): Likewise.
13225 (expand_builtin_memset_with_bounds): Likewise.
13226 (expand_builtin_memset_args): Likewise.
13227 (std_expand_builtin_va_start): Likewise.
13228 (expand_builtin): Likewise.
13229 (expand_builtin_with_bounds): Likewise.
13230 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
13231 (DEF_LIB_BUILTIN_CHKP): Likewise.
13232 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
13233 (DEF_CHKP_BUILTIN): Likewise.
13234 (BUILT_IN_MEMCPY): Likewise.
13235 (BUILT_IN_MEMMOVE): Likewise.
13236 (BUILT_IN_MEMPCPY): Likewise.
13237 (BUILT_IN_MEMSET): Likewise.
13238 (BUILT_IN_STPCPY): Likewise.
13239 (BUILT_IN_STRCAT): Likewise.
13240 (BUILT_IN_STRCHR): Likewise.
13241 (BUILT_IN_STRCPY): Likewise.
13242 (BUILT_IN_STRLEN): Likewise.
13243 (BUILT_IN_MEMCPY_CHK): Likewise.
13244 (BUILT_IN_MEMMOVE_CHK): Likewise.
13245 (BUILT_IN_MEMPCPY_CHK): Likewise.
13246 (BUILT_IN_MEMSET_CHK): Likewise.
13247 (BUILT_IN_STPCPY_CHK): Likewise.
13248 (BUILT_IN_STRCAT_CHK): Likewise.
13249 (BUILT_IN_STRCPY_CHK): Likewise.
13250 * calls.c (store_bounds): Likewise.
13251 (emit_call_1): Likewise.
13252 (special_function_p): Likewise.
13253 (maybe_warn_nonstring_arg): Likewise.
13254 (initialize_argument_information): Likewise.
13255 (finalize_must_preallocate): Likewise.
13256 (compute_argument_addresses): Likewise.
13257 (expand_call): Likewise.
13258 * cfgexpand.c (expand_call_stmt): Likewise.
13259 (expand_return): Likewise.
13260 (expand_gimple_stmt_1): Likewise.
13261 (pass_expand::execute): Likewise.
13262 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
13263 (cgraph_node::remove): Likewise.
13264 (cgraph_node::dump): Likewise.
13265 (cgraph_node::verify_node): Likewise.
13266 * cgraph.h (chkp_function_instrumented_p): Likewise.
13267 (symtab_node::get_alias_target): Likewise.
13268 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
13269 (cgraph_local_p): Likewise.
13270 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
13271 (cgraph_edge::rebuild_references): Likewise.
13272 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
13273 (walk_polymorphic_call_targets): Likewise.
13274 (cgraph_node::expand_thunk): Likewise.
13275 (symbol_table::output_weakrefs): Likewise.
13276 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
13277 (ix86_handle_option): Likewise.
13278 * config/i386/constraints.md: Likewise.
13279 * config/i386/i386-builtin-types.def (BND): Likewise.
13283 * config/i386/i386-builtin.def (BDESC_END): Likewise.
13284 (BDESC_FIRST): Likewise.
13286 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
13287 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
13288 * config/i386/i386.c (enum reg_class): Likewise.
13289 (ix86_target_string): Likewise.
13290 (ix86_option_override_internal): Likewise.
13291 (ix86_conditional_register_usage): Likewise.
13292 (ix86_valid_target_attribute_inner_p): Likewise.
13293 (ix86_set_indirect_branch_type): Likewise.
13294 (ix86_set_current_function): Likewise.
13295 (ix86_function_arg_regno_p): Likewise.
13296 (init_cumulative_args): Likewise.
13297 (ix86_function_arg_advance): Likewise.
13298 (ix86_function_arg): Likewise.
13299 (ix86_pass_by_reference): Likewise.
13300 (ix86_function_value_regno_p): Likewise.
13301 (ix86_function_value_1): Likewise.
13302 (ix86_function_value_bounds): Likewise.
13303 (ix86_return_in_memory): Likewise.
13304 (ix86_setup_incoming_vararg_bounds): Likewise.
13305 (ix86_va_start): Likewise.
13306 (indirect_thunk_need_prefix): Likewise.
13307 (print_reg): Likewise.
13308 (ix86_print_operand): Likewise.
13309 (ix86_expand_call): Likewise.
13310 (ix86_output_function_return): Likewise.
13311 (reg_encoded_number): Likewise.
13312 (BDESC_VERIFYS): Likewise.
13313 (ix86_init_mpx_builtins): Likewise.
13314 (ix86_init_builtins): Likewise.
13315 (ix86_emit_cmove): Likewise.
13316 (ix86_emit_move_max): Likewise.
13317 (ix86_expand_builtin): Likewise.
13318 (ix86_builtin_mpx_function): Likewise.
13319 (ix86_get_arg_address_for_bt): Likewise.
13320 (ix86_load_bounds): Likewise.
13321 (ix86_store_bounds): Likewise.
13322 (ix86_load_returned_bounds): Likewise.
13323 (ix86_store_returned_bounds): Likewise.
13324 (ix86_class_likely_spilled_p): Likewise.
13325 (ix86_hard_regno_mode_ok): Likewise.
13326 (x86_order_regs_for_local_alloc): Likewise.
13327 (ix86_mitigate_rop): Likewise.
13328 (ix86_bnd_prefixed_insn_p): Likewise.
13329 (ix86_mpx_bound_mode): Likewise.
13330 (ix86_make_bounds_constant): Likewise.
13331 (ix86_initialize_bounds): Likewise.
13332 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
13333 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
13334 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
13335 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
13336 (TARGET_CHKP_BOUND_MODE): Likewise.
13337 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
13338 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
13339 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
13340 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
13341 * config/i386/i386.h (TARGET_MPX): Likewise.
13342 (TARGET_MPX_P): Likewise.
13343 (VALID_BND_REG_MODE): Likewise.
13344 (FIRST_BND_REG): Likewise.
13345 (LAST_BND_REG): Likewise.
13346 (enum reg_class): Likewise.
13347 (BND_REG_P): Likewise.
13348 (BND_REGNO_P): Likewise.
13349 (BNDmode): Likewise.
13350 (ADJUST_INSN_LENGTH): Likewise.
13351 * config/i386/i386.md: Likewise.
13352 * config/i386/i386.opt: Likewise.
13353 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
13354 (defined): Likewise.
13355 (LINK_MPX): Likewise.
13356 (MPX_SPEC): Likewise.
13357 (LIBMPX_SPEC): Likewise.
13358 (LIBMPXWRAPPERS_SPEC): Likewise.
13359 (CHKP_SPEC): Likewise.
13360 * config/i386/predicates.md: Likewise.
13361 * dbxout.c (dbxout_type): Likewise.
13362 * doc/extend.texi: Likewise.
13363 * doc/invoke.texi: Likewise.
13364 * doc/md.texi: Likewise.
13365 * doc/tm.texi: Likewise.
13366 * doc/tm.texi.in: Likewise.
13367 * dwarf2out.c (is_base_type): Likewise.
13368 (gen_formal_types_die): Likewise.
13369 (gen_subprogram_die): Likewise.
13370 (gen_type_die_with_usage): Likewise.
13371 (gen_decl_die): Likewise.
13372 (dwarf2out_late_global_decl): Likewise.
13373 * expr.c (expand_assignment): Likewise.
13374 (emit_storent_insn): Likewise.
13375 (store_expr_with_bounds): Likewise.
13376 (store_expr): Likewise.
13377 (expand_expr_real_1): Likewise.
13378 * expr.h (store_expr_with_bounds): Likewise.
13379 * function.c (use_register_for_decl): Likewise.
13380 (struct bounds_parm_data): Likewise.
13381 (assign_parms_augmented_arg_list): Likewise.
13382 (assign_parm_find_entry_rtl): Likewise.
13383 (assign_parm_is_stack_parm): Likewise.
13384 (assign_parm_load_bounds): Likewise.
13385 (assign_bounds): Likewise.
13386 (assign_parms): Likewise.
13387 (expand_function_start): Likewise.
13388 * gcc.c (CHKP_SPEC): Likewise.
13389 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
13390 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
13391 (wrestrict_dom_walker::check_call): Likewise.
13392 * gimple.c (gimple_build_call_from_tree): Likewise.
13393 * gimple.h (enum gf_mask): Likewise.
13394 (gimple_call_with_bounds_p): Likewise.
13395 (gimple_call_set_with_bounds): Likewise.
13396 * gimplify.c (gimplify_init_constructor): Likewise.
13397 * ipa-cp.c (initialize_node_lattices): Likewise.
13398 (propagate_constants_across_call): Likewise.
13399 (find_more_scalar_values_for_callers_subset): Likewise.
13400 * ipa-hsa.c (process_hsa_functions): Likewise.
13401 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
13402 * ipa-icf.c (sem_function::merge): Likewise.
13403 * ipa-inline.c (early_inliner): Likewise.
13404 * ipa-pure-const.c (warn_function_noreturn): Likewise.
13405 (warn_function_cold): Likewise.
13406 (propagate_pure_const): Likewise.
13407 * ipa-ref.h (enum GTY): Likewise.
13408 * ipa-split.c (find_retbnd): Likewise.
13409 (consider_split): Likewise.
13410 (split_function): Likewise.
13411 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
13412 * ipa.c (walk_polymorphic_call_targets): Likewise.
13413 (symbol_table::remove_unreachable_nodes): Likewise.
13414 (process_references): Likewise.
13415 (cgraph_build_static_cdtor_1): Likewise.
13416 * lto-cgraph.c (lto_output_node): Likewise.
13417 (output_refs): Likewise.
13418 (compute_ltrans_boundary): Likewise.
13419 (input_overwrite_node): Likewise.
13420 (input_node): Likewise.
13421 (input_cgraph_1): Likewise.
13422 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
13423 * passes.c (pass_manager::execute_early_local_passes): Likewise.
13424 (class pass_chkp_instrumentation_passes): Likewise.
13425 (make_pass_chkp_instrumentation_passes): Likewise.
13426 * passes.def: Likewise.
13427 * rtl.h (struct GTY): Likewise.
13428 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
13429 * stor-layout.c (layout_type): Likewise.
13430 * symtab.c: Likewise.
13431 * target.def: Likewise.
13432 * targhooks.c (default_chkp_bound_type): Likewise.
13433 (default_chkp_bound_mode): Likewise.
13434 (default_builtin_chkp_function): Likewise.
13435 (default_chkp_function_value_bounds): Likewise.
13436 (default_chkp_make_bounds_constant): Likewise.
13437 (default_chkp_initialize_bounds): Likewise.
13438 * targhooks.h (default_chkp_bound_type): Likewise.
13439 (default_chkp_bound_mode): Likewise.
13440 (default_builtin_chkp_function): Likewise.
13441 (default_chkp_function_value_bounds): Likewise.
13442 (default_chkp_make_bounds_constant): Likewise.
13443 (default_chkp_initialize_bounds): Likewise.
13444 * toplev.c (compile_file): Likewise.
13445 (process_options): Likewise.
13446 * tree-core.h (DEF_BUILTIN): Likewise.
13447 (DEF_BUILTIN_CHKP): Likewise.
13448 * tree-inline.c (declare_return_variable): Likewise.
13449 (remap_gimple_stmt): Likewise.
13450 (copy_bb): Likewise.
13451 (initialize_inlined_parameters): Likewise.
13452 (expand_call_inline): Likewise.
13453 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
13454 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
13455 (make_pass_ipa_chkp_produce_thunks): Likewise.
13456 (make_pass_chkp): Likewise.
13457 (make_pass_chkp_opt): Likewise.
13458 (make_pass_chkp_instrumentation_passes): Likewise.
13459 * tree-pretty-print.c (dump_generic_node): Likewise.
13460 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
13461 * tree-ssa-dce.c (propagate_necessity): Likewise.
13462 (eliminate_unnecessary_stmts): Likewise.
13463 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
13464 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
13465 * tree-ssa-sccvn.h: Likewise.
13466 * tree-ssa-strlen.c (get_string_length): Likewise.
13467 (valid_builtin_call): Likewise.
13468 (adjust_last_stmt): Likewise.
13469 (handle_builtin_strchr): Likewise.
13470 (handle_builtin_strcpy): Likewise.
13471 (handle_builtin_stxncpy): Likewise.
13472 (handle_builtin_memcpy): Likewise.
13473 (handle_builtin_strcat): Likewise.
13474 (strlen_check_and_optimize_stmt): Likewise.
13475 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
13476 * tree-streamer-in.c: Likewise.
13477 * tree-streamer.c (record_common_node): Likewise.
13478 * tree.c (tree_code_size): Likewise.
13479 (wide_int_to_tree_1): Likewise.
13480 (type_contains_placeholder_1): Likewise.
13481 (build_common_tree_nodes): Likewise.
13482 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
13483 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
13484 (POINTER_BOUNDS_P): Likewise.
13485 (BOUNDED_TYPE_P): Likewise.
13486 (BOUNDED_P): Likewise.
13487 (CALL_WITH_BOUNDS_P): Likewise.
13488 (pointer_bounds_type_node): Likewise.
13489 * value-prof.c (gimple_ic): Likewise.
13490 * var-tracking.c (vt_add_function_parameters): Likewise.
13491 * varasm.c (make_decl_rtl): Likewise.
13492 (assemble_start_function): Likewise.
13493 (output_constant): Likewise.
13494 (maybe_assemble_visibility): Likewise.
13495 * varpool.c (ctor_for_folding): Likewise.
13496 * chkp-builtins.def: Remove.
13497 * ipa-chkp.c: Remove.
13498 * ipa-chkp.h: Remove.
13499 * rtl-chkp.c: Remove.
13500 * rtl-chkp.h: Remove.
13501 * tree-chkp-opt.c: Remove.
13502 * tree-chkp.c: Remove.
13503 * tree-chkp.h: Remove.
13505 2018-06-07 Carl Love <cel@us.ibm.com>
13507 * config/rs6000/vsx.md (vextract_fp_from_shorth,
13508 vextract_fp_from_shortl): Add BE support.
13510 2018-06-07 Paul Koning <ni1d@arrl.net>
13512 * compare-elim.c (try_merge_compare): Don't merge compare if
13513 address contains a side effect.
13514 (try_eliminate_compare): Likewise.
13516 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
13518 * config.gcc: Support "tremont".
13519 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
13520 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13522 * config/i386/i386.c (m_TREMONT): Define.
13523 (processor_target_table): Add "tremont".
13524 (PTA_TREMONT): Define.
13525 (ix86_lea_outperforms): Add TARGET_TREMONT.
13526 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
13527 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
13528 and M_INTEL_GOLDMONT_PLUS.
13529 (fold_builtin_cpu): Add "tremont".
13530 (ix86_add_stmt_cost): Add TARGET_TREMONT.
13531 (ix86_option_override_internal): Add "tremont".
13532 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
13533 (processor_type): Add PROCESSOR_TREMONT.
13534 * config/i386/x86-tune.def: Add m_TREMONT.
13535 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
13537 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13539 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
13540 symbol defined for msp430i* devices to be lower case.
13542 2018-06-07 Richard Biener <rguenther@suse.de>
13544 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
13545 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
13546 Properly wrap signed arithmetic if overflow wraps.
13548 2018-06-07 Jakub Jelinek <jakub@redhat.com>
13550 PR tree-optimization/69615
13551 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
13552 of a cast from a same precision integral SSA_NAME in a bb dominated
13553 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
13554 cast to utype if rhs2 has already a compatible type.
13556 2018-06-07 Richard Biener <rguenther@suse.de>
13558 PR tree-optimization/85935
13559 * graphite-scop-detection.c (find_params_in_bb): Analyze
13560 condition operands with respect to the correct loop. Assert
13561 the analysis doesn't fail.
13563 2018-06-04 Carl Love <cel@us.ibm.com>
13565 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
13566 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
13567 as it is slightly cheaper.
13568 (first_match_or_eos_index_<mode>):
13569 Calculate index using natural element order.
13570 (first_match_index_<mode>):
13571 Calculate index using natural element order.
13572 (first_match_or_eos_index_<mode>):
13573 Calculate index using natural order.
13574 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
13575 for BE and LE modes.
13576 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
13577 P9V_BUILTIN_VCLZLSBB_V16QI.
13578 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
13581 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
13583 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
13584 indentation and line wrap for many prototypes. Add missing
13585 @smallexample directives around block of prototypes for vec_xl and
13588 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
13590 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
13591 track if we pass or return IEEE 128-bit floating point.
13592 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
13593 C++ mangling that is compatible with GCC 8.1.
13594 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
13595 (init_cumulative_args): Note if we pass or return IEEE 128-bit
13596 floating point types.
13597 (rs6000_function_arg_advance_1): Likewise.
13598 (rs6000_mangle_type): Optionally generate mangled names that match
13599 what GCC 8.1 generated for IEEE 128-bit floating point types.
13600 (rs6000_globalize_decl_name): If we have an external function that
13601 passes or returns IEEE 128-bit types, generate a weak reference
13602 from the mangled name used in GCC 8.1 to the current mangled
13604 (rs6000_init_builtins): Make __ibm128 use the long double type if
13605 long double is IBM extended double. Make __float128 use the long
13606 double type if long double is IEEE 128-bit.
13609 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
13610 macro for __ibm128 built-in functions.
13611 (PACK_IF): Add __ibm128 pack/unpack functions.
13612 (UNPACK_IF): Likewise.
13613 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
13614 enable long double built-in functions if long double is IEEE
13615 128-bit floating point.
13616 (rs6000_invalid_builtin): Update long double built-in function
13618 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
13619 functions, adjust the built-in function to use the long double
13620 built-in function if __ibm128 and long double are the same type.
13621 * doc/extend.texi (PowerPC builtins): Update documention for
13622 __builtin_{,un}pack_longdouble. Add documentation for
13623 __builtin_{,un}pack_ibm128.
13625 2018-06-06 Jim Wilson <jimw@sifive.com>
13627 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
13628 (struct machine_function): New field interrupt_mode.
13629 (riscv_handle_type_attribute): New function. Add forward declaration.
13630 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
13631 (riscv_expand_epilogue): Check interrupt_mode field.
13632 (riscv_set_current_function): Check interrupt attribute args and
13633 set interrupt_mode field.
13634 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
13635 (riscv_sret, riscv_uret): New.
13636 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
13637 new arguments to interrupt attribute.
13639 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
13642 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
13643 Don't handle -mcpu=power8 if -mpower9-vector is also used.
13645 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
13647 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
13648 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
13649 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13650 several redundant entries.
13652 2018-06-06 David Malcolm <dmalcolm@redhat.com>
13654 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
13655 type from "rtx" to "rtx_insn *".
13656 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
13657 for local "call_insn", removing cast.
13658 (ix86_expand_call): Likewise, introducing a "call_insn" local.
13660 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
13662 PR tree-optimization/86066
13663 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
13664 for BIT_INSERT_EXPR stores.
13666 2018-06-06 Richard Biener <rguenther@suse.de>
13668 PR tree-optimization/86062
13669 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13670 component refs ontop
13671 of to be offsetted base.
13673 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13675 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
13676 to be static and remove check on interrupt attribute name.
13678 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
13680 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13681 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
13683 2018-06-05 Steve Ellcey <sellcey@cavium.com>
13686 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
13688 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
13689 Remove second argument, change how error is called.
13690 (aarch64_layout_arg): Remove second argument from
13691 aarch64_err_no_fpadvsimd call.
13692 (aarch64_init_cumulative_args): Ditto.
13693 (aarch64_gimplify_va_arg_expr): Ditto.
13694 * config/aarch64/aarch64.md (mov<mode>): Ditto.
13696 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
13698 * config/i386/i386.md (simple_return_indirect_internal): New expander.
13699 (*simple_return_indirect_internal<mode>): Rename from
13700 simple_return_indirect_internal. Use W mode iterator.
13701 (rstorssp): New expander.
13702 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
13703 (clrssbsy): New expander.
13704 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
13706 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
13708 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
13710 (cmse_check_pointed_object): Likewise.
13712 2018-06-05 Martin Liska <mliska@suse.cz>
13714 PR gcov-profile/47618
13715 * doc/invoke.texi: Document how -fprofile-dir format
13718 2018-06-05 Richard Biener <rguenther@suse.de>
13720 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
13721 removal pretend DOM info isn't available so we do not update
13722 it and only remove edges, not dominated blocks. Actually free
13723 DOM info in case we removed something. Remove unreachable blocks.
13724 (mfb_keep_latches): Work with either DOM info or marked backedges.
13725 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
13726 first. Mark backedges if DOM info isn't available.
13727 (Re-)compute DOM info after cleanup_control_flow_pre.
13729 2018-06-05 Richard Biener <rguenther@suse.de>
13731 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
13732 (locus_discrim_hasher::hash): Adjust.
13733 (locus_discrim_hasher::equal): Likewise.
13734 (next_discriminator_for_locus): Work on line directly.
13735 (same_line_p): Pass in expanded locus1 as well.
13736 (assign_discriminators): Avoid redundant location expansions.
13738 2018-06-05 Richard Biener <rguenther@suse.de>
13740 PR tree-optimization/86046
13741 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
13742 if required after clearing TREE_ADDRESSABLE.
13744 2018-06-05 Richard Biener <rguenther@suse.de>
13746 PR tree-optimization/86047
13747 * tree-ssa-loop.c (for_each_index): Glob handling of all
13748 decls and constants and really handle all of them.
13750 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13753 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
13754 qualifier_void_pointer and qualifier_const_void_pointer.
13755 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
13756 (arm_init_builtins): Handle the above.
13757 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
13758 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
13759 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
13762 2018-06-05 Martin Liska <mliska@suse.cz>
13764 * auto-profile.c (read_autofdo_file): Do not use
13765 gcov_ctr_summary struct.
13766 (afdo_callsite_hot_enough_for_early_inline): Likewise.
13767 * coverage.c (struct counts_entry): Likewise.
13768 (read_counts_file): Read just single summary entry.
13769 (get_coverage_counts): Use gcov_summary struct.
13770 * coverage.h (get_coverage_counts): Likewise.
13771 * gcov-dump.c (dump_working_sets): Likewise.
13772 (tag_summary): Dump just single summary.
13773 * gcov-io.c (gcov_write_summary): Write just histogram
13775 (gcov_read_summary): Read just single summary.
13776 (compute_working_sets): Use gcov_summary struct.
13777 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
13778 of GCOV_COUNTERS_SUMMABLE.
13779 (GCOV_COUNTERS_SUMMABLE): Remove.
13780 (GCOV_FIRST_VALUE_COUNTER): Replace with
13781 GCOV_COUNTER_V_INTERVAL.
13782 (struct gcov_ctr_summary): Remove.
13783 (struct gcov_summary): Directly use fields of former
13785 (compute_working_sets): Use gcov_summary struct.
13786 * gcov.c (read_count_file): Do not use ctrs fields.
13787 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
13789 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
13791 * profile.c: Likewise.
13792 * profile.h: Likewise.
13794 2018-06-05 Martin Liska <mliska@suse.cz>
13796 PR gcov-profile/84846
13797 * gcov.c (output_lines): Print working directory only
13798 in intermediate format.
13800 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
13802 * config/s390/s390-builtin-types.def: Add void function type.
13803 * config/s390/s390-builtins.def: Use the function type for the
13806 2018-06-04 Jim Wilson <jimw@sifive.com>
13808 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
13810 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
13811 and maybe_eh_return. Change regno to unsigned int. Use new args to
13812 handle EH_RETURN_DATA_REGNO registers properly.
13813 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
13814 (riscv_expand_epilogue): Update comment. Change argument name and
13815 type. Update code to use new name and type. Pass new args to
13816 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
13818 * config/riscv/riscv.md (NORMAL_RETURN): New.
13819 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
13820 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
13821 (eh_return): Call gen_eh_return_internal and emit barrier.
13822 (eh_return_internal): Call riscv_expand_epilogue.
13824 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
13826 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
13827 bit_insertion field and declare can_be_merged_into method.
13828 (merged_store_group::can_be_merged_into): New method.
13829 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
13830 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
13831 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
13833 2018-06-04 Richard Biener <rguenther@suse.de>
13835 PR tree-optimization/85955
13836 * builtins.c (fold_builtin_sincos): Convert pointers to
13837 destination to appropriate type before dereferencing.
13839 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
13841 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
13843 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
13845 * expr.c (expand_expr_real_1): Force the operand into memory if
13846 its TYPE_MODE is BLKmode and if there is no integer mode for
13847 the number of bits being extracted.
13849 2018-06-04 Jakub Jelinek <jakub@redhat.com>
13853 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13854 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
13856 2018-06-04 Richard Biener <rguenther@suse.de>
13858 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
13859 (cleanup_tree_cfg_noloop): ... single caller. Do
13860 start_recording_case_labels later.
13862 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
13864 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
13865 to _IMMINTRIN_H_INCLUDED.
13866 * config/i386/pconfigintrin.h: Ditto.
13867 * config/i386/waitpkgintrin.h: Ditto.
13868 * config/i386/immintrin.h: Add includes for sgxintrin.h,
13869 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
13870 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
13871 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
13872 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
13873 waitpkgintrin.h and cldemoteintrin.h.
13875 2018-06-04 Richard Biener <rguenther@suse.de>
13877 PR tree-optimization/86038
13878 * tracer.c (find_best_successor): Check probability for
13879 being initialized, bail out if not.
13881 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
13884 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
13885 of bits to ignore when comparing architectures.
13887 2018-06-04 Jakub Jelinek <jakub@redhat.com>
13889 PR tree-optimization/69615
13890 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
13891 maximum or minimum of the type, try to merge it also as if
13892 range1 is + [-, x - 1] or + [x + 1, -].
13895 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
13897 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
13899 PR tree-optimization/86034
13900 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
13901 the unsigned bitfield type in a bit insertion sequence if it does not
13902 have a larger precision than the bitfield size.
13903 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
13905 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
13907 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
13909 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13911 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
13912 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13913 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13914 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13916 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13918 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13919 Disable -fdelete-null-pointer-checks for ELF toolchain.
13921 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13922 Kito Cheng <kito.cheng@gmail.com>
13924 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13925 (nds32le-*-*, nds32be-*-*): Integrate checking process.
13926 (nds32*-*-*): Add glibc and uclibc conditions.
13927 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13928 (TARGET_EXCEPT_UNWIND_INFO): Define.
13929 * config/nds32/elf.h: New file.
13930 * config/nds32/linux.h: New file.
13931 * config/nds32/nds32-elf.opt: New file.
13932 * config/nds32/nds32-linux.opt: New file.
13933 * config/nds32/nds32-fp-as-gp.c
13934 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13935 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13937 (nds32_asm_file_end): Ditto.
13938 (nds32_print_operand): Ditto.
13939 (nds32_insert_attributes): Ditto.
13940 (nds32_init_libfuncs): New function.
13941 (TARGET_HAVE_TLS): Define.
13942 (TARGET_INIT_LIBFUNCS): Define.
13943 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13945 (TARGET_ELF): Apply different mcmodel setting.
13946 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13947 been migrated into elf.h and linux.h files.
13948 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13949 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13950 (mcmodel): The content has been migrated into nds32-elf.opt and
13951 nds32-linux.opt files.
13952 * config/nds32/t-elf: New file.
13953 * config/nds32/t-linux: New file.
13955 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13956 Shiva Chen <shiva0217@gmail.com>
13958 * config/nds32/constants.md (unspec_volatile_element): Add
13959 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13960 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13962 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13963 (make_pass_nds32_fp_as_gp): Declare.
13964 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13966 (nds32_asm_function_end_prologue): Remove unused asm output.
13967 (nds32_asm_function_begin_epilogue): Remove unused asm output.
13968 (nds32_asm_file_start): Output necessary fp_as_gp information.
13969 (nds32_option_override): Adjust register usage.
13970 (nds32_expand_prologue): Consider fp_as_gp situation.
13971 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13972 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13975 (simple_return): Ditto.
13976 (omit_fp_begin): Output special directive for fp_as_gp.
13977 (omit_fp_end): Output special directive for fp_as_gp.
13978 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13979 mforbid-fp-as-gp): New options.
13981 2018-06-01 Mark Wielaard <mark@klomp.org>
13983 * dwarf2out.c (dwarf2out_finish): Remove generation of
13984 DW_AT_loclists_base.
13986 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
13988 * gimple-ssa-store-merging.c: Include gimple-fold.h.
13989 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13990 (struct merged_store_group): Add bit_insertion field.
13991 (dump_char_array): Use standard hexadecimal format.
13992 (merged_store_group::merged_store_group): Set bit_insertion to false.
13993 (merged_store_group::apply_stores): Use optimal buffer size. Deal
13994 with BIT_INSERT_EXPR stores. Move up code updating the mask and
13995 also print the mask in the dump file.
13996 (pass_store_merging::gate): Minor tweak.
13997 (imm_store_chain_info::coalesce_immediate): Fix wrong association
13998 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
13999 stores with INTEGER_CST stores.
14000 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
14001 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
14002 and use it throughout. Generate bit insertion sequences if need be.
14003 (pass_store_merging::process_store): Remove redundant condition.
14004 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
14006 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
14008 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
14009 the 128-bit floating point types. Fix function comment.
14011 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14013 * config/aarch64/aarch64-simd.md
14014 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
14016 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
14019 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
14021 PR tree-optimization/85989
14022 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
14024 (backprop::intersect_uses): Check it when deciding whether this
14025 is a backedge reference.
14026 (backprop::process_block): Add each phi to m_visited_phis
14027 after visiting it, then clear it at the end.
14029 2018-06-01 Richard Biener <rguenther@suse.de>
14031 * tree-vectorizer.h (vect_dr_stmt): New function.
14032 (vect_get_load_cost): Adjust.
14033 (vect_get_store_cost): Likewise.
14034 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
14035 Use vect_dr_stmt instead of DR_SMTT.
14036 (vect_record_base_alignments): Likewise.
14037 (vect_calculate_target_alignment): Likewise.
14038 (vect_compute_data_ref_alignment): Likewise and make static.
14039 (vect_update_misalignment_for_peel): Likewise.
14040 (vect_verify_datarefs_alignment): Likewise.
14041 (vector_alignment_reachable_p): Likewise.
14042 (vect_get_data_access_cost): Likewise. Pass down
14043 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
14044 (vect_get_peeling_costs_all_drs): Likewise.
14045 (vect_peeling_hash_get_lowest_cost): Likewise.
14046 (vect_enhance_data_refs_alignment): Likewise.
14047 (vect_find_same_alignment_drs): Likewise.
14048 (vect_analyze_data_refs_alignment): Likewise.
14049 (vect_analyze_group_access_1): Likewise.
14050 (vect_analyze_group_access): Likewise.
14051 (vect_analyze_data_ref_access): Likewise.
14052 (vect_analyze_data_ref_accesses): Likewise.
14053 (vect_vfa_segment_size): Likewise.
14054 (vect_small_gap_p): Likewise.
14055 (vectorizable_with_step_bound_p): Likewise.
14056 (vect_prune_runtime_alias_test_list): Likewise.
14057 (vect_analyze_data_refs): Likewise.
14058 (vect_supportable_dr_alignment): Likewise.
14059 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14060 (vect_gen_prolog_loop_niters): Likewise.
14061 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14062 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
14064 (vect_recog_mask_conversion_pattern): Likewise.
14065 (vect_try_gather_scatter_pattern): Likewise.
14066 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
14067 to vect_get_store_cost.
14068 (vect_get_store_cost): Get stmt_info instead of DR.
14069 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
14070 (vect_get_load_cost): Get stmt_info instead of DR.
14072 2018-06-01 Richard Biener <rguenther@suse.de>
14074 PR middle-end/86017
14075 * gimple-fold.c (var_decl_component_p): Also allow offsetted
14076 vars wrapped in MEM_REFs.
14078 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
14080 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14081 Fix subreg tests so that we only return a choice between
14082 GENERAL_REGS and FP_REGS if the original classes included both.
14084 2018-06-01 Richard Biener <rguenther@suse.de>
14087 * tree-ssa-structalias.c (get_function_part_constraint):
14088 Handle NULL fi->decl.
14089 (find_func_aliases_for_call): Properly handle indirect
14090 fi from direct call.
14091 (find_func_clobbers): Likewise.
14092 (ipa_pta_execute): Likewise.
14093 (create_variable_info_for): For functions that are ifunc_resolver
14094 resolve to a varinfo that contains the result of the resolver call.
14095 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
14098 2018-05-31 Michael Collison <michael.collison@arm.com>
14100 * config/aarch64/aarch64.md:
14101 (*fix_to_zero_extenddfdi2): New pattern.
14102 * gcc.target/aarch64/fix_extend1.c: New testcase.
14104 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
14106 PR middle-end/78809
14107 PR middle-end/83026
14108 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14109 and BUILT_IN_STRNCMP_EQ.
14110 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14111 BUILT_IN_STRNCMP_EQ.
14112 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14113 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14114 (gimple_fold_builtin): Likewise.
14115 * tree-ssa-strlen.c (compute_string_length): New function.
14116 (determine_min_obsize): New function.
14117 (handle_builtin_string_cmp): New function to handle calls to
14118 string compare functions.
14119 (strlen_optimize_stmt): Add handling to builtin string compare
14121 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14122 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14123 * tree.c (build_common_builtin_nodes): Add new defines of
14124 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14126 2018-05-31 Jakub Jelinek <jakub@redhat.com>
14129 * bb-reorder.c (pass_partition_blocks::gate): Return false for
14130 functions with naked attribute.
14132 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
14134 * config/i386/sse.md (avx_vec_concat<mode>):
14135 Substitute concat_tg_mode mode attribute with xtg_mode.
14136 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14137 (concat_tg_mode): Remove mode attribute.
14139 2018-05-31 Martin Sebor <msebor@redhat.com>
14142 * calls.c (alloc_max_size): Correct a logic error/typo.
14143 Treat excessive arguments as infinite. Warn for invalid arguments.
14144 * doc/invoke.texi (-Walloc-size-larger-than): Update.
14146 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
14149 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14150 and movx for Haswell.
14152 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
14153 Cesar Philippidis <cesar@codesourcery.com>
14155 PR middle-end/85879
14156 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
14157 when emitting error on private/firstprivate reductions.
14158 * omp-low.c (lower_omp_target): Avoid reference-type processing
14159 on pointers for firstprivate clause.
14161 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
14163 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14166 * config/aarch64/aarch64-simd.md
14167 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14168 (aarch64_ld1_x3_<mode>): Likewise
14169 (aarch64_st1x2<VALLDIF:mode>): Likewise
14170 (aarch64_st1_x2_<mode>): Likewise
14171 (aarch64_st1x3<VALLDIF:mode>): Likewise
14172 (aarch64_st1_x3_<mode>): Likewise
14173 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14174 (vld1_s8_x3): Likewise.
14175 (vld1_u16_x3): Likewise.
14176 (vld1_s16_x3): Likewise.
14177 (vld1_u32_x3): Likewise.
14178 (vld1_s32_x3): Likewise.
14179 (vld1_u64_x3): Likewise.
14180 (vld1_s64_x3): Likewise.
14181 (vld1_f16_x3): Likewise.
14182 (vld1_f32_x3): Likewise.
14183 (vld1_f64_x3): Likewise.
14184 (vld1_p8_x3): Likewise.
14185 (vld1_p16_x3): Likewise.
14186 (vld1_p64_x3): Likewise.
14187 (vld1q_u8_x3): Likewise.
14188 (vld1q_s8_x3): Likewise.
14189 (vld1q_u16_x3): Likewise.
14190 (vld1q_s16_x3): Likewise.
14191 (vld1q_u32_x3): Likewise.
14192 (vld1q_s32_x3): Likewise.
14193 (vld1q_u64_x3): Likewise.
14194 (vld1q_s64_x3): Likewise.
14195 (vld1q_f16_x3): Likewise.
14196 (vld1q_f32_x3): Likewise.
14197 (vld1q_f64_x3): Likewise.
14198 (vld1q_p8_x3): Likewise.
14199 (vld1q_p16_x3): Likewise.
14200 (vld1q_p64_x3): Likewise.
14201 (vst1_s64_x2): Likewise.
14202 (vst1_u64_x2): Likewise.
14203 (vst1_f64_x2): Likewise.
14204 (vst1_s8_x2): Likewise.
14205 (vst1_p8_x2): Likewise.
14206 (vst1_s16_x2): Likewise.
14207 (vst1_p16_x2): Likewise.
14208 (vst1_s32_x2): Likewise.
14209 (vst1_u8_x2): Likewise.
14210 (vst1_u16_x2): Likewise.
14211 (vst1_u32_x2): Likewise.
14212 (vst1_f16_x2): Likewise.
14213 (vst1_f32_x2): Likewise.
14214 (vst1_p64_x2): Likewise.
14215 (vst1q_s8_x2): Likewise.
14216 (vst1q_p8_x2): Likewise.
14217 (vst1q_s16_x2): Likewise.
14218 (vst1q_p16_x2): Likewise.
14219 (vst1q_s32_x2): Likewise.
14220 (vst1q_s64_x2): Likewise.
14221 (vst1q_u8_x2): Likewise.
14222 (vst1q_u16_x2): Likewise.
14223 (vst1q_u32_x2): Likewise.
14224 (vst1q_u64_x2): Likewise.
14225 (vst1q_f16_x2): Likewise.
14226 (vst1q_f32_x2): Likewise.
14227 (vst1q_f64_x2): Likewise.
14228 (vst1q_p64_x2): Likewise.
14229 (vst1_s64_x3): Likewise.
14230 (vst1_u64_x3): Likewise.
14231 (vst1_f64_x3): Likewise.
14232 (vst1_s8_x3): Likewise.
14233 (vst1_p8_x3): Likewise.
14234 (vst1_s16_x3): Likewise.
14235 (vst1_p16_x3): Likewise.
14236 (vst1_s32_x3): Likewise.
14237 (vst1_u8_x3): Likewise.
14238 (vst1_u16_x3): Likewise.
14239 (vst1_u32_x3): Likewise.
14240 (vst1_f16_x3): Likewise.
14241 (vst1_f32_x3): Likewise.
14242 (vst1_p64_x3): Likewise.
14243 (vst1q_s8_x3): Likewise.
14244 (vst1q_p8_x3): Likewise.
14245 (vst1q_s16_x3): Likewise.
14246 (vst1q_p16_x3): Likewise.
14247 (vst1q_s32_x3): Likewise.
14248 (vst1q_s64_x3): Likewise.
14249 (vst1q_u8_x3): Likewise.
14250 (vst1q_u16_x3): Likewise.
14251 (vst1q_u32_x3): Likewise.
14252 (vst1q_u64_x3): Likewise.
14253 (vst1q_f16_x3): Likewise.
14254 (vst1q_f32_x3): Likewise.
14255 (vst1q_f64_x3): Likewise.
14256 (vst1q_p64_x3): Likewise.
14258 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14260 * config/msp430/msp430.c (msp430_output_labelref): Prepend
14261 user_label_prefix to name.
14263 * tree-core.h: Update comment about the format of NAME string
14264 passed to handler in attribute_spec.
14266 * config/msp430/msp430.md: Remove erroneous subreg expression from
14267 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
14268 zero_extend{q,h}isi2.
14270 2018-05-30 Borislav Petkov <bp@suse.de>
14272 * doc/extend.texi: Document some architecture specific
14273 constraints and sort entries.
14275 2018-05-30 Martin Sebor <msebor@redhat.com>
14277 PR middle-end/85369
14278 * builtins.c (expand_builtin_stpcpy_1): New function.
14279 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
14280 only if the former succeeds.
14282 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
14284 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
14287 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14289 * doc/invoke.texi (-flinker-output): Document
14291 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14293 * passes.c (ipa_write_summaries): Only modify statements if body
14295 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
14296 incrementally linking.
14297 (ipa_passes): Likewise.
14298 * lto-cgraph.c (lto_output_node): When incrementally linking do not
14299 pass down resolution info.
14300 * common.opt (flag_incremental_link): Update info.
14301 * gcc.c (plugin specs): Turn flinker-output=* to
14302 -plugin-opt=-linker-output-known
14303 * toplev.c (compile_file): Also cut compilation when doing incremental
14305 * flag-types. (enum lto_partition_model): Add
14306 LTO_LINKER_OUTPUT_NOLTOREL.
14307 (invoke.texi): Add -flinker-output docs.
14308 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
14309 link same way as WPA; do not stream in dead initializers.
14311 * dwarf2out.c (dwarf2out_die_ref_for_decl,
14312 darf2out_register_external_decl): Support incremental link.
14314 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14316 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
14318 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14320 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
14321 it down to simple_object_copy_lto_debug_sections.
14322 (run_gcc): Determine incremental LTO link time and configure
14323 lto1 into non-wpa mode, disable renaming of debug sections.
14325 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
14327 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14328 descriptions of various incorrectly documented functions.
14330 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
14333 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14334 address check not strict.
14336 2018-05-30 Richard Biener <rguenther@suse.de>
14338 PR tree-optimization/85964
14339 * tracer.c (better_p): Drop initialized count check, we only
14340 call the function with initialized counts now.
14341 (find_best_successor): Do find a best edge if one
14342 has uninitialized count.
14343 (find_best_predecessor): Likewise. Do BB frequency check only
14344 if count is initialized.
14346 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
14348 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14349 (aarch64_ldrstr_offset_compare): New.
14350 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
14351 load/store orderings.
14352 (aarch64_gen_adjusted_ldpstp): Likewise.
14354 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
14356 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14357 Check for subset of GENERAL_REGS and FP_REGS.
14358 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14361 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
14363 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
14364 and wi::to_poly_offset. Add the current offset and then check
14365 whether the sum fits, rather than using an unchecked addition of
14366 a checked term. Check for a shwi rather than a uhwi.
14367 * expr.c (get_bit_range): Use tree_to_poly_uint64.
14368 (store_constructor): Use poly_int_tree_p.
14369 (expand_expr_real_1): Likewise.
14370 * function.c (assign_temp): Likewise.
14371 * fold-const.c (const_binop): Use poly_int_tree_p and
14372 wi::to_poly_offset.
14373 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
14375 * ipa-icf-gimple.c (func_checker::compare_operand): Use
14376 to_poly_offset for MEM offsets.
14377 * ipa-icf.c (sem_variable::equals): Likewise.
14378 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
14379 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
14380 wi::to_poly_offset for BIT_FIELD_REF offsets.
14381 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
14382 wi::to_poly_offset.
14383 * var-tracking.c (emit_note_insn_var_location): Use
14384 tree_to_poly_uint64.
14386 2018-05-29 Jim Wilson <jimw@sifive.com>
14388 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
14390 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
14393 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
14394 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
14396 (sse4_1_round<mode>2): Use nonimmediate_operand
14397 for operand 1 predicate.
14399 2018-05-29 Martin Sebor <msebor@redhat.com>
14400 Richard Biener <rguenther@suse.de>
14403 * calls.c (get_size_range): Call determine_value_range instead
14404 of get_value_range..
14405 * tree-vrp.h (determine_value_range): Declared new function.
14406 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
14408 2018-05-29 Richard Biener <rguenther@suse.de>
14410 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
14411 sure to use non-pattern stmts for get_earlier_stmt arguments.
14412 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
14413 called on pattern stmts.
14414 (get_later_stmt): Likewise.
14416 2018-05-29 Martin Liska <mliska@suse.cz>
14418 PR gcov-profile/85759
14419 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
14422 2018-05-29 Jakub Jelinek <jakub@redhat.com>
14424 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
14426 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
14430 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14431 VEC_PACK_FLOAT_EXPR): New tree codes.
14432 * tree-pretty-print.c (op_code_prio): Handle
14433 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14434 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14435 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14436 * tree-inline.c (estimate_operator_cost): Likewise.
14437 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
14438 * fold-const.c (const_binop): Likewise.
14439 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
14440 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14441 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14442 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
14443 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14444 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14445 * expr.c (expand_expr_real_2): Likewise.
14446 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
14447 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
14448 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
14450 * optabs.c (expand_widen_pattern_expr): For
14451 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
14452 sign from result type rather than operand's type.
14453 (expand_binop_directly): For vec_packu_float_optab and
14454 vec_packs_float_optab allow result type to be different from operand's
14456 * optabs-tree.c (optab_for_tree_code): Handle
14457 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14458 VEC_PACK_FLOAT_EXPR. Formatting fixes.
14459 * tree-vect-generic.c (expand_vector_operations_1): Handle
14460 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14461 VEC_PACK_FLOAT_EXPR.
14462 * tree-vect-stmts.c (supportable_widening_operation): Handle
14464 (supportable_narrowing_operation): Handle FLOAT_EXPR.
14465 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
14466 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
14467 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
14468 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
14470 (vec_pack<floatprefix>_float_<mode>): New expander.
14471 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
14473 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
14474 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
14475 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
14476 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
14477 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
14479 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
14480 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
14481 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14482 VEC_PACK_FLOAT_EXPR): Document.
14484 2018-05-29 Richard Biener <rguenther@suse.de>
14486 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
14488 (stmt_vec_info_vec): Make pointer.
14489 (init_stmt_vec_info_vec): Remove.
14490 (free_stmt_vec_info_vec): Likewise.
14491 (set_stmt_vec_info_vec): New function.
14492 (free_stmt_vec_infos): Likewise.
14493 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
14494 (set_vinfo_for_stmt): Likewise.
14495 (get_earlier_stmt): Likewise.
14496 (get_later_stmt): Likewise.
14497 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
14498 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
14499 (vec_info::~vec_info): Free stmt_vec_infos.
14500 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
14501 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
14502 (pass_slp_vectorize::execute): Likewise.
14503 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
14504 (free_stmt_vec_info_vec): Likewise.
14505 (set_stmt_vec_info_vec): New function.
14506 (free_stmt_vec_infos): Likewise.
14507 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
14508 the global stmt_vec_info_vec.
14509 * tree-parloops.c (gather_scalar_reductions): Use
14510 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
14513 2018-05-29 Richard Biener <rguenther@suse.de>
14515 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
14517 2018-05-29 Martin Liska <mliska@suse.cz>
14518 David Malcolm <dmalcolm@redhat.com>
14520 * vec.c (test_reverse): New.
14521 (vec_c_tests): Add new test.
14522 * vec.h (vl_ptr>::reverse): New function.
14524 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
14526 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
14528 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
14531 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14533 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
14535 2018-05-28 Richard Biener <rguenther@suse.de>
14537 PR tree-optimization/85933
14538 * tree-vect-data-refs.c (vect_record_base_alignments): Only
14539 look at stmts marked as vectorizable.
14541 2018-05-28 Richard Biener <rguenther@suse.de>
14543 PR tree-optimization/85934
14544 * tree-vect-generic.c (expand_vector_operations_1): Hoist
14545 vector boolean check before scalar optimization.
14547 2018-05-28 Jakub Jelinek <jakub@redhat.com>
14549 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
14552 2018-05-28 Mark Wielaard <mark@klomp.org>
14554 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
14555 if it is an expression containing a minus sign.
14557 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
14559 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
14561 2018-05-27 Paul Koning <ni1d@arrl.net>
14563 * config/pdp11/pdp11.md (truncsihi2): Remove.
14565 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
14566 Chung-Ju Wu <jasonwucj@gmail.com>
14568 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
14570 (unaligned_store_dw): Ditto.
14571 * config/nds32/nds32-memory-manipulation.c
14572 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
14573 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
14574 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
14575 (emit_setmem_word_loop): Rename to ...
14576 (emit_setmem_doubleword_loop): ... this.
14577 (nds32_gen_dup_4_byte_to_word_value): New function.
14578 (nds32_gen_dup_8_byte_to_double_word_value): New function.
14579 (nds32_expand_setmem_loop): Refine implementation.
14580 (nds32_expand_setmem_loop_v3m): Ditto.
14581 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
14584 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
14586 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
14588 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
14590 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
14591 (nds32_init_machine_status): Initialize machine->attr_naked_p and
14592 machine->attr_no_prologue_p.
14593 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
14594 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
14595 (nds32_expand_epilogue): Consider attr_naked_p.
14596 (nds32_expand_epilogue_v3pop): Likewise.
14597 (nds32_can_use_return_insn): Likewise.
14598 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
14599 attr_no_prologue_p fields.
14600 * config/nds32/nds32.opt (mret-in-naked-func): New option.
14602 2018-05-27 Jakub Jelinek <jakub@redhat.com>
14605 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
14607 * config/i386/sse.md
14608 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
14610 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
14612 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
14614 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
14616 (*<floatsuffix>floatv2div2sf2): Rename to ...
14617 (*float<floatunssuffix>v2div2sf2): ... this.
14618 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
14619 (float<floatunssuffix>v2div2sf2_mask): ... this.
14620 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
14621 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
14622 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
14624 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
14626 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14628 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14630 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14632 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14634 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
14635 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
14636 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
14637 gen_ufix_truncv8dfv8si2.
14638 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
14639 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
14640 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
14641 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
14642 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
14643 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
14644 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
14645 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
14647 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
14651 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
14653 2018-05-25 Jim Wilson <jimw@sifive.com>
14655 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
14656 * config/riscv/riscv.c (struct machine_function): Add
14657 interrupt_handler_p and attribute_checked_p fields.
14658 (riscv_attribute_table): Add interrupt.
14659 (riscv_interrupt_type_p): New.
14660 (riscv_save_reg_p): Save extra regs for interrupt handler.
14661 (riscv_use_save_libcall): Return false for interrupt handler.
14662 (riscv_first_stack_step): Add forward declaration.
14663 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
14664 for interrupt handler with large frame. Use it for saved reg list.
14665 (riscv_expand_prologue): Move flag_stack_usage_info support to
14666 eliminate duplication.
14667 (riscv_expand_epilogue): Generate mret for interrupt handler.
14668 (riscv_epilogue_uses): New.
14669 (riscv_can_use_return_insn): Return false for interrupt handler.
14670 (riscv_function_ok_for_sibcall): Likewise.
14671 (riscv_set_current_function): Add interrupt handler support.
14672 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
14673 * config/riscv/riscv.md (UNSPECV_MRET): New.
14675 (riscv_frflags, riscv_fsflags): Use tab after opcode.
14677 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
14679 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
14681 PR tree-optimization/85712
14682 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
14683 this candidate has already been replaced in-situ by a copy.
14685 2018-05-25 Jason Merrill <jason@redhat.com>
14687 PR c++/80485 - inline function non-zero address.
14688 * symtab.c (nonzero_address): Check DECL_COMDAT.
14690 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
14693 * config/alpha/alpha.md (ashlsi3): New insn pattern.
14694 (*ashlsi_se): Rename from *ashldi_se. Define as sign
14695 extension of SImode operation. Use const123_operand predicate.
14696 (*saddsi_1): Remove.
14697 (*saddl_se_1): Ditto.
14698 (*ssubsi_1): Ditto.
14699 (*ssubl_se_1): Ditto.
14700 * config/alpha/predicates.md (const123_operand): New predicate.
14701 * config/alpha/constraints.md (P): Use IN_RANGE.
14703 2018-05-25 Richard Biener <rguenther@suse.de>
14705 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
14707 (ref_maybe_used_by_stmt_p): Likewise.
14708 (stmt_may_clobber_ref_p): Likewise.
14709 (stmt_may_clobber_ref_p_1): Likewise.
14710 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
14712 (ref_maybe_used_by_stmt_p): Likewise.
14713 (stmt_may_clobber_ref_p): Likewise.
14714 (stmt_may_clobber_ref_p_1): Likewise.
14715 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
14716 the alias oracle to disambiguate DRs with stmts DR analysis
14718 (vect_analyze_data_refs): Do not give up on not analyzable
14719 DRs for BB vectorization. Remove code truncating the dataref
14722 2018-05-25 Jakub Jelinek <jakub@redhat.com>
14725 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14726 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
14727 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
14729 2018-05-25 Richard Biener <rguenther@suse.de>
14731 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
14732 function, combining stmt data ref gathering and fatal analysis
14734 (vect_analyze_data_refs): Remove now redudnant code and simplify.
14735 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
14736 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
14737 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
14738 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
14740 2018-05-25 Bin Cheng <bin.cheng@arm.com>
14742 PR tree-optimization/85720
14743 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
14744 SCC if all partitions are builtins.
14745 (version_loop_by_alias_check): New parameter. Generate cancelable
14746 runtime alias check if all partitions are builtins.
14747 (distribute_loop): Update call to above function.
14749 2018-05-25 Bin Cheng <bin.cheng@arm.com>
14751 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
14752 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
14753 (parm_default_def_partition_arg): Ditto.
14754 (set_parm_default_def_partition): Ditto.
14755 (get_parm_default_def_partitions): Ditto and make it static.
14756 (get_undefined_value_partitions): Ditto and make it static.
14757 (remove_ssa_form): Refactor call to init_var_map here.
14758 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
14759 computation for loop region.
14760 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
14761 (register_default_def): Delete.
14762 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
14763 (parm_default_def_partition_arg): Ditto.
14764 (set_parm_default_def_partition): Ditto.
14765 (get_parm_default_def_partitions): Ditto and make it static.
14766 (get_undefined_value_partitions): Ditto and make it static.
14767 (coalesce_with_default, coalesce_with_default): Update comment.
14768 (create_coalesce_list_for_region): New func factored out from
14769 create_outofssa_var_map.
14770 (populate_coalesce_list_for_outofssa): New func factored out from
14771 create_outofssa_var_map and coalesce_ssa_name.
14772 (create_outofssa_var_map): Delete.
14773 (coalesce_ssa_name): Refactor to support live range computation.
14774 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
14775 (get_parm_default_def_partitions): Delete.
14776 (get_undefined_value_partitions): Ditto.
14777 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
14778 computation for loop region.
14779 (new_tree_live_info, loe_visit_block): Ditto.
14780 (live_worklist, set_var_live_on_entry): Ditto.
14781 (calculate_live_on_exit, verify_live_on_entry): Ditto.
14782 * tree-ssa-live.h (struct _var_map): New fields.
14783 (init_var_map): Change decl.
14784 (region_contains_p): New.
14786 2018-05-25 Bin Cheng <bin.cheng@arm.com>
14788 * tree-ssa-live.h (live_merge_and_clear): Delete.
14790 2018-05-25 Richard Biener <rguenther@suse.de>
14793 * tree-dump.c (dequeue_and_dump): Remove access to removed
14794 operand 2 of a SWITCH_EXPR.
14796 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14798 * doc/sourcebuild.texi (vect_double_cond_arith): Include
14799 multiplication and division.
14800 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
14801 (cond_udiv@var{m}, cond_umod@var{m}): Document.
14802 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
14803 (cond_udiv_optab, cond_umod_optab): New optabs.
14804 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
14805 (IFN_COND_RDIV): New internal functions.
14806 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
14807 TRUNC_MOD_EXPR and RDIV_EXPR.
14808 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
14809 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
14811 (SVE_INT_BINARY): Include mult.
14812 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
14813 (optab, sve_int_op): Handle mult.
14814 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
14816 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
14817 for SVE_INT_BINARY_SD.
14819 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14821 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
14822 (optab, sve_int_op): Handle div and udiv.
14823 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
14824 for SVE_INT_BINARY_SD.
14825 (*<optab><mode>3): New insn for the same.
14827 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14829 * tree-vect-patterns.c: Include predict.h.
14830 (vect_recog_divmod_pattern): Restrict check for division support
14831 to when optimizing for size.
14833 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14835 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
14836 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
14837 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
14838 (gimple_match_op::set_op): Likewise.
14839 (gimple_resimplify4): Declare.
14840 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
14841 (expr::gen_transform): Likewise.
14842 (decision_tree::gen): Generate a simplification routine for 4 operands.
14843 * gimple-match-head.c (gimple_simplify): Add an overload for
14844 4 operands. In the top-level function, handle up to 4 call
14845 arguments and call gimple_resimplify4.
14846 (gimple_resimplify4): New function.
14847 (build_call_internal): Pass a fourth operand.
14848 (maybe_push_to_seq): Likewise.
14849 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
14850 Fold VEC_COND_EXPRs of an operation and a default value into
14851 an IFN_COND_* function if possible.
14852 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
14854 (SVE_COND_FP_BINARY): Include them.
14855 (optab, sve_fp_op): Handle them.
14856 (SVE_INT_BINARY_REV): New code iterator.
14857 (SVE_COND_FP_BINARY_REV): New int iterator.
14858 (commutative): New int attribute.
14859 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
14861 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
14863 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
14864 (*cond_<optab><mode>): New patterns for reversed operands.
14866 2018-05-25 Richard Biener <rguenther@suse.de>
14868 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
14869 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
14870 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
14871 (STMT_VINFO_GROUPED_ACCESS): Adjust.
14872 * tree-vect-data-refs.c (everywhere): Adjust users.
14873 * tree-vect-loop.c (everywhere): Likewise.
14874 * tree-vect-slp.c (everywhere): Likewise.
14875 * tree-vect-stmts.c (everywhere): Likewise.
14876 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14878 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14880 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
14882 (gcc_cv_as_section_exclude): ... this.
14883 Try Solaris as #exclude syntax.
14884 * configure: Regenerate.
14885 * config.in: Regenerate.
14886 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
14888 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
14889 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
14891 * varasm.c (default_elf_asm_named_section): Don't check if
14892 HAVE_GAS_SECTION_EXCLUDE is defined.
14894 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14896 * doc/md.texi: Update the documentation of the cond_* optabs
14897 to mention the new final operand. Fix GET_MODE_NUNITS call.
14898 Describe the scalar case too.
14899 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
14900 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
14902 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
14903 (get_conditional_internal_fn): Update comment.
14904 * tree-vect-loop.c (vectorizable_reduction): Pass the original
14905 accumulator value as a final argument to conditional functions.
14906 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
14907 a define_expand and add an "else" operand. Assert for now that
14908 the else operand is equal to operand 2. Use SVE_INT_BINARY and
14909 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
14910 (*cond_<optab><mode>): New patterns.
14911 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
14912 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14913 (UNSPEC_COND_EOR): Delete.
14914 (optab): Remove associated mappings.
14915 (SVE_INT_BINARY): New code iterator.
14916 (sve_int_op): Remove int attribute and add "minus" to the code
14918 (SVE_COND_INT_OP): Delete.
14919 (SVE_COND_FP_OP): Rename to...
14920 (SVE_COND_FP_BINARY): ...this.
14922 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14924 * optabs.c (can_reuse_operands_p): New function.
14925 (maybe_legitimize_operands): Try to reuse the results for
14928 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14930 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14931 Add {q} suffix to insn mnemonic.
14933 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14935 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14936 (msp430_warn_func_return): New.
14938 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
14940 * fold-const.c (tree_nonzero_bits): New function.
14941 * fold-const.h (tree_nonzero_bits): Likewise.
14942 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14943 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14945 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
14949 * varasm.c (assemble_alias): Check ifunc_resolver only on
14952 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14955 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14956 when memory input operand is handled.
14958 2018-05-24 Luis Machado <luis.machado@linaro.org>
14960 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14962 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14964 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14966 * match.pd: Delay FMA folds until after vectorization.
14968 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
14971 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14972 address check not strict.
14974 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14976 * gimple-match.h (gimple_match_op): New class.
14977 (mprts_hook): Replace parameters with a gimple_match_op *.
14978 (maybe_build_generic_op): Likewise.
14979 (gimple_simplified_result_is_gimple_val): Replace parameters with
14980 a const gimple_match_op *.
14981 (gimple_simplify): Replace code_helper * and tree * parameters with
14982 a gimple_match_op * parameter.
14983 (gimple_resimplify1): Replace code_helper *, tree and tree *
14984 parameters with a gimple_match_op * parameter.
14985 (gimple_resimplify2): Likewise.
14986 (gimple_resimplify3): Likewise.
14987 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14988 parameters with a gimple_match_op * parameter.
14989 * gimple-match-head.c (gimple_simplify): Change prototypes of
14990 auto-generated functions to take a gimple_match_op * instead of
14991 separate code_helper * and tree * parameters. Make the same
14992 change in the top-level overload and update calls to the
14993 gimple_resimplify routines. Update calls to the auto-generated
14994 functions and to maybe_push_res_to_seq in the publicly-facing
14995 operation-specific gimple_simplify overloads.
14996 (gimple_match_op::MAX_NUM_OPS): Define.
14997 (gimple_resimplify1): Replace rcode and ops with a single res_op
14998 parameter. Update call to gimple_simplify.
14999 (gimple_resimplify2): Likewise.
15000 (gimple_resimplify3): Likewise.
15001 (mprts_hook): Replace parameters with a gimple_match_op *.
15002 (maybe_build_generic_op): Likewise.
15003 (build_call_internal): Replace type, nargs and ops with
15004 a gimple_match_op *.
15005 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
15006 with a single gimple_match_op *. Update calls to mprts_hook,
15007 build_call_internal and gimple_simplified_result_is_gimple_val.
15008 Factor out code that is common to the tree_code and combined_fn cases.
15009 * genmatch.c (expr::gen_transform): Replace tem_code and
15010 tem_ops with a gimple_match_op called tem_op. Update calls
15011 to the gimple_resimplify functions and maybe_push_res_to_seq.
15012 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
15013 res_ops. Update call to the gimple_resimplify functions.
15014 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
15015 (decision_tree::gen): Make the functions take a gimple_match_op *
15016 called res_op instead of separate res_code and res_ops parameters.
15017 Update call accordingly.
15018 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
15019 and ops with a single res_op parameter. Update calls to
15020 maybe_build_generic_op and maybe_push_res_to_seq.
15021 (fold_stmt_1): Update calls to gimple_simplify and
15022 replace_stmt_with_simplification.
15023 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
15024 and gimple_simplified_result_is_gimple_val.
15025 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
15027 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
15028 with a gimple_match_op *.
15029 (vn_nary_build_or_lookup): Likewise. Update call to
15030 vn_nary_build_or_lookup_1.
15031 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
15032 gimple_match_op *. Update calls to the gimple_resimplify routines
15033 and to gimple_simplified_result_is_gimple_val.
15034 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
15035 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
15036 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
15037 (visit_nary_op): Likewise.
15038 (visit_reference_op_load): Likewise.
15040 2018-05-23 Luis Machado <luis.machado@linaro.org>
15042 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
15043 modifier for printing the step amount.
15045 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
15048 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
15051 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
15053 * doc/sourcebuild.texi (Endianness): New subsubsection.
15055 2018-05-23 Luis Machado <luis.machado@linaro.org>
15057 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15058 <prefetch_dynamic_strides>: New const bool field.
15059 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15060 prefetch_dynamic_strides.
15061 (exynosm1_prefetch_tune): Likewise.
15062 (thunderxt88_prefetch_tune): Likewise.
15063 (thunderx_prefetch_tune): Likewise.
15064 (thunderx2t99_prefetch_tune): Likewise.
15065 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
15067 (aarch64_override_options_internal): Update to set
15068 PARAM_PREFETCH_DYNAMIC_STRIDES.
15069 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15070 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15071 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15072 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15073 prefetch-dynamic-strides setting.
15075 2018-05-23 Luis Machado <luis.machado@linaro.org>
15077 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15078 <minimum_stride>: New const int field.
15079 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15080 minimum_stride field defaulting to -1.
15081 (exynosm1_prefetch_tune): Likewise.
15082 (thunderxt88_prefetch_tune): Likewise.
15083 (thunderx_prefetch_tune): Likewise.
15084 (thunderx2t99_prefetch_tune): Likewise.
15085 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
15086 <default_opt_level>: Set to 3.
15087 (aarch64_override_options_internal): Update to set
15088 PARAM_PREFETCH_MINIMUM_STRIDE.
15089 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15090 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15091 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15092 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15093 stride is constant and is below the minimum stride threshold.
15095 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15097 * config/arm/arm-cpus.in (mode26): Delete.
15098 (armv4): Delete mode26 reference.
15099 * config/arm/arm.c (arm_configure_build_target): Delete use of
15102 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
15104 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
15106 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15107 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
15109 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
15110 (floatunsdidf2): Ditto.
15112 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
15114 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15115 (fixuns_trunc<mode>si2_avx512f): Ditto.
15116 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15117 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15118 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15120 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
15122 PR rtl-optimization/79985
15123 * df-scan.c (df_insn_refs_collect): Remove special case for
15124 global registers and asm statements.
15126 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
15128 * extend.texi (Global Register Variables): Rewrite the bullet list.
15129 Note that the register is available for allocation. Note that access
15130 via inline asm must use constraints. Add note about async-signal
15131 handlers. Remove paragraph about automagic register selection.
15133 2018-05-23 Richard Biener <rguenther@suse.de>
15135 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15136 of fixed offset from memset VN.
15138 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
15140 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15141 first_interp field.
15142 (alloc_cand_and_find_basis): Initialize first_interp field.
15143 (slsr_process_mul): Modify first_interp field.
15144 (slsr_process_add): Likewise.
15145 (slsr_process_cast): Modify first_interp field for each new
15147 (slsr_process_copy): Likewise.
15148 (dump_candidate): Dump first_interp field.
15149 (replace_mult_candidate): Process all interpretations, not just
15151 (replace_rhs_if_not_dup): Likewise.
15152 (replace_one_candidate): Likewise.
15154 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
15156 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
15158 (aarch64_needs_frame_chain): New function.
15159 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
15161 2018-05-23 Sudakshina Das <sudi.das@arm.com>
15164 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
15165 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15166 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15167 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15168 as true for strict-align.
15169 (aarch64_can_inline_p): Perform checks even when callee has no
15170 attributes to check for strict alignment.
15171 * doc/extend.texi (AArch64 Function Attributes): Document
15173 * doc/invoke.texi: (AArch64 Options): Likewise.
15175 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
15177 PR tree-optimization/85853
15178 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15179 the handling of the root of the node to...
15180 (vect_slp_analyze_node_operations_1): ...this new function,
15181 and run the whole thing with the child nodes' def types
15182 set according to their SLP node's def type.
15184 2018-05-23 Richard Biener <rguenther@suse.de>
15186 PR middle-end/85874
15187 * tree-data-ref.c (create_runtime_alias_checks): Defer
15188 and ignore overflow warnings.
15190 2018-05-23 Yury Gribov <tetra2005@gmail.com>
15192 PR tree-optimization/85822
15193 * tree-vrp.c (is_masked_range_test): Fix handling of negative
15196 2018-05-23 Richard Biener <rguenther@suse.de>
15198 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15199 memset constants via native_interpret_expr.
15201 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
15204 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15206 (cgraph_node::create_alias): Likewise.
15207 (cgraph_node::get_availability): Check ifunc_resolver instead
15208 of looking up ifunc attribute.
15209 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15210 * varasm.c (do_assemble_alias): Likewise.
15211 (assemble_alias): Likewise.
15212 (default_binds_local_p_3): Likewise.
15213 * cgraph.h (cgraph_node): Add ifunc_resolver.
15214 (cgraph_node::only_called_directly_or_aliased_p): Return false
15215 for IFUNC resolver.
15216 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
15218 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
15219 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
15220 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
15221 instead of looking up ifunc attribute.
15223 2018-05-22 Luis Machado <luis.machado@linaro.org>
15225 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
15227 2018-05-22 Martin Sebor <msebor@redhat.com>
15229 PR middle-end/85359
15230 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
15231 only when expasion succeeds.
15232 (expand_builtin_strcmp): Same.
15233 (expand_builtin_strncmp): Same.
15235 2018-05-22 Martin Sebor <msebor@redhat.com>
15237 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
15239 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
15240 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15242 * config/aarch64/aarch64-ldpstp.md: Replace uses of
15243 aarch64_mem_pair_operand with memory_operand and delete operand swapping
15245 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15246 Add check for legitimate_address.
15247 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
15248 (aarch64_swap_ldrstr_operands): New.
15249 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
15252 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
15253 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15255 * config/aarch64/aarch64.md: New patterns to generate stp
15257 (store_pair_sw, store_pair_dw): New patterns to generate stp for
15258 single words and double words.
15259 (load_pair_sw, load_pair_dw): Likewise.
15260 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
15262 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
15264 * config/aarch64/aarch64-ldpstp.md: Modify peephole
15265 for different mode ldpstp and add peephole for merged zero stores.
15266 Likewise for loads.
15267 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15269 (aarch64_gen_store_pair): Rename calls to match new patterns.
15270 (aarch64_gen_load_pair): Rename calls to match new patterns.
15271 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
15272 (load_pair<DREG:mode><DREG2:mode>): ... This.
15273 (store_pair<mode>): Rename to...
15274 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
15275 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
15276 New mode iterators.
15277 (V_INT_EQUIV): Handle SImode.
15278 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
15281 2018-05-22 Martin Sebor <msebor@redhat.com>
15284 * calls.c (maybe_warn_nonstring_arg): Use string length to set
15285 or ajust the presumed bound on an operation to avoid unnecessary
15288 2018-05-22 Martin Sebor <msebor@redhat.com>
15290 PR tree-optimization/85826
15291 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
15292 assuming that a DECL necesarily has a constant size.
15294 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
15296 PR middle-end/85862
15297 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
15299 2018-05-22 Richard Biener <rguenther@suse.de>
15301 PR tree-optimization/85834
15302 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
15303 non-constant and non-zero memset arguments.
15305 2018-05-22 Martin Liska <mliska@suse.cz>
15308 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
15310 2018-05-22 Richard Biener <rguenther@suse.de>
15312 PR tree-optimization/85863
15313 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
15314 comparisons when vectype is specified.
15315 (vectorizable_condition): Do not specify vectype for
15316 vect_is_simple_cond when SLP vectorizing.
15318 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
15321 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
15322 define __ibm128 as long double.
15323 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
15324 as a distinct type when IEEE 128-bit support is enabled.
15325 (init_float128_ieee): Fix up conversions between IFmode and IEEE
15326 128-bit types to use the correct functions.
15327 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
15328 convert between 128-bit floating point types that have different
15329 modes but the same representation, instead of using gen_lowpart to
15331 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
15333 (IFKF_reg): New attributes to give the register constraints for
15335 (extend<mode>tf2_internal): New insns to mark an explicit
15336 conversion between 128-bit floating point types that have a
15337 different mode but share the same representation.
15339 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
15341 PR tree-optimization/85814
15342 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
15343 a null return from get_strinfo when unsharing the next
15344 strinfo in the chain.
15346 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
15349 * varasm.c (weak_finish): Clean up weak_decls.
15351 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15353 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
15354 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
15355 UNSPEC_UADALP values.
15356 * config/aarch64/iterators.md (ABAL): New int iterator.
15359 (sur): Add mappings for the above.
15360 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
15362 (aarch64_<sur>abal<mode>_4): Likewise.
15363 (aarch64_<sur>adalp<mode>_3): Likewise.
15364 (<sur>sadv16qi): New define_expand.
15366 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
15368 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
15369 (*movdf_internal): Ditto.
15370 (*rcpsf2_sse): Ditto.
15371 (*rsqrtsf2_sse): Ditto.
15372 (*sqrt<mode>2_sse): Ditto.
15374 2018-05-21 Tamar Christina <tamar.christina@arm.com>
15376 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
15378 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
15379 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15380 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
15381 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
15383 * config/aarch64/arm_neon.h: Likewise.
15384 * config/aarch64/iterators.md (VQ_I): New.
15386 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
15388 * config.gcc: Add arc/t-multilib-linux to tmake_file for
15390 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
15393 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
15395 * config/nds32/constraints.md (S): New constraint.
15396 * config/nds32/nds32.md (call_internal): Use constraint S.
15397 (call_value_internal): Likewise.
15398 (sibcall_internal): Likewise.
15399 (sibcall_value_internal): Likewise.
15401 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
15402 Chung-Ju Wu <jasonwucj@gmail.com>
15404 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15405 into consideration.
15407 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
15408 Chung-Ju Wu <jasonwucj@gmail.com>
15410 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
15411 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
15412 (nds32_rtx_costs_impl): Simplify.
15413 (nds32_address_cost_impl): Simplify.
15414 (nds32_init_rtx_costs): New function.
15415 (nds32_rtx_costs_speed_prefer): Likewise.
15416 (nds32_rtx_costs_size_prefer): Likewise.
15417 (nds32_address_cost_speed_prefer): Likewise.
15418 (nds32_address_cost_speed_fwprop): Likewise.
15419 (nds32_address_cost_size_prefer): Likewise.
15420 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
15421 * config/nds32/nds32.c (nds32_option_override): Use
15422 nds32_init_rtx_costs function.
15424 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
15426 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
15427 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
15428 (TARGET_PIPELINE_N8): Likewise.
15429 (TARGET_PIPELINE_N10): Likewise.
15430 (TARGET_PIPELINE_N13): Likewise.
15431 (TARGET_PIPELINE_GRAYWOLF): Likewise.
15433 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
15435 * config/nds32/nds32-fpu.md: Update copyright year.
15437 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15439 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
15441 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15443 * config/nds32/nds32.c
15444 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
15445 * config/nds32/nds32.opt (minline-asm-r15): New option.
15447 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15449 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
15451 * config/nds32/nds32.md (abssi2): New pattern.
15453 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
15455 * config/i386/i386.md (rex64namesuffix): New mode attribute.
15456 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
15457 Merge insn pattern from sse_cvtsi2ss<round_name> and
15458 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
15459 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
15460 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
15461 using SWI48 mode iterator.
15462 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
15463 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
15464 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
15465 pattern from sse_cvttss2si<round_saeonly_name>
15466 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
15467 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
15468 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
15469 using SWI48 mode iterator.
15470 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
15471 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
15472 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
15473 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
15474 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
15475 using SWI48 mode iterator.
15476 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
15477 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
15478 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
15479 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
15480 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
15481 SWI48 mode iterator.
15482 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
15483 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
15484 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
15485 pattern from sse_cvttsd2si<round_saeonly_name>
15486 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
15488 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15490 * config/nds32/nds32-md-auxiliary.c
15491 (nds32_valid_smw_lwm_base_p): Refine.
15492 (nds32_output_smw_single_word): Refine.
15493 (nds32_output_smw_double_word): New.
15494 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
15496 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15498 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
15499 (nds32_output_stack_pop): Refine.
15500 (nds32_expand_unaligned_load): Refine.
15501 (nds32_expand_unaligned_store): Refine.
15503 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15504 Chung-Ju Wu <jasonwucj@gmail.com>
15506 * config/nds32/constants.md: Add TP_REGNUM constant.
15507 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
15508 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
15510 * config/nds32/nds32-doubleword.md: Consider flag_pic.
15511 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
15512 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
15513 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
15514 and PIC code generation.
15515 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
15517 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
15519 * config/nds32/nds32.md: Support TLS and PIC.
15520 * config/nds32/nds32.c: Support TLS and PIC.
15521 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
15522 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
15525 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15527 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
15528 mode with E_ prefix.
15530 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15531 Chung-Ju Wu <jasonwucj@gmail.com>
15533 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
15534 * config/nds32/nds32-md-auxiliary.c
15535 (symbolic_reference_mentioned_p): New.
15536 (nds32_legitimize_ict_address): New.
15537 (nds32_expand_ict_move): New.
15538 (nds32_indirect_call_referenced_p): New.
15539 (nds32_symbol_binds_local_p): Delete.
15540 (nds32_long_call_p): Modify.
15541 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
15542 * config/nds32/nds32-protos.h
15543 (symbolic_reference_mentioned_p): Declare.
15544 (nds32_legitimize_ict_address): Declare.
15545 (nds32_expand_ict_move): Declare.
15546 (nds32_indirect_call_referenced_p): Declare.
15547 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
15548 (nds32_relax_group): Use nds32_ict_const_p as condition.
15549 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
15550 (nds32_asm_file_start): Output ict_model directive in asm code.
15551 (nds32_legitimate_address_p): Consider indirect call.
15552 (nds32_print_operand): Consider indirect call.
15553 (nds32_print_operand_address): Consider indirect call.
15554 (nds32_insert_attributes): Handle "indirect_call" attribute.
15555 (TARGET_LEGITIMATE_ADDRESS_P): Define.
15556 (TARGET_LEGITIMATE_CONSTANT_P): Define.
15557 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
15558 (TARGET_DELEGITIMIZE_ADDRESS): Define.
15559 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
15560 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
15561 (TARGET_ICT_MODEL_SMALL): Define.
15562 (TARGET_ICT_MODEL_LARGE): Define.
15563 * config/nds32/nds32.md (movsi): Consider ict model.
15564 (call, call_value): Consider ict model.
15565 (sibcall, sibcall_value): Consider ict model.
15566 * config/nds32/nds32.opt (mict-model): New option.
15567 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
15570 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
15571 Monk Chiang <sh.chiang04@gmail.com>
15572 Jim Wilson <jimw@sifive.com>
15574 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
15575 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
15576 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
15577 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
15578 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
15579 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
15580 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
15581 compute save_libcall_adjustment properly.
15582 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
15583 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
15584 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
15585 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
15586 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
15587 (ABI_SPEC): Handle mabi=ilp32e.
15588 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
15589 (RVE): Add RVE mask.
15590 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
15591 <-march>: Add rv32e as an example.
15593 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
15596 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
15597 (intra_create_variable_infos): Handle C++ constructors.
15599 2018-05-18 Martin Liska <mliska@suse.cz>
15601 * passes.def: Remove a redundant pass.
15603 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
15606 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
15608 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15610 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
15612 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
15614 (armv2, armv2a, armv3, armv3m): Delete architectures.
15615 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
15616 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
15617 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
15619 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
15620 (*mulsidi3adddi): Likewise.
15621 (mulsidi3): Likewise.
15622 (*mulsidi3_nov6): Likewise.
15623 (umulsidi3): Likewise.
15624 (umulsidi3_nov6): Likewise.
15625 (umaddsidi4): Likewise.
15626 (*umulsidi3adddi): Likewise.
15627 (smulsi3_highpart): Likewise.
15628 (*smulsi3_highpart_nov6): Likewise.
15629 (umulsi3_highpart): Likewise.
15630 (*umulsi3_highpart_nov6): Likewise.
15631 * config/arm/arm.h (arm_arch3m): Delete.
15632 * config/arm/arm.c (arm_arch3m): Delete.
15633 (arm_option_override_internal): Update armv3-related comment.
15634 (arm_configure_build_target): Delete use of isa_bit_mode32.
15635 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
15636 (arm_rtx_costs_internal): Delete check of arm_arch3m.
15637 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
15638 (mulsa3): Likewise.
15639 (mulusa3): Likewise.
15640 * config/arm/arm-protos.h (arm_arch3m): Delete.
15641 * config/arm/arm-tables.opt: Regenerate.
15642 * config/arm/arm-tune.md: Likewise.
15643 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
15644 deleted architectures.
15646 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15648 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
15649 (armv5t, armv5te): New features.
15650 (ARMv5, ARMv5e): Delete fgroups.
15651 (ARMv5t, ARMv5te): Adjust for above changes.
15652 (ARMv6m): Likewise.
15653 (armv5, armv5e): Delete arches.
15654 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
15656 (*call_reg_arm): Likewise.
15657 (*call_value_reg_armv5): Likewise.
15658 (*call_value_reg_arm): Likewise.
15659 (*call_symbol): Likewise.
15660 (*call_value_symbol): Likewise.
15661 (*sibcall_insn): Likewise.
15662 (*sibcall_value_insn): Likewise.
15663 (clzsi2): Likewise.
15664 (prefetch): Likewise.
15665 (define_split and define_peephole2 dependent on arm_arch5):
15667 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
15669 (TARGET_ARM_QBIT): Likewise.
15670 (TARGET_DSP_MULTIPLY): Likewise.
15671 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
15672 (arm_arch5, arm_arch5e): Delete.
15673 (arm_arch5t, arm_arch5te): Declare.
15674 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
15675 (arm_arch5t): Declare.
15676 (arm_option_reconfigure_globals): Update for the above.
15677 (arm_options_perform_arch_sanity_checks): Update comment, replace
15678 use of arm_arch5 with arm_arch5t.
15679 (use_return_insn): Likewise.
15680 (arm_emit_call_insn): Likewise.
15681 (output_return_instruction): Likewise.
15682 (arm_final_prescan_insn): Likewise.
15683 (arm_coproc_builtin_available): Likewise.
15684 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
15685 arm_arch5e with arm_arch5t and arm_arch5te.
15686 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
15687 (arm_arch5t, arm_arch5te): Declare.
15688 * config/arm/arm-tables.opt: Regenerate.
15689 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
15690 * config/arm/t-multilib: Likewise.
15691 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
15692 instead of arm_arch5.
15693 (*call_reg_thumb1): Likewise.
15694 (*call_value_reg_thumb1_v5): Likewise.
15695 (*call_value_reg_thumb1): Likewise.
15696 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
15698 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
15700 2018-05-18 Martin Liska <mliska@suse.cz>
15702 PR gcov-profile/84846
15703 * doc/gcov.texi: Document -t option of gcov tool.
15705 2018-05-18 Martin Liska <mliska@suse.cz>
15707 PR gcov-profile/84846
15708 * gcov.c (print_usage): Add new -t option.
15709 (process_args): Handle the option.
15710 (generate_results): Use stdout as output when requested by
15713 2018-05-18 Martin Liska <mliska@suse.cz>
15715 PR gcov-profile/84846
15716 * coverage.c (coverage_init): Write PWD to .gcno file.
15717 * doc/gcov.texi: Document how working directory is printed.
15718 * gcov-dump.c (dump_gcov_file): Print PWD.
15719 * gcov.c (output_intermediate_file): Likewise.
15720 (read_graph_file): Read PWD string.
15721 (output_lines): Print PWD.
15723 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15725 PR middle-end/85817
15726 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
15727 for retval and return false if all args to phi are zero.
15729 2018-05-18 Richard Biener <rguenther@suse.de>
15731 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
15733 (evrp_dom_walker::before_dom_children): Call it.
15735 2018-05-18 Richard Biener <rguenther@suse.de>
15737 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
15738 results when processing array refs with variable index.
15740 2018-05-18 Toon Moene <toon@moene.org>
15742 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
15743 directly after that of -floop-interchange. Indicate that both
15744 options are enabled by default when specifying -O3.
15746 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15748 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
15749 iterator. Delete separate integer-mode vec_set<mode> expander.
15750 (aarch64_simd_vec_setv2di): Delete.
15751 (vec_setv2di): Delete.
15752 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
15753 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
15754 the "w, r" alternative.
15756 2018-05-18 Martin Liska <mliska@suse.cz>
15758 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
15759 * tree-pass.h (make_pass_lower_switch_O0): New function.
15760 * tree-switch-conversion.c (node_has_low_bound): Remove.
15761 (node_has_high_bound): Likewise.
15762 (node_is_bounded): Likewise.
15763 (class pass_lower_switch): Make it a template type and create
15765 (pass_lower_switch::execute): Add template argument.
15766 (make_pass_lower_switch): New function.
15767 (make_pass_lower_switch_O0): New function.
15768 (do_jump_if_equal): Remove.
15769 (emit_case_nodes): Simplify to just handle all 3 cases and leave
15770 all the hard work to tree optimization passes.
15772 2018-05-18 Martin Liska <mliska@suse.cz>
15774 * dbgcnt.c (limit_low): Renamed from limit.
15775 (limit_high): New variable.
15776 (dbg_cnt_is_enabled): Check for upper limit.
15777 (dbg_cnt): Adjust dumping.
15778 (dbg_cnt_set_limit_by_index): Add new argument for high
15780 (dbg_cnt_set_limit_by_name): Likewise.
15781 (dbg_cnt_process_single_pair): Parse new format.
15782 (dbg_cnt_process_opt): Use strtok.
15783 (dbg_cnt_list_all_counters): Remove 'value' and add
15785 * doc/invoke.texi: Document changes.
15787 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
15789 * doc/sourcebuild.texi (scalar_all_fma): Document.
15790 * tree.def (FMA_EXPR): Delete.
15791 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
15792 * internal-fn.c (ternary_direct): New macro.
15793 (expand_ternary_optab_fn): Likewise.
15794 (direct_ternary_optab_supported_p): Likewise.
15795 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
15796 * builtins.c (fold_builtin_fma): Delete.
15797 (fold_builtin_3): Don't call it.
15798 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
15799 * expr.c (expand_expr_real_2): Likewise.
15800 * fold-const.c (operand_equal_p): Likewise.
15801 (fold_ternary_loc): Likewise.
15802 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
15803 * gimple.c (DEFTREECODE): Likewise.
15804 * gimplify.c (gimplify_expr): Likewise.
15805 * optabs-tree.c (optab_for_tree_code): Likewise.
15806 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
15807 * tree-eh.c (operation_could_trap_p): Likewise.
15808 (stmt_could_throw_1_p): Likewise.
15809 * tree-inline.c (estimate_operator_cost): Likewise.
15810 * tree-pretty-print.c (dump_generic_node): Likewise.
15811 (op_code_prio): Likewise.
15812 * tree-ssa-loop-im.c (stmt_cost): Likewise.
15813 * tree-ssa-operands.c (get_expr_operands): Likewise.
15814 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
15815 * fold-const-call.h (fold_fma): Delete.
15816 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
15817 CFN_FNMA and CFN_FNMS.
15818 (fold_fma): Delete.
15819 * genmatch.c (combined_fn): New enum.
15820 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
15821 (commutative_op): New function.
15822 (commutate): Use it. Handle more than 2 operands.
15823 (dt_operand::gen_gimple_expr): Use commutative_op.
15824 (parser::parse_expr): Allow :c to be used with non-binary
15825 operators if the commutative operand is known.
15826 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
15827 CFN_FMS, CFN_FNMA and CFN_FNMS.
15828 (backprop::process_assign_use): Remove FMA_EXPR handling.
15829 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
15830 (gen_hsa_fma): New function.
15831 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
15832 IFN_FNMA and IFN_FNMS.
15833 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
15834 * gimple-fold.h (follow_all_ssa_edges): Declare.
15835 * gimple-fold.c (follow_all_ssa_edges): New function.
15836 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
15837 gimple_build interface and use follow_all_ssa_edges to fold the result.
15838 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
15839 instead of checking for optabs directly.
15840 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
15841 rather than FMA_EXPRs.
15842 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
15843 call to IFN_FMA instead of an FMA_EXPR.
15845 2018-05-17 Jim Wilson <jimw@sifive.com>
15847 * expr.c (do_tablejump): When converting index to Pmode, if we have a
15848 sign extended promoted subreg, and the range does not have the sign bit
15849 set, then do a sign extend.
15851 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
15852 test, check for sign extended subreg and/or constant operands, and
15853 do a sign extend in that case.
15855 2018-05-17 Steve Ellcey <sellcey@cavium.com>
15857 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
15858 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
15860 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
15861 Change logics_shift_reg to logics_shift_imm.
15862 (thunderx2t99_fp_loadpair_basic): Delete.
15863 (thunderx2t99_fp_storepair_basic): Delete.
15864 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
15865 (thunderx2t99_asimd_polynomial): Delete.
15866 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
15867 and neon_fp_mul_d_scalar_q.
15868 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15869 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
15870 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
15871 (thunderx2t99_asimd_lut): Add missing tbl types.
15872 (thunderx2t99_asimd_ext): Delete.
15873 (thunderx2t99_asimd_load1_1_mult): Delete.
15874 (thunderx2t99_asimd_load1_2_mult): Delete.
15875 (thunderx2t99_asimd_load1_ldp): New.
15876 (thunderx2t99_asimd_load1): New.
15877 (thunderx2t99_asimd_load2): Add missing *load2* types.
15878 (thunderx2t99_asimd_load3): New.
15879 (thunderx2t99_asimd_load4): New.
15880 (thunderx2t99_asimd_store1_1_mult): Delete.
15881 (thunderx2t99_asimd_store1_2_mult): Delete.
15882 (thunderx2t99_asimd_store2_mult): Delete.
15883 (thunderx2t99_asimd_store2_onelane): Delete.
15884 (thunderx2t99_asimd_store_stp): New.
15885 (thunderx2t99_asimd_store1): New.
15886 (thunderx2t99_asimd_store2): New.
15887 (thunderx2t99_asimd_store3): New.
15888 (thunderx2t99_asimd_store4): New.
15890 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
15892 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
15893 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
15895 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
15896 Segher Boessenkool <segher@kernel.crashing.org>
15899 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
15902 2018-05-17 Richard Biener <rguenther@suse.de>
15904 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
15905 for pruning loop and prune defs feeding only already visited PHIs.
15907 2018-05-17 Richard Biener <rguenther@suse.de>
15909 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
15911 2018-05-17 Bin Cheng <bin.cheng@arm.com>
15912 Richard Biener <rguenther@suse.de>
15914 PR tree-optimization/85793
15915 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15916 for VMAT_ELEMENTWISE.
15918 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15920 * internal-fn.h (lookup_internal_fn): Declare
15921 * internal-fn.c (lookup_internal_fn): New function.
15922 * gimple.c (gimple_build_call_from_tree): Handle calls to
15923 internal functions.
15924 * gimple-pretty-print.c (dump_gimple_call): Print "." before
15925 internal function names.
15926 * tree-pretty-print.c (dump_generic_node): Likewise.
15927 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15929 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15931 * gimple-fold.h (gimple_build): Make the function forms take
15932 combined_fn rather than built_in_function.
15933 (gimple_simplify): Likewise.
15934 * gimple-match-head.c (gimple_simplify): Likewise.
15935 * gimple-fold.c (gimple_build): Likewise.
15936 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15937 rather than gimple_build_call_internal.
15938 (get_initial_defs_for_reduction): Likewise.
15939 (vect_create_epilog_for_reduction): Likewise.
15940 (vectorizable_live_operation): Likewise.
15942 2018-05-17 Martin Liska <mliska@suse.cz>
15944 * gimple-ssa-sprintf.c (format_directive): Do not use
15945 space in between 'G_' and '('.
15947 2018-05-17 Jakub Jelinek <jakub@redhat.com>
15950 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15951 even if the mask is not all ones.
15954 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15956 (ix86_gimple_fold_builtin): Likewise.
15959 * config/i386/i386.c: Include tree-vector-builder.h.
15960 (ix86_vector_shift_count): New function.
15961 (ix86_fold_builtin): Fold shift builtins by scalar count.
15962 (ix86_gimple_fold_builtin): Likewise.
15964 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15965 _mm512_setzero): New intrinsics.
15967 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
15968 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15970 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15971 code generation for cases where splatting a value is not useful.
15972 * simplify-rtx.c (simplify_ternary_operation): Simplify
15973 vec_merge across a vec_duplicate and a paradoxical subreg forming
15974 a vector mode to a vec_concat.
15976 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
15978 * config.gcc: Support "goldmont-plus".
15979 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15981 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15982 PROCESSOR_GOLDMONT_PLUS.
15983 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15984 (processor_target_table): Add "goldmont-plus".
15985 (PTA_GOLDMONT_PLUS): Define.
15986 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15987 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15988 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15989 (fold_builtin_cpu): Add "goldmont-plus".
15990 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15991 (ix86_option_override_internal): Add "goldmont-plus".
15992 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15993 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15994 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15995 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15997 2018-05-17 Richard Biener <rguenther@suse.de>
15999 PR tree-optimization/85757
16000 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
16001 remove defs that only feed that PHI from further processing.
16003 2018-05-16 Jim Wilson <jimw@sifive.com>
16005 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
16007 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
16009 2018-05-16 Mark Wielaard <mark@klomp.org>
16011 * dwarf2out.c (count_index_strings): New function.
16012 (output_indirect_strings): Call count_index_strings and generate
16013 header for dwarf_version >= 5.
16015 2018-05-16 Mark Wielaard <mark@klomp.org>
16017 * dwarf2out.c (dwarf_FORM): New function.
16018 (set_indirect_string): Use dwarf_FORM.
16019 (reset_indirect_string): Likewise.
16020 (size_of_die): Likewise.
16021 (value_format): Likewise.
16022 (output_die): Likewise.
16023 (add_skeleton_AT_string): Likewise.
16024 (output_macinfo_op): Likewise.
16025 (index_string): Likewise.
16026 (output_index_string_offset): Likewise.
16027 (output_index_string): Likewise.
16028 (count_index_strings): Likewise.
16030 2018-05-16 Carl Love <cel@us.ibm.com>
16032 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16033 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
16035 2018-05-16 Martin Jambor <mjambor@suse.cz>
16037 * ipa-prop.c (ipa_free_all_edge_args): Remove.
16038 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
16040 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
16042 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
16043 (fnma<mode>4): Likewise.
16044 (fms<mode>4): Likewise.
16045 (fnms<mode>4): Likewise.
16046 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
16047 (aarch64_fnma<mode>4): Likewise.
16048 (aarch64_fms<mode>4): Likewise.
16049 (aarch64_fnms<mode>4): Likewise.
16050 (aarch64_fnmadd<mode>4): Likewise.
16052 2018-05-16 Jason Merrill <jason@redhat.com>
16054 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
16056 2018-05-16 Richard Biener <rguenther@suse.de>
16058 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
16059 (dump_stmt_cost): Declare.
16060 (add_stmt_cost): Dump cost we add.
16061 (add_stmt_costs): New function.
16062 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
16063 No longer exported.
16064 (vect_analyze_stmt): Adjust prototype.
16065 (vectorizable_condition): Likewise.
16066 (vectorizable_live_operation): Likewise.
16067 (vectorizable_reduction): Likewise.
16068 (vectorizable_induction): Likewise.
16069 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
16070 cost vector to pass to vectorizable_ and record afterwards.
16071 (vect_model_reduction_cost): Take cost vector argument and adjust.
16072 (vect_model_induction_cost): Likewise.
16073 (vectorizable_reduction): Likewise.
16074 (vectorizable_induction): Likewise.
16075 (vectorizable_live_operation): Likewise.
16076 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
16077 SLP_TREE_NUMBER_OF_VEC_STMTS.
16078 (vect_analyze_slp_cost_1): Remove.
16079 (vect_analyze_slp_cost): Likewise.
16080 (vect_slp_analyze_node_operations): Take visited args and
16081 a target cost vector. Avoid processing already visited stmt sets.
16082 (vect_slp_analyze_operations): Use a local cost vector to gather
16083 costs and register those of non-discarded instances.
16084 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
16085 (vect_schedule_slp_instance): Remove copying of
16086 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
16088 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
16089 adding cost. Record cost entry location.
16090 (vect_prologue_cost_for_slp_op): Function to compute cost of
16091 a constant or invariant generated for SLP vect in the prologue,
16092 split out from vect_analyze_slp_cost_1.
16093 (vect_model_simple_cost): Make static. Adjust for SLP costing.
16094 (vect_model_promotion_demotion_cost): Likewise.
16095 (vect_model_store_cost): Likewise, make static.
16096 (vect_model_load_cost): Likewise.
16097 (vectorizable_bswap): Add cost vector arg and adjust.
16098 (vectorizable_call): Likewise.
16099 (vectorizable_simd_clone_call): Likewise.
16100 (vectorizable_conversion): Likewise.
16101 (vectorizable_assignment): Likewise.
16102 (vectorizable_shift): Likewise.
16103 (vectorizable_operation): Likewise.
16104 (vectorizable_store): Likewise.
16105 (vectorizable_load): Likewise.
16106 (vectorizable_condition): Likewise.
16107 (vectorizable_comparison): Likewise.
16108 (can_vectorize_live_stmts): Likewise.
16109 (vect_analyze_stmt): Likewise.
16110 (vect_transform_stmt): Adjust calls to vectorizable_*.
16111 * tree-vectorizer.c: Include gimple-pretty-print.h.
16112 (dump_stmt_cost): New function.
16114 2018-05-16 Richard Biener <rguenther@suse.de>
16116 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16117 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16118 * tree-ssa-dse.c: Include tree-ssa-loop.h.
16119 (check_name): New callback.
16120 (dse_classify_store): Track cycles via a visited bitmap of PHI
16121 defs and simplify handling of in-loop and across loop dead stores
16122 and properly fail for loop-variant refs. Handle byte-tracking with
16123 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16126 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
16128 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16129 (vect_get_mask_type_for_stmt): Likewise.
16130 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16132 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
16133 to determine the statement's vector type and the vector type that
16134 should be used for calculating nunits. Deal with cases in which
16135 the type has to be deferred.
16136 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16137 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16138 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16139 (vect_determine_vf_for_stmt): New functions, split out from...
16140 (vect_determine_vectorization_factor): ...here.
16141 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16142 (vect_get_mask_type_for_stmt): New functions, split out from
16143 vect_determine_vectorization_factor.
16145 2018-05-16 Richard Biener <rguenther@suse.de>
16147 * tree-cfg.c (verify_gimple_assign_ternary): Properly
16148 verify the [VEC_]COND_EXPR embedded comparison.
16150 2018-05-15 Martin Sebor <msebor@redhat.com>
16152 PR tree-optimization/85753
16153 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
16154 RECORD_TYPE in addition to ARRAY_TYPE.
16156 2018-05-15 Martin Sebor <msebor@redhat.com>
16158 PR middle-end/85643
16159 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
16161 2018-05-15 Richard Biener <rguenther@suse.de>
16163 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
16164 add by_clobber_p one. Change algorithm to collect all defs
16165 representing uses we need to walk and try reducing them to
16166 a single one before failing.
16167 (dse_dom_walker::dse_optimize_stmt): Adjust.
16169 2018-05-13 Mark Wielaard <mark@klomp.org>
16171 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16172 (size_of_loc_descr): Likewise.
16173 (output_loc_operands): Likewise.
16174 (output_loc_operands_raw): Likewise.
16175 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16176 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16177 (hash_loc_operands): Likewise.
16178 (compare_loc_operands): Likewise.
16180 2018-05-14 Mark Wielaard <mark@klomp.org>
16182 * dwarf2out.c (count_index_addrs): New function.
16183 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16185 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16187 PR tree-optimization/83648
16188 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16189 return value as malloc candidate.
16191 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16194 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16195 param as true in call to suggest_attribute.
16197 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
16199 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16200 -mreadonly-in-sdata.
16202 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16204 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16206 (aarch64_crypto_aesd_fused): Likewise.
16208 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
16210 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
16211 (movsi_aarch64): Likewise.
16212 (load_pairsi): Likewise.
16213 (load_pairdi): Likewise.
16214 (store_pairsi): Likewise.
16215 (store_pairdi): Likewise.
16216 (load_pairsf): Likewise.
16217 (load_pairdf): Likewise.
16218 (store_pairsf): Likewise.
16219 (store_pairdf): Likewise.
16220 (zero_extend): Likewise.
16221 (trunc): Swap alternatives.
16222 (fcvt_target): Add '?' to prefer w over r.
16224 2018-05-14 Jakub Jelinek <jakub@redhat.com>
16227 * config/i386/i386.md: Disallow non-commutative arithmetics in
16228 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
16229 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
16230 in the peephole2 before it.
16232 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
16234 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
16235 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
16236 (ix86_handle_option): Handle -mcldemote.
16237 * config.gcc: New header.
16238 * config/i386/cldemoteintrin.h: New file.
16239 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
16240 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16242 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16243 OPTION_MASK_ISA_CLDEMOTE.
16244 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
16245 (ix86_valid_target_attribute_inner_p): Ditto.
16246 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
16247 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
16248 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
16249 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
16250 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
16252 * config/i386/i386.opt: Add -mcldemote.
16253 * config/i386/x86intrin.h: New header.
16254 * doc/invoke.texi: Add -mcldemote.
16256 2018-05-14 Richard Biener <rguenther@suse.de>
16258 * doc/match-and-simplify.texi: Adjust :s documentation.
16260 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
16262 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
16263 intended memcpy size.
16264 (REORDER_45): Likewise.
16266 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
16268 * sort.cc: New file.
16269 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
16270 * vec.c (qsort_chk): Use gcc_qsort.
16271 * Makefile.in (OBJS-libcommon): Add sort.o.
16272 (build/sort.o): New target. Use it...
16273 (BUILD_RTL): ... here, and...
16274 (build/gencfn-macros): ... here, and...
16275 (build/genmatch): ... here.
16277 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16278 Chung-Ju Wu <jasonwucj@gmail.com>
16280 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
16281 * config/nds32/nds32-graywolf.md: New file.
16282 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
16283 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
16285 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
16286 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
16287 * config/nds32/nds32.md (pipeline_model): Add graywolf.
16288 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
16289 * config/nds32/pipelines.md: Include n15 settings.
16291 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16292 Chung-Ju Wu <jasonwucj@gmail.com>
16294 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
16295 * config/nds32/nds32-n13.md: New file.
16296 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
16297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
16299 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
16300 * config/nds32/nds32.md (pipeline_model): Add n13.
16301 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
16302 * config/nds32/pipelines.md: Include n13 settings.
16304 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16305 Chung-Ju Wu <jasonwucj@gmail.com>
16307 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
16308 * config/nds32/nds32-n10.md: New file.
16309 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
16310 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
16312 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
16313 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
16314 * config/nds32/nds32.md (pipeline_model): Add n10.
16315 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
16316 * config/nds32/pipelines.md: Include n10 settings.
16318 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
16319 Kito Cheng <kito.cheng@gmail.com>
16320 Chung-Ju Wu <jasonwucj@gmail.com>
16322 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
16323 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16324 Add enum values for DSP extension instructions.
16325 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
16327 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
16328 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
16329 New code iterators.
16330 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
16331 * config/nds32/nds32-dspext.md: New file for DSP implementation.
16332 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
16333 * config/nds32/nds32-intrinsic.md: Likewise.
16334 * config/nds32/nds32_intrinsic.h: Likewise.
16335 * config/nds32/nds32-md-auxiliary.c: Likewise.
16336 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
16337 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
16338 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
16339 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
16340 * config/nds32/nds32-protos.h: New declarations for DSP extension.
16341 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
16342 TYPE_DMAC in switch statement.
16343 * config/nds32/nds32.c: New checking and implementation for DSP
16344 extension instructions.
16345 * config/nds32/nds32.h: Likewise.
16346 * config/nds32/nds32.md: Likewise.
16347 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
16348 * config/nds32/predicates.md: Implement new predicates for DSP
16351 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16353 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
16354 Reformat alternatives and attributes so it is easier to identify
16355 which constraints/attributes go with which instruction.
16356 (mov<mode>_hardfloat32, FMOVE64): Likewise.
16357 (mov<mode>_softfloat32, FMOVE64): Likewise.
16358 (mov<mode>_hardfloat64, FMOVE64): Likewise.
16359 (mov<mode>_softfloat64, FMOVE64): Likewise.
16361 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
16363 * doc/extend.texi (PowerPC Built-in Functions): Rename this
16365 (Basic PowerPC Built-in Functions): The new name of the
16366 subsection previously known as "PowerPC Built-in Functions".
16367 (Basic PowerPC Built-in Functions Available on all Configurations):
16369 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
16370 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
16371 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
16372 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
16374 2018-05-11 Martin Jambor <mjambor@suse.cz>
16377 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
16380 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
16383 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
16385 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
16387 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
16388 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
16389 (ix86_handle_option): Handle -mwaitpkg.
16390 * config.gcc: New header.
16391 * config/i386/cpuid.h (bit_WAITPKG): New bit.
16392 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
16393 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
16395 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16396 OPTION_MASK_ISA_WAITPKG.
16397 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
16398 (ix86_option_override_internal): Add PTA_WAITPKG.
16399 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
16400 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
16401 IX86_BUILTIN_TPAUSE.
16402 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
16403 __builtin_ia32_umwait and __builtin_ia32_tpause.
16404 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
16405 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
16406 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
16407 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
16408 UNSPECV_TPAUSE): New.
16409 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
16410 * config/i386/i386.opt: Add -mwaitpkg.
16411 * config/i386/waitpkgintrin.h: New file.
16412 * config/i386/x86intrin.h: New header.
16413 * doc/invoke.texi: Add -mwaitpkg.
16415 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
16418 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
16420 (cortex-m0): Use armv6s-m isa.
16421 (cortex-m0plus): Likewise.
16422 (cortex-m1): Likewise.
16423 (cortex-m0.small-multiply): Likewise.
16424 (cortex-m0plus.small-multiply): Likewise.
16425 (cortex-m1.small-multiply): Likewise.
16427 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
16428 Jakub Jelinek <jakub@redhat.com>
16430 PR tree-optimization/85692
16431 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
16432 source permute as well.
16434 2018-05-11 Martin Liska <mliska@suse.cz>
16437 * doc/extend.texi: Document LLVM style format for no_sanitize
16440 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16442 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
16443 mode_supports_vsx_dform_quad to mode_supports_dq_form.
16444 (mode_supports_vsx_dform_quad): Likewise.
16445 (mode_supports_vmx_dform): Move these functions to be next to the
16446 other mode_supports functions.
16447 (mode_supports_dq_form): Likewise.
16448 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
16449 mode_supports_dq_form.
16450 (reg_offset_addressing_ok_p): Likewise.
16451 (offsettable_ok_by_alignment): Likewise.
16452 (rs6000_legitimate_offset_address_p): Likewise.
16453 (legitimate_lo_sum_address_p): Likewise.
16454 (rs6000_legitimize_address): Likewise.
16455 (rs6000_legitimize_reload_address): Likewise.
16456 (rs6000_secondary_reload_inner): Likewise.
16457 (rs6000_preferred_reload_class): Likewise.
16458 (rs6000_output_move_128bit): Likewise.
16460 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
16462 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
16463 Generate SImode target register for null target.
16464 <case IX86_BUILTIN_XGETBV>: Ditto.
16465 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
16466 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
16468 2018-05-10 Carl Love <cel@us.ibm.com>
16470 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16471 dcbtt and dcbtstt if operands[2] is 0.
16473 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
16476 * config/i386/sse.md (usadv64qi): New expander.
16478 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
16480 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
16481 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
16482 -maltivec=be support.
16483 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
16484 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
16485 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
16486 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
16487 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
16488 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
16489 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
16490 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
16491 altivec_vsumsws): Adjust.
16492 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
16493 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
16494 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
16496 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
16497 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
16498 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
16499 (altivec_lve<VI_char>x): Delete expand.
16500 (*altivec_lve<VI_char>x_internal): Rename to...
16501 (altivec_lve<VI_char>x): ... this.
16502 (altivec_lvxl_<mode>): Delete expand.
16503 (*altivec_lvxl_<mode>_internal): Rename to ...
16504 (altivec_lvxl_<mode>): ... this.
16505 (altivec_stvxl_<mode>): Delete expand.
16506 (*altivec_stvxl_<mode>_internal): Rename to ...
16507 (altivec_stvxl_<mode>): ... this.
16508 (altivec_stve<VI_char>x): Delete expand.
16509 (*altivec_stve<VI_char>x_internal): Rename to ...
16510 (altivec_stve<VI_char>x): ... this.
16511 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
16512 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
16513 reduc_plus_scal_<mode>): Adjust.
16514 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
16516 (rs6000_cpu_cpp_builtins): Adjust.
16517 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
16518 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
16519 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
16520 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
16521 -maltivec=be support.
16522 (rs6000_split_vec_extract_var): Adjust.
16523 (rs6000_split_v4si_init): Adjust.
16524 (swap_selector_for_mode): Delete.
16525 (altivec_expand_lvx_be, altivec_expand_stvx_be,
16526 altivec_expand_stvex_be): Delete.
16527 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
16528 -maltivec=be support.
16529 (rs6000_gimple_fold_builtin): Ditto.
16530 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
16532 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
16533 (TARGET_DIRECT_MOVE_64BIT): Adjust.
16534 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
16535 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
16536 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
16537 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
16538 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
16539 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
16540 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
16541 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
16542 anonymous split): Adjust.
16543 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
16544 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
16546 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
16548 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
16549 when --with-gxx-include-dir is also specified.
16550 * configure: Regenerate.
16552 2018-05-09 Jim Wilson <jimw@sifive.com>
16555 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
16556 * config/riscv/t-withmultilib: New.
16557 * config/riscv/withmultilib.h: New.
16558 * doc/install.texi: Document RISC-V --with-multilib-list support.
16560 2018-05-09 Richard Biener <rguenther@suse.de>
16562 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
16564 (vect_bb_vectorization_profitable_p): Adjust. Compute
16565 actual scalar cost using the cost vector and the add_stmt_cost
16568 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16570 PR rtl-optimization/85645
16571 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
16572 in the REG_CFA_REGISTER note for LR, don't leave it empty.
16574 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16576 PR rtl-optimization/85645
16577 * shrink-wrap.c (spread_components): Return a boolean saying if
16578 anything was changed.
16579 (try_shrink_wrapping_separate): Iterate spread_components until
16580 nothing changes anymore.
16582 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16584 PR rtl-optimization/85645
16585 * regrename.c (build_def_use): Also kill the chains that include the
16586 destination of a REG_CFA_REGISTER note.
16588 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16590 PR rtl-optimization/85645
16591 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
16592 insn that has a REG_CFA_REGISTER note.
16594 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
16596 * cfgexpand.c (expand_clobber): New function.
16597 (expand_gimple_stmt_1): Use it.
16598 * tree-vect-stmts.c (vect_clobber_variable): New function,
16600 (vectorizable_simd_clone_call): ...here.
16601 (vectorizable_store): Emit a clobber either side of an
16602 IFN_STORE_LANES sequence.
16603 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
16605 2018-05-09 Tom de Vries <tom@codesourcery.com>
16608 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
16609 (define_insn "trap_if_false"): Add exit after trap.
16611 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
16613 PR rtl-optimization/85638
16614 * bb-reorder.c: Include common/common-target.h.
16615 (create_forwarder_block): New function extracted from...
16616 (fix_up_crossing_landing_pad): ...here. Rename into...
16617 (dw2_fix_up_crossing_landing_pad): ...this.
16618 (sjlj_fix_up_crossing_landing_pad): New function.
16619 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
16620 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
16621 from both partitions and exit the loop after one iteration.
16623 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
16626 * doc/extend.texi (PowerPC Built-in Functions): Rename this
16628 (Basic PowerPC Built-in Functions): The new name of the
16629 subsection previously known as "PowerPC Built-in Functions".
16630 (Basic PowerPC Built-in Functions Available on all Configurations):
16632 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16634 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16636 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16638 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16641 2018-05-08 Jim Wilson <jimw@sifive.com>
16643 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
16644 (LD_EMUL_SUFFIX): New.
16645 (LINK_SPEC): Use it.
16647 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
16649 * doc/extend.texi (PowerPC Built-in Functions): Rename this
16651 (Basic PowerPC Built-in Functions): The new name of the
16652 subsection previously known as "PowerPC Built-in Functions".
16653 (Basic PowerPC Built-in Functions Available on all Configurations):
16655 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16657 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16659 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16661 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16664 2018-05-08 Jakub Jelinek <jakub@redhat.com>
16667 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
16668 after cmpelim optimization.
16670 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
16672 * config.gcc: Support "goldmont".
16673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
16674 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16675 PROCESSOR_GOLDMONT.
16676 * config/i386/i386.c (m_GOLDMONT): Define.
16677 (processor_target_table): Add "goldmont".
16678 (PTA_GOLDMONT): Define.
16679 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
16680 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
16681 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
16682 (fold_builtin_cpu): Add "goldmont".
16683 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
16684 (ix86_option_override_internal): Add "goldmont".
16685 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
16686 (processor_type): Add PROCESSOR_GOLDMONT.
16687 * config/i386/i386.md: Add CPU "glm".
16688 * config/i386/glm.md: New file.
16689 * config/i386/x86-tune.def: Add m_GOLDMONT.
16690 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
16692 2018-05-08 Jakub Jelinek <jakub@redhat.com>
16695 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
16697 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
16698 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
16699 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
16702 * config/i386/i386.c (ix86_fold_builtin): Handle
16703 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
16706 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
16707 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
16709 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
16712 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
16713 (check_arch): Likewise.
16714 (check_fpu): Return the result rather than printing it.
16715 (end arch): Fix operator precedence.
16716 (end cpu): Likewise.
16717 (END): Print the result from check_fpu.
16719 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
16720 Alan Hayward <alan.hayward@arm.com>
16721 David Sherwood <david.sherwood@arm.com>
16723 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
16724 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
16725 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
16726 (*fcmuo<mode>_and): New patterns.
16728 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
16730 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
16731 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
16732 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
16733 (cmp_op, sve_imm_con): New code attributes.
16734 (SVE_COND_INT_CMP, imm_con): Delete.
16735 (cmp_op): Remove above unspecs from int attribute.
16736 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
16738 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
16739 comparison-specific unspecs.
16740 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
16741 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
16742 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
16743 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
16744 (*vec_fcm<cmp_op><mode>): Rename to...
16745 (*fcm<cmp_op><mode>): ...this and adjust likewise.
16746 (*vec_fcmuo<mode>): Rename to...
16747 (*fcmuo<mode>): ...this and adjust likewise.
16748 (*pred_fcm<cmp_op><mode>): New pattern.
16749 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
16750 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
16752 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
16754 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
16755 (aarch64_emit_sve_predicated_cond): New function.
16756 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
16757 (aarch64_emit_unspec_cond_or): Replace with...
16758 (aarch64_emit_sve_or_conds): ...this new function. Use
16759 aarch64_emit_sve_ptrue_op for the individual comparisons and
16760 aarch64_emit_binop to OR them together.
16761 (aarch64_emit_inverted_unspec_cond): Replace with...
16762 (aarch64_emit_sve_inverted_cond): ...this new function. Use
16763 aarch64_emit_sve_ptrue_op for the comparison and
16764 aarch64_emit_unop to invert the result.
16765 (aarch64_expand_sve_vec_cmp_float): Update after the above
16766 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
16768 2018-05-07 Nathan Sidwell <nathan@acm.org>
16770 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
16771 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
16772 (Backwards Compatibility): Likewise.
16774 2018-05-07 Luis Machado <luis.machado@linaro.org>
16778 2018-05-07 Luis Machado <luis.machado@linaro.org>
16780 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16781 <prefetch_dynamic_strides>: New const bool field.
16782 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16783 prefetch_dynamic_strides.
16784 (exynosm1_prefetch_tune): Likewise.
16785 (thunderxt88_prefetch_tune): Likewise.
16786 (thunderx_prefetch_tune): Likewise.
16787 (thunderx2t99_prefetch_tune): Likewise.
16788 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16790 (aarch64_override_options_internal): Update to set
16791 PARAM_PREFETCH_DYNAMIC_STRIDES.
16792 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16793 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16794 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16795 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16796 prefetch-dynamic-strides setting.
16798 2018-05-07 Luis Machado <luis.machado@linaro.org>
16800 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16801 <minimum_stride>: New const int field.
16802 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16803 minimum_stride field.
16804 (exynosm1_prefetch_tune): Likewise.
16805 (thunderxt88_prefetch_tune): Likewise.
16806 (thunderx_prefetch_tune): Likewise.
16807 (thunderx2t99_prefetch_tune): Likewise.
16808 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16809 (aarch64_override_options_internal): Update to set
16810 PARAM_PREFETCH_MINIMUM_STRIDE.
16811 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16812 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16813 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16814 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16815 stride is constant and is below the minimum stride threshold.
16817 2018-05-07 Luis Machado <luis.machado@linaro.org>
16819 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
16822 2018-05-07 Luis Machado <luis.machado@linaro.org>
16824 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16825 <prefetch_dynamic_strides>: New const bool field.
16826 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16827 prefetch_dynamic_strides.
16828 (exynosm1_prefetch_tune): Likewise.
16829 (thunderxt88_prefetch_tune): Likewise.
16830 (thunderx_prefetch_tune): Likewise.
16831 (thunderx2t99_prefetch_tune): Likewise.
16832 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16834 (aarch64_override_options_internal): Update to set
16835 PARAM_PREFETCH_DYNAMIC_STRIDES.
16836 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16837 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16838 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16839 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16840 prefetch-dynamic-strides setting.
16842 2018-05-07 Luis Machado <luis.machado@linaro.org>
16844 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16845 <minimum_stride>: New const int field.
16846 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16847 minimum_stride field.
16848 (exynosm1_prefetch_tune): Likewise.
16849 (thunderxt88_prefetch_tune): Likewise.
16850 (thunderx_prefetch_tune): Likewise.
16851 (thunderx2t99_prefetch_tune): Likewise.
16852 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16853 (aarch64_override_options_internal): Update to set
16854 PARAM_PREFETCH_MINIMUM_STRIDE.
16855 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16856 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16857 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16858 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16859 stride is constant and is below the minimum stride threshold.
16861 2018-05-06 Jakub Jelinek <jakub@redhat.com>
16864 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
16865 the type is addressable. Don't force op into register if it has
16868 2018-05-05 Roland McGrath <mcgrathr@google.com>
16871 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
16872 any section for which we don't know a specific type it should have,
16873 regardless of name. Previously this was done only for the exact
16874 names ".init_array", ".fini_array", and ".preinit_array".
16875 (default_elf_asm_named_section): Add comment about
16876 relationship with default_section_type_flags and SECTION_NOTYPE.
16877 (get_section): Don't consider it a type conflict if one side has
16878 SECTION_NOTYPE and the other doesn't, as long as neither has the
16879 SECTION_BSS et al used in the default_section_type_flags logic.
16881 2018-05-05 Tom de Vries <tom@codesourcery.com>
16884 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
16885 (workaround_barsyncs): New function.
16886 (nvptx_reorg): Use workaround_barsyncs.
16887 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16888 (define_expand "nvptx_membar_cta"): New define_expand.
16889 (define_insn "*nvptx_membar_cta"): New insn.
16891 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
16893 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
16894 To improve optimization opportunities.
16895 * builtin-types.def: The new needed builtin types for the above.
16897 2018-05-04 Richard Biener <rguenther@suse.de>
16899 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
16900 * gimple-ssa-store-merging.c
16901 (imm_store_chain_info::output_merged_store): Remove redundant create,
16902 release split_store vector contents on failure.
16903 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
16904 scalar stmt vector on cache hit.
16906 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16908 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
16910 * config.gcc (powerpc-xilinx-eabi*): Remove.
16911 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16913 (fusion_addis_mem_combo_load): Ditto.
16914 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16916 (rs6000_cpu_cpp_builtins): Ditto.
16917 * config/rs6000/rs6000-linux.c
16918 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16919 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16920 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16922 (rs6000_setup_reg_addr_masks): Ditto.
16923 (rs6000_init_hard_regno_mode_ok): Ditto.
16924 (rs6000_option_override_internal): Ditto.
16925 (legitimate_lo_sum_address_p): Ditto.
16926 (rs6000_legitimize_address): Ditto.
16927 (rs6000_legitimize_reload_address): Ditto.
16928 (rs6000_legitimate_address_p): Ditto.
16929 (abi_v4_pass_in_fpr): Ditto.
16930 (setup_incoming_varargs): Ditto.
16931 (rs6000_gimplify_va_arg): Ditto.
16932 (rs6000_split_multireg_move): Ditto.
16933 (rs6000_savres_strategy): Ditto.
16934 (rs6000_emit_prologue_components): Ditto.
16935 (rs6000_emit_epilogue_components): Ditto.
16936 (rs6000_emit_prologue): Ditto.
16937 (rs6000_emit_epilogue): Ditto.
16938 (rs6000_elf_file_end): Ditto.
16939 (rs6000_function_value): Ditto.
16940 (rs6000_libcall_value): Ditto.
16941 * config/rs6000/rs6000.h: Ditto.
16942 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16943 (TARGET_MINMAX): ... this. New.
16944 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16945 * config/rs6000/rs6000.md: Remove Xilinx FP support.
16946 (*movsi_internal1_single): Delete.
16947 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16948 mfpu=, mxilinx-fpu): Delete.
16949 * config/rs6000/singlefp.h: Delete.
16950 * config/rs6000/sysv4.h: Remove Xilinx FP support.
16951 * config/rs6000/t-rs6000: Ditto.
16952 * config/rs6000/t-xilinx: Delete.
16953 * config/rs6000/titan.md: Adjust for fp_type removal.
16954 * config/rs6000/vsx.md: Remove Xilinx FP support.
16955 (VStype_simple): Delete.
16956 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16957 * config/rs6000/xfpu.h: Delete.
16958 * config/rs6000/xfpu.md: Delete.
16959 * config/rs6000/xilinx.h: Delete.
16960 * config/rs6000/xilinx.opt: Delete.
16961 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16962 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16964 2018-05-04 Tom de Vries <tom@codesourcery.com>
16967 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16970 2018-05-04 Richard Biener <rguenther@suse.de>
16972 PR middle-end/85627
16973 * tree-complex.c (update_complex_assignment): We are always in SSA form.
16974 (expand_complex_div_wide): Likewise.
16975 (expand_complex_operations_1): Likewise.
16976 (expand_complex_libcall): Preserve EH info of the original stmt.
16977 (tree_lower_complex): Handle removed blocks.
16978 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16979 on complex multiplication and division libcall builtins.
16981 2018-05-04 Richard Biener <rguenther@suse.de>
16983 PR middle-end/85574
16984 * fold-const.c (negate_expr_p): Restrict negation of operand
16985 zero of a division to when we know that can happen without
16987 (fold_negate_expr_1): Likewise.
16989 2018-05-04 Jakub Jelinek <jakub@redhat.com>
16992 * real.h (real_nextafter): Declare.
16993 * real.c (real_nextafter): New function.
16994 * fold-const-call.c (fold_const_nextafter): New function.
16995 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16996 CASE_CFN_NEXTTOWARD.
16997 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16998 even when arg1_mode is different from arg0_mode.
17000 2018-05-03 Nathan Sidwell <nathan@acm.org>
17002 * doc/extend.texi (Deprecated Features): Remove
17003 -ffriend-injection.
17004 (Backwards Compatibility): Likewise.
17005 * doc/invoke.texi (C++ Language Options): Likewise.
17006 (C++ Dialect Options): Likewise.
17008 2018-05-03 Jakub Jelinek <jakub@redhat.com>
17011 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
17012 _mm512_mask_mullox_epi64): New intrinsics.
17014 2018-05-03 Tom de Vries <tom@codesourcery.com>
17017 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17018 dump files): Add offload-tree.
17020 2018-05-03 Richard Biener <rguenther@suse.de>
17022 PR tree-optimization/85615
17023 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17024 to loops not nested in BBs loop father to avoid creating multi-entry
17027 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17029 PR tree-optimization/70291
17030 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
17031 arguments. Change return type to tree. Emit libcall as a new
17032 statement rather than replacing existing one when inplace_p is true.
17033 (expand_complex_multiplication_components): New function.
17034 (expand_complex_multiplication): Expand floating-point complex
17035 multiplication using the above.
17036 (expand_complex_division): Rename inner_type parameter to type.
17037 Update expand_complex_libcall call-site.
17038 (expand_complex_operations_1): Update expand_complex_multiplication
17039 and expand_complex_division call-sites.
17041 2018-05-02 Jakub Jelinek <jakub@redhat.com>
17044 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17045 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17046 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
17047 the highest significant bit of the shift count mask is clear. In
17048 check whether and[sq]i3 is needed verify that all significant bits
17049 of the shift count other than the highest are set.
17051 2018-05-02 Tom de Vries <tom@codesourcery.com>
17054 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
17055 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
17056 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
17057 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
17058 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
17059 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
17060 __builtin_goacc_parlevel_size.
17062 2018-05-02 Richard Biener <rguenther@suse.de>
17064 PR tree-optimization/85597
17065 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
17066 do not use split vect_get_vec_defs call but call vect_get_slp_defs
17069 2018-05-02 Tom de Vries <tom@codesourcery.com>
17072 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17073 dump files): Add ltrans-tree.
17075 2018-05-02 Tom de Vries <tom@codesourcery.com>
17078 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17079 dump files): Add wpa-ipa.
17081 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
17083 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
17084 powerpc*-*-linux*paired* target.
17085 * config/rs6000/750cl.h: Delete.
17086 * config/rs6000/paired.h: Delete.
17087 * config/rs6000/paired.md: Delete.
17088 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
17090 * config/rs6000/rs6000-builtin.def: Remove paired float support.
17091 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
17092 comment. Remove paired float support.
17093 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
17094 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
17096 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
17097 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
17099 * config/rs6000/rs6000.c: Remove paired float support.
17100 (paired_expand_vector_init, paired_expand_vector_move,
17101 paired_emit_vector_compare, paired_emit_vector_cond_expr,
17102 (paired_expand_lv_builtin, paired_expand_stv_builtin,
17103 paired_expand_builtin, paired_expand_predicate_builtin,
17104 paired_init_builtins): Delete.
17105 * config/rs6000/rs6000.h: Remove paired float support.
17106 * config/rs6000/rs6000.md: Remove paired float support.
17107 (move_from_CR_ov_bit): Delete.
17108 * config/rs6000/rs6000.opt (mpaired): Delete.
17109 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17110 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17112 2018-05-02 Richard Biener <rguenther@suse.de>
17114 PR middle-end/85567
17115 * gimplify.c (gimplify_save_expr): When in SSA form allow
17116 SAVE_EXPRs to compute to SSA vars.
17118 2018-05-02 Jakub Jelinek <jakub@redhat.com>
17121 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17122 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17123 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17124 clobber operands[2], instead use a new pseudo. Formatting fixes.
17126 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
17128 PR tree-optimization/85586
17129 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17130 exit early for statements in the same group if the accesses are
17133 2018-05-02 Tom de Vries <tom@codesourcery.com>
17136 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17139 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
17141 PR tree-optimization/85143
17142 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17144 2018-05-01 Tom de Vries <tom@codesourcery.com>
17147 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17148 not found" error message.
17150 2018-05-01 Tom de Vries <tom@codesourcery.com>
17153 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
17154 * vec.c (test_ordered_remove_if): New function.
17155 (vec_c_tests): Call test_ordered_remove_if.
17156 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
17157 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
17158 * tree-vect-patterns.c (vect_pattern_recog_1): Use
17159 VEC_ORDERED_REMOVE_IF.
17161 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17163 PR tree-optimization/82665
17164 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
17165 pointer subtraction where arguments come from a memchr call.
17167 2018-05-01 Jakub Jelinek <jakub@redhat.com>
17169 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17170 --push-state --as-needed and --pop-state instead of --as-needed and
17171 --no-as-needed if ld supports it.
17172 * configure: Regenerated.
17175 * doc/install.texi2html: Replace _002d with - and _002a with * in
17176 generated html files using sed.
17178 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17181 * gcc-rich-location.c (blank_line_before_p): New function.
17182 (use_new_line): New function.
17183 (gcc_rich_location::add_fixit_insert_formatted): New function.
17184 * gcc-rich-location.h
17185 (gcc_rich_location::add_fixit_insert_formatted): New function.
17187 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17189 * selftest.c (assert_streq): Rename "expected" and "actual" to
17190 "val1" and "val2". Extend NULL-handling to cover both inputs
17191 symmetrically, while still requiring both to be non-NULL for a pass.
17192 * selftest.h (assert_streq): Rename "expected" and "actual" to
17194 (ASSERT_EQ): Likewise.
17195 (ASSERT_EQ_AT): Likewise.
17196 (ASSERT_KNOWN_EQ): Likewise.
17197 (ASSERT_KNOWN_EQ_AT): Likewise.
17198 (ASSERT_NE): Likewise.
17199 (ASSERT_MAYBE_NE): Likewise.
17200 (ASSERT_MAYBE_NE_AT): Likewise.
17201 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
17202 the assertion to pass.
17203 (ASSERT_STREQ_AT): Likewise.
17205 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
17207 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17208 interaction with -pie.
17210 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17212 * selftest.h: Fix alphabetization of per-source-file selftest
17215 2018-04-30 Jason Merrill <jason@redhat.com>
17217 PR c++/61982 - dead stores to destroyed objects.
17218 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
17221 2018-04-30 Jason Merrill <jason@redhat.com>
17223 * tree.c (build_clobber): New.
17224 * tree.h: Declare it.
17225 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
17227 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17229 * diagnostic-show-locus.c (layout::layout): Update for
17230 location_get_source_line returning a char_span.
17231 (struct char_span): Move to input.h.
17232 (struct correction): Update for fields in char_span becoming
17234 (struct source_line): Update for location_get_source_line
17235 returning a char_span.
17236 (layout::print_line): Likewise.
17237 * edit-context.c (edited_file::print_content): Likewise.
17238 (edited_file::print_diff_hunk): Likewise.
17239 (edited_file::print_run_of_changed_lines): Likewise.
17240 (edited_file::get_num_lines): Likewise.
17241 (edited_line::edited_line): Likewise.
17242 * final.c (asm_show_source): Likewise.
17243 * input.c (location_get_source_line): Convert return type
17244 from const char * to char_span, losing the final "line_len"
17246 (dump_location_info): Update for the above.
17247 (get_substring_ranges_for_loc): Likewise. Use a char_span
17248 when handling the literal within the line.
17249 (test_reading_source_line): Update for location_get_source_line
17250 returning a char_span.
17251 * input.h (class char_span): Move here from
17252 diagnostic-show-locus.c, converting from a struct to a class.
17253 Make data members private.
17254 (char_span::operator bool): New.
17255 (char_span::length): New.
17256 (char_span::get_buffer): New.
17257 (char_span::operator[]): New.
17258 (char_span::subspan): Make const.
17259 (char_span::xstrdup): New.
17260 (location_get_source_line): Convert return type from const char *
17261 to char_span, losing the final "line_size" param.
17263 2018-04-30 Jan Hubicka <jh@suse.cz>
17265 * lto-wrapper.c (ltrans_priorities): New static var.
17266 (cmp_priority): New.
17267 (run_gcc): Read priorities and if doing parallel build order
17268 the Makefile by them.
17270 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17272 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
17274 2018-04-30 Richard Biener <rguenther@suse.de>
17276 * tree-cfg.c (verify_address): Remove base argument, add
17277 flag whether to check TREE_ADDRESSABLE and do that.
17278 (verify_expr): Remove.
17279 (verify_types_in_gimple_reference): Add pieces from verify_expr.
17280 (verify_gimple_assign_single): Likewise.
17281 (verify_gimple_switch): Likewise.
17282 (verify_expr_location_1): Dereference tp once. Add (disabled)
17283 piece from verify_expr.
17284 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
17286 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
17288 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17290 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
17292 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
17293 (small_data_pattern): Likewise.
17294 (arc_rewrite_small_data): Likewise.
17295 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
17296 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
17297 (get_symbol_alignment): New function.
17298 (legitimate_small_data_address_p): Likewise.
17299 (legitimate_scaled_address): Update, call
17300 legitimate_small_data_address_p.
17301 (output_sdata): New static variable.
17302 (arc_print_operand): Update how we handle small data operands.
17303 (arc_print_operand_address): Likewise.
17304 (arc_legitimate_address_p): Update, use
17305 legitimate_small_data_address_p.
17306 (arc_rewrite_small_data_p): Remove.
17307 (arc_rewrite_small_data_1): Likewise.
17308 (arc_rewrite_small_data): Likewise.
17309 (small_data_pattern): Likewise.
17310 (compact_sda_memory_operand): Update to use
17311 legitimate_small_data_address_p and get_symbol_alignment.
17312 (prepare_move_operands): Don't rewite sdata pattern.
17313 (prepare_extend_operands): Remove.
17314 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
17316 (zero_extendqisi2): Likewise.
17317 (zero_extendhisi2): Likewise.
17318 (extendqihi2): Likewise.
17319 (extendqisi2): Likewise.
17320 (extendhisi2): Likewise.
17321 (addsi3): Likewise.
17322 (subsi3): Likewise.
17323 (andsi3): Likewise.
17324 * config/arc/constraints.md (Usd): Change it to memory constraint.
17326 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
17328 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
17329 as source of std instructions.
17330 * config/arc/arc.md (movsi_insn): Update pattern predicate to
17331 allow 6-bit constants as source for store instructions.
17332 (movdi_insn): Update instruction pattern to allow 6-bit constants
17333 as source for store instructions.
17335 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
17337 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
17339 2018-04-30 Nathan Sidwell <nathan@acm.org>
17340 Sandra Loosemore <sandra@codesourcery.com>
17342 * dumpfile.c (dump_open): Allow '-' for stdout.
17343 * doc/invoke.texi (Developer Options): Document dump filename
17344 determination early. Document stdin/stdout selection.
17346 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
17348 Microblaze Target: PIC data text relative
17350 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
17351 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
17353 * config/microblaze/microblaze.h (microblaze_constant_address_p):
17354 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
17355 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
17356 New addressing mode for data-text relative position indepenedent code.
17357 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
17358 'ADDRESS_SYMBOLIC_TXT_REL'.
17359 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
17360 (microblaze_legitimate_pic_operand): Exclude function calls from
17361 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
17362 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
17364 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17365 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17366 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
17367 for 'address + offset'.
17368 (microblaze_expand_prologue): Add new function prologue call for
17370 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
17371 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
17372 table in case of TARGET_PIC_DATA_TEXT_REL.
17373 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
17374 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
17375 Add new macros 'UNSPEC_TEXT',
17376 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
17377 + exclude function calls from 'UNSPEC_PLT' in case of data text
17379 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17380 new target hook for generating address diff vector tables in case of
17382 * doc/tm.texi : Regenerate.
17383 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
17384 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
17385 of addr diff vector generation.
17386 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17387 target hook definition.
17388 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
17389 Add default function for generate_pic_addr_diff_vec -> flag_pic.
17390 * doc/invoke.texi (Add new pic option): Add new microblaze pic
17391 option for data text relative.
17393 2018-04-30 Richard Biener <rguenther@suse.de>
17395 * tree-chrec.h (evolution_function_is_constant_p): Remove
17397 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
17399 2018-04-30 Richard Biener <rguenther@suse.de>
17402 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
17404 2018-04-30 Richard Biener <rguenther@suse.de>
17406 PR tree-optimization/28364
17407 PR tree-optimization/85275
17408 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17409 copying first exit test.
17411 2018-04-28 Mark Wielaard <mark@klomp.org>
17413 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
17414 dwarf_version >= 5.
17415 (dwarf_AT): Handle DW_AT_addr_base.
17416 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
17418 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
17421 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
17422 (*ashl<dwi>3_doubleword_mask_1): Ditto.
17423 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
17424 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
17426 2018-04-28 Richard Biener <rguenther@suse.de>
17428 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
17429 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
17430 to reflect use. Only add interesting stmts.
17432 2018-04-27 Martin Jambor <mjambor@suse.cz>
17435 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
17436 the jump function allows for passing through aggregate values.
17438 2018-04-27 David Malcolm <dmalcolm@redhat.com>
17440 * input.h (in_system_header_at): Convert from macro to inline
17442 (from_macro_expansion_at): Likewise.
17443 (from_macro_definition_at): Likewise.
17445 2018-04-27 Jeff Law <law@redhat.com>
17447 * config.gcc: Mark tile* targets as deprecated/obsolete.
17449 2018-04-27 Richard Biener <rguenther@suse.de>
17451 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
17454 2018-04-27 Richard Biener <rguenther@suse.de>
17456 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
17458 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
17460 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
17461 with Yd constraint. Set "preferred_for_speed" attribute from
17462 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
17463 with Yd constraint.
17464 (*movdi_internal): Ditto.
17465 (movti_interunit splitters): Remove
17466 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
17467 (movdi_interunit splitters): Ditto.
17468 * config/i386/constraints.md (Ye): Remove.
17469 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
17471 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17474 * config/aarch64/constraints.md (Usg): Limit to 31.
17475 (Usj): Limit to 63.
17477 2018-04-27 Jakub Jelinek <jakub@redhat.com>
17479 PR tree-optimization/85529
17480 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
17481 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
17482 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
17483 zero extension or masking of the MSB bit.
17484 (optimize_range_tests): Add FIRST_BB argument, pass it through
17485 to optimize_range_tests_var_bound.
17486 (maybe_optimize_range_tests, reassociate_bb): Adjust
17487 optimize_range_tests callers.
17489 2018-04-26 Richard Biener <rguenther@suse.de>
17490 Jakub Jelinek <jakub@redhat.com>
17492 * cgraph.h (symbol_table): Just declare debug method here.
17493 * symtab.c (symbol_table::debug): Define.
17495 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
17497 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
17499 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
17501 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
17502 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
17503 (*movdi_internal): Substitute Yi and Yj constraint with x
17504 and Ym and Yn constraint with y constraint. Update "isa"
17505 attribute and set "preferred_for_speed" attribute from
17506 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
17507 (*movsi_internal): Ditto.
17508 (*movdf_internal): Ditto.
17509 (*movsf_internal): Ditto.
17510 (*zero_extendsidi2): Ditto.
17511 * config/i386/sse.md (vec_set<mode>_0): Ditto.
17512 (sse2_loadld): Ditto.
17513 (*vec_extract<ssevecmodelower>_0): Ditto.
17514 (*vec_extractv4si_0_zext_sse4): Ditto.
17515 (vec_concatv2di): Ditto.
17516 (*vec_dup<mode>): Ditto.
17517 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
17518 * config/i386/constraints.md (Yi): Remove.
17523 2018-04-26 Nathan Sidwell <nathan@acm.org>
17525 * dumpfile.c (dump_open): New.
17526 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
17527 (dump_finish): Detect stdio/stderr by value not name.
17529 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
17531 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
17533 2018-04-26 Tom de Vries <tom@codesourcery.com>
17536 * config/nvptx/nvptx.c (verify_neutering_jumps)
17537 (verify_neutering_labels): New function
17538 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
17540 2018-04-26 Tom de Vries <tom@codesourcery.com>
17543 * config/nvptx/nvptx.c (needs_neutering_p): New function.
17544 (nvptx_single): Use needs_neutering_p to skip over insns that do not
17547 2018-04-26 Richard Biener <rguenther@suse.de>
17548 Tom de Vries <tom@codesourcery.com>
17551 * lto-streamer-out.c (output_function): Fixup loops if required to match
17552 discovery done in the reader.
17554 2018-04-26 Richard Biener <rguenther@suse.de>
17556 PR tree-optimization/85116
17557 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
17558 have a loop exit from the single latch predecessor. Remove
17559 case of header with just condition.
17560 (ch_base::copy_headers): Exclude infinite loops from any
17562 (pass_ch::execute): Record exits.
17564 2018-04-26 Richard Biener <rguenther@suse.de>
17566 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
17567 prologue cost vector and pass it to vect_get_load_cost.
17568 (vect_get_peeling_costs_all_drs): Likewise.
17569 (vect_peeling_hash_get_lowest_cost): Likewise.
17570 (vect_enhance_data_refs_alignment): Likewise.
17572 2018-04-26 Richard Biener <rguenther@suse.de>
17574 PR middle-end/85450
17575 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
17576 checking of integer<->pointer conversions.
17577 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
17578 sign-/zero-extending pointer types.
17579 (expand_omp_for_static_chunk): Likewise.
17581 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
17582 Jean Lee <xiaoyur347@gmail.com>
17584 * config/mips/mips.c (mips_asan_shadow_offset): New function.
17585 (TARGET_ASAN_SHADOW_OFFSET): Define.
17586 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
17587 true for -fsanitize=address.
17589 2018-04-25 Mark Wielaard <mark@klomp.org>
17591 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
17594 2018-04-25 Jakub Jelinek <jakub@redhat.com>
17596 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
17597 than "alu", remove explicit "memory" and "imm_disp" attributes.
17598 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
17600 PR middle-end/85414
17601 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
17602 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
17603 gen_lowpart_no_emit.
17605 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
17608 * config/i386/i386.c (ix86_expand_builtin): Change memory
17609 operand to XI, extend p0 to Pmode.
17610 * config/i386/i386.md: Change unspec volatile and operand
17611 1 mode to XI, change operand 0 mode to P.
17613 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17615 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
17616 GET_MODE_MASK before any checking.
17617 (nds32_can_use_bset_p): Likewise.
17618 (nds32_can_use_btgl_p): Likewise.
17620 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17622 * config/nds32/nds32-doubleword.md: New define_split pattern for
17623 illegal register number.
17625 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17627 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
17629 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17631 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
17633 2018-04-25 Richard Biener <rguenther@suse.de>
17635 * lto-streamer.h (LTO_major_version): Bump to 8.
17637 2018-04-25 Jakub Jelinek <jakub@redhat.com>
17639 * BASE-VER: Set to 9.0.0.
17641 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
17643 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
17644 in __abskf2 and __powikf2.
17646 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17649 * config/aarch64/constraints.md (Usg, Usj): New constraints.
17650 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
17651 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
17652 Use the above on operand 2. Reindent.
17653 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
17655 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
17658 * common/config/i386/i386-common.c (ix86_handle_option): Don't
17660 * config/i386/i386.opt (mcet): Removed.
17661 * doc/install.texi: Remove -mcet documentation.
17662 * doc/invoke.texi: Likewise.
17664 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
17667 * doc/install.texi: Remove -mcet from bootstrap-cet.
17669 2018-04-24 Jakub Jelinek <jakub@redhat.com>
17672 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
17673 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
17677 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
17678 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
17679 containing a CONST_VECTOR.
17681 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
17683 * doc/install.texi: Update newlib dependency for nvptx.
17685 2018-04-24 Jakub Jelinek <jakub@redhat.com>
17688 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
17689 instead of INTVAL when shifting x left.
17691 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
17693 PR tree-optimization/85478
17694 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
17695 vect_grouped_store_supported for single element vectors.
17697 2018-04-24 Richard Biener <rguenther@suse.de>
17700 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
17701 load cost increase to the case of non-constant step.
17703 2018-04-24 Jakub Jelinek <jakub@redhat.com>
17706 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
17707 destination if any_malformed_asm.
17709 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
17711 PR middle-end/85496
17712 * expr.c (store_field): In the bitfield case, if the value comes from
17713 a function call and is returned in registers by means of a PARALLEL,
17714 do not change the mode of the temporary unless BLKmode and VOIDmode.
17716 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
17718 PR rtl-optimization/85423
17719 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
17720 dependencies to debug insns when the previous insn is non-debug.
17722 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
17724 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
17725 enums into a single definition.
17726 (fls): Fix predicates and printing.
17729 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
17731 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
17732 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
17733 and short u6 immediate.
17734 (check_if_valid_sleep_operand): Remove.
17735 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
17737 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17739 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
17740 flag_always_save_lp condition.
17741 * config/nds32/nds32.opt (malways-save-lp): New option.
17743 2018-04-22 Shiva Chen <shiva0217@gmail.com>
17745 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
17746 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
17747 * config/nds32/nds32.h
17748 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
17749 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
17751 2018-04-22 Shiva Chen <shiva0217@gmail.com>
17753 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
17754 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
17756 2018-04-22 Shiva Chen <shiva0217@gmail.com>
17757 Chung-Ju Wu <jasonwucj@gmail.com>
17759 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
17761 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
17762 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
17764 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17766 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
17768 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17770 * config/nds32/nds32-protos.h (nds32_data_alignment,
17771 nds32_local_alignment): Declare.
17772 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
17773 nds32_local_alignment): New functions.
17774 (TARGET_CONSTANT_ALIGNMENT): Define.
17775 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
17777 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17779 * config/nds32/nds32.c
17780 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
17781 (TARGET_MODES_TIEABLE_P): Likewise.
17783 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17785 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
17786 level Ofast and Og.
17788 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
17789 Chung-Ju Wu <jasonwucj@gmail.com>
17791 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17792 for unaligned access.
17793 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17795 * config/nds32/nds32-intrinsic.md: Likewise.
17796 * config/nds32/nds32_intrinsic.h: Likewise.
17797 * config/nds32/nds32.h (nds32_builtins): Likewise.
17798 * config/nds32/nds32.opt (munaligned-access): New option.
17799 * config/nds32/nds32.c (nds32_asm_file_start): Display
17800 flag_unaligned_access status.
17802 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
17804 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
17805 -mno-relax is present.
17806 * config/riscv/linux.h (LINK_SPEC): Ditto.
17808 2018-04-20 Martin Sebor <msebor@redhat.com>
17811 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
17813 (gimple_fold_builtin_stxcpy_chk): Same.
17814 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
17816 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
17819 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
17820 __powikf2 when long double is IEEE 128-bit.
17822 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
17824 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
17825 step to make sure stack always aligned.
17827 2018-04-20 Carl Love <cel@us.ibm.com>
17830 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
17831 size check for arg0.
17833 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
17834 Tom de Vries <tom@codesourcery.com>
17837 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17838 Emit insns for calls too.
17839 (nvptx_find_par): Always look for worker-level predecessor insn.
17840 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
17842 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
17843 (nvptx_process_pars): Propagate frames for calls.
17845 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
17848 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
17850 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
17851 (ix86_handle_option): Don't handle OPT_mibt.
17852 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
17854 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
17856 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
17857 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
17858 (ix86_target_macros): Define __CET__ with flag_cf_protection
17859 for -fcf-protection.
17860 * config/i386/i386.c (isa2_opts): Remove -mibt.
17861 * config/i386/i386.h (TARGET_IBT): Removed.
17862 (TARGET_IBT_P): Likewise.
17863 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
17864 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
17865 * config/i386/i386.opt (mcet): Update help message.
17866 (mshstk): Likewise.
17868 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
17869 -mcet as an alias for -mshstk.
17871 2018-04-20 Richard Biener <rguenther@suse.de>
17873 PR middle-end/85475
17874 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
17875 complexity by forcing a single use of the multiply operand.
17877 2018-04-20 Martin Jambor <mjambor@suse.cz>
17880 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
17881 recursion dependency to only apply to non-clones.
17883 2018-04-20 Martin Jambor <mjambor@suse.cz>
17886 * ipa-cp.c (create_specialized_node): Check that clones of
17887 self-recursive edges exist during IPA-CP.
17889 2018-04-19 Toon Moene <toon@moene.org>
17891 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
17894 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17896 PR tree-optimization/85467
17897 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
17898 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
17899 VECTOR_CST element to type.
17901 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17904 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
17905 * config/i386/i386.md (builtin_setjmp_setup): Removed.
17906 (builtin_longjmp): Likewise.
17907 (save_stack_nonlocal): New pattern.
17908 (restore_stack_nonlocal): Likewise.
17910 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17913 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17914 Replace ASM_OUTPUT_LABEL with fprintf.
17916 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17919 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17920 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17921 * config/i386/i386-c.c (ix86_target_macros_internal): Also
17922 define __IBT__ and __SHSTK__ for -fcf-protection.
17923 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17925 (ix86_trampoline_init): Likewise.
17926 (x86_output_mi_thunk): Likewise.
17927 (ix86_notrack_prefixed_insn_p): Likewise.
17928 (ix86_option_override_internal): Don't disallow -fcf-protection.
17929 * config/i386/i386.md (rdssp<mode>): Also enable for
17931 (incssp<mode>): Likewise.
17932 (nop_endbr): Likewise.
17933 * config/i386/i386.opt (mcet): Change help message to built-in
17936 (mshstk): Likewise.
17937 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17938 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
17939 enable CET built-in functions.
17941 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
17943 * common/config/i386/i386-common.c
17944 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17945 OPTION_MASK_ISA_MOVDIRI_UNSET,
17946 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17947 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17948 * config.gcc (movdirintrin.h): New header.
17949 * config/i386/cpuid.h (bit_MOVDIRI,
17950 bit_MOVDIR64B): New bits.
17951 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17953 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17954 (VOID, PVOID, PCVOID)): New function types.
17955 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17956 __builtin_ia32_directstoreu_u64,
17957 __builtin_ia32_movdir64b): New builtins.
17958 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17959 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17961 (ix86_valid_target_attribute_inner_p): Ditto.
17962 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17963 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17964 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17965 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17966 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17967 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17968 (movdiri<mode>, movdir64b_<mode>): New.
17969 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17970 * config/i386/immintrin.h: Include movdirintrin.h.
17971 * config/i386/movdirintrin.h: New file.
17972 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17974 2018-04-19 Richard Biener <rguenther@suse.de>
17976 PR middle-end/85455
17977 * cfg.c (clear_bb_flags): When loop state says we have
17978 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17980 2018-04-19 Richard Biener <rguenther@suse.de>
17982 PR tree-optimization/84737
17983 * tree-vect-data-refs.c (vect_copy_ref_info): New function
17984 copying restrict info.
17985 (vect_setup_realignment): Use it.
17986 * tree-vectorizer.h (vect_copy_ref_info): Declare.
17987 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17988 the first DR to all generated stores.
17989 (vectorizable_load): Likewise for loads.
17991 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17993 PR tree-optimization/85446
17994 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17995 the integral and pointer types to have the same precision.
17997 * doc/install.texi: Document --disable-cet being the default and
18000 2018-04-18 Martin Liska <mliska@suse.cz>
18002 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
18005 2018-04-18 Martin Liska <mliska@suse.cz>
18008 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
18011 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18012 arguments if they are comparable.
18014 2018-04-18 Martin Liska <mliska@suse.cz>
18017 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
18020 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18023 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
18026 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
18027 ENDBR after calling __morestack.
18029 2018-04-18 David Malcolm <dmalcolm@redhat.com>
18032 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
18033 by using gcc_base_ver to generate a gcc_driver_version, and use
18034 it when generating GCC_DRIVER_NAME.
18035 * configure: Regenerate.
18037 2018-04-18 Jakub Jelinek <jakub@redhat.com>
18040 * config.gcc: Obsolete powerpc*-*-*spe*.
18042 2018-04-17 Jakub Jelinek <jakub@redhat.com>
18045 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
18046 (stabstr_D): Change type of unum from unsigned int to
18047 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
18050 2018-04-17 Jim Wilson <jimw@sifive.com>
18053 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
18054 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
18055 Set arg_pointer_offset after using pretend_args_size.
18057 2018-04-17 Jakub Jelinek <jakub@redhat.com>
18059 PR rtl-optimization/85431
18060 * dse.c (record_store): Ignore zero width stores.
18063 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
18064 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
18065 __builtin_stack_restore rather than after it.
18066 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18067 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
18068 argument instead of virtual_dynamic_stack_rtx.
18070 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
18072 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
18074 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
18075 Add note to error message to explain internal mapping of overloaded
18076 built-in function name to non-overloaded built-in function name.
18077 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
18080 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18083 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
18084 where the inputs overlap with the output.
18086 2018-04-17 Jakub Jelinek <jakub@redhat.com>
18088 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
18089 (=v, v) alternative and explicit "memory" attribute.
18090 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
18091 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18093 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
18094 "sselog1" type instead of "sselog".
18095 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
18096 "sselog". Remove explicit "memory" attribute.
18097 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
18098 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18100 (vec_extract_hi_v32hi): Merge all alternatives into one, use
18101 "sselog1" type instead of "sselog". Remove explicit "memory"
18103 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
18104 use "sselog1" type instead of "sselog". Remove explicit "memory"
18106 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18107 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18109 (vec_extract_hi_v64qi): Merge all alternatives into one, use
18110 "sselog1" type instead of "sselog". Remove explicit "memory"
18112 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18113 use "sselog1" type instead of "sselog". Remove explicit "memory"
18117 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18119 PR middle-end/85414
18120 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18123 2018-04-17 Martin Jambor <mjambor@suse.cz>
18126 * ipa-cp.c (create_specialized_node): Call
18127 expand_all_artificial_thunks if necessary.
18129 2018-04-17 Martin Liska <mliska@suse.cz>
18132 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18133 in message, remote space in between '_G' and '('.
18135 2018-04-17 Jakub Jelinek <jakub@redhat.com>
18138 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18139 avx512f_vmcmp<mode>3<round_saeonly_name>,
18140 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18141 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18142 avx512f_rndscale<mode><round_saeonly_name>,
18143 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18144 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18145 Use %<iptr>2 instead of %2 for -masm=intel.
18146 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18147 avx512f_vcvttss2usi<round_saeonly_name>,
18148 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18150 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18151 avx512f_vcvttsd2usi<round_saeonly_name>,
18152 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18153 Use %q1 instead of %1 for -masm=intel.
18154 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
18155 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
18156 of %3 for -masm=intel.
18157 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
18159 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
18161 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
18163 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
18164 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18166 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18168 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18169 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18170 %g1 and one with %0 and %1.
18171 (avx512er_vmrcp28<mode><round_saeonly_name>,
18172 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18173 %1 for -masm=intel.
18174 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18175 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18176 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18177 of %0 and %{%4%} for -masm=intel.
18178 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18179 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18180 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18181 order of %0 and %{%5%}%{z%} for -masm=intel.
18183 2018-04-17 Jan Hubicka <jh@suse.cz>
18186 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18188 2018-04-17 Martin Liska <mliska@suse.cz>
18191 * multiple_target.c (create_dispatcher_calls): Set apostrophes
18192 for target_clone error message. Make default implementation
18193 clone to be a local declaration.
18194 (separate_attrs): Add new argument and check for an empty
18196 (expand_target_clones): Handle it.
18197 (ipa_target_clone): Make redirection just for target_clones
18200 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
18201 Tom de Vries <tom@codesourcery.com>
18203 PR middle-end/84955
18204 * omp-expand.c (expand_oacc_for): Add dummy false branch for
18205 tiled basic blocks without omp continue statements.
18207 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
18210 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
18211 vec_extract expression as having side effects to make sure it gets
18214 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
18217 * config/i386/i386.c (get_builtin_code_for_version): Check
18220 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
18223 * config.gcc: Support "skylake".
18224 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18226 * config/i386/i386.c (m_SKYLAKE): Define.
18227 (processor_target_table): Add "skylake".
18228 (ix86_option_override_internal): Add "skylake".
18229 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
18230 PROCESSOR_CANNONLAKE.
18231 (get_builtin_code_for_version): Fix priority for
18232 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
18233 PROCESSOR_SKYLAKE-AVX512.
18234 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
18235 (processor_type): Add PROCESSOR_SKYLAKE.
18237 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
18238 Jason Merrill <jason@redhat.com>
18241 * convert.c (convert_to_integer_1): Use direct recursion for
18242 enumeral types and types with a precision less than the number
18243 of bits in their mode.
18245 2018-04-16 Julia Koval <julia.koval@intel.com>
18248 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
18249 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
18251 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
18254 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
18255 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
18256 and -mno-direct-move.
18258 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
18261 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
18262 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
18263 Ensure negative shifts result in {0}.
18265 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
18267 PR rtl-optimization/79916
18268 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
18269 regs (if any) to define how to gnerate SD moves when LRA is in
18272 2018-04-13 Jakub Jelinek <jakub@redhat.com>
18274 PR rtl-optimization/85393
18275 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
18276 * except.c (expand_dw2_landing_pad_for_region): Make static.
18277 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
18278 a label and unconditional jump to old_bb, rather than
18279 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
18282 PR rtl-optimization/85376
18283 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
18284 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
18285 instead of a specific value.
18287 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
18288 Bin Cheng <bin.cheng@arm.com>
18290 PR tree-optimization/82965
18291 PR tree-optimization/83991
18292 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
18293 by_profile_only parameter.
18294 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
18295 information if the loop was predicted to iterate too many times.
18296 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
18298 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
18301 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
18304 2018-04-13 Martin Liska <mliska@suse.cz>
18305 Jakub Jelinek <jakub@redhat.com>
18307 PR middle-end/81657
18308 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
18309 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
18310 * builtins.c (expand_builtin_memory_copy_args): Use
18311 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
18312 handle dest_addr == pc_rtx.
18314 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
18317 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
18318 asked to not generate direct moves.
18319 (fix_trunc<mode>si2_stfiwx): Similar.
18320 (fix_trunc<mode>si2_internal): Similar.
18322 2018-04-12 Jakub Jelinek <jakub@redhat.com>
18325 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
18326 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
18327 lookup if dest in some wider mode is known to be const0_rtx and
18328 if so, record permanent equivalence for it to be ZERO_EXTEND of
18329 the narrower mode destination.
18331 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
18333 * lto-streamer-out.c (output_function): Revert 259346.
18334 * omp-expand.c (expand_oacc_for): Likewise.
18336 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
18338 PR rtl-optimization/85354
18339 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
18340 * sel-sched.c (sel_global_init): ... here.
18342 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
18345 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
18346 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
18347 mode for PE-COFF targets.
18348 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
18349 (i386_pe_asm_lto_end): Likewise.
18350 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
18351 (TARGET_ASM_LTO_END): Likewise.
18352 * config/i386/winnt.c (saved_debug_info_level): New static variable.
18353 (i386_pe_asm_lto_start): New function.
18354 (i386_pe_asm_lto_end): Likewise.
18356 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
18357 Richard Biener <rguenther@suse.de>
18359 PR middle-end/84955
18360 * lto-streamer-out.c (output_function): Fix CFG loop state before
18362 * omp-expand.c (expand_oacc_for): Handle calls to internal
18363 functions like regular functions.
18365 2018-04-12 Richard Biener <rguenther@suse.de>
18368 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
18369 for the early LTO debug to properly generate references to it
18370 during DIE emission. Do not re-use that for the skeleton for
18372 (dwarf2out_early_finish): Likewise.
18374 2018-04-12 Jakub Jelinek <jakub@redhat.com>
18377 * config/i386/sse.md
18378 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
18379 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
18380 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
18381 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
18382 and output is a reg, avoid creating invalid lowpart subreg, but
18383 instead split into a 512-bit move. Don't split if not AVX512VL,
18384 input is xmm16+ reg and output is a mem.
18385 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
18386 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
18387 xmm16+ reg and output is a mem.
18389 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18391 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
18392 also for flag_dwarf2_cfi_asm.
18394 2018-04-12 Jakub Jelinek <jakub@redhat.com>
18396 PR rtl-optimization/85342
18397 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
18398 a bool scalar var inside of the loop instead. Don't try to update
18399 recog_data.operand after failed apply_change_group.
18401 2018-04-12 Tom de Vries <tom@codesourcery.com>
18404 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
18405 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
18406 array with flexible array member as array without given dimension.
18407 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
18408 argument for undefined param to true.
18410 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
18413 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18414 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
18415 from PowerPC section.
18416 * config/rs6000/sysv4.opt (mcall-): Improve help text.
18417 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
18418 help text that is too long.
18419 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
18420 help text that is too long.
18421 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
18422 help text that is too long.
18424 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
18426 * config/alpha/alpha.md (stack_probe_internal): Rename
18427 from "probe_stack". Update all callers.
18429 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
18431 PR rtl-optimization/84566
18432 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
18433 sched_macro_fuse_insns.
18435 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
18438 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
18439 (compute_block_dependences): ... from here.
18441 2018-04-11 Jakub Jelinek <jakub@redhat.com>
18443 PR tree-optimization/85331
18444 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
18445 from int to HOST_WIDE_INT.
18447 2018-04-11 Martin Jambor <mjambor@suse.cz>
18450 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
18451 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
18452 not the same as the source val.
18453 (cgraph_edge_brings_value_p): New parameter.
18454 (gather_edges_for_value): Pass destination value to
18455 cgraph_edge_brings_value_p.
18456 (perhaps_add_new_callers): Likewise.
18457 (get_info_about_necessary_edges): Likewise and exclude values brought
18458 only by self-recursive edges.
18459 (create_specialized_node): Redirect only clones of self-calling edges.
18460 (+self_recursive_pass_through_p): New function.
18461 (find_more_scalar_values_for_callers_subset): Use it.
18462 (find_aggregate_values_for_callers_subset): Likewise.
18463 (known_aggs_to_agg_replacement_list): Removed.
18464 (decide_whether_version_node): Re-calculate known constants for all
18465 remaining context clones.
18467 2018-04-11 Richard Biener <rguenther@suse.de>
18470 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
18471 from early DWARF output.
18472 (dwarf2out_early_finish): Output line info unconditionally into
18473 early DWARF and add reference to it.
18475 2018-04-11 Jakub Jelinek <jakub@redhat.com>
18478 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
18479 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
18480 other than V2DFmode using iptr mode attribute.
18481 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
18483 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
18485 PR rtl-optimization/84659
18486 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
18488 2018-04-11 Jakub Jelinek <jakub@redhat.com>
18491 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
18493 (output_loc_list): Pass address of a dummy size variable even in the
18494 locview handling loop.
18495 (index_location_lists): Add comment on why skip_loc_list_entry can't
18498 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
18501 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
18504 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
18507 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18508 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
18509 and -mstring-compare-inline-limit.
18511 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
18514 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
18515 for stack clash protection in a register whenever we need it to be in
18518 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
18520 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
18521 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
18523 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
18526 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
18528 (mlong-double-): Ditto.
18529 * config/rs6000/sysv4.opt (msdata=): Ditto.
18530 (mtls-size=): Ditto.
18532 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
18534 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18535 erroneous entries for
18536 "vector int vec_ldl (int, long int *)", and
18537 "vector unsigned int vec_ldl (int, unsigned long int *)".
18538 Add comments and entries for
18539 "vector bool char vec_ldl (int, bool char *)",
18540 "vector bool short vec_ldl (int, bool short *)",
18541 "vector bool int vec_ldl (int, bool int *)",
18542 "vector bool long long vec_ldl (int, bool long long *)",
18543 "vector pixel vec_ldl (int, pixel *)",
18544 "vector long long vec_ldl (int, long long *)",
18545 "vector unsigned long long vec_ldl (int, unsigned long long *)".
18546 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
18547 type tree bool_long_long_type_node and correct definition of
18548 bool_V2DI_type_node to make reference to this new type tree.
18549 (rs6000_mangle_type): Replace erroneous reference to
18550 bool_long_type_node with bool_long_long_type_node.
18551 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
18552 comments to emphasize sign distinctions for char and int types and
18553 replace RS6000_BTI_bool_long constant with
18554 RS6000_BTI_bool_long_long constant. Also add comment to restrict
18555 use of RS6000_BTI_pixel.
18556 (bool_long_type_node): Remove this macro definition.
18557 (bool_long_long_type_node): New macro definition
18559 2018-04-10 Jakub Jelinek <jakub@redhat.com>
18561 PR rtl-optimization/85300
18562 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
18563 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
18564 simplify_unary_operation fails.
18566 2018-04-10 Martin Liska <mliska@suse.cz>
18568 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
18569 cgraph_edge and ipa_ref.
18571 2018-04-10 Jakub Jelinek <jakub@redhat.com>
18575 * config/i386/sse.md
18576 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
18577 computation of the VEC_MERGE selector from mask.
18578 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
18579 Fix decoding of the VEC_MERGE selector into mask.
18581 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
18583 PR tree-optimization/85286
18584 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
18586 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
18588 * final.c (final_1): Set insn_last_address as well as
18589 insn_current_address.
18591 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18594 * explow.c (emit_stack_probe): Call validize_mem on memory location
18595 before passing it to gen_probe_stack. Create address operand and
18596 legitimize it for the probe_stack_address case.
18598 2018-04-09 Jan Hubicka <jh@suse.cz>
18601 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
18602 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
18603 * tree.c (free_lang_data_in_type): Fix handling of binfos;
18605 (free_lang_data): Rebuild type inheritance graph.
18607 2018-04-09 Martin Sebor <msebor@redhat.com>
18609 * invoke.texi (-finline-small-functions): Mention other optimization
18611 (-findirect-inlining, -fpartial-inlining): Same.
18612 (-finline-functions-called-once): Same.
18613 (-freorder-blocks-and-partition): Same.
18615 2018-04-09 Jan Hubicka <jh@suse.cz>
18618 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
18619 jumps; choose last target that matches the criteria (i.e.
18620 no partition changes for non-crossing jumps).
18621 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
18622 support for redirecting crossing jumps to non-crossing.
18624 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
18626 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
18627 also for naked functions.
18629 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
18631 * config/arc/arc.md (add_shift): New pattern.
18632 (add_shift2): Likewise.
18633 (sub_shift): Likewise.
18634 (sub_shift_cmp0_noout): Likewise.
18635 (compare_si_ashiftsi): Likewise.
18636 (xbfu_cmp0_noout): New combine pattern.
18637 (xbfu_cmp0"): Likewise.
18638 (movsi_set_cc_insn): Place the predicable variant first.
18639 (commutative_binary_cmp0_noout): Remove clobber.
18640 (commutative_binary_cmp0): New pattern.
18641 (noncommutative_binary_cmp0): Likewise.
18642 (noncommutative_binary_cmp0_noout): Likewise.
18643 (noncommutative_binary_comparison_result_used): Removed.
18644 (rsub_cmp0): New pattern.
18645 (rsub_cmp0_noout): Likewise.
18646 (extzvsi): Changed, keep only meaningful variants.
18647 (SQH, SEZ): New iterators.
18648 (SQH_postfix): New mode attribute.
18649 (SEZ_prefix): New code attribute.
18650 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
18651 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
18652 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
18653 of numerical value.
18654 (noncommutative_operator): Check the availability of barrel
18657 2018-04-09 Richard Biener <rguenther@suse.de>
18659 PR tree-optimization/85284
18660 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
18661 Only use the niter constraining form of simple_iv when the exit
18662 is always executed.
18664 2018-04-09 Tom de Vries <tom@codesourcery.com>
18667 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18668 (define_expand "*memory_barrier"): New define_expand.
18669 (define_insn "memory_barrier"): New insn.
18671 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18673 PR rtl-optimization/80463
18674 PR rtl-optimization/83972
18675 PR rtl-optimization/83480
18677 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
18678 correct producer for the insn.
18679 (tidy_control_flow): Fixup seqnos in case of debug insns.
18681 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18683 PR rtl-optimization/83913
18685 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
18686 different sched-times when merging exprs.
18688 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18690 PR rtl-optimization/83962
18692 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
18693 tidy_fallthru_edge and tidy_control_flow.
18695 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18697 PR rtl-optimization/83530
18699 * sel-sched.c (force_next_insn): New global variable.
18700 (remove_insn_for_debug): When force_next_insn is true, also leave only
18701 next insn in the ready list.
18702 (sel_sched_region): When the region wasn't scheduled, make another pass
18703 over it with force_next_insn set to 1.
18705 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
18707 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
18709 * config/nds32/constants.md (unspec_volatile_element): Add enum values
18710 for interrupt control.
18711 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18712 functions for interrupt control.
18713 * config/nds32/nds32-intrinsic.md: Likewise.
18714 * config/nds32/nds32_intrinsic.h: Likewise.
18715 * config/nds32/nds32.h (nds32_builtins): Likewise.
18717 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
18719 * config/nds32/nds32.c (nds32_init_machine_status,
18720 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
18721 strict_aligned_p field.
18722 (nds32_expand_to_rtl_hook): New function.
18723 (TARGET_EXPAND_TO_RTL_HOOK): Define.
18724 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
18726 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18727 Chung-Ju Wu <jasonwucj@gmail.com>
18729 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
18730 * config/nds32/nds32-n7.md: New file.
18731 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
18732 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
18734 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
18735 * config/nds32/nds32.md (pipeline_model): Add n7.
18736 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
18737 * config/nds32/pipelines.md: Include n7 settings.
18739 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18740 Chung-Ju Wu <jasonwucj@gmail.com>
18742 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
18743 * config/nds32/nds32-e8.md: New file.
18744 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
18745 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
18747 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
18748 * config/nds32/nds32.md (pipeline_model): Add e8.
18749 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
18750 * config/nds32/pipelines.md: Include e8 settings.
18752 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18753 Chung-Ju Wu <jasonwucj@gmail.com>
18755 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
18756 * config/nds32/nds32-n8.md: New file.
18757 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
18758 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
18760 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
18761 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
18762 * config/nds32/nds32.md (pipeline_model): Add n8.
18763 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
18764 * config/nds32/pipelines.md: Include n8 settings.
18766 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18767 Chung-Ju Wu <jasonwucj@gmail.com>
18769 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
18770 * config/nds32/nds32-n9-2r1w.md: New file.
18771 * config/nds32/nds32-n9-3r2w.md: New file.
18772 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
18773 nds32_register_ports): New or modify for cpu n9.
18774 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
18776 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
18777 * config/nds32/nds32-utils.c: New file.
18778 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
18779 TARGET_MUL_SLOW): Define.
18780 * config/nds32/nds32.md (pipeline_model): New attribute.
18781 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
18782 New options that support cpu n9.
18783 * config/nds32/pipelines.md: Include n9 settings.
18784 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
18786 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
18788 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
18789 information if necessary.
18790 (output_cond_branch_compare_zero): Likewise.
18791 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
18792 (nds32_target_alignment): Refine for alignment.
18793 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
18794 (FUNCTION_BOUNDARY): Modify.
18795 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
18797 * config/nds32/nds32.opt (malways-align, malign-functions): New.
18799 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
18801 * config/nds32/constants.md (unspec_volatile_element): Add values for
18802 TLB operation and data prefetch.
18803 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18804 functions for TLB operation and data prefetch.
18805 * config/nds32/nds32-intrinsic.md: Likewise.
18806 * config/nds32/nds32_intrinsic.h: Likewise.
18807 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
18808 (nds32_print_operand): Likewise.
18809 * config/nds32/nds32.h (nds32_builtins): Likewise.
18811 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
18812 Andrew Pinski <pinsika@gcc.gnu.org>
18814 PR middle-end/82976
18815 * match.pd: Use constant_boolean_node of correct type instead of
18816 boolean_true_node or boolean_false_node for simplifying
18817 pointer comparisons to zero.
18819 2018-04-07 Jakub Jelinek <jakub@redhat.com>
18821 PR tree-optimization/80021
18822 * tree.c (verify_type_variant): Make error call in verify_variant_match
18823 translatable and remove final full stop.
18825 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18827 * config/nds32/constants.md (unspec_volatile_element): Add
18828 UNSPEC_VOLATILE_EH_RETURN.
18829 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
18830 nds32_output_stack_pop): Support dwarf exception handling process.
18831 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
18832 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
18833 exception handling process.
18834 (nds32_compute_stack_frame): Likewise.
18835 (nds32_return_addr_rtx): Likewise.
18836 (nds32_initial_elimination_offset): Likewise.
18837 (nds32_expand_prologue): Likewise.
18838 (nds32_expand_epilogue): Likewise.
18839 (nds32_dynamic_chain_address): New function.
18840 * config/nds32/nds32.h (machine_function): Add fields for dwarf
18841 exception handling.
18842 (DYNAMIC_CHAIN_ADDRESS): Define.
18843 (EH_RETURN_DATA_REGNO): Define.
18844 (EH_RETURN_STACKADJ_RTX): Define.
18845 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
18846 patterns for dwarf exception handling.
18848 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18850 * config/nds32/nds32.h: Clean up obsolete macros.
18852 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18854 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18855 Add enum values for particular instructions.
18856 * config/nds32/nds32-intrinsic.c: Implementation of expanding
18857 particular intrinsic functions.
18858 * config/nds32/nds32-intrinsic.md: Likewise.
18859 * config/nds32/nds32_intrinsic.h: Likewise.
18860 * config/nds32/nds32.h (nds32_builtins): Likewise.
18861 * config/nds32/nds32.md (type): Add pbsad and pbsada.
18862 (btst, ave): New patterns for particular instructions.
18864 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18866 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18867 Add enum values for atomic load/store and memory sync.
18868 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
18870 * config/nds32/nds32-intrinsic.md: Likewise.
18871 * config/nds32/nds32_intrinsic.h: Likewise.
18872 * config/nds32/nds32.h (nds32_builtins): Likewise.
18874 2018-04-07 Jakub Jelinek <jakub@redhat.com>
18876 PR tree-optimization/85257
18877 * fold-const.c (native_encode_vector): If not all elts could fit
18878 and off is -1, return 0 rather than offset.
18879 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
18880 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
18881 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
18882 adjust buffer in native_interpret_expr call.
18884 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18886 * config/nds32/constants.md (unspec_volatile_element): Add cache
18887 control enum values.
18888 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
18889 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
18890 * config/nds32/nds32.c (nds32_cctl_names): New.
18891 (nds32_print_operand): Handle cache control register names.
18892 * config/nds32/nds32.h (nds32_builtins): New enum values.
18893 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
18895 * config/nds32/nds32.md (type): Add mmu.
18896 * config/nds32/pipelines.md (simple_insn): Add mmu.
18898 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18900 * config/nds32/nds32.md (type): Remove call.
18901 * config/nds32/pipelines.md (simple_insn): Likewise.
18903 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18905 * config/nds32/constants.md (unspec_volatile_element): Add
18906 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
18907 UNSPEC_VOLATILE_FMFCFG.
18908 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
18909 description for fmfcfg and fmfcsr.
18910 (bdesc_1arg): Add fmtcsr.
18911 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
18912 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18913 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18914 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18915 unspec_fmfcfg): New patterns.
18916 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18917 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18918 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18919 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18920 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18921 __nds32__fmfcfg): Define.
18923 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18925 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18926 intrinsic register names.
18927 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18928 intrinsic register enum values and macros.
18930 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18932 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18933 for load/store addressing form.
18934 (nds32_print_operand_address): Likewise.
18936 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
18939 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18940 based on LABEL_REF. Remove useless assertion.
18941 (pic_address_needs_scratch): Fix formatting.
18942 (sparc_legitimize_pic_address): Minor tweaks.
18943 (sparc_delegitimize_address): Adjust assertion accordingly.
18944 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18945 into symbolic_operand.
18946 (movsi_high_pic_label_ref): Likewise.
18947 (movsi_lo_sum_pic_label_ref): Likewise.
18948 (movdi_pic_label_ref): Likewise.
18949 (movdi_high_pic_label_ref): Likewise.
18950 (movdi_lo_sum_pic_label_ref): Likewise.
18952 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
18954 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18955 custom LIB_SPEC setup.
18957 2018-04-06 Ruslan Bukin <br@bsdpad.com>
18958 Kito Cheng <kito.cheng@gmail.com>
18960 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18961 * config/riscv/freebsd.h: New.
18963 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18965 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18966 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18969 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18970 Kito Cheng <kito.cheng@gmail.com>
18972 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18973 nds32_output_call, nds32_symbol_binds_local_p): New functions.
18974 * config/nds32/nds32-protos.h (nds32_output_call,
18975 nds32_output_return): Declare.
18976 * config/nds32/nds32.md: Refine all the call and return patterns.
18978 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18981 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18982 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18984 PR rtl-optimization/84872
18985 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18986 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18987 EDGE_CROSSING edge.
18989 2018-04-06 Tamar Christina <tamar.christina@arm.com>
18991 * expr.c (copy_blkmode_to_reg): Revert 254862.
18992 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18994 2018-04-06 Richard Biener <rguenther@suse.de>
18996 PR middle-end/85244
18997 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18998 after seeing a component reference with an adjacent field. Treat
18999 refs to arrays at struct end of external decls similar to
19000 refs to unconstrained commons.
19002 2018-04-06 Jakub Jelinek <jakub@redhat.com>
19005 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
19006 look through SAVE_EXPRs with non-side-effects argument. Adjust
19008 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
19011 2018-04-06 Richard Biener <rguenther@suse.de>
19013 PR middle-end/85180
19014 * alias.c (find_base_term): New wrapper around find_base_term
19015 unwinding CSELIB_VAL_PTR changes.
19016 (find_base_term): Do not restore CSELIB_VAL_PTR during the
19019 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19021 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
19023 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
19024 constant definitions.
19025 ("nop"): lr 0,0 -> nopr r0
19026 ("nop_lr0", "nop_lr1"): New insn definitions.
19028 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
19030 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
19031 NDS32_V3PUSH_AVAILABLE_P macro.
19033 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
19034 Chung-Ju Wu <jasonwucj@gmail.com>
19036 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
19037 (nds32*-*-*): Add float and fpu_config into supported_defaults.
19038 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
19039 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
19040 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
19041 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
19042 * config/nds32/constraints.md: New constraints and checking for hard
19043 float configuration.
19044 * config/nds32/iterators.md: New mode iterator and attribute for hard
19045 float configuration.
19046 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
19048 * config/nds32/nds32-fpu.md: New file.
19049 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
19050 deal with hard float code generation.
19051 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
19053 (abi_type, float_reg_number): New enum type.
19054 * config/nds32/nds32-predicates.c: New predicates for hard float.
19055 * config/nds32/nds32-protos.h: Declare functions for hard float.
19056 * config/nds32/nds32.c: Implementation for hard float configuration.
19057 * config/nds32/nds32.h: Definitions for hard float configuration.
19058 * config/nds32/nds32.md: Include hard float machine description and
19059 modify patterns for hard float configuration.
19060 * config/nds32/nds32.opt: New options for hard float configuration.
19061 * config/nds32/predicates.md: New predicates for hard float
19064 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
19066 * common/config/nds32/nds32-common.c
19067 (nds32_option_optimization_table): Enable -mreleax-hint by default.
19069 2018-04-05 Jakub Jelinek <jakub@redhat.com>
19071 PR middle-end/85195
19072 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
19073 CONSTRUCTOR_ELT (ctor, ...)->value.
19075 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
19078 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
19080 2018-04-05 Tom de Vries <tom@codesourcery.com>
19083 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
19086 2018-04-05 Shiva Chen <shiva0217@gmail.com>
19087 Kito Cheng <kito.cheng@gmail.com>
19089 * config/nds32/constraints.md (U33): Fine-tune checking condition.
19090 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
19091 * config/nds32/nds32.h (nds32_16bit_address_type): Add
19092 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
19094 2018-04-05 Shiva Chen <shiva0217@gmail.com>
19095 Kito Cheng <kito.cheng@gmail.com>
19097 * config/nds32/constraints.md (Ufe): New memory constraint.
19098 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
19099 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
19100 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
19102 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
19103 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
19105 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19107 * config/nds32/nds32.md: Use optimize_size in the condition for
19108 alu-shift instructions.
19110 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19112 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19114 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19116 * config/nds32/nds32.md (negsi2): Refine pattern.
19118 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
19119 Chung-Ju Wu <jasonwucj@gmail.com>
19121 * config/nds32/iterators.md (shift_rotate): New code iterator.
19122 (shift): New code attribute.
19123 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19124 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19125 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19126 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19127 bit-wise operations.
19128 (andsi3, *andsi3): Ditto.
19129 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19130 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19131 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19132 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19133 nds32_ior_operand, nds32_xor_operand): New predicates.
19135 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19137 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19138 (addsi3, subsi3): ... this.
19140 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19142 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19144 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19146 * config/nds32/nds32.md: Adjust indention.
19148 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
19150 * config/nds32/nds32.md (feature): New attribute.
19152 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
19154 * config/nds32/nds32.md (subtype): New attribute.
19156 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
19159 * config/arm/arm-builtins.c (arm_expand_builtin): Change
19160 expansion to perform a bitwise AND of the argument followed by a
19161 boolean negation of the result.
19163 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
19165 PR rtl-optimization/84878
19166 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19167 the basic block. Assert the use reference is not artificial and that
19168 it has an associated insn.
19170 2018-04-04 Michael Matz <matz@suse.de>
19172 * builtins.c (compute_objsize): Pass correct operand
19173 to array_at_struct_end_p.
19175 2018-04-04 Richard Biener <rguenther@suse.de>
19178 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19179 from contexts for DINFO_LEVEL_TERSE and below.
19181 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
19183 * config/nds32/nds32-doubleword.md (move_<mode>): Require
19184 resiter_operand condition.
19185 * config/nds32/nds32.md (*move<mode>): Ditto.
19187 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
19188 Monk Chiang <sh.chiang04@gmail.com>
19190 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19192 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19194 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19196 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19197 Kito Cheng <kito.cheng@gmail.com>
19199 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19200 nds32_cond_code_str, output_cond_branch,
19201 output_cond_branch_compare_zero, nds32_expand_cbranch,
19202 nds32_expand_cstore, nds32_expand_movcc,
19203 nds32_output_cbranchsi4_equality_zero,
19204 nds32_output_cbranchsi4_equality_reg,
19205 nds32_output_cbranchsi4_equality_reg_or_const_int,
19206 nds32_output_cbranchsi4_greater_less_zero: New functions.
19207 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19208 nds32_expand_cstore, nds32_expand_movcc,
19209 nds32_output_cbranchsi4_equality_zero,
19210 nds32_output_cbranchsi4_equality_reg,
19211 nds32_output_cbranchsi4_equality_reg_or_const_int,
19212 nds32_output_cbranchsi4_greater_less_zero): Declare.
19213 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19214 nds32_rimm11s_operand): New predicates.
19215 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
19216 * config/nds32/nds32.md: Rewrite all the branch and conditional move
19219 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
19221 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
19222 * config/nds32/nds32.md: Ditto.
19223 * config/nds32/pipelines.md: Ditto.
19225 2018-04-04 Richard Biener <rguenther@suse.de>
19227 PR tree-optimization/85168
19228 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
19229 propagating abnormals.
19231 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19233 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
19235 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19236 Kito Cheng <kito.cheng@gmail.com>
19238 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
19239 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
19240 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
19241 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
19242 * config/nds32/nds32.md (sibcall_internal): New.
19243 (sibcall_register): Remove.
19244 (sibcall_immediate): Remove.
19245 (sibcall_value_internal): New.
19246 (sibcall_value_register): Remove.
19247 (sibcall_value_immediate): Remove.
19248 * config/nds32/predicates.md (nds32_general_register_operand): New.
19249 (nds32_call_address_operand): New.
19251 2018-04-03 Jakub Jelinek <jakub@redhat.com>
19253 PR rtl-optimization/85167
19254 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
19255 bb_defs if *split_p, instead preinitialize it to NULL.
19257 PR tree-optimization/85156
19258 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
19259 evaluating the argument multiple times.
19261 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
19263 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
19265 (_mm_cvtpd_ps): Likewise.
19266 (_mm_cvttpd_epi32): Likewise.
19267 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
19268 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
19269 vector, pixel, and bool following altivec.h include.
19271 2018-04-03 Martin Sebor <msebor@redhat.com>
19273 * doc/extend.texi (Common Function Attributes): Clarify.
19274 (const attribute): Likewise.
19275 (pure attribute): Likewise.
19277 2018-04-03 Jakub Jelinek <jakub@redhat.com>
19280 * config/i386/i386.c (ix86_expand_vector_set): Use
19281 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
19283 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
19285 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
19286 instructions when changing rounding bits to preserve precision bits
19287 in the x87 control word.
19289 2018-04-03 Martin Liska <mliska@suse.cz>
19291 PR tree-optimization/82491
19292 * rtl.h (strip_offset_and_add): Replace += suboffset with
19293 poly_uint64 () + suboffset.
19295 2018-03-29 Martin Liska <mliska@suse.cz>
19296 Martin Jambor <mjambor@suse.cz>
19299 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
19300 param_type is not an integral or pointer type.
19302 2018-04-03 Richard Biener <rguenther@suse.de>
19304 * sese.h (recompute_all_dominators): Remove.
19306 2018-04-02 Martin Sebor <msebor@redhat.com>
19308 * doc/invoke.texi (-Wrestrict): Fix typos.
19310 2018-04-02 Jim Wilson <jimw@sifive.com>
19312 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
19313 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
19314 (<optab>di3, <optab>si3_extend): Likewise.
19315 (<optab>si3_mask, <optab>si3_mask_1): New.
19316 (<optab>di3_mask, <optab>di3_mask_1): New.
19317 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
19318 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
19319 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
19321 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
19323 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
19326 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
19328 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
19329 (nds32_canonicalize_comparison): New function.
19331 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
19332 Kito Cheng <kito.cheng@gmail.com>
19333 Kuan-Lin Chen <kuanlinchentw@gmail.com>
19335 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
19336 * config/nds32/constants.md (unspec_volatile_element): Add
19337 UNSPEC_VOLATILE_RELAX_GROUP.
19338 * config/nds32/nds32-relax-opt.c: New file.
19339 * config/nds32/nds32-predicates.c
19340 (nds32_symbol_load_store_p): New function.
19341 * config/nds32/nds32-protos.h
19342 (nds32_symbol_load_store_p): Declare function.
19343 (make_pass_nds32_relax_opt): Declare new rtl pass function.
19344 * config/nds32/nds32.c
19345 (nds32_register_pass): New function to register pass.
19346 (nds32_register_passes): New function to register passes.
19347 * config/nds32/nds32.md (relax_group): New pattern.
19348 * config/nds32/nds32.opt (mrelax-hint): New option.
19349 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
19351 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
19353 * config/nds32/t-nds32: Modify files dependency.
19355 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
19357 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
19358 (PROFILE_HOOK): Define its implementation.
19360 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
19362 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
19363 type and 32-bit size.
19365 2018-04-01 Jakub Jelinek <jakub@redhat.com>
19367 PR middle-end/85090
19368 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
19369 (V_128_256): New mode iterator.
19370 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
19371 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
19372 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
19374 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
19375 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
19377 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
19380 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
19381 NaN inputs correctly.
19383 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
19386 * config/rs6000/vsx.md (??r): New mode attribute.
19387 (*vsx_mov<mode>_64bit): Use it.
19388 (*vsx_mov<mode>_32bit): Likewise.
19390 2018-03-30 Martin Sebor <msebor@redhat.com>
19392 PR tree-optimization/84818
19393 * builtins.c (check_access): Use warning_n.
19395 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19398 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
19400 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
19403 2018-03-30 Julia Koval <julia.koval@intel.com>
19406 * x86-tune.def (movx, partial_reg_dependency): Enable for
19409 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
19411 PR inline-asm/84985
19412 * lra-constraints.c (process_alt_operands): Move setting
19413 this_alternative_matches below.
19415 2018-03-29 Martin Liska <mliska@suse.cz>
19418 * doc/invoke.texi: Document how LTO works with debug info.
19419 Describe auto-load support of binutils. Mention 'x86-64'
19420 as valid option value of -march option.
19422 2018-03-29 Jakub Jelinek <jakub@redhat.com>
19424 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
19427 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
19428 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
19429 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
19432 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
19435 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
19436 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
19437 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
19438 for RS6000_BTM_POWERPC64.
19439 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
19440 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
19441 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
19444 (DIVDEU): Likewise.
19446 2018-03-28 Carl Love <cel@us.ibm.com>
19449 2017-09-27 Carl Love <cel@us.ibm.com>
19451 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
19452 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
19453 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
19456 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19458 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
19459 instead of __vector bool.
19460 (_mm_max_pu8): Likewise.
19461 (_mm_min_pi16): Likewise.
19463 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
19466 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
19467 (DIVWEUO): Likewise.
19468 (DIVDEO): Likewise.
19469 (DIVDEUO): Likewise.
19470 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
19471 DIVWEUO and DIVDEUO.
19472 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
19473 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
19474 (div_extend): Likewise.
19475 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
19477 (__builtin_divweuo): Likewise.
19478 (__builtin_divdeo): Likewise.
19479 (__builtin_divdeuo): Likewise.
19481 2018-03-28 Jakub Jelinek <jakub@redhat.com>
19484 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
19485 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
19487 PR tree-optimization/82004
19488 * gimple-match-head.c (optimize_pow_to_exp): New function.
19489 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
19490 Don't fold to exp if optimize_pow_to_exp is false.
19492 2018-03-28 Martin Liska <mliska@suse.cz>
19495 * calls.c (initialize_argument_information): Fix trailing space.
19496 * common.opt: Fix typo and provide better explanation for
19497 -fsanitize-coverage option.
19498 * config/i386/i386.opt: Fix typo.
19500 2018-03-28 Jakub Jelinek <jakub@redhat.com>
19501 Martin Liska <mliska@suse.cz>
19504 * gimplify.c (asan_poison_variable): Don't do the check for
19505 gimplify_omp_ctxp here.
19506 (gimplify_decl_expr): Do it here.
19507 (gimplify_target_expr): Likewise.
19509 2018-03-28 Martin Liska <mliska@suse.cz>
19512 * config/i386/i386.c (ix86_function_arg_advance): Do not call
19513 chkp_type_bounds_count if MPX is not enabled.
19515 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
19517 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
19519 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
19522 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
19523 function to create the function decl for complex long double
19524 multiply and divide for -mabi=ieeelongdouble.
19525 (init_float128_ieee): Call it.
19527 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
19530 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
19531 -fcf-protection=branch -mibt.
19532 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
19534 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19537 * config/arm/arm.c (arm_valid_symbolic_address): Handle
19538 arm_word_relocations.
19540 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
19543 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
19544 extern array declarations.
19546 2018-03-27 Richard Biener <rguenther@suse.de>
19548 PR middle-end/84067
19549 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
19550 explicit single_use checks.
19552 2018-03-27 Richard Biener <rguenther@suse.de>
19554 PR tree-optimization/85082
19555 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
19558 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19560 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
19561 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
19562 Turn on fasynchronous-unwind-tables and funwind-tables.
19564 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
19567 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
19568 (*bmi_blsr_<mode>_ccz): Ditto.
19570 2018-03-26 Tom de Vries <tom@codesourcery.com>
19572 PR tree-optimization/85063
19573 * omp-general.c (offloading_function_p): New function. Factor out
19575 * omp-offload.c (pass_omp_target_link::gate): ... here.
19576 * omp-general.h (offloading_function_p): Declare.
19577 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
19578 with attribute omp declare target for offloading functions.
19580 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
19582 PR tree-optimization/84005
19583 * tree-data-ref.h (get_base_for_alignment): Declare.
19584 * tree-data-ref.c (get_base_for_alignment_1): New function.
19585 (get_base_for_alignment): Likewise.
19586 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
19587 get_base_for_alignment to find a suitable base object, instead
19588 of always using drb->base_address.
19590 2018-03-23 Jakub Jelinek <jakub@redhat.com>
19592 PR inline-asm/85022
19593 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
19594 known size by default.
19596 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
19598 PR inline-asm/85030
19599 * lra-constraints.c (process_alt_operands): Don't match BLKmode
19600 and non BLKmode operands.
19602 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19605 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
19606 Clean up attributes.
19608 2018-03-23 Richard Biener <rguenther@suse.de>
19611 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
19612 we are going to emit early debug for LTO.
19614 2018-03-23 Jakub Jelinek <jakub@redhat.com>
19616 PR inline-asm/85034
19617 * function.c (match_asm_constraints_1): Don't optimize if input
19618 doesn't satisfy general_operand predicate for output's mode.
19620 PR inline-asm/85022
19621 * alias.c (write_dependence_p): Don't require for x_canonicalized
19622 non-VOIDmode if x has VOIDmode.
19625 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
19626 just don't try to optimize it rather than assert it never happens.
19628 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
19630 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
19631 macro expansions for definition of ST_INTERNAL_<mode> and
19632 LD_INTERNAL_<mode> builtins.
19633 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
19635 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
19637 (altivec_expand_st_builtin): Likewise.
19638 (altivec_expand_builtin): Remove calls to deleted functions.
19639 (rs6000_address_for_altivec): Delete this function.
19640 * config/rs6000/vector.md: Remove expands for
19641 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
19643 2018-03-22 Sudakshina Das <sudi.das@arm.com>
19646 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
19647 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
19648 re-computing once computed.
19649 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
19650 (arm_init_machine_status): Initialize
19651 machine->static_chain_stack_bytes.
19653 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
19656 * doc/extend.texi: Add four new prototypes for vec_ld.
19657 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
19658 definitions for more logical presentation.
19659 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
19660 entries for V1TI variants of __builtin_altivec_ld builtin.
19661 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
19662 handling of V1TI variant of LVX icode pattern.
19663 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
19664 (rs6000_gimple_fold_builtin): Likewise.
19665 (altivec_init_builtins): Add code to define
19666 __builtin_altivec_lvx_v1ti function.
19668 2018-03-22 Jakub Jelinek <jakub@redhat.com>
19670 PR inline-asm/84941
19671 * function.c (match_asm_constraints_1): Don't do the optimization
19672 if input isn't a REG, SUBREG, MEM or constant.
19674 2018-03-22 Tom de Vries <tom@codesourcery.com>
19676 PR tree-optimization/84956
19677 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
19678 bb_has_abnormal_pred.
19680 2018-03-22 Jakub Jelinek <jakub@redhat.com>
19683 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
19684 DECL_INITIAL (decl) to decl at the end.
19685 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
19686 adjust the comment.
19688 2018-03-21 Joseph Myers <joseph@codesourcery.com>
19690 * doc/extend.texi (__builtin_tgmath): Document when complex
19691 integer types are treated as _Complex _Float64.
19693 2018-03-21 Tom de Vries <tom@codesourcery.com>
19695 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
19697 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19699 PR tree-optimization/84960
19700 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
19701 if it is ENTRY block, move them into single succ of ENTRY in that case.
19703 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
19705 PR tree-optimization/84811
19706 * poly-int.h (poly_span_traits): Remove the T3 parameter and
19707 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
19708 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
19709 (known_subrange_p): Update accordingly. Cast each value involved
19710 in the size comparison, rather than casting the result of the
19713 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19715 PR tree-optimization/84982
19716 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
19717 by flipping the least significant bit rather than all bits from
19718 bitpos to bitpos + bitsize - 1.
19720 2018-03-21 Nathan Sidwell <nathan@acm.org>
19722 * doc/extend.texi (Deprecated Features): Remove mention of
19723 long-deleted deprecations.
19725 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19728 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
19729 * configure: Regenerate.
19731 2018-03-21 Tom de Vries <tom@codesourcery.com>
19733 PR tree-optimization/83126
19734 * tree-parloops.c (num_phis): New function.
19735 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
19737 2018-03-21 Nathan Sidwell <nathan@acm.org>
19739 * doc/extend.texi (Deprecated Features): Update deprecated flags,
19740 mention anon-struct/union members and trailing attributes.
19742 2018-03-21 Bin Cheng <bin.cheng@arm.com>
19744 PR tree-optimization/84969
19745 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
19746 builtin memset partitions if they set different rhs values.
19748 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19750 PR rtl-optimization/84989
19751 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
19752 VEC_DUPLICATE with scalar result mode.
19754 2018-03-21 Martin Liska <mliska@suse.cz>
19757 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
19758 not intended return statement.
19760 2018-03-21 Martin Liska <mliska@suse.cz>
19763 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
19764 (chkp_find_bound_slots_1): Limit number of iterations.
19766 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
19769 * Minor grammar fixes for x86 options.
19771 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19774 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
19775 holding REG_CFA_RESTORE notes, instead turn them into a USE.
19777 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
19780 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
19781 (altivec_lvx_<mode>_1op): Likewise.
19782 (altivec_stvx_<mode>_2op): Likewise.
19783 (altivec_stvx_<mode>_1op): Likewise.
19784 (altivec_lvx_<VM2:mode>): New define_expand.
19785 (altivec_stvx_<VM2:mode>): Likewise.
19786 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
19787 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19788 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
19789 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19790 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
19791 (rs6000_gen_lvx): Likewise.
19792 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
19793 (altivec_expand_stv_builtin): Likewise.
19794 (altivec_expand_builtin): Likewise.
19795 * config/rs6000/vector.md: Likewise.
19797 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19800 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
19803 2018-03-20 Richard Biener <rguenther@suse.de>
19806 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
19807 sign-conversions as zero, fall back to standard scalar_stmt
19810 2018-03-20 Martin Liska <mliska@suse.cz>
19813 * predict.c (rebuild_frequencies): Handle case when we have
19814 PROFILE_ABSENT, but flag_guess_branch_prob is false.
19816 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19819 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
19820 flag_section_anchors.
19821 * varasm.c (use_blocks_for_decl_p): Remove hack for
19822 dw2_force_const_mem.
19825 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
19827 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
19828 be created, use lowpart_subreg of operands[0] rather than operands[0]
19830 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
19831 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
19832 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
19833 and n constraint instead of aarch64_shift_imm_di and Usd.
19834 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
19835 (*aarch64_<optab>_reg_minus<mode>3): ... this.
19837 2018-03-20 Sudakshina Das <sudi.das@arm.com>
19840 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
19841 to favor GPR over NEON registers.
19842 (<shift>di3_neon): Likewise.
19844 2018-03-20 Tom de Vries <tom@codesourcery.com>
19847 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
19848 (nvptx_process_pars): Emit bar.sync asap and alap.
19850 2018-03-20 Tom de Vries <tom@codesourcery.com>
19853 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
19854 seen_label if seen_label is already set.
19856 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19859 * config/i386/i386.c (fold_builtin_cpu): For features above 31
19860 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
19861 Use 1U instead of 1. Formatting fixes.
19864 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
19865 instead of TREE_TYPE (s1) for the return value.
19867 2018-03-19 Jakub Jelinek <jakub@redhat.com>
19869 PR tree-optimization/84946
19870 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
19871 bitsize + bitsize in poly_uint64 rather than poly_int64.
19874 * dwarf2asm.c: Include fold-const.c.
19875 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
19876 of decl rather than decl itself.
19878 PR rtl-optimization/84643
19879 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
19881 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
19884 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
19885 calling assemble_variable.
19887 2018-03-19 Sudakshina Das <sudi.das@arm.com>
19890 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
19891 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
19893 2018-03-19 Jim Wilson <jimw@sifive.com>
19896 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
19897 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
19898 (riscv_first_stack_step): Likewise.
19899 (riscv_option_override): Use STACK_BOUNDARY instead of
19900 MIN_STACK_BOUNDARY.
19901 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
19902 MIN_STACK_BOUNDARY.
19903 (BIGGEST_ALIGNMENT): Set to 128.
19904 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19905 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
19908 2018-03-19 Richard Biener <rguenther@suse.de>
19910 PR tree-optimization/84933
19911 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
19912 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19914 2018-03-19 Richard Biener <rguenther@suse.de>
19916 PR tree-optimization/84859
19917 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19918 (cond_if_else_store_replacement): Perform sinking operation on
19919 single-store BBs regardless of MAX_STORES_TO_SINK setting.
19920 Generalize what a BB with a single eligible store is.
19922 2018-03-19 Richard Biener <rguenther@suse.de>
19924 PR tree-optimization/84929
19925 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19926 chrec_is_positive against non-chrec arg.
19928 2018-03-19 Tamar Christina <tamar.christina@arm.com>
19931 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19933 2018-03-18 Martin Liska <mliska@suse.cz>
19935 PR rtl-optimization/84635
19936 * regrename.c (build_def_use): Use matches_mode only when
19939 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
19941 PR tree-optimization/84913
19942 * tree-vect-loop.c (vectorizable_reduction): Don't try to
19943 vectorize chains of COND_EXPRs.
19945 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19947 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19949 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19951 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19953 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19955 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19957 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19958 Kito Cheng <kito.cheng@gmail.com>
19960 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19961 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19962 (nds32_adjust_reg_alloc_order): New function.
19963 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19965 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
19967 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19968 nds32_print_operand, nds32_print_operand_address): Use
19969 HOST_WIDE_INT_PRINT_DEC instead.
19971 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19973 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19975 2018-03-17 Jakub Jelinek <jakub@redhat.com>
19978 * config/i386/i386.c (initial_ix86_tune_features,
19979 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19980 unsigned long long.
19981 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19982 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19983 rather than 1u << ix86_tune. Formatting fix.
19984 (ix86_option_override_internal): Change ix86_arch_mask from
19985 unsigned int to unsigned HOST_WIDE_INT, initialize to
19986 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19987 (ix86_function_specific_restore): Likewise.
19989 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19992 * postreload.c (reload_combine_recognize_pattern): Perform
19993 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19994 truncate_int_for_mode the result for the destination's mode.
19997 * hsa-gen.c (mem_type_for_type): Fix comment typo.
19998 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
20000 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20003 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
20006 * lra-assigns.c (lra_split_hard_reg_for): Don't use
20007 regno_allocno_class_array and sorted_pseudos.
20008 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
20009 insns where regno is used.
20011 2018-03-16 Martin Liska <mliska@suse.cz>
20014 * multiple_target.c (create_dispatcher_calls): Redirect
20015 reference in the symbol table.
20017 2018-03-16 Martin Liska <mliska@suse.cz>
20020 * multiple_target.c (create_dispatcher_calls): Redirect also
20023 2018-03-16 Jakub Jelinek <jakub@redhat.com>
20027 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
20028 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
20029 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
20031 2018-03-16 Julia Koval <julia.koval@intel.com>
20033 * doc/invoke.texi (Skylake Server): Add CLWB.
20034 Cannonlake): Remove CLWB.
20036 2018-03-16 Jakub Jelinek <jakub@redhat.com>
20038 PR tree-optimization/84841
20039 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
20041 (FLOAT_ONE_CONST_TYPE): Define.
20042 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
20043 (sort_by_operand_rank): Put entries with higher constant_type last
20044 rather than first to match comments.
20046 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
20048 * config/nios2/nios2.md (movsi_internal): Fix thinko in
20051 2018-03-15 Jakub Jelinek <jakub@redhat.com>
20054 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
20055 check and use address of target always.
20057 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
20060 * config/i386/i386.c (indirect_thunk_needed): Update comments.
20061 (indirect_thunk_bnd_needed): Likewise.
20062 (indirect_thunks_used): Likewise.
20063 (indirect_thunks_bnd_used): Likewise.
20064 (indirect_return_needed): New.
20065 (indirect_return_bnd_needed): Likewise.
20066 (output_indirect_thunk_function): Add a bool argument for
20068 (output_indirect_thunk_function): Don't generate alias for
20069 function return thunk.
20070 (ix86_code_end): Call output_indirect_thunk_function to generate
20071 function return thunks.
20072 (ix86_output_function_return): Set indirect_return_bnd_needed
20073 and indirect_return_needed instead of indirect_thunk_bnd_needed
20074 and indirect_thunk_needed.
20076 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
20078 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
20079 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
20080 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
20082 2018-03-15 David Malcolm <dmalcolm@redhat.com>
20083 Paul Hua <paul.hua.gm@gmail.com>
20086 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
20088 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
20090 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
20091 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
20092 resp. SFmode cases.
20094 2018-03-15 Tamar Christina <tamar.christina@arm.com>
20097 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
20098 instead of GET_MODE_SIZE when comparing Units.
20100 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
20103 * varasm.c (hash_section): Return an unchangeble hash value
20104 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
20105 Return !aarch64_can_use_per_function_literal_pools_p ().
20107 2018-03-15 Jakub Jelinek <jakub@redhat.com>
20110 * optabs.c (emit_conditional_move): Pass address of cmode's copy
20111 rather than address of cmode as last argument to prepare_cmp_insn.
20113 2018-03-15 Julia Koval <julia.koval@intel.com>
20115 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20116 F_AVX512VNNI, F_AVX512BITALG): New.
20118 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
20121 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
20122 insn for floating-point loads and stores.
20124 2018-03-14 Carl Love <cel@us.ibm.com>
20126 * config/rs6000/rs6000-c.c: Add macro definitions for
20127 ALTIVEC_BUILTIN_VEC_PERMXOR.
20128 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20129 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20130 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20131 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20133 * config/doc/extend.texi: Add prototypes for vec_permxor.
20135 2018-03-14 David Malcolm <dmalcolm@redhat.com>
20138 * diagnostic-show-locus.c (class layout_point): Convert m_line
20139 from int to linenum_type.
20140 (line_span::comparator): Use linenum "compare" function when
20141 comparing line numbers.
20142 (test_line_span): New function.
20143 (layout_range::contains_point): Convert param "row" from int to
20145 (layout_range::intersects_line_p): Likewise.
20146 (layout::will_show_line_p): Likewise.
20147 (layout::print_source_line): Likewise.
20148 (layout::should_print_annotation_line_p): Likewise.
20149 (layout::print_annotation_line): Likewise.
20150 (layout::print_leading_fixits): Likewise.
20151 (layout::annotation_line_showed_range_p): Likewise.
20152 (struct line_corrections): Likewise for field m_row.
20153 (line_corrections::line_corrections): Likewise for param "row".
20154 (layout::print_trailing_fixits): Likewise.
20155 (layout::get_state_at_point): Likewise.
20156 (layout::get_x_bound_for_row): Likewise.
20157 (layout::print_line): Likewise.
20158 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
20159 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
20160 * input.c (selftest::test_linenum_comparisons): New function.
20161 (selftest::input_c_tests): Call it.
20162 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
20163 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
20164 * selftest.h (ASSERT_GT): New macro.
20165 (ASSERT_GT_AT): New macro.
20166 (ASSERT_LT): New macro.
20167 (ASSERT_LT_AT): New macro.
20169 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
20171 PR rtl-optimization/84780
20172 * combine.c (distribute_links): Don't make a link based on pc_rtx.
20174 2018-03-14 Martin Liska <mliska@suse.cz>
20176 * tree.c (record_node_allocation_statistics): Use
20177 get_stats_node_kind.
20178 (get_stats_node_kind): New function extracted from
20179 record_node_allocation_statistics.
20180 (free_node): Use get_stats_node_kind.
20182 2018-03-14 Richard Biener <rguenther@suse.de>
20184 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20185 that the value-set of ANTIC_IN doesn't grow.
20188 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20190 (BB_VISITED_WITH_VISITED_SUCCS): New define.
20191 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20193 2018-03-14 Julia Koval <julia.koval@intel.com>
20195 * config.gcc (icelake-client, icelake-server): New.
20197 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20198 (initial_ix86_arch_features): Ditto.
20199 (PTA_SKYLAKE): Add SGX.
20200 (PTA_ICELAKE): Remove.
20201 (PTA_ICELAKE_CLIENT): New.
20202 (PTA_ICELAKE_SERVER): New.
20203 (ix86_option_override_internal): Split up icelake on icelake client and
20205 (get_builtin_code_for_version): Ditto.
20206 (fold_builtin_cpu): Ditto.
20207 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20208 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20209 * config/i386/i386.h (processor_type): Ditto.
20210 * doc/invoke.texi: Ditto.
20212 2018-03-14 Jakub Jelinek <jakub@redhat.com>
20215 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
20216 INTEGER_CST offset, add it together with bitpos / 8 and
20217 sign extend based on POINTER_SIZE.
20221 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
20224 * config/i386/constraints.md (Yc): New register constraint.
20225 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
20226 Use Yc constraint for alternative 2 of operand 0. Remove
20227 preferred_for_speed attribute.
20229 2018-03-14 Richard Biener <rguenther@suse.de>
20231 PR tree-optimization/84830
20232 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
20233 with the old one to avoid oscillations.
20235 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
20238 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
20240 (assign_by_spills): Return a flag of reload assignment failure.
20241 Do not process the reload assignment failures. Do not spill other
20242 reload pseudos if they has the same reg class. Update n if
20244 (lra_assign): Add a return arg. Set up from the result of
20245 assign_by_spills call.
20246 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20247 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
20248 usage_insns if it is not NULL.
20249 (spill_hard_reg_in_range): New function.
20250 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20251 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20252 function prototypes.
20253 (lra_assign): Change prototype.
20254 * lra.c (lra): Add code to deal with fails by splitting hard reg
20257 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
20259 * config/riscv/riscv.opt (mrelax): New option.
20260 * config/riscv/riscv.c (riscv_file_start): Emit ".option
20261 "norelax" when riscv_mrelax is disabled.
20262 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
20264 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20267 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
20268 reassociation for int modes.
20270 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
20272 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
20273 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
20275 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
20276 * config/aarch64/aarch64-sve.md
20277 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
20278 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
20279 (*extend<mode><Vwide>2): Rename to...
20280 (aarch64_sve_extend<mode><Vwide>2): ...this.
20281 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
20282 renaming the old pattern to...
20283 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
20285 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
20286 define_expand, renaming the old pattern to...
20287 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
20288 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
20289 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
20290 account when deciding which SVE instruction the optab should use.
20291 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
20293 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
20295 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
20296 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
20297 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
20298 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
20299 (tlsdesc_small_<mode>): Turn a define_expand and use
20300 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
20301 (tlsdesc_small_advsimd_<mode>): ...this.
20302 (tlsdesc_small_sve_<mode>): New pattern.
20304 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
20306 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
20307 (UNSPEC_UMUL_HIGHPART): New constants.
20308 (MUL_HIGHPART): New int iteraor.
20309 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
20310 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20312 (*<su>mul<mode>3_highpart): New define_insn.
20314 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
20317 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20320 2018-03-13 Martin Liska <mliska@suse.cz>
20323 * (sem_item_optimizer::sem_item_optimizer): Initialize new
20325 (sem_item_optimizer::~sem_item_optimizer): Release it.
20326 (sem_item_optimizer::merge_classes): Register variable aliases.
20327 (sem_item_optimizer::fixup_pt_set): New function.
20328 (sem_item_optimizer::fixup_points_to_sets): Likewise.
20329 * ipa-icf.h: Declare new variables and functions.
20331 2018-03-13 Jakub Jelinek <jakub@redhat.com>
20333 PR middle-end/84834
20334 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
20335 integer_pow2p@2 and test integer_pow2p in condition.
20336 (A < 0 ? C : 0): Similarly for @1.
20338 PR middle-end/84831
20339 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
20340 characters starting at p contain '\0' character, don't look beyond
20344 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
20345 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
20348 * reg-stack.c (change_stack): Change update_end var from int to
20349 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
20350 also call set_block_for_insn on the newly added insns and rescan.
20353 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
20354 on the last operand.
20357 * tree.c (stabilize_reference_1): Return save_expr (e) for
20358 STATEMENT_LIST even if it doesn't have side-effects.
20360 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
20362 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
20364 2018-03-12 Renlin Li <renlin.li@arm.com>
20366 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
20367 aarch64_output_scalar_simd_mov_immediate.
20369 2018-03-12 Martin Sebor <msebor@redhat.com>
20371 PR tree-optimization/83456
20372 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
20373 for perfectly overlapping calls to memcpy.
20374 (gimple_fold_builtin_memory_chk): Same.
20375 (gimple_fold_builtin_strcpy): Handle no-warning.
20376 (gimple_fold_builtin_stxcpy_chk): Same.
20377 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
20379 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
20381 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
20382 parameter. Use it for SFmode.
20383 (rs6000_function_arg_advance_1): Adjust.
20384 (rs6000_function_arg): Adjust.
20385 (rs6000_gimplify_va_arg): Pass false for that new parameter.
20387 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
20389 PR rtl-optimization/84169
20390 PR rtl-optimization/84780
20391 * combine.c (can_combine_p): Check for a 2-insn combination whether
20392 the destination register is used between the two insns, too.
20394 2018-03-12 Richard Biener <rguenther@suse.de>
20396 PR tree-optimization/84803
20397 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20398 for refs DR analysis didn't process.
20400 2018-03-12 Richard Biener <rguenther@suse.de>
20402 PR tree-optimization/84777
20403 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
20404 force-vectorize loops ignore whether we are optimizing for size.
20406 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
20408 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
20409 (TARGET_MD_ASM_ADJUST): Define.
20411 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
20412 Kito Cheng <kito.cheng@gmail.com>
20413 Chung-Ju Wu <jasonwucj@gmail.com>
20415 * config/nds32/nds32.c (nds32_compute_stack_frame,
20416 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
20417 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
20418 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
20419 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
20420 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
20421 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
20422 * config/nds32/nds32.md (prologue, epilogue): Use macro
20423 NDS32_V3PUSH_AVAILABLE_P to do checking.
20425 2018-03-11 Jakub Jelinek <jakub@redhat.com>
20428 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
20429 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
20430 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
20431 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
20432 addition of most attributes on !orig_type_die or the attribute not
20433 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
20435 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
20436 Chung-Ju Wu <jasonwucj@gmail.com>
20438 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
20439 __NDS32_VH__ macro.
20440 * config/nds32/nds32.opt (mvh): New option.
20442 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
20443 Chung-Ju Wu <jasonwucj@gmail.com>
20445 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
20447 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
20448 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
20451 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
20452 Chung-Ju Wu <jasonwucj@gmail.com>
20454 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
20456 * config/nds32/nds32-multiple.md (strlensi): New pattern.
20457 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
20459 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
20460 Kito Cheng <kito.cheng@gmail.com>
20461 Chung-Ju Wu <jasonwucj@gmail.com>
20463 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
20464 UNSPEC_FFMISM and UNSPEC_FLMISM.
20465 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
20466 for ffb, ffmism and flmism.
20467 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
20468 (unspec_ffmism): Ditto.
20469 (unspec_flmism): Ditto.
20470 (nds32_expand_builtin_impl): Check if string extension is available.
20471 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
20472 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
20474 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
20477 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
20480 * lra-assigns.c (assign_by_spills): Return a flag of reload
20481 assignment failure. Do not process the reload assignment
20482 failures. Do not spill other reload pseudos if they has the same
20484 (lra_assign): Add a return arg. Set up from the result of
20485 assign_by_spills call.
20486 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20487 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
20488 usage_insns if it is not NULL.
20489 (spill_hard_reg_in_range): New function.
20490 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20491 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20492 function prototypes.
20493 (lra_assign): Change prototype.
20494 * lra.c (lra): Add code to deal with fails by splitting hard reg
20497 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
20500 * config/i386/i386.opt: Replace Enforcment with Enforcement.
20502 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
20505 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
20506 (dw_val_node): Add val_symbolic_view.
20507 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
20508 (symview_upper_bound): New.
20509 (new_line_info_table): Initialize symviews_since_reset.
20510 (dwarf2out_source_line): Count symviews_since_reset and set
20511 symview_upper_bound.
20512 (dw_val_equal_p): Handle symview.
20513 (add_AT_symview): New.
20514 (print_dw_val): Handle symview.
20515 (attr_checksum, attr_checksum_ordered): Likewise.
20516 (same_dw_val_p, size_of_die): Likewise.
20517 (value_format, output_die): Likewise.
20518 (add_high_low_attributes): Use add_AT_symview for entry_view.
20519 (dwarf2out_finish): Reset symview_upper_bound, clear
20522 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
20525 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
20526 Add strict argument and use it.
20527 (rs6000_split_multireg_move): Update for new strict argument.
20528 (mem_operand_gpr): Disallow all non-offsettable addresses.
20529 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
20531 2018-03-09 Jakub Jelinek <jakub@redhat.com>
20534 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
20535 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
20536 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
20539 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
20540 decl, use remap_type if we want to use the type.
20542 2018-03-09 Martin Sebor <msebor@redhat.com>
20544 PR tree-optimization/84526
20545 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20547 (builtin_access::generic_overlap): Be prepared to handle non-array
20550 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
20552 PR rtl-optimization/84682
20553 * lra-constraints.c (process_address_1): Check is_address flag
20554 for address constraints.
20555 (process_alt_operands): Likewise.
20556 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
20557 preprocess_constraints.
20558 * recog.h (preprocess_constraints): Add oploc parameter.
20560 * recog.c (preprocess_constraints): Test address_operand for
20561 CT_ADDRESS constraints.
20563 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
20566 * lra-assigns.c (assign_by_spills): Return a flag of reload
20567 assignment failure. Do not process the reload assignment
20568 failures. Do not spill other reload pseudos if they has the same
20570 (lra_assign): Add a return arg. Set up from the result of
20571 assign_by_spills call.
20572 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20573 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
20574 usage_insns if it is not NULL.
20575 (spill_hard_reg_in_range): New function.
20576 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20577 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20578 function prototypes.
20579 (lra_assign): Change prototype.
20580 * lra.c (lra): Add code to deal with fails by splitting hard reg
20583 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20586 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
20587 Accept complain bool parameter. Only emit errors if it is true.
20588 (arm_parse_cpu_option_name): Likewise.
20589 (arm_target_thumb_only): Adjust callers of the above.
20590 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
20591 prototype to take a default true bool parameter.
20592 (arm_parse_arch_option_name): Likewise.
20594 2018-03-09 David Malcolm <dmalcolm@redhat.com>
20595 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
20599 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
20600 * configure: Regenerate.
20601 * configure.ac ("linker --version-script option"): New.
20602 ("linker soname option"): New.
20604 2018-03-09 Richard Biener <rguenther@suse.de>
20606 PR tree-optimization/84775
20607 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
20608 immediate uses of predicate stmts and mark them modified.
20611 PR tree-optimization/84178
20612 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20614 (version_loop_for_if_conversion): Delay update_ssa call.
20615 (tree_if_conversion): Delay update_ssa until after predicate
20618 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
20621 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
20622 when the function accesses prior frames.
20624 2018-03-08 Jakub Jelinek <jakub@redhat.com>
20627 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
20628 gen_llsym, otherwise call maybe_gen_llsym.
20630 PR inline-asm/84742
20631 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
20632 has ',' character inside of it.
20634 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20637 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
20638 as clobbering CC_REGNUM.
20640 2018-03-08 Richard Biener <rguenther@suse.de>
20642 PR middle-end/84552
20643 * tree-scalar-evolution.c: Include tree-into-ssa.h.
20644 (follow_copies_to_constant): Do not follow SSA names registered
20647 2018-03-08 Richard Biener <rguenther@suse.de>
20649 PR tree-optimization/84178
20650 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20652 (version_loop_for_if_conversion): Delay update_ssa call.
20653 (tree_if_conversion): Delay update_ssa until after predicate
20656 2018-03-08 David Malcolm <dmalcolm@redhat.com>
20658 PR tree-optimization/84178
20659 * tree-if-conv.c (release_bb_predicate): Remove the
20660 the assertion that the stmts have NULL use_ops.
20661 Discard the statements, asserting that they haven't
20662 yet been added to a BB.
20664 2018-03-08 Richard Biener <rguenther@suse.de>
20666 PR tree-optimization/84746
20667 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
20668 (phi_translate): Pass in destination ANTIC_OUT set.
20669 (phi_translate_1): Likewise. For a simplified result lookup
20670 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
20671 (phi_translate_set): Adjust.
20672 (do_pre_regular_insertion): Likewise.
20673 (do_pre_partial_partial_insertion): Likewise.
20675 2018-03-08 Martin Liska <mliska@suse.cz>
20677 PR gcov-profile/84735
20678 * doc/gcov.texi: Document usage of profile files.
20679 * gcov-io.h: Document changes in the format.
20681 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
20685 * dwarf2out.c (struct dw_line_info_table): Update comments for
20687 (FORCE_RESET_NEXT_VIEW): New.
20688 (FORCE_RESETTING_VIEW_P): New.
20689 (RESETTING_VIEW_P): Check for -1 too.
20690 (ZERO_VIEW_P): Likewise.
20691 (new_line_info_table): Force-reset next view.
20692 (dwarf2out_begin_function): Likewise.
20693 (dwarf2out_source_line): Simplify zero_view_p initialization.
20694 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
20695 view directly. Omit view when omitting .loc at line 0.
20697 2018-03-08 Jakub Jelinek <jakub@redhat.com>
20699 PR tree-optimization/84740
20700 * tree-switch-conversion.c (process_switch): Call build_constructors
20701 only if info.phi_count is non-zero.
20703 PR tree-optimization/84739
20704 * tree-tailcall.c (find_tail_calls): Check call arguments against
20705 DECL_ARGUMENTS (current_function_decl) rather than
20706 DECL_ARGUMENTS (func) when checking for tail recursion.
20708 2018-03-07 Jakub Jelinek <jakub@redhat.com>
20710 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
20711 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
20712 Volker Reichelt's entry and add entries for people that perform
20713 GCC fuzzy testing and report numerous bugs.
20715 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
20718 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
20719 readonly data in sdata, if that is disabled.
20720 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
20721 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
20722 -mreadonly-in-sdata option.
20724 2018-03-07 Martin Sebor <msebor@redhat.com>
20726 PR tree-optimization/84468
20727 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
20728 basic block when looking for nul assignment.
20730 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
20733 * except.h (output_function_exception_table): Adjust prototype.
20734 * except.c (output_function_exception_table): Remove FNNAME parameter
20735 and add SECTION parameter. Ouput one part of the table at a time.
20736 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
20737 the first part of the exception table and emit unwind directives.
20738 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
20739 (i386_pe_seh_cold_init): Likewise.
20740 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
20741 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
20742 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
20743 (ix86_output_call_insn): Emit a nop in one more case for SEH.
20744 * config/i386/winnt.c: Include except.h.
20745 (struct seh_frame_state): Add reg_offset, after_prologue and
20746 in_cold_section fields.
20747 (i386_pe_seh_end_prologue): Set seh->after_prologue.
20748 (i386_pe_seh_cold_init): New function.
20749 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
20750 to seh->in_cold_section.
20751 (seh_emit_push): Record the offset of the push.
20752 (seh_emit_save): Record the offet of the save.
20753 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
20754 Test seh->after_prologue to disregard the epilogue.
20755 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
20756 (i386_pe_end_cold_function): New function.
20758 2018-03-07 Jakub Jelinek <jakub@redhat.com>
20761 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
20762 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
20765 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
20767 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
20768 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
20770 PR middle-end/84723
20771 * multiple_target.c: Include tree-inline.h and intl.h.
20772 (expand_target_clones): Diagnose and fail if node->definition and
20773 !tree_versionable_function_p (node->decl).
20775 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
20777 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
20779 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
20780 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20781 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
20783 2018-03-06 Jakub Jelinek <jakub@redhat.com>
20786 * combine.c (try_combine): Use reg_or_subregno instead of handling
20787 just paradoxical SUBREGs and REGs.
20789 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
20791 * config/arc/arc.c (arc_finalize_pic): Remove function.
20792 (arc_must_save_register): We use single base PIC register, remove
20793 checks to save/restore the PIC register.
20794 (arc_expand_prologue): Likewise.
20795 * config/arc/arc-protos.h (arc_set_default_type_attributes):
20797 (arc_verify_short): Likewise.
20798 (arc_attr_type): Likewise.
20799 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
20800 (walk_stores): Likewise.
20801 (arc_address_cost): Make it static.
20802 (arc_verify_short): Likewise.
20803 (branch_dest): Likewise.
20804 (arc_attr_type): Likewise.
20805 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
20806 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
20807 (arc_final_prescan_insn): Remove inserting the nops due to
20808 hardware hazards. It is done in reorg step.
20809 (insn_length_variant_t): Remove.
20810 (insn_length_parameters_t): Likewise.
20811 (arc_insn_length_parameters): Likewise.
20812 (arc_get_insn_variants): Likewise.
20813 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
20815 2018-03-06 Jakub Jelinek <jakub@redhat.com>
20817 PR inline-asm/84683
20818 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
20821 PR tree-optimization/84687
20822 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
20824 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
20826 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20828 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
20829 Rename to ppc_speculation_barrier.
20830 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
20831 __builtin_ppc_speculation_barrier.
20833 2018-03-05 Jakub Jelinek <jakub@redhat.com>
20836 * combine.c (combine_simplify_rtx): Don't try to simplify if
20837 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
20840 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
20842 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
20843 to 32 bytes when compiling for POWER9.
20845 2018-03-05 Jakub Jelinek <jakub@redhat.com>
20848 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
20849 regparm >= 3 with no arg reg available also for calls with
20850 flag_force_indirect_call. Pass decl to ix86_function_regparm.
20853 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
20855 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
20857 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
20860 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
20862 2018-03-05 Richard Biener <rguenther@suse.de>
20864 PR tree-optimization/84486
20865 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
20866 When inserting a __builtin_assume_aligned call set the LHS
20867 SSA name alignment info accordingly.
20869 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
20871 PR tree-optimization/84114
20872 * config/aarch64/aarch64.c (aarch64_reassociation_width)
20873 Avoid reassociation of FLOAT_MODE addition.
20875 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
20877 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
20878 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
20879 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
20880 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
20881 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
20882 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
20883 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
20885 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
20886 __builtin_ia32_wbinvd): New builtins.
20887 (SPECIAL_ARGS2): New.
20888 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
20889 (SPECIAL_ARGS2): New.
20890 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
20891 (ix86_valid_target_attribute_inner_p): Ditto.
20892 (ix86_init_mmx_sse_builtins): Add special_args2.
20893 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
20894 TARGET_WBNOINVD_P): New.
20895 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
20896 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
20897 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
20898 * config/i386/immintrin.h (_wbinvd): New intrinsic.
20899 * config/i386/pconfigintrin.h: New file.
20900 * config/i386/wbnoinvdintrin.h: Ditto.
20901 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
20903 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
20905 2018-03-05 Richard Biener <rguenther@suse.de>
20907 PR tree-optimization/84670
20908 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20910 (BB_VISITED_WITH_VISITED_SUCCS): New define.
20911 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20912 (compute_antic_aux): Only assert the number of values in ANTIC_IN
20913 doesn't grow if all successors (recursively) were visited at least
20916 2018-03-05 Richard Biener <rguenther@suse.de>
20918 PR tree-optimization/84650
20919 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20920 if executed in the loop pipeline.
20922 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
20924 * doc/configfiles.texi (Configuration Files): Move info about
20925 conditionalizing $target-protos.h to...
20926 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
20927 differs from $target-protos.h.
20929 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
20930 Chung-Ju Wu <jasonwucj@gmail.com>
20932 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20933 * config/nds32/nds32-multiple.md (setmemsi): Define.
20934 * config/nds32/nds32-memory-manipulation.c
20935 (nds32_gen_dup_4_byte_to_word_value): New.
20936 (emit_setmem_word_loop): New.
20937 (emit_setmem_byte_loop): New.
20938 (nds32_expand_setmem_loop): New.
20939 (nds32_expand_setmem_loop_v3m): New.
20940 (nds32_expand_setmem_unroll): New.
20941 (nds32_expand_setmem): New.
20943 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20944 Chung-Ju Wu <jasonwucj@gmail.com>
20946 * config/nds32/nds32-memory-manipulation.c
20947 (nds32_emit_load_store): New.
20948 (nds32_emit_post_inc_load_store): New.
20949 (nds32_emit_mem_move): New.
20950 (nds32_emit_mem_move_block): New.
20951 (nds32_expand_movmemsi_loop_unknown_size): New.
20952 (nds32_expand_movmemsi_loop_known_size): New.
20953 (nds32_expand_movmemsi_loop): New.
20954 (nds32_expand_movmemsi_unroll): New.
20955 (nds32_expand_movmemqi): Rename ...
20956 (nds32_expand_movmemsi): ... to this.
20957 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20958 (movmemsi): ... to this.
20959 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20960 (nds32_expand_movmemsi): ... to this.
20962 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20963 Monk Chiang <sh.chiang04@gmail.com>
20964 Chung-Ju Wu <jasonwucj@gmail.com>
20966 * config/nds32/nds32-protos.h
20967 (nds32_expand_load_multiple): New arguments.
20968 (nds32_expand_store_multiple): Ditto.
20969 (nds32_valid_multiple_load_store): Rename ...
20970 (nds32_valid_multiple_load_store_p): ... to this.
20971 * config/nds32/nds32-memory-manipulation.c
20972 (nds32_expand_load_multiple): Refine implementation.
20973 (nds32_expand_store_multiple): Ditto.
20974 * config/nds32/nds32-multiple.md
20975 (load_multiple): Update nds32_expand_load_multiple interface.
20976 (store_multiple): Update nds32_expand_store_multiple interface.
20977 * config/nds32/nds32-predicates.c
20978 (nds32_valid_multiple_load_store): Rename ...
20979 (nds32_valid_multiple_load_store_p): ... to this and refine
20981 * config/nds32/predicates.md
20982 (nds32_load_multiple_and_update_address_operation): New predicate.
20983 (nds32_store_multiple_and_update_address_operation): New predicate.
20985 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20986 Chung-Ju Wu <jasonwucj@gmail.com>
20988 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20989 (combo): New attribute.
20990 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20992 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
20994 * config/nds32/nds32.opt: Change -mcmodel= default value.
20996 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
20997 Monk Chiang <sh.chiang04@gmail.com>
20998 Chung-Ju Wu <jasonwucj@gmail.com>
21000 * config/nds32/constants.md (unspec_element): New enum.
21001 * config/nds32/constraints.md (Umw): New constraint.
21002 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
21003 * config/nds32/nds32-intrinsic.md: Likewise.
21004 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
21005 (nds32_valid_smw_lwm_base_p): New.
21006 (nds32_output_smw_single_word): New.
21007 (nds32_output_lmw_single_word): New.
21008 (nds32_expand_unaligned_load): New.
21009 (nds32_expand_unaligned_store): New.
21010 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
21011 (nds32_output_smw_single_word): Declare.
21012 (nds32_output_lmw_single_word): Declare.
21013 (nds32_expand_unaligned_load): Declare.
21014 (nds32_expand_unaligned_store): Declare.
21015 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
21016 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
21017 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
21018 NDS32_BUILTIN_UASTORE_DW.
21019 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
21022 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
21023 Kito Cheng <kito.cheng@gmail.com>
21024 Chung-Ju Wu <jasonwucj@gmail.com>
21026 * config/nds32/nds32-intrinsic.c
21027 (nds32_expand_builtin_null_ftype_reg): Delete.
21028 (nds32_expand_builtin_reg_ftype_imm): Ditto.
21029 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
21030 (nds32_read_argument): New.
21031 (nds32_legitimize_target): Ditto.
21032 (nds32_legitimize_argument): Ditto.
21033 (nds32_check_constant_argument): Ditto.
21034 (nds32_expand_unop_builtin): Ditto.
21035 (nds32_expand_unopimm_builtin): Ditto.
21036 (nds32_expand_binop_builtin): Ditto.
21037 (nds32_builtin_decl_impl): Ditto.
21038 (builtin_description): Ditto.
21039 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
21040 (nds32_init_builtins_impl): Ditto.
21041 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
21042 (nds32_builtin_decl): New.
21043 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
21044 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
21046 2018-03-02 Jeff Law <law@redhat.com>
21048 * reorg.c (stop_search_p): Handle DEBUG_INSN.
21049 (redundant_insn, fill_simple_delay_slots): Likewise.
21050 (fill_slots_from_thread): Likewise.
21051 * resource.c (mark_referenced_resources): Likewise.
21052 (mark_set_resources, find_dead_or_set_registers): Likewise.
21054 2018-03-02 Jakub Jelinek <jakub@redhat.com>
21056 * substring-locations.h (format_warning_va): Formatting fix for
21057 ATTRIBUTE_GCC_DIAG.
21058 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
21060 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
21061 * substring-locations.c: Include intl.h.
21062 (format_warning_va): Turned into small wrapper around
21063 format_warning_n_va, renamed to ...
21064 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
21065 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
21067 (format_warning_at_substring_n): New function.
21068 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
21069 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
21070 format_warning_at_substring with just a shorter name instead of
21071 const function pointer.
21072 (fmtwarn_n): New function.
21073 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
21074 appropriate, get rid of all the fmtstr temporaries, move conditionals
21075 with G_() wrapped string literals directly into fmtwarn arguments,
21076 cast dir.len to (int), formatting fixes.
21078 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
21080 * doc/invoke.texi: Remove "Cilk Plus" references.
21082 2018-03-02 Jakub Jelinek <jakub@redhat.com>
21083 Richard Biener <rguenther@suse.de>
21086 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
21087 for error or warning attributes if CALL_FROM_THUNK_P is set.
21090 2018-03-02 Jakub Jelinek <jakub@redhat.com>
21093 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
21094 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
21097 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
21098 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
21100 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
21101 instead of -1U in last predictors element's probability member.
21103 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
21106 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21107 arguments if they are comparable.
21109 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
21111 PR tree-optimization/84634
21112 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21113 masks and masked_loop_p with a single loop_masks, making sure it's
21114 null for bb vectorization.
21116 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
21118 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21119 (vect_analyze_data_ref_access): Use loop->safe_len rather than
21120 loop->force_vectorize to check whether there is no alias.
21122 2018-03-02 Jakub Jelinek <jakub@redhat.com>
21125 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21127 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21129 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21130 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21131 instead of a loop around prev_real_insn.
21132 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21135 PR inline-asm/84625
21136 * config/i386/i386.c (ix86_print_operand): Use conditional
21137 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21140 2018-03-02 Richard Biener <rguenther@suse.de>
21142 PR tree-optimization/84427
21143 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21144 (bitmap_set_subtract_values): Rewrite to handle multiple
21147 (prune_clobbered_mems): Likewise.
21148 (phi_translate): Take edge instead of pred/phiblock.
21149 (phi_translate_1): Likewise.
21150 (phi_translate_set): Likewise. Insert all translated
21151 exprs for a value into the set, keeping possibly multiple
21152 expressions per value.
21153 (compute_antic_aux): Adjust for phi_translate changes.
21154 When intersecting union the expressions and prune those
21155 not in the final value set, keeping possibly multiple
21156 expressions per value. Do not use value-insertion
21157 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
21158 all expressions. Add verification that the value-sets
21159 only shrink during iteration.
21160 (compute_partial_antic_aux): Adjust for the phi_translate changes.
21161 (do_pre_regular_insertion): Likewise.
21162 (do_pre_partial_partial_insertion): Likewise.
21164 2018-03-02 Richard Biener <rguenther@suse.de>
21167 * config/darwin.c (saved_debug_info_level): New static global.
21168 (darwin_asm_lto_start): Disable debug info generation for LTO out.
21169 (darwin_asm_lto_end): Restore debug info generation settings.
21171 2018-03-01 Martin Liska <mliska@suse.cz>
21174 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21175 volatile arguments.
21177 2018-03-01 Richard Biener <rguenther@suse.de>
21180 * dwarf2out.c (gen_variable_die): Properly handle late VLA
21181 type annotation with LTO when debug was disabled at compile-time.
21183 2018-03-01 Matthew Fortune <mfortune@gmail.com>
21185 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21187 (mips_final_postscan_insn): Likewise.
21189 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
21191 PR rtl-optimization/84528
21192 * alias.c (init_alias_target): Add commentary.
21193 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21194 a unique base value if the frame pointer is not eliminated
21195 to the stack pointer.
21197 2018-03-01 Tom de Vries <tom@codesourcery.com>
21199 PR rtl-optimization/83327
21200 * lra-int.h (hard_regs_spilled_into): Declare.
21201 * lra.c (hard_regs_spilled_into): Define.
21202 (init_reg_info): Init hard_regs_spilled_into.
21203 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21204 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21205 (process_bb_lives): Handle hard_regs_spilled_into.
21206 (lra_create_live_ranges_1): Before doing liveness propagation, clear
21207 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21209 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
21211 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
21212 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
21213 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
21214 * config/rs6000/aix72.h: New file.
21216 2018-02-28 Jakub Jelinek <jakub@redhat.com>
21218 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
21219 instead of warning_at with conditional singular and plural messages
21223 * stor-layout.c (update_alignment_for_field): For
21224 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
21225 && !DECL_PACKED (field), do the alignment update, just use
21226 only desired_align instead of MAX (type_align, desired_align)
21228 (place_field): Don't do known_align < desired_align handling
21229 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
21230 is non-NULL, instead do it after rli->prev_field handling and
21231 only if not within a bitfield word. For DECL_PACKED (field)
21232 use type_align of BITS_PER_UNIT.
21234 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
21236 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
21237 superfluous parentheses and trailing spaces.
21239 2018-02-28 Richard Biener <rguenther@suse.de>
21241 PR tree-optimization/84584
21242 * graphite-scop-detection.c (scop_detection::add_scop): Discard
21243 SCoPs with fake exit edge.
21245 2018-02-28 Martin Liska <mliska@suse.cz>
21248 * timevar.c (timer::print): Fix format to properly print 100%
21251 2018-02-28 Richard Biener <rguenther@suse.de>
21253 PR middle-end/84607
21254 * genmatch.c (capture_info::walk_match): Do not mark
21255 captured expressions without operands as expr_p given
21256 they act more like predicates and should be subject to
21257 "lost tail" side-effect preserving.
21259 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
21261 PR rtl-optimization/81611
21262 * auto-inc-dec.c (attempt_change): Move dead note from
21263 mem_insn if it's the next use of regno
21264 (find_address): Take address use of reg holding
21265 non-incremented value. Add parm to limit search to the named
21267 (merge_in_block): Attempt to use a mem insn that is the next
21268 use of the original regno.
21270 2018-02-27 Martin Sebor <msebor@redhat.com>
21273 * doc/invoke.texi (-Wmissing-attributes): New option.
21274 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
21276 2018-02-27 Martin Sebor <msebor@redhat.com>
21278 PR translation/84207
21279 * diagnostic-core.h (warning_n, error_n, inform_n): Change
21280 n argument to unsigned HOST_WIDE_INT.
21281 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
21282 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
21283 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
21284 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
21286 2018-02-27 Richard Biener <rguenther@suse.de>
21288 PR tree-optimization/84512
21289 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21290 Do not use the estimate returned from record_stmt_cost for
21291 the scalar iteration cost but sum properly using add_stmt_cost.
21293 2018-02-27 Richard Biener <rguenther@suse.de>
21295 PR tree-optimization/84466
21296 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21297 Adjust last change to less strictly validate use operands.
21299 2018-02-27 Martin Liska <mliska@suse.cz>
21301 PR gcov-profile/84548
21302 * gcov.c (process_file): Allow partial overlap and consider it
21303 also as group functions.
21304 (output_lines): Properly calculate range of lines for a group.
21306 2018-02-27 Martin Liska <mliska@suse.cz>
21308 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
21309 'ggc' suffixes. Change first column width.
21310 (timer::print): Fix formatting of the column.
21312 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
21314 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
21315 preserve inline entry blocks for the sake of debug inline
21316 entry point markers alone.
21317 (remove_unused_locals): Suggest in comments a better place to
21318 force the preservation of inline entry blocks that are
21319 otherwise unused, but do not preserve them.
21321 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
21323 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
21325 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
21328 * config/i386/constraints.md (Bs): Replace
21329 ix86_indirect_branch_register with
21330 TARGET_INDIRECT_BRANCH_REGISTER.
21332 * config/i386/i386.md (indirect_jump): Likewise.
21333 (tablejump): Likewise.
21334 (*sibcall_memory): Likewise.
21335 (*sibcall_value_memory): Likewise.
21336 Peepholes of indirect call and jump via memory: Likewise.
21337 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
21338 (*sibcall_value_GOT_32): Likewise.
21339 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21340 (GOT_memory_operand): Likewise.
21341 (call_insn_operand): Likewise.
21342 (sibcall_insn_operand): Likewise.
21343 (GOT32_symbol_operand): Likewise.
21344 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
21346 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
21348 PR rtl-optimization/83496
21349 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
21350 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
21351 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
21352 redundant insn, if any.
21353 (relax_delay_slots): Likewise.
21354 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
21356 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
21358 PR tree-optimization/83965
21359 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
21360 that grouped statements are part of a reduction chain. Return
21361 true if the statement is not marked as a reduction itself but
21362 is part of a group.
21363 (vect_recog_dot_prod_pattern): Don't check whether the statement
21364 is part of a group here.
21365 (vect_recog_sad_pattern): Likewise.
21366 (vect_recog_widen_sum_pattern): Likewise.
21368 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
21371 * final.c (rest_of_clean_state): Also look for calls inside sequences.
21373 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
21376 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
21378 (ix86_output_indirect_function_return): New prototype.
21379 (ix86_split_simple_return_pop_internal): Likewise.
21380 * config/i386/i386.c (indirect_return_via_cx): New.
21381 (indirect_return_via_cx_bnd): Likewise.
21382 (indirect_thunk_name): Handle return va CX_REG.
21383 (output_indirect_thunk_function): Create alias for
21384 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
21385 (ix86_output_indirect_jmp): Remove the bool argument.
21386 (ix86_output_indirect_function_return): New function.
21387 (ix86_split_simple_return_pop_internal): Likewise.
21388 * config/i386/i386.md (*indirect_jump): Don't pass false
21389 to ix86_output_indirect_jmp.
21390 (*tablejump_1): Likewise.
21391 (simple_return_pop_internal): Change it to define_insn_and_split.
21392 Call ix86_split_simple_return_pop_internal to split it for
21393 -mfunction-return=.
21394 (simple_return_indirect_internal): Call
21395 ix86_output_indirect_function_return instead of
21396 ix86_output_indirect_jmp.
21398 2018-02-26 Jakub Jelinek <jakub@redhat.com>
21401 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
21402 memset and value initialization afterwards.
21404 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
21406 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
21408 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
21411 * common/config/aarch64/aarch64-common.c
21412 (aarch_option_optimization_table[]): Switch
21413 off fomit-frame-pointer
21415 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
21416 Chung-Ju Wu <jasonwucj@gmail.com>
21418 * config/nds32/nds32-multiple.md (load_multiple): Disallow
21420 (store_multiple): Ditto.
21422 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
21424 * config.gcc: Add --with-cpu support for nds32 target.
21425 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
21426 * config/nds32/nds32.opt: Add -mcpu= option.
21428 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
21430 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
21431 isel=yes): Warn for these deprecated options.
21433 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
21435 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
21436 ISA_2_5_MASKS_EMBEDDED.
21438 2018-02-23 Jakub Jelinek <jakub@redhat.com>
21440 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
21441 p->max as pointers rather than using iterative_hash_expr.
21443 2018-02-23 Carl Love <cel@us.ibm.com>
21445 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
21446 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
21448 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
21449 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
21450 P8V_BUILTIN_VEC_VUNSIGNED2.
21452 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
21455 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
21456 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
21458 * lra-constraints.c (curr_insn_transform): Set up
21459 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
21461 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
21462 * lra-eliminations.c (spill_pseudos): Ditto.
21463 (process_insn_for_elimination): Ditto.
21464 * lra-lives.c (reg_early_clobber_p): Use the new macros.
21465 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
21466 LRA_NON_CLOBBERED_ALT.
21468 2018-02-22 Martin Sebor <msebor@redhat.com>
21470 PR tree-optimization/84480
21471 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
21472 to maybe_diag_stxncpy_trunc. Call it.
21473 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
21474 from gimple_fold_builtin_strcpy. Print inlining stack.
21475 (handle_builtin_stxncpy): Print inlining stack.
21476 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
21478 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
21481 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
21482 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
21483 and -fcheck-pointer-bounds are used together.
21484 (indirect_thunk_prefix): New enum.
21485 (indirect_thunk_need_prefix): New function.
21486 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
21487 "_nt" instead of "_bnd" for NOTRACK prefix.
21488 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
21489 (output_indirect_thunk_function): Likewise.
21491 (ix86_code_end): Update output_indirect_thunk_function calls.
21492 (ix86_output_indirect_branch_via_reg): Replace
21493 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
21494 (ix86_output_indirect_branch_via_push): Likewise.
21495 (ix86_output_function_return): Likewise.
21496 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
21497 incompatible with -fcf-protection=branch and
21498 -fcheck-pointer-bounds.
21500 2018-02-22 Steve Ellcey <sellcey@cavium.com>
21503 * config/aarch64/aarch64.c (aarch64_print_address_internal):
21504 Change gcc_assert call to output_operand_lossage.
21506 2018-02-22 Steve Ellcey <sellcey@cavium.com>
21508 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
21510 2018-02-22 DJ Delorie <dj@redhat.com>
21511 Sebastian Perta <sebastian.perta@renesas.com>
21512 Oleg Endo <olegendo@gcc.gnu.org>
21514 * config/rx/rx.c (rx_rtx_costs): New function.
21515 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
21517 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
21519 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
21521 2018-02-22 Martin Liska <mliska@suse.cz>
21524 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
21525 Add "native" as a possible value.
21527 2018-02-22 Martin Liska <mliska@suse.cz>
21530 * config/i386/i386.c (ix86_option_override_internal):
21531 Add "native" as a possible value for -march and -mtune.
21533 2018-02-22 Jakub Jelinek <jakub@redhat.com>
21536 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
21537 to all type variants.
21539 PR tree-optimization/84503
21540 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
21541 width as info->bitpos + info->bitsize - start.
21542 (merged_store_group::merge_overlapping): Simplify width computation.
21543 (check_no_overlap): New function.
21544 (imm_store_chain_info::try_coalesce_bswap): Compute expected
21545 start + width and last_order of the group, fail if check_no_overlap
21547 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
21548 to group if check_no_overlap fails.
21550 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
21552 * config/rs6000/altivec.md: Delete contraint arguments to
21553 define_expand, define_split, and define_peephole2, and in
21554 define_insn_and_split if always unused.
21555 * config/rs6000/darwin.md: Ditto.
21556 * config/rs6000/dfp.md: Ditto.
21557 * config/rs6000/rs6000.md: Ditto.
21558 * config/rs6000/sync.md: Ditto.
21559 * config/rs6000/vector.md: Ditto.
21560 * config/rs6000/vsx.md: Ditto.
21562 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
21564 * config/rs6000/altivec.md: Write output control strings as braced
21565 blocks instead of double-quoted strings.
21566 * config/rs6000/darwin.md: Ditto.
21567 * config/rs6000/rs6000.md: Ditto.
21568 * config/rs6000/vector.md: Ditto.
21569 * config/rs6000/vsx.md: Ditto.
21571 2018-02-21 Jason Merrill <jason@redhat.com>
21573 PR c++/84314 - ICE with templates and fastcall attribute.
21574 * attribs.c (build_type_attribute_qual_variant): Remove assert.
21576 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
21578 * ipa-cp.c (determine_versionability): Fix comment typos.
21580 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
21583 * ipa-cp.c (determine_versionability): Do not version functions caling
21586 2018-02-21 Martin Liska <mliska@suse.cz>
21589 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
21590 Add "native" as a possible value.
21591 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
21592 the macro when native cpu detection is available.
21594 2018-02-21 Martin Liska <mliska@suse.cz>
21597 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
21598 Add "native" as a possible value.
21599 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
21600 when native cpu detection is available.
21602 2018-02-21 Jakub Jelinek <jakub@redhat.com>
21603 Martin Sebor <msebor@redhat.com>
21605 PR tree-optimization/84478
21606 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
21608 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
21609 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
21610 support which is conservatively correct, for 2 only stay conservative
21611 for maxlen. Formatting and comment capitalization fixes. Add STRICT
21612 argument to the 2 argument get_range_strlen, adjust 6 arg
21613 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
21615 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
21616 (gimple_fold_builtin_strlen): Pass true as last argument to
21619 2018-02-20 Martin Sebor <msebor@redhat.com>
21621 PR middle-end/84095
21622 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
21623 (builtin_memref::set_base_and_offset): Same. Handle inner references.
21624 (builtin_memref::builtin_memref): Factor out parts into
21625 set_base_and_offset and call it.
21627 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
21629 PR middle-end/84406
21630 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
21631 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
21632 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
21633 search at the associated MODE_INT.
21635 2018-02-20 Jeff Law <law@redhat.com>
21637 PR middle-end/82123
21638 PR tree-optimization/81592
21639 PR middle-end/79257
21640 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
21641 for range data rather than using global data.
21642 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
21643 range data rather than using global data.
21644 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
21645 pass it to children as needed.
21646 (struct directive::fmtresult): Similarly.
21647 (struct directive::set_width): Similarly.
21648 (struct directive::set_precision): Similarly.
21649 (format_integer, format_directive, parse_directive): Similarly.
21650 (format_none): Accept unnamed vr_values parameter.
21651 (format_percent, format_floating, format_character): Similarly.
21652 (format_string, format_plain): Similarly.
21653 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
21654 the EVRP range analyzer for range data rather than using global data.
21655 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
21656 gimple-ssa-evrp-analyze.h
21657 (class sprintf_dom_walker): Add after_dom_children member function.
21658 Add evrp_range_analyzer member.
21659 (sprintf_dom_walker::before_dom_children): Call into the EVRP
21660 range analyzer as needed.
21661 (sprintf_dom_walker::after_dom_children): New member function.
21662 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
21664 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
21665 (evrp_range_analyzer::pop_to_marker): Likewise.
21667 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
21669 PR tree-optimization/84419
21670 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
21671 with the required type if its current type is compatible but
21674 2018-02-20 Jakub Jelinek <jakub@redhat.com>
21676 PR middle-end/82004
21677 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
21678 after vectorization.
21680 2018-02-20 Martin Liska <mliska@suse.cz>
21683 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
21684 possible values if we don't have a hint.
21686 2018-02-20 Martin Liska <mliska@suse.cz>
21690 * final.c (shorten_branches): Build align_tab array with one
21692 * opts.c (finish_options): Add alignment option limit check.
21693 (MAX_CODE_ALIGN): Likewise.
21694 (MAX_CODE_ALIGN_VALUE): Likewise.
21695 * doc/invoke.texi: Document maximum allowed option value for
21696 all -falign-* options.
21698 2018-02-19 Jakub Jelinek <jakub@redhat.com>
21701 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
21702 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
21703 * var-tracking.c (emit_note_insn_var_location): Remove all references
21704 to NOTE_INSN_CALL_ARG_LOCATION.
21705 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
21706 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
21707 Use copy_rtx_if_shared.
21708 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
21709 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
21710 (dwarf2out_var_location): Remove handling of
21711 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
21713 * final.c (final_scan_insn): Remove all references to
21714 NOTE_INSN_CALL_ARG_LOCATION.
21715 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
21716 before dumping final insns.
21717 * except.c (emit_note_eh_region_end): Remove all references to
21718 NOTE_INSN_CALL_ARG_LOCATION.
21719 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
21720 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
21721 * config/arc/arc.c (hwloop_optimize): Likewise.
21722 * config/arm/arm.c (create_fix_barrier): Likewise.
21723 * config/s390/s390.c (s390_chunkify_start): Likewise.
21724 * config/sh/sh.c (find_barrier): Likewise.
21725 * config/i386/i386.c (rest_of_insert_endbranch,
21726 ix86_seh_fixup_eh_fallthru): Likewise.
21727 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
21728 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
21729 * config/frv/frv.c (frv_function_prologue): Likewise.
21730 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
21732 (note_outside_basic_block_p): Remove all references to
21733 NOTE_INSN_CALL_ARG_LOCATION.
21734 * gengtype.c (adjust_field_rtx_def): Likewise.
21735 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
21737 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
21738 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
21741 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
21744 PR tree-optimization/84452
21745 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
21746 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
21749 2018-02-19 Martin Liska <mliska@suse.cz>
21752 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
21754 2018-02-19 Martin Liska <mliska@suse.cz>
21755 Richard Sandiford <richard.sandiford@linaro.org>
21757 PR tree-optimization/82491
21758 * gimple-fold.c (get_base_constructor): Make earlier bail out
21761 2018-02-19 Carl Love <cel@us.ibm.com>
21763 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
21764 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
21766 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
21767 P8V_BUILTIN_VEC_NEG.
21769 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
21771 * config/rl78/rl78.md (movdf): New define expand.
21773 2018-02-19 Martin Liska <mliska@suse.cz>
21776 * doc/invoke.texi: Fix typo.
21777 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
21779 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
21781 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
21782 handle rs6000_single_float and rs6000_double_float specially for
21785 2018-02-16 Jeff Law <law@redhat.com>
21787 * config/rx/rx.c (add_pop_cfi_notes): New function.;
21788 (pop_regs): Use it.
21790 2018-02-16 Jakub Jelinek <jakub@redhat.com>
21793 * ipa-inline.c (inline_small_functions): Fix a typo.
21795 2018-02-16 Nathan Sidwell <nathan@acm.org>
21797 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
21799 2018-02-16 Carl Love <cel@us.ibm.com>
21801 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
21802 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
21803 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
21804 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
21805 expansion to P8V_BUILTIN_VEC_FLOAT2.
21807 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
21809 PR rtl-optimization/70023
21810 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
21811 src_regno into account.
21813 2018-02-16 Carl Love <cel@us.ibm.com>
21815 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
21816 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
21817 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
21818 * config/rs6000/rs6000.c: Remove case statements for
21819 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
21820 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
21821 and P9V_BUILTIN_VEC_VINSERT4B.
21822 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
21823 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
21824 * config/rs6000/vsx.md:
21825 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
21828 2018-02-16 Carl Love <cel@us.ibm.com>
21830 * config/rs6000/altivec.h: Add builtin names vec_extract4b
21832 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
21834 * config/rs6000/rs6000-c.c: Add the definitions for
21835 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
21836 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
21837 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
21838 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
21839 definition for insert4b and define insn *insert3b_internal.
21840 * doc/extend.texi: Add documentation for vec_extract4b.
21842 2018-02-16 Nathan Sidwell <nathan@acm.org>
21844 * doc/extend.texi (Backwards Compatibility): Mention friend
21845 injection. Note for-scope is deprecated.
21846 * doc/invoke.texi (-ffriend-injection): Deprecate.
21848 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
21850 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
21851 that moved to I2, also allow destinations that are a paradoxical
21852 subreg (instead of a normal reg).
21854 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
21857 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
21860 2018-02-16 Richard Biener <rguenther@suse.de>
21862 PR tree-optimization/84037
21863 PR tree-optimization/84016
21865 * config/i386/i386.c (ix86_builtin_vectorization_cost):
21866 Adjust vec_construct for the fact we need additional higher latency
21867 128bit inserts for AVX256 and AVX512 vector builds.
21868 (ix86_add_stmt_cost): Scale vector construction cost for
21871 2018-02-16 Richard Biener <rguenther@suse.de>
21873 PR tree-optimization/84417
21874 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
21875 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
21876 (non_rewritable_lvalue_p): Likewise, use poly-ints.
21878 2018-02-16 Martin Liska <mliska@suse.cz>
21881 * internal-fn.def (ASAN_CHECK): Set proper flags.
21882 (ASAN_MARK): Likewise.
21884 2018-02-16 Julia Koval <julia.koval@intel.com>
21886 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
21887 from PTA_CANNONLAKE.
21889 2018-02-16 Jakub Jelinek <jakub@redhat.com>
21892 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
21893 Use ++iter rather than iter++ for std::list iterators.
21894 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
21895 defer deleting them until all nodes in the forest are processed. Do
21896 free even leaf nodes. Change to_process into auto_vec.
21899 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
21900 * vec.h (vec_default_construct): Use memset instead of placement new
21901 if BROKEN_VALUE_INITIALIZATION is defined.
21902 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
21903 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
21906 PR rtl-optimization/83723
21907 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
21908 * lra.c (lra_substitute_pseudo): Likewise. If true, use
21909 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
21911 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21913 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21915 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
21917 PR rtl-optimization/81443
21918 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21919 from inner REGs to paradoxical SUBREGs.
21921 2018-02-16 Richard Biener <rguenther@suse.de>
21923 PR tree-optimization/84399
21924 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21925 For operands we can analyze at their definition make sure we can
21926 analyze them at each use as well.
21928 2018-02-16 Richard Biener <rguenther@suse.de>
21930 PR tree-optimization/84190
21931 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21932 volatile accesses if the decl isn't volatile.
21934 2018-02-15 Jason Merrill <jason@redhat.com>
21936 PR c++/84314 - ICE with templates and fastcall attribute.
21937 * attribs.c (build_type_attribute_qual_variant): Don't clobber
21938 TYPE_CANONICAL on an existing type.
21940 2018-02-15 Jakub Jelinek <jakub@redhat.com>
21942 PR tree-optimization/84383
21943 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21944 dstoff nor call operand_equal_p if dstbase is NULL.
21946 PR tree-optimization/84334
21947 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21948 also a CONSTANT_CLASS_P, punt.
21950 2018-02-14 Jim Wilson <jimw@sifive.com>
21952 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21953 first SMALL_OPERAND check. New local min_second_step. Move assert
21954 to where locals are set. Add TARGET_RVC support.
21955 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21957 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
21959 * doc/invoke.texi: Correct -Wformat-overflow code sample.
21961 2018-02-14 Martin Sebor <msebor@redhat.com>
21963 PR tree-optimization/83698
21964 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21965 arrays constrain the offset range to their bounds.
21966 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21967 (builtin_access::overlap): Avoid setting the size of overlap if it's
21969 (maybe_diag_overlap): Also consider arrays when deciding what values
21970 of offsets to include in diagnostics.
21972 2018-02-14 Martin Sebor <msebor@redhat.com>
21975 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21976 that correspond to the kind of a declaration.
21978 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
21981 * config/pa/pa.md: Load address of PIC label using the linkage table
21982 if the label is nonlocal.
21984 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
21986 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21987 warning message if user requests -maltivec=be.
21988 * doc/invoke.texi: Document deprecation of -maltivec=be.
21990 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
21993 * config/rs6000/rs6000-c.c: Update definitions for
21994 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21995 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21997 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22000 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
22001 add _get_ssp intrinsics. Remove argument from
22002 __builtin_ia32_rdssp[d|q].
22003 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
22004 * config/i386/i386-builtin.def: Remove argument from
22005 __builtin_ia32_rdssp[d|q].
22006 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
22007 ix86_expand_special_args_builtin for _rdssp[d|q].
22008 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
22009 Clear register before usage.
22010 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
22011 Add documentation for new _get_ssp and _inc_ssp intrinsics.
22013 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
22015 PR tree-optimization/84357
22016 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
22017 operand 1 of an ARRAY_REF too.
22019 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
22022 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
22023 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22025 (set_of_reg): New struct.
22026 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
22027 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
22028 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22030 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
22031 Split into bitclr, bitset, bitinvert patterns if appropriate.
22032 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
22033 use rx_fuse_in_memory_bitop.
22034 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
22035 to named insn, correct maximum insn length.
22037 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
22040 * machmode.def: Define a complex mode for PARTIAL_INT.
22041 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
22043 * doc/rtl.texi: Document CSPImode.
22044 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
22046 (msp430_hard_regno_nregs_with_padding): Likewise.
22048 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
22051 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
22053 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
22055 PR rtl-optimization/84169
22056 * combine.c (try_combine): New variable split_i2i3. Set it to true if
22057 we generated a parallel as new i3 and we split that to new i2 and i3
22058 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
22059 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
22060 those to i2, not i1. Partially rewrite this scan code.
22062 2018-02-13 Jakub Jelinek <jakub@redhat.com>
22065 * stor-layout.c (place_field): For variable length fields, adjust
22066 offset_align afterwards not just based on the field's alignment,
22067 but also on the size.
22069 PR middle-end/84309
22070 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
22071 of exps and logs in the use_exp2 case.
22073 2018-02-13 Jeff Law <law@redhat.com>
22075 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
22076 entry for "vector".
22078 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
22081 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
22085 * common.opt (gas-loc-support, gas-locview-support): New.
22086 (ginline-points, ginternal-reset-location-views): New.
22087 * doc/invoke.texi: Document them. Use @itemx where intended.
22088 (gvariable-location-views): Adjust.
22089 * target.def (reset_location_view): New.
22090 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
22091 (TARGET_RESET_LOCATION_VIEW): New.
22092 * doc/tm.texi: Rebuilt.
22093 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
22094 (dwarf2out_default_as_locview_support): New.
22095 (output_asm_line_debug_info): Use option variables.
22096 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
22097 (output_loc_list): Likewise.
22098 (add_high_low_attributes): Check option variables.
22099 Don't output entry view attribute in strict mode.
22100 (gen_inlined_subroutine_die): Check option variables.
22101 (dwarf2out_inline_entry): Likewise.
22102 (init_sections_and_labels): Likewise.
22103 (dwarf2out_early_finish): Likewise.
22104 (maybe_reset_location_view): New, from...
22105 (dwarf2out_var_location): ... here. Call it.
22106 * debug.h (dwarf2out_default_as_loc_support): Declare.
22107 (dwarf2out_default_as_locview_support): Declare.
22108 * hooks.c (hook_int_rtx_insn_0): New.
22109 * hooks.h (hook_int_rtx_insn_0): Declare.
22110 * toplev.c (process_options): Take -gas-loc-support and
22111 -gas-locview-support from dwarf2out. Enable
22112 -gvariable-location-views by default only with locview
22113 assembler support. Enable -ginternal-reset-location-views by
22114 default only if the target defines the corresponding hook.
22115 Enable -ginline-points by default if location views are
22116 enabled; force it disabled if statement frontiers are
22118 * tree-inline.c (expand_call_inline): Check option variables.
22119 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22121 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
22123 PR tree-optimization/84321
22124 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22125 handling. Also check whether the anti-range contains any values
22126 that satisfy the mask; switch to a VR_RANGE if not.
22128 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
22131 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22133 2018-02-13 Martin Jambor <mjambor@suse.cz>
22136 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22137 of call statements, also set location of a load to a temporary.
22139 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
22141 * config/rl78/rl78.c (add_vector_labels): New function.
22142 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22143 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22144 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22145 which checks that no arguments are passed.
22146 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22147 * doc/extend.texi: Documentation for the new attribute.
22149 2018-02-13 Andreas Schwab <schwab@suse.de>
22151 * config/riscv/linux.h (CPP_SPEC): Define.
22153 2018-02-13 Jakub Jelinek <jakub@redhat.com>
22156 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
22157 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
22158 OPTION_MASK_ISA_AES as first argument to def_builtin_const
22159 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
22160 instead of OPTION_MASK_ISA_PCLMUL as first argument to
22161 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
22162 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
22163 temporarily for AES and PCLMUL builtins.
22165 PR tree-optimization/84339
22166 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22167 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22170 PR middle-end/84309
22171 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22172 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22173 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22175 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22177 * omp-simd-clone.h: New file.
22178 * omp-simd-clone.c: Include omp-simd-clone.h.
22179 (expand_simd_clones): No longer static.
22180 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22181 cgraph.h and omp-simd-clone.h.
22182 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22183 (vect_recog_widen_shift_pattern): Formatting fix.
22184 (vect_pattern_recog_1): Don't check optab for calls.
22187 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22188 operands[2] into a REG before using gen_lowpart on it.
22190 2018-02-12 Jeff Law <law@redhat.com>
22193 * config/sh/sh.c (find_barrier): Consider a sibling call
22196 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22197 successfully back substituting a reg.
22199 2018-02-12 Richard Biener <rguenther@suse.de>
22201 PR tree-optimization/84037
22202 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22203 parameter, move visited init to caller.
22204 (vect_slp_analyze_operations): Separate cost from validity
22205 check, initialize visited once for all instances.
22206 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22208 * tree-vect-stmts.c (vect_model_simple_cost): Make early
22210 (vect_model_promotion_demotion_cost): Likewise.
22211 (vectorizable_bswap): Guard cost modeling with !slp_node
22212 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
22214 (vectorizable_call): Likewise.
22215 (vectorizable_conversion): Likewise.
22216 (vectorizable_assignment): Likewise.
22217 (vectorizable_shift): Likewise.
22218 (vectorizable_operation): Likewise.
22219 (vectorizable_store): Likewise.
22220 (vectorizable_load): Likewise.
22221 (vectorizable_condition): Likewise.
22222 (vectorizable_comparison): Likewise.
22224 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
22227 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
22228 (ASAN_MARK): Fix fnspec to account for return value, change pointer
22229 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
22231 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
22233 PR middle-end/83665
22234 * params.def (inline-min-speedup): Increase from 8 to 15.
22235 (max-inline-insns-auto): Decrease from 40 to 30.
22236 * ipa-split.c (consider_split): Add some buffer for function to
22237 be considered inlining candidate.
22238 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
22241 2018-02-12 Richard Biener <rguenther@suse.de>
22243 PR tree-optimization/84037
22244 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
22245 matched stmts if we cannot swap the non-matched ones.
22247 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
22249 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
22250 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
22251 _mm_maskz_scalef_round_ss): New intrinsics.
22252 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
22253 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
22254 __builtin_ia32_scalefss_round): Remove.
22255 (__builtin_ia32_scalefsd_mask_round,
22256 __builtin_ia32_scalefss_mask_round): New intrinsics.
22257 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
22258 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
22259 ((match_operand:VF_128 2 "<round_nimm_predicate>"
22260 "<round_constraint>")): Changed to ...
22261 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
22262 "<round_scalar_constraint>")): ... this.
22263 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
22264 %0, %1, %2<round_op3>}"): Changed to ...
22265 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
22266 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
22267 %2<round_scalar_mask_op3>}"): ... this.
22268 * config/i386/subst.md (round_scalar_nimm_predicate): New.
22270 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
22272 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
22273 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
22274 (_mm_maskz_sqrt_round_ss): New intrinsics.
22275 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
22276 (__builtin_ia32_sqrtsd_mask_round)
22277 (__builtin_ia32_sqrtss_mask_round): New builtins.
22278 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
22279 (__builtin_ia32_sqrtss_round): Remove.
22280 (__builtin_ia32_sqrtsd_mask_round)
22281 (__builtin_ia32_sqrtss_mask_round): New builtins.
22282 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
22283 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
22284 ((match_operand:VF_128 1 "vector_operand"
22285 "xBm,<round_constraint>")): Changed to ...
22286 ((match_operand:VF_128 1 "vector_operand"
22287 "xBm,<round_scalar_constraint>")): ... this.
22288 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
22289 %0, %2, %<iptr>1<round_op3>}): Changed to ...
22290 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
22291 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
22292 %<iptr>1<round_scalar_mask_op3>}): ... this.
22293 ((set_attr "prefix" "<round_prefix>")): Changed to ...
22294 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
22296 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
22299 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
22300 Cast vec_cmpeq result to correct type.
22301 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
22302 Cast vec_cmpgt result to correct type.
22304 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
22306 * final.c (final_scan_insn_1): Renamed from...
22307 (final_scan_insn): ... this. New wrapper, to recover
22308 seen from the outermost call in recursive ones.
22309 * config/sparc/sparc.c (output_return): Drop seen from call.
22310 (output_sibcall): Likewise.
22311 * config/visium/visium.c (output_branch): Likewise.
22313 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
22315 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
22318 2018-02-10 Alan Modra <amodra@gmail.com>
22321 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
22322 Specify LR as an input.
22324 2018-02-10 Jakub Jelinek <jakub@redhat.com>
22327 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
22328 remove_member_access_dummy_vars): New functions.
22329 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
22330 lower_omp_1, execute_lower_omp): Use them.
22332 PR rtl-optimization/84308
22333 * shrink-wrap.c (spread_components): Release todo vector.
22335 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
22337 PR rtl-optimization/57193
22338 * ira-color.c (struct allocno_color_data): Add member
22339 conflict_allocno_hard_prefs.
22340 (update_conflict_allocno_hard_prefs): New.
22341 (bucket_allocno_compare_func): Add a preference based on
22342 conflict_allocno_hard_prefs.
22343 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
22344 (color_allocnos): Remove a dead code. Initiate
22345 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
22347 2018-02-09 Jakub Jelinek <jakub@redhat.com>
22350 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
22351 constraint from =wa to wa. Avoid a subreg on the output operand,
22352 instead use a pseudo and subreg it in a move.
22353 (p9_xxbrd_<mode>): Changed to ...
22354 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
22355 (p9_xxbrd_v2df): New expander.
22356 (p9_xxbrw_<mode>): Changed to ...
22357 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
22358 (p9_xxbrw_v4sf): New expander.
22360 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
22362 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
22364 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
22367 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
22368 multiply in 32-bit mode.
22369 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
22370 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
22373 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
22375 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
22376 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
22377 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
22378 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
22380 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
22383 * dwarf2out.c (is_trivial_indirect_ref): New function.
22384 (dwarf2out_late_global_decl): Do not generate a location
22385 attribute for variables that have a non-trivial DECL_VALUE_EXPR
22386 and that are not defined in the current unit.
22388 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
22390 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
22391 instead of a libcall for UNORDERED.
22393 2018-02-09 Tamar Christina <tamar.christina@arm.com>
22396 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
22397 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
22399 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22402 * config/s390/s390.c (s390_set_current_function): Invoke
22403 s390_indirect_branch_settings also if fndecl didn't change.
22405 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
22407 * config/rs6000/rs6000.md (blockage): Set length to zero.
22409 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
22411 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
22413 2018-02-09 Jakub Jelinek <jakub@redhat.com>
22416 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
22417 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
22421 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
22422 PARALLEL incoming that failed vt_get_decl_and_offset check.
22424 PR middle-end/84237
22425 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
22426 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
22428 (get_variable_section): For decls in named .bss* sections pass true as
22429 second argument to bss_initializer_p.
22431 2018-02-09 Marek Polacek <polacek@redhat.com>
22432 Jakub Jelinek <jakub@redhat.com>
22435 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
22436 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
22437 Sync some changes from cxx_fold_indirect_ref.
22439 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
22441 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
22443 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
22444 (BLOCK_INLINE_ENTRY_LABEL): New.
22445 (dwarf2out_var_location): Disregard inline entry markers.
22446 (inline_entry_data): New struct.
22447 (inline_entry_data_hasher): New hashtable type.
22448 (inline_entry_data_hasher::hash): New.
22449 (inline_entry_data_hasher::equal): New.
22450 (inline_entry_data_table): New variable.
22451 (add_high_low_attributes): Add DW_AT_entry_pc and
22452 DW_AT_GNU_entry_view attributes if a pending entry is found
22453 in inline_entry_data_table. Add old entry_pc attribute only
22454 if debug nonbinding markers are disabled.
22455 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
22456 markers are enabled.
22457 (block_within_block_p, dwarf2out_inline_entry): New.
22458 (dwarf2out_finish): Check that no entries remained in
22459 inline_entry_data_table.
22460 * final.c (reemit_insn_block_notes): Handle inline entry notes.
22461 (final_scan_insn, notice_source_line): Likewise.
22462 (rest_of_clean_state): Skip inline entry markers.
22463 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
22465 * gimple.c (gimple_build_debug_inline_entry): New.
22466 * gimple.h (enum gimple_debug_subcode): Add
22467 GIMPLE_DEBUG_INLINE_ENTRY.
22468 (gimple_build_debug_inline_entry): Declare.
22469 (gimple_debug_inline_entry_p): New.
22470 (gimple_debug_nonbind_marker_p): Adjust.
22471 * insn-notes.def (INLINE_ENTRY): New.
22472 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
22473 inline entry marker notes.
22474 (print_insn): Likewise.
22475 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
22476 (INSN_DEBUG_MARKER_KIND): Likewise.
22477 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
22478 * tree-inline.c (expand_call_inline): Build and insert
22479 debug_inline_entry stmt.
22480 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
22481 inline entry blocks early, if nonbind markers are enabled.
22482 (dump_scope_block): Dump fragment info.
22483 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
22484 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
22485 (gimple_build_debug_inline_entry): New.
22486 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
22487 Enable/disable inline entry points too.
22488 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
22489 (DEBUG_INSN): Describe inline entry markers.
22491 * common.opt (gvariable-location-views): New.
22492 (gvariable-location-views=incompat5): New.
22493 * config.in: Rebuilt.
22494 * configure: Rebuilt.
22495 * configure.ac: Test assembler for view support.
22496 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
22497 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
22498 * dwarf2out.c (var_loc_view): New typedef.
22499 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
22500 (dwarf2out_locviews_in_attribute): New.
22501 (dwarf2out_locviews_in_loclist): New.
22502 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
22503 (enum dw_line_info_opcode): Add LI_adv_address.
22504 (struct dw_line_info_table): Add view.
22505 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
22506 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
22507 (zero_view_p): New variable.
22508 (ZERO_VIEW_P): New macro.
22509 (output_asm_line_debug_info): New.
22510 (struct var_loc_node): Add view.
22511 (add_AT_view_list, AT_loc_list): New.
22512 (add_var_loc_to_decl): Add view param. Test it against last.
22513 (new_loc_list): Add view params. Record them.
22514 (AT_loc_list_ptr): Handle loc and view lists.
22515 (view_list_to_loc_list_val_node): New.
22516 (print_dw_val): Handle dw_val_class_view_list.
22517 (size_of_die): Likewise.
22518 (value_format): Likewise.
22519 (loc_list_has_views): New.
22520 (gen_llsym): Set vl_symbol too.
22521 (maybe_gen_llsym, skip_loc_list_entry): New.
22522 (dwarf2out_maybe_output_loclist_view_pair): New.
22523 (output_loc_list): Output view list or entries too.
22524 (output_view_list_offset): New.
22525 (output_die): Handle dw_val_class_view_list.
22526 (output_dwarf_version): New.
22527 (output_compilation_unit_header): Use it.
22528 (output_skeleton_debug_sections): Likewise.
22529 (output_rnglists, output_line_info): Likewise.
22530 (output_pubnames, output_aranges): Update version comments.
22531 (output_one_line_info_table): Output view numbers in asm comments.
22532 (dw_loc_list): Determine current endview, pass it to new_loc_list.
22533 Call maybe_gen_llsym.
22534 (loc_list_from_tree_1): Adjust.
22535 (add_AT_location_description): Create view list attribute if
22536 needed, check it's absent otherwise.
22537 (convert_cfa_to_fb_loc_list): Adjust.
22538 (maybe_emit_file): Call output_asm_line_debug_info for test.
22539 (dwarf2out_var_location): Reset views as needed. Precompute
22540 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
22541 attribute. Set view.
22542 (new_line_info_table): Reset next view.
22543 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
22544 (dwarf2out_source_line): Likewise. Output view resets and labels to
22545 the assembler, or select appropriate line info opcodes.
22546 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
22547 (optimize_string_length): Catch it. Adjust.
22548 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
22549 dw_val_class_view_list, and remove it if no longer needed.
22550 (hash_loc_list): Hash view numbers.
22551 (loc_list_hasher::equal): Compare them.
22552 (optimize_location_lists): Check whether a view list symbol is
22553 needed, and whether the locview attribute is present, and
22554 whether they match. Remove the locview attribute if no longer
22556 (index_location_lists): Call skip_loc_list_entry for test.
22557 (dwarf2out_finish): Call output_asm_line_debug_info for test.
22558 Use output_dwarf_version.
22559 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
22560 (struct dw_val_node): Add val_view_list.
22561 * final.c (SEEN_NEXT_VIEW): New.
22562 (set_next_view_needed): New.
22563 (clear_next_view_needed): New.
22564 (maybe_output_next_view): New.
22565 (final_start_function): Rename to...
22566 (final_start_function_1): ... this. Take pointer to FIRST,
22567 add SEEN parameter. Emit param bindings in the initial view.
22568 (final_start_function): Reintroduce SEEN-less interface.
22569 (final): Rename to...
22570 (final_1): ... this. Take SEEN parameter. Output final pending
22571 next view at the end.
22572 (final): Reintroduce seen-less interface.
22573 (final_scan_insn): Output pending next view before switching
22574 sections or ending a block. Mark the next view as needed when
22575 outputting variable locations. Notify debug backend of section
22576 changes, and of location view changes.
22577 (rest_of_handle_final): Adjust.
22578 * toplev.c (process_options): Autodetect value for debug variable
22579 location views option. Warn on incompat5 without -gdwarf-5.
22580 * doc/invoke.texi (gvariable-location-views): New.
22581 (gvariable-location-views=incompat5): New.
22582 (gno-variable-location-views): New.
22584 2018-02-08 David Malcolm <dmalcolm@redhat.com>
22586 PR tree-optimization/84136
22587 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
22588 that the result of find_edge is non-NULL.
22590 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
22593 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
22594 storing integer register in SImode. Fix cost of 256 and 512
22595 byte aligned SSE register store.
22597 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
22599 * config/i386/i386.c (ix86_multiplication_cost): Fix
22600 multiplication cost for TARGET_AVX512DQ.
22602 2018-02-08 Marek Polacek <polacek@redhat.com>
22604 PR tree-optimization/84238
22605 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
22608 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
22610 PR tree-optimization/84265
22611 * tree-vect-stmts.c (vectorizable_store): Don't treat
22612 VMAT_CONTIGUOUS accesses as grouped.
22613 (vectorizable_load): Likewise.
22615 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
22617 PR tree-optimization/81635
22618 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
22619 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
22620 (test_round_for_mask): New functions.
22621 (wide_int_cc_tests): Call test_round_for_mask.
22622 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
22623 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
22624 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
22625 range returned by get_range_info.
22627 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
22630 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
22631 * symtab.c: Include builtins.h
22632 (symtab_node::output_to_lto_symbol_table_p): Move here
22633 from lto-streamer-out.c:output_symbol_p.
22634 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
22635 (output_symbol_p): Move all logic to symtab.c
22636 (produce_symtab): Update.
22638 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22640 * config/s390/s390-opts.h (enum indirect_branch): Define.
22641 * config/s390/s390-protos.h (s390_return_addr_from_memory)
22642 (s390_indirect_branch_via_thunk)
22643 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
22644 (enum s390_indirect_branch_type): Define.
22645 * config/s390/s390.c (struct s390_frame_layout, struct
22646 machine_function): Remove.
22647 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
22648 (indirect_branch_table_label_no, indirect_branch_table_name):
22650 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
22651 (enum s390_indirect_branch_option): Define.
22652 (s390_return_addr_from_memory): New function.
22653 (s390_handle_string_attribute): New function.
22654 (s390_attribute_table): Add new attribute handler.
22655 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
22656 (s390_indirect_branch_via_thunk): New function.
22657 (s390_indirect_branch_via_inline_thunk): New function.
22658 (s390_function_ok_for_sibcall): When jumping via thunk disallow
22659 sibling call optimization for non z10 compiles.
22660 (s390_emit_call): Force indirect branch target to be a single
22661 register. Add r1 clobber for non-z10 compiles.
22662 (s390_emit_epilogue): Emit return jump via return_use expander.
22663 (s390_reorg): Handle JUMP_INSNs as execute targets.
22664 (s390_option_override_internal): Perform validity checks for the
22665 new command line options.
22666 (s390_indirect_branch_attrvalue): New function.
22667 (s390_indirect_branch_settings): New function.
22668 (s390_set_current_function): Invoke s390_indirect_branch_settings.
22669 (s390_output_indirect_thunk_function): New function.
22670 (s390_code_end): Implement target hook.
22671 (s390_case_values_threshold): Implement target hook.
22672 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
22674 * config/s390/s390.h (struct s390_frame_layout)
22675 (struct machine_function): Move here from s390.c.
22676 (TARGET_INDIRECT_BRANCH_NOBP_RET)
22677 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
22678 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
22679 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
22680 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
22681 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
22682 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
22683 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
22684 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
22685 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
22686 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
22687 (mnemonic attribute): Add values which aren't recognized
22689 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
22690 pattern for branch conversion. Fix mnemonic attribute.
22691 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
22692 indirect branch via thunk if requested.
22693 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
22694 ("*indirect_jump"): Disable for branch conversion using out of
22696 ("indirect_jump_via_thunk<mode>_z10")
22697 ("indirect_jump_via_thunk<mode>")
22698 ("indirect_jump_via_inlinethunk<mode>_z10")
22699 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
22700 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
22701 ("casesi_jump_via_inlinethunk<mode>_z10")
22702 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
22703 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
22704 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
22705 ("*indirect2_jump"): Disable for branch conversion.
22706 ("casesi_jump"): Turn into expander and expand patterns for branch
22708 ("return_use"): New expander.
22709 ("*return"): Emit return via thunk and rename it to ...
22710 ("*return<mode>"): ... this one.
22711 * config/s390/s390.opt: Add new options and and enum for the
22714 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
22716 * lra-constraints.c (match_reload): Unconditionally use
22717 gen_lowpart_SUBREG, rather than selecting between that
22718 and equivalent gen_rtx_SUBREG code.
22720 2018-02-08 Richard Biener <rguenther@suse.de>
22722 PR tree-optimization/84233
22723 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
22724 changed flag instead of boguously re-using phi_inserted.
22726 2018-02-08 Martin Jambor <mjambor@suse.cz>
22728 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
22729 static local variables.
22731 2018-02-08 Richard Biener <rguenther@suse.de>
22733 PR tree-optimization/84278
22734 * tree-vect-stmts.c (vectorizable_store): When looking for
22735 smaller vector types to perform grouped strided loads/stores
22736 make sure the mode is supported by the target.
22737 (vectorizable_load): Likewise.
22739 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
22741 * config/aarch64/aarch64.c (aarch64_components_for_bb):
22742 Increase LDP/STP opportunities by adding adjacent callee-saves.
22744 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
22746 PR rtl-optimization/84068
22747 PR rtl-optimization/83459
22748 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
22750 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
22752 PR tree-optimization/84224
22753 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
22754 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
22755 non-zero arguments.
22757 2018-02-07 Iain Sandoe <iain@codesourcery.com>
22760 * config/rs6000/altivec.md (*restore_world): Remove LR use.
22761 * config/rs6000/predicates.md (restore_world_operation): Adjust op
22762 count, remove one USE.
22764 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
22766 * doc/install.texi (Configuration): Document the
22767 --with-long-double-format={ibm,ieee} PowerPC configuration
22771 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
22772 Convert from define_expand to be define_insn_and_split. Rework
22773 float/double/_Float128 conversions to QI/HI/SImode to work with
22774 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
22775 conversions to QI/HImode types did a store and then a load to
22776 truncate the value. For conversions to VSX registers, don't split
22777 the insn, instead emit the code directly. Use the code iterator
22778 any_fix to combine signed and unsigned conversions.
22779 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
22780 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
22781 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22782 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
22783 (fix_<mode>di2_hw): Likewise.
22784 (fixuns_<mode>di2_hw): Likewise.
22785 (fix_<mode>si2_hw): Likewise.
22786 (fixuns_<mode>si2_hw): Likewise.
22787 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
22788 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22789 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
22790 fix<uns>_trunc<SFDF:mode>si2_p8.
22791 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
22792 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
22793 (fix<uns>_<mode>_mem): Likewise.
22794 (fctiw<u>z_<mode>_mem): Likewise.
22795 (fix<uns>_<mode>_mem): Likewise.
22796 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
22797 the register allocator from doing a direct move to the GPRs to do
22798 a store, and instead use the ISA 3.0 store byte/half-word from
22799 vector register instruction. For IEEE 128-bit floating point,
22800 also optimize stores of 32-bit ints.
22801 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
22803 2018-02-07 Alan Hayward <alan.hayward@arm.com>
22805 * genextract.c (push_pathstr_operand): New function to support
22807 (walk_rtx): Call push_pathstr_operand.
22808 (print_path): Support [a-zA-Z].
22810 2018-02-07 Richard Biener <rguenther@suse.de>
22812 PR tree-optimization/84037
22813 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
22814 (cse_and_gimplify_to_preheader): Declare.
22815 (vect_get_place_in_interleaving_chain): Likewise.
22816 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22818 (_loop_vec_info::~_loop_vec_info): Delete it.
22819 (cse_and_gimplify_to_preheader): New function.
22820 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
22821 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
22822 (vectorizable_load): Likewise. For grouped stores always base
22823 the IV on the first element.
22824 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
22825 condition before gimplifying.
22827 2018-02-07 Jakub Jelinek <jakub@redhat.com>
22829 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
22830 *DIV_EXPR and *MOD_EXPR.
22832 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
22835 * config/i386/i386.c (ix86_option_override_internal): Mask out
22836 the CF_SET bit when checking -fcf-protection.
22838 2018-02-07 Tom de Vries <tom@codesourcery.com>
22841 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
22844 2018-02-07 Richard Biener <rguenther@suse.de>
22846 PR tree-optimization/84204
22847 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
22850 PR tree-optimization/84205
22851 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
22852 special-case isl_ast_op_zdiv_r.
22854 PR tree-optimization/84223
22855 * graphite-scop-detection.c (gather_bbs::before_dom_children):
22856 Only add conditions from within the region.
22857 (gather_bbs::after_dom_children): Adjust.
22859 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
22862 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
22863 * config/avr/avr.md: Only post-reload split REG-REG moves if
22864 either register is GENERAL_REG_P.
22866 2018-02-07 Jakub Jelinek <jakub@redhat.com>
22868 PR tree-optimization/84235
22869 * tree-ssa-scopedtables.c
22870 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
22871 if the subtraction is performed in floating point type where NaNs are
22872 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
22873 build 1. Formatting fix.
22875 2018-02-06 Jakub Jelinek <jakub@redhat.com>
22878 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
22879 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
22880 and skip it regardless of bb boundaries. Use CALL_P macro,
22881 don't test INSN_P (insn) together with CALL_P or JUMP_P check
22882 unnecessarily, formatting fix.
22884 2018-02-06 Michael Collison <michael.collison@arm.com>
22886 * config/arm/thumb2.md:
22887 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
22888 (*thumb_mov_notscc): Ditto.
22890 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
22893 * config/rs6000/rs6000.md (su code attribute): Use "u" for
22894 unsigned_fix, not "s".
22896 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22898 * configure.ac (gcc_fn_eh_frame_ro): New function.
22899 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
22900 correct .eh_frame permissions.
22901 * configure: Regenerate.
22903 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
22905 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
22906 irrelevant options.
22908 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22910 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22911 Display warning message for -mno-speculate-indirect-jumps.
22913 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
22915 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22917 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22919 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
22921 PR tree-optimization/84225
22922 * tree-eh.c (find_trapping_overflow): Only call
22923 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22925 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22928 * config/i386/i386.c: Reimplement the check of possible options
22929 -mibt/-mshstk conbination. Change error messages.
22930 * doc/invoke.texi: Fix a typo: remove extra '='.
22932 2018-02-06 Marek Polacek <polacek@redhat.com>
22934 PR tree-optimization/84228
22935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22937 2018-02-06 Tamar Christina <tamar.christina@arm.com>
22940 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22941 emitted arch directives.
22942 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22943 __ARM_FEATURE_COPROC before changing architectures.
22945 2018-02-06 Richard Biener <rguenther@suse.de>
22947 * config/i386/i386.c (print_reg): Fix typo.
22948 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22950 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
22952 * configure: Regenerate.
22954 2018-02-05 Martin Sebor <msebor@redhat.com>
22956 PR tree-optimization/83369
22957 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22960 2018-02-05 Martin Liska <mliska@suse.cz>
22962 * doc/invoke.texi: Cherry-pick upstream r323995.
22964 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
22966 * ira.c (ira_init_register_move_cost): Adjust comment.
22968 2018-02-05 Martin Liska <mliska@suse.cz>
22970 PR gcov-profile/84137
22971 * doc/gcov.texi: Fix typo in documentation.
22973 2018-02-05 Martin Liska <mliska@suse.cz>
22975 PR gcov-profile/83879
22976 * doc/gcov.texi: Document necessity of --dynamic-list-data when
22977 using dlopen functionality.
22979 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
22981 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22982 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22983 _mm_maskz_range_ss, _mm_mask_range_round_ss,
22984 _mm_maskz_range_round_ss): New intrinsics.
22985 (__builtin_ia32_rangesd128_round)
22986 (__builtin_ia32_rangess128_round): Remove.
22987 (__builtin_ia32_rangesd128_mask_round,
22988 __builtin_ia32_rangess128_mask_round): New builtins.
22989 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22990 __builtin_ia32_rangess128_round): Remove.
22991 (__builtin_ia32_rangesd128_mask_round,
22992 __builtin_ia32_rangess128_mask_round): New builtins.
22993 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22994 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22995 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22996 "<round_saeonly_constraint>")): Changed to ...
22997 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22998 "<round_saeonly_scalar_constraint>")): ... this.
22999 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
23000 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
23001 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
23002 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
23003 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
23005 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
23007 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
23009 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
23010 Remove all values except native, 8540 and 8548.
23012 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
23014 * config/i386/i386.c (ix86_output_function_return): Pass
23015 INVALID_REGNUM, instead of -1, as invalid register number to
23016 indirect_thunk_name and output_indirect_thunk.
23018 2018-02-02 Julia Koval <julia.koval@intel.com>
23020 * config.gcc: Add -march=icelake.
23021 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
23022 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
23023 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
23024 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
23025 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
23026 (processor_target_table): Add icelake.
23027 (ix86_option_override_internal): Handle new PTAs.
23028 (get_builtin_code_for_version): Handle icelake.
23029 (M_INTEL_COREI7_ICELAKE): New.
23030 (fold_builtin_cpu): Handle icelake.
23031 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
23032 * doc/invoke.texi: Add -march=icelake.
23034 2018-02-02 Julia Koval <julia.koval@intel.com>
23036 * config/i386/i386.c (ix86_option_override_internal): Change flags type
23037 to wide_int_bitmask.
23038 * wide-int-bitmask.h: New.
23040 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
23043 * config/i386/i386.md: Replace Pmode with word_mode in
23044 builtin_setjmp_setup and builtin_longjmp to support x32.
23046 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
23050 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
23052 (rs6000_supported_cpu_names): New static variable.
23053 (linux_cpu_translation_table): Likewise.
23054 (elf_platform) <cpu>: Define new static variable and use it.
23055 Translate kernel AT_PLATFORM name to canonical name if needed.
23056 Error if platform name is unknown.
23058 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
23061 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
23063 2018-02-01 Jeff Law <law@redhat.com>
23066 * config/i386/i386.c (release_scratch_register_on_entry): Add new
23067 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
23068 the scratch if RELEASE_VIA_POP is false.
23069 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
23070 If we have to save a temporary register, decrement SIZE appropriately.
23071 Pass new arguments to release_scratch_register_on_entry.
23072 (ix86_adjust_stack_and_probe): Likewise.
23073 (ix86_emit_probe_stack_range): Pass new arguments to
23074 release_scratch_register_on_entry.
23076 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
23078 PR rtl-optimization/84157
23079 * combine.c (change_zero_ext): Use REG_P predicate in
23080 front of HARD_REGISTER_P predicate.
23082 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
23084 * config/avr/avr.c (avr_option_override): Move disabling of
23085 -fdelete-null-pointer-checks to...
23086 * common/config/avr/avr-common.c (avr_option_optimization_table):
23089 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
23091 PR tree-optimization/81635
23092 * tree-data-ref.c (split_constant_offset_1): For types that
23093 wrap on overflow, try to use range info to prove that wrapping
23096 2018-02-01 Renlin Li <renlin.li@arm.com>
23099 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
23100 TAILCALL_ADDR_REGS.
23101 (aarch64_register_move_cost): Likewise.
23102 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
23103 TAILCALL_ADDR_REGS.
23104 (REG_CLASS_NAMES): Likewise.
23105 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
23106 TAILCALL_ADDR_REGS. Remove IP registers.
23107 * config/aarch64/aarch64.md (Ucs): Update register constraint.
23109 2018-02-01 Richard Biener <rguenther@suse.de>
23111 * domwalk.h (dom_walker::dom_walker): Add additional constructor
23112 for specifying RPO order and allow NULL for that.
23113 * domwalk.c (dom_walker::dom_walker): Likewise.
23114 (dom_walker::walk): Handle NULL RPO order.
23115 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23117 (rewrite_update_dom_walker): Likewise.
23118 (mark_def_dom_walker): Likewise.
23120 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
23122 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23123 (aarch64_maybe_expand_sve_subreg_move): Declare.
23124 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23125 * config/aarch64/predicates.md (aarch64_any_register_operand): New
23127 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23128 that are semantically a reverse operation.
23129 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23130 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23131 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23133 (aarch64_can_change_mode_class): For big-endian, forbid changes
23134 between two SVE modes if they have different element sizes.
23136 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
23138 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23139 the TImode handling for big-endian targets.
23141 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
23143 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23144 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
23146 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23147 Remove BSWAP handing for big-endian targets and use the form of
23148 LD1RQ appropariate for the mode.
23150 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
23152 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23153 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
23154 duplicated element.
23156 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
23159 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
23160 check for operands that need to go through aarch64_sve_reload_be.
23162 2018-02-01 Jakub Jelinek <jakub@redhat.com>
23164 PR tree-optimization/81661
23165 PR tree-optimization/84117
23166 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23167 * tree-eh.c: Include gimplify.h.
23168 (find_trapping_overflow, replace_trapping_overflow,
23169 rewrite_to_non_trapping_overflow): New functions.
23170 * tree-vect-loop.c: Include tree-eh.h.
23171 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23172 * tree-data-ref.c: Include tree-eh.h.
23173 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23175 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
23177 PR rtl-optimization/84123
23178 * combine.c (change_zero_ext): Check if hard register satisfies
23179 can_change_dest_mode before calling gen_lowpart_SUBREG.
23181 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
23184 * ira.c (ira_init_register_move_cost): Remove assert.
23186 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
23188 PR rtl-optimization/84071
23189 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23190 * doc/tm.texi: Regenerate.
23192 2018-01-31 Richard Biener <rguenther@suse.de>
23194 PR tree-optimization/84132
23195 * tree-data-ref.c (analyze_miv_subscript): Properly
23196 check whether evolution_function_is_affine_multivariate_p
23197 before calling gcd_of_steps_may_divide_p.
23199 2018-01-31 Julia Koval <julia.koval@intel.com>
23202 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23203 * config/i386/i386.md (rdpid_rex64) New.
23204 (rdpid): Make 32bit only.
23206 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
23209 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23210 an IDENTIFIER_NODE for FUNCTION_TYPE's.
23212 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
23215 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23217 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23219 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
23221 PR rtl-optimization/84071
23222 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
23223 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
23225 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
23227 * config/arc/arc.c (arc_handle_aux_attribute): New function.
23228 (arc_attribute_table): Add 'aux' attribute.
23229 (arc_in_small_data_p): Consider aux like variables.
23230 (arc_is_aux_reg_p): New function.
23231 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
23232 (arc_get_aux_arg): New function.
23233 (prepare_move_operands): Handle aux-register access.
23234 (arc_handle_aux_attribute): New function.
23235 * doc/extend.texi (ARC Variable attributes): Add subsection.
23237 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
23239 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
23240 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
23241 (arc_attribute_table): Add 'uncached' attribute.
23242 (arc_print_operand): Print '.di' flag for uncached memory
23244 (arc_in_small_data_p): Do not consider for small data the uncached
23246 (arc_is_uncached_mem_p): New function.
23247 * config/arc/predicates.md (compact_store_memory_operand): Check
23248 for uncached memory accesses.
23249 (nonvol_nonimm_operand): Likewise.
23250 * doc/extend.texi (ARC Type Attribute): New subsection.
23252 2018-01-31 Jakub Jelinek <jakub@redhat.com>
23255 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
23256 falign-loops=): Add Optimization flag.
23258 2018-01-30 Jeff Law <law@redhat.com>
23261 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
23262 INT_REGISTERS_SAVED. Check it prior to calling
23263 get_scratch_register_on_entry.
23264 (ix86_adjust_stack_and_probe): Similarly.
23265 (ix86_emit_probe_stack_range): Similarly.
23266 (ix86_expand_prologue): Corresponding changes.
23268 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23271 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
23272 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
23274 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
23277 * lra-constraints.c (curr_insn_transform): Process AND in the
23280 2018-01-30 Jakub Jelinek <jakub@redhat.com>
23282 PR rtl-optimization/83986
23283 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
23284 dependence against last_pending_memory_flush in addition to
23285 pending_jump_insns.
23287 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
23289 PR tree-optimization/81611
23290 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
23293 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23296 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
23299 2018-01-30 Richard Biener <rguenther@suse.de>
23300 Jakub Jelinek <jakub@redhat.com>
23302 PR tree-optimization/84111
23303 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
23304 inner loops added during recursion, as they don't have up-to-date
23307 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
23310 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
23311 (can_inline_edge_by_limits_p): ... here.
23312 (can_early_inline_edge_p, check_callers,
23313 update_caller_keys, update_callee_keys, recursive_inlining,
23314 add_new_edges_to_heap, speculation_useful_p,
23315 inline_small_functions,
23316 inline_small_functions, flatten_function,
23317 inline_to_all_callers_1): Update.
23319 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
23321 * profile-count.c (profile_count::combine_with_ipa_count): Handle
23324 2018-01-30 Richard Biener <rguenther@suse.de>
23326 PR tree-optimization/83008
23327 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
23328 invariant and constant vector uses in stmts when they need
23329 more than one stmt.
23331 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23334 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
23335 * configure: Regenerate.
23337 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
23339 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
23341 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
23342 Use gen_rtx_REG rather than gen_lowpart.
23344 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
23346 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
23347 rather than 0 when creating partial subregs.
23349 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
23351 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
23354 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
23357 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
23358 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
23359 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
23360 flags. This restores the settings used before the 2017-07-24.
23361 Turning off pre increment/decrement/modify allows IVOPTS to
23362 optimize DF/SF loops where the index is an int.
23364 2018-01-29 Richard Biener <rguenther@suse.de>
23365 Kelvin Nilsen <kelvin@gcc.gnu.org>
23368 * tree-vect-stmts.c (vectorizable_call): Don't call
23369 targetm.vectorize_builtin_md_vectorized_function if callee is
23372 2018-01-22 Carl Love <cel@us.ibm.com>
23374 * doc/extend.tex: Fix typo in second arg in
23375 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
23377 2018-01-29 Richard Biener <rguenther@suse.de>
23379 PR tree-optimization/84086
23380 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
23381 (flush_ssaname_freelist): When SSA names were released reset
23382 the SCEV hash table.
23384 2018-01-29 Richard Biener <rguenther@suse.de>
23386 PR tree-optimization/84057
23387 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
23388 removed paths when removing edges.
23390 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
23392 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
23393 -mfunction-return=@var{choice}.
23395 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
23397 PR diagnostic/84034
23398 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
23399 Handle CR like TAB.
23400 (layout::print_source_line): Likewise.
23401 (test_get_line_width_without_trailing_whitespace): Add test cases.
23403 2018-01-27 Jakub Jelinek <jakub@redhat.com>
23405 PR middle-end/84040
23406 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
23409 2018-01-26 Jim Wilson <jimw@sifive.com>
23411 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
23413 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
23416 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23418 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
23419 and CMP + SUB-immediate -> SUBS.
23421 2018-01-26 Martin Sebor <msebor@redhat.com>
23423 PR tree-optimization/83896
23424 * tree-ssa-strlen.c (get_string_len): Rename...
23425 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
23426 Avoid assuming length is constant.
23427 (handle_char_store): Use HOST_WIDE_INT for string length.
23429 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
23432 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
23433 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
23435 2018-01-26 Richard Biener <rguenther@suse.de>
23437 PR rtl-optimization/84003
23438 * dse.c (record_store): Only record redundant stores when
23439 the earlier store aliases at least all accesses the later one does.
23441 2018-01-26 Jakub Jelinek <jakub@redhat.com>
23443 PR rtl-optimization/83985
23444 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
23445 REG_CFA_RESTORE insns.
23446 (delete_unmarked_insns): Don't ignore separate shrink wrapping
23447 REG_CFA_RESTORE insns here.
23450 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
23451 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
23453 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23455 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
23456 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
23457 (arc_init): Likewise.
23458 (arc_override_options): Likewise.
23459 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
23461 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
23463 * config/arc/arc.h (TARGET_DBNZ): Define.
23464 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
23465 properly set the tune attribute.
23466 (dbnz): Use TARGET_DBNZ guard.
23467 * config/arc/arc.opt (mtune): Add core3 option.
23469 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23471 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
23472 recognize new pic like addresses.
23473 (arc_delegitimize_address): Clean up.
23475 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23477 * config/arc/arc-arches.def: Option mrf16 valid for all
23479 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
23480 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
23481 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
23482 * config/arc/arc-tables.opt: Regenerate.
23483 * config/arc/arc.c (arc_conditional_register_usage): Handle
23484 reduced register file case.
23485 (arc_file_start): Set must have build attributes.
23486 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
23487 mrf16 option value.
23488 * config/arc/arc.opt (mrf16): Add new option.
23489 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
23490 * config/arc/genmultilib.awk: Handle new mrf16 option.
23491 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
23492 * config/arc/t-multilib: Regenerate.
23493 * doc/invoke.texi (ARC Options): Document mrf16 option.
23495 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23497 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
23498 * config/arc/arc.c (arc_handle_secure_attribute): New function.
23499 (arc_attribute_table): Add 'secure_call' attribute.
23500 (arc_print_operand): Print secure call operand.
23501 (arc_function_ok_for_sibcall): Don't optimize tail calls when
23503 (arc_is_secure_call_p): New function. * config/arc/arc.md
23504 (call_i): Add support for sjli instruction.
23505 (call_value_i): Likewise.
23506 * config/arc/constraints.md (Csc): New constraint.
23508 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23509 John Eric Martin <John.Martin@emmicro-us.com>
23511 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
23512 * config/arc/arc.c (_arc_jli_section): New struct.
23513 (arc_jli_section): New type.
23514 (rc_jli_sections): New static variable.
23515 (arc_handle_jli_attribute): New function.
23516 (arc_attribute_table): Add jli_always and jli_fixed attribute.
23517 (arc_file_end): New function.
23518 (TARGET_ASM_FILE_END): Define.
23519 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
23520 (arc_add_jli_section): New function.
23521 (jli_call_scan): Likewise.
23522 (arc_reorg): Call jli_call_scan.
23523 (arc_output_addsi): Remove 'S' from printing asm operand.
23524 (arc_is_jli_call_p): New function.
23525 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
23527 (movhi_insn): Likewise.
23528 (movsi_insn): Likewise.
23529 (movsi_set_cc_insn): Likewise.
23530 (loadqi_update): Likewise.
23531 (load_zeroextendqisi_update): Likewise.
23532 (load_signextendqisi_update): Likewise.
23533 (loadhi_update): Likewise.
23534 (load_zeroextendhisi_update): Likewise.
23535 (load_signextendhisi_update): Likewise.
23536 (loadsi_update): Likewise.
23537 (loadsf_update): Likewise.
23538 (movsicc_insn): Likewise.
23539 (bset_insn): Likewise.
23540 (bxor_insn): Likewise.
23541 (bclr_insn): Likewise.
23542 (bmsk_insn): Likewise.
23543 (bicsi3_insn): Likewise.
23544 (cmpsi_cc_c_insn): Likewise.
23545 (movsi_ne): Likewise.
23546 (movsi_cond_exec): Likewise.
23547 (clrsbsi2): Likewise.
23548 (norm_f): Likewise.
23558 (call_i): Remove 'S' asm letter, add jli instruction.
23559 (call_value_i): Likewise.
23560 * config/arc/arc.op (mjli-always): New option.
23561 * config/arc/constraints.md (Cji): New constraint.
23562 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
23564 (subsf3_fpx): Likewise.
23565 (mulsf3_fpx): Likewise.
23566 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
23568 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
23569 function attrbutes.
23570 * doc/invoke.texi (ARC): Document mjli-always option.
23572 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
23574 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
23575 avoid addition with 0 and use incw and decw where possible.
23577 2018-01-26 Richard Biener <rguenther@suse.de>
23579 PR tree-optimization/81082
23580 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
23581 association if it requires casting to unsigned.
23582 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
23583 from fold_plusminus_mult_expr to catch important cases late when
23584 range info is available.
23586 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23588 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
23589 * configure.ac (hidden_linkonce): New test.
23590 * configure: Regenerate.
23591 * config.in: Regenerate.
23593 2018-01-26 Julia Koval <julia.koval@intel.com>
23595 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
23596 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
23597 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
23598 _mm_mask_bitshuffle_epi64_mask): Fix type.
23599 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
23600 USI_FTYPE_V4DI_V4DI_USI): Remove.
23601 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
23602 __builtin_ia32_vpshufbitqmb256_mask,
23603 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
23604 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
23605 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
23607 2018-01-26 Alan Modra <amodra@gmail.com>
23610 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
23611 UNSPEC_VBPERMQ. Sort other unspecs.
23613 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
23615 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
23617 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
23619 PR middle-end/83055
23620 * predict.c (drop_profile): Do not push/pop cfun; update also
23622 (handle_missing_profiles): Fix logic looking for zero profiles.
23624 2018-01-25 Jakub Jelinek <jakub@redhat.com>
23626 PR middle-end/83977
23627 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
23628 on functions with #pragma omp declare simd or functions with simd
23630 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
23631 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
23632 Remove trailing \n from warning_at calls.
23634 2018-01-25 Tom de Vries <tom@codesourcery.com>
23637 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23638 for neutered workers.
23640 2018-01-24 Joseph Myers <joseph@codesourcery.com>
23643 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
23644 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
23646 2018-01-24 Jeff Law <law@redhat.com>
23649 * i386.c (get_probe_interval): Move to earlier point.
23650 (ix86_compute_frame_layout): If -fstack-clash-protection and
23651 the frame is larger than the probe interval, then use pushes
23652 to save registers rather than reg->mem moves.
23653 (ix86_expand_prologue): Remove conditional for int_registers_saved
23656 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
23659 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
23660 min/max for never referenced object.
23662 2018-01-24 Jakub Jelinek <jakub@redhat.com>
23664 PR middle-end/83977
23665 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
23667 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
23668 attributes from DECL_ATTRIBUTES (decl) without affecting
23669 DECL_ATTRIBUTES (current_function_decl).
23670 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
23671 functions with non-NULL DECL_ABSTRACT_ORIGIN.
23673 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
23675 PR tree-optimization/83979
23676 * fold-const.c (fold_comparison): Use constant_boolean_node
23677 instead of boolean_{true,false}_node.
23679 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
23681 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
23684 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23686 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23687 Simplify the clause that sets the length attribute.
23688 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23689 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
23690 clause that sets the length attribute.
23691 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23693 2018-01-24 Tom de Vries <tom@codesourcery.com>
23696 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
23697 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
23698 Add strict parameter.
23699 (prevent_branch_around_nothing): Insert dummy insn between branch to
23700 label and label with no ptx insn inbetween.
23701 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
23703 2018-01-24 Tom de Vries <tom@codesourcery.com>
23706 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23707 for neutered threads in warp.
23708 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
23710 2018-01-24 Richard Biener <rguenther@suse.de>
23712 PR tree-optimization/83176
23713 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
23716 2018-01-24 Richard Biener <rguenther@suse.de>
23718 PR tree-optimization/82819
23719 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
23720 code generating pluses that are no-ops in the target precision.
23722 2018-01-24 Richard Biener <rguenther@suse.de>
23724 PR middle-end/84000
23725 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
23727 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23729 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
23730 to merge probabilities.
23731 * predict.c (probably_never_executed): Also mark as cold functions
23732 with global 0 profile and guessed local profile.
23733 * profile-count.c (profile_probability::combine_with_count): New
23735 * profile-count.h (profile_probability::operator*,
23736 profile_probability::operator*=, profile_probability::operator/,
23737 profile_probability::operator/=): Reduce precision to adjusted
23738 and set value to guessed on contradictory divisions.
23739 (profile_probability::combine_with_freq): Remove.
23740 (profile_probability::combine_wiht_count): Declare.
23741 (profile_count::force_nonzero):: Set to adjusted.
23742 (profile_count::probability_in):: Set quality to adjusted.
23743 * tree-ssa-tail-merge.c (replace_block_by): Use
23744 combine_with_count.
23746 2018-01-23 Andrew Waterman <andrew@sifive.com>
23747 Jim Wilson <jimw@sifive.com>
23749 * config/riscv/riscv.c (riscv_stack_boundary): New.
23750 (riscv_option_override): Set riscv_stack_boundary. Handle
23751 riscv_preferred_stack_boundary_arg.
23752 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
23753 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
23754 (STACK_BOUNDARY): Set to riscv_stack_boundary.
23755 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
23756 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
23757 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
23759 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
23762 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
23763 of struct ix86_frame.
23764 (ix86_expand_epilogue): Likewise. Add a local variable for
23765 the reg_save_offset field in struct ix86_frame.
23767 2018-01-23 Bin Cheng <bin.cheng@arm.com>
23769 PR tree-optimization/82604
23770 * tree-loop-distribution.c (enum partition_kind): New enum item
23771 PKIND_PARTIAL_MEMSET.
23772 (partition_builtin_p): Support above new enum item.
23773 (generate_code_for_partition): Ditto.
23774 (compute_access_range): Differentiate cases that equality can be
23775 proven at all loops, the innermost loops or no loops.
23776 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
23777 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
23778 (finalize_partitions, distribute_loop): Don't fuse partition of
23779 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
23780 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
23781 parloop is enabled.
23783 2018-01-23 Martin Liska <mliska@suse.cz>
23785 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
23786 order to ignore the predictor.
23787 (PRED_POLYMORPHIC_CALL): Likewise.
23788 (PRED_RECURSIVE_CALL): Likewise.
23790 2018-01-23 Martin Liska <mliska@suse.cz>
23792 * tree-profile.c (tree_profiling): Print function header to
23793 aware reader which function we are working on.
23794 * value-prof.c (gimple_find_values_to_profile): Do not print
23795 not interesting value histograms.
23797 2018-01-23 Martin Liska <mliska@suse.cz>
23799 * profile-count.h (enum profile_quality): Add
23800 profile_uninitialized as the first value. Do not number values
23801 as they are zero based.
23802 (profile_count::verify): Update sanity check.
23803 (profile_probability::verify): Likewise.
23805 2018-01-23 Nathan Sidwell <nathan@acm.org>
23807 * doc/invoke.texi (ffor-scope): Deprecate.
23809 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23811 PR tree-optimization/83510
23812 * domwalk.c (set_all_edges_as_executable): New function.
23813 (dom_walker::dom_walker): Convert bool param
23814 "skip_unreachable_blocks" to enum reachability. Move setup of
23815 edge flags to set_all_edges_as_executable and only do it when
23816 reachability is REACHABLE_BLOCKS.
23817 * domwalk.h (enum dom_walker::reachability): New enum.
23818 (dom_walker::dom_walker): Convert bool param
23819 "skip_unreachable_blocks" to enum reachability.
23820 (set_all_edges_as_executable): New decl.
23821 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
23822 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
23824 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
23825 but converting true to REACHABLE_BLOCKS.
23826 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
23828 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
23829 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
23830 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
23832 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
23833 if check_all_array_refs will be called.
23835 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23837 * tree.c (selftest::test_location_wrappers): Add more test
23840 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23842 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
23843 (selftest::test_bit_in_range): Likewise.
23845 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
23848 * doc/sourcebuild.texi (vect_float): Say that the selector
23849 only describes the situation when -funsafe-math-optimizations is on.
23850 (vect_float_strict): Document.
23852 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
23854 PR tree-optimization/83965
23855 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
23856 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
23857 instead of checking only for a reduction.
23858 (vect_recog_widen_sum_pattern): Likewise.
23860 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23862 * predict.c (probably_never_executed): Only use precise profile info.
23863 (compute_function_frequency): Skip after inlining hack since we now
23864 have quality checking.
23866 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23868 * profile-count.h (profile_probability::very_unlikely,
23869 profile_probability::unlikely, profile_probability::even): Set
23870 precision to guessed.
23872 2018-01-23 Richard Biener <rguenther@suse.de>
23874 PR tree-optimization/83963
23875 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
23876 Properly terminate dominator walk when crossing the exit edge not
23877 when visiting its source block.
23879 2018-01-23 Jakub Jelinek <jakub@redhat.com>
23882 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
23883 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
23885 2018-01-22 Jakub Jelinek <jakub@redhat.com>
23887 PR tree-optimization/83957
23888 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
23889 semicolon after for body surrounded by braces.
23891 PR tree-optimization/83081
23892 * profile-count.h (profile_probability::split): New method.
23893 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
23894 Use profile_probability::split.
23895 (do_compare_rtx_and_jump): Fix adjustment of probabilities
23896 when splitting a single conditional jump into 2.
23898 2018-01-22 David Malcolm <dmalcolm@redhat.com>
23900 PR tree-optimization/69452
23901 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
23904 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23906 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
23907 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
23908 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
23910 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23912 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23914 * config/rl78/rl78.md (movdi): New define_expand.
23915 * config/rl78/rl78.c (rl78_split_movdi): New function.
23917 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
23920 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23922 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23923 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23924 128-bit to produce an UNSPEC move to get the double word with the
23925 signbit and then a shift directly to do signbit.
23926 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23927 implementation with a new version that just does either a direct
23928 move or a regular move. Move memory interface to separate insns.
23929 Move insns so they are next to the expander.
23930 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23931 with signbit move. Split big and little endian case.
23932 (signbit<mode>2_dm_mem_le): Likewise.
23933 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23934 (signbit<mode>2_dm2): Likewise.
23936 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23938 * config/rl78/rl78.md (anddi3): New define_expand.
23940 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23942 * config/rl78/rl78.md (umindi3): New define_expand.
23944 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23946 * config/rl78/rl78.md (smindi3): New define_expand.
23948 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23950 * config/rl78/rl78.md (smaxdi3): New define_expand.
23952 2018-01-22 Carl Love <cel@us.ibm.com>
23954 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23955 LVX_V1TI): Add macro expansion.
23956 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23957 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23958 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23959 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23960 Change check to determine if the instruction is a byte reversing
23961 entry. Fix typo in comment.
23962 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23963 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23964 Add def_builtin calls for new builtins.
23965 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23966 Add define_insn expansion.
23968 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23970 * config/rl78/rl78.md (umaxdi3): New define_expand.
23972 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23974 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23975 for non-QImode registers.
23977 2018-01-22 Richard Biener <rguenther@suse.de>
23979 PR tree-optimization/83963
23980 * graphite-scop-detection.c (scop_detection::get_sese): Delay
23981 including the loop exit block.
23982 (scop_detection::merge_sese): Likewise.
23983 (scop_detection::add_scop): Do it here instead.
23985 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23987 * doc/sourcebuild.texi (arm_softfloat): Document.
23989 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
23992 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23993 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23994 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23996 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23997 David Edelsohn <dje.gcc@gmail.com>
24000 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24001 Change "crset eq" to "crset 2".
24002 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24003 (*call_indirect_aix<mode>_nospec): Likewise.
24004 (*call_value_indirect_aix<mode>_nospec): Likewise.
24005 (*call_indirect_elfv2<mode>_nospec): Likewise.
24006 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
24007 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
24008 change assembly output from . to $.
24009 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24010 (indirect_jump<mode>_nospec): Change assembly output from . to $.
24011 (*tablejump<mode>_internal1_nospec): Likewise.
24013 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
24016 * config/sh/sh_optimize_sett_clrt.cc:
24017 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
24019 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
24021 PR tree-optimization/83940
24022 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
24023 offset_dt to vect_constant_def rather than vect_unknown_def_type.
24024 (vect_check_load_store_mask): Add a mask_dt_out parameter and
24025 use it to pass back the definition type.
24026 (vect_check_store_rhs): Likewise rhs_dt_out.
24027 (vect_build_gather_load_calls): Add a mask_dt argument and use
24028 it instead of a call to vect_is_simple_use.
24029 (vectorizable_store): Update calls to vect_check_load_store_mask
24030 and vect_check_store_rhs. Use the dt returned by the latter instead
24031 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
24032 instead of calls to vect_is_simple_use. Pass the scalar rather
24033 than the vector operand to vect_is_simple_use when handling
24034 second and subsequent copies of an rhs value.
24035 (vectorizable_load): Update calls to vect_check_load_store_mask
24036 and vect_build_gather_load_calls. Use the cached mask_dt and
24037 gs_info.offset_dt instead of calls to vect_is_simple_use.
24039 2018-01-20 Jakub Jelinek <jakub@redhat.com>
24041 PR middle-end/83945
24042 * tree-emutls.c: Include gimplify.h.
24043 (lower_emutls_2): New function.
24044 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
24045 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
24046 it before further processing.
24049 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
24050 UINTVAL (trueop1) instead of INTVAL (op1).
24052 2018-01-19 Jakub Jelinek <jakub@redhat.com>
24056 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
24057 INCOMING_FRAME_SP_OFFSET if not defined.
24058 (scan_trace): Add ENTRY argument. If true and
24059 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
24060 emit a note to adjust the CFA offset.
24061 (create_cfi_notes): Adjust scan_trace callers.
24062 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
24063 INCOMING_FRAME_SP_OFFSET in the CIE.
24064 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
24065 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
24067 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
24068 * doc/tm.texi: Regenerated.
24070 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
24072 PR rtl-optimization/83147
24073 * lra-constraints.c (remove_inheritance_pseudos): Use
24074 lra_substitute_pseudo_within_insn.
24076 2018-01-19 Tom de Vries <tom@codesourcery.com>
24077 Cesar Philippidis <cesar@codesourcery.com>
24080 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
24082 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
24085 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
24086 spaces for function labels.
24088 2018-01-19 Martin Liska <mliska@suse.cz>
24090 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
24091 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
24092 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
24093 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
24094 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
24095 (PRED_CONST_RETURN): Change from 69 to 65.
24096 (PRED_NULL_RETURN): Change from 91 to 71.
24097 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
24098 (PRED_LOOP_GUARD): Change from 66 to 73.
24100 2018-01-19 Martin Liska <mliska@suse.cz>
24102 * predict.c (predict_insn_def): Add new assert.
24103 (struct branch_predictor): Change type to signed integer.
24104 (test_prediction_value_range): Amend test to cover
24105 PROB_UNINITIALIZED.
24106 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24107 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24108 (PRED_LOOP_ITERATIONS_MAX): Likewise.
24109 (PRED_LOOP_IV_COMPARE): Likewise.
24110 * predict.h (PROB_UNINITIALIZED): Define new constant.
24112 2018-01-19 Martin Liska <mliska@suse.cz>
24114 * predict.c (dump_prediction): Add new format for
24115 analyze_brprob.py script which is enabled with -details
24117 * profile-count.h (precise_p): New function.
24119 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
24121 PR tree-optimization/83922
24122 * tree-vect-loop.c (vect_verify_full_masking): Return false if
24123 there are no statements that need masking.
24124 (vect_active_double_reduction_p): New function.
24125 (vect_analyze_loop_operations): Use it when handling phis that
24126 are not in the loop header.
24128 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
24130 PR tree-optimization/83914
24131 * tree-vect-loop.c (vectorizable_induction): Don't convert
24132 init_expr or apply the peeling adjustment for inductions
24133 that are nested within the vectorized loop.
24135 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24137 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24140 2018-01-18 Jakub Jelinek <jakub@redhat.com>
24144 * function.h (gimplify_parameters): Add gimple_seq * argument.
24145 * function.c: Include gimple.h and options.h.
24146 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24147 for the added local temporaries if needed.
24148 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24149 if there are any parameter cleanups, wrap whole body into a
24150 try/finally with the cleanups.
24152 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
24155 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
24156 Use GET_MODE_CLASS for scalar floating point.
24158 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
24162 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
24163 Fix call of call_cgraph_insertion_hooks.
24165 2018-01-18 Martin Sebor <msebor@redhat.com>
24167 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24169 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
24172 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24175 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
24178 * common.opt: (-ffile-prefix-map): New option.
24179 * opts.c (common_handle_option): Defer it.
24180 * opts-global.c (handle_common_deferred_options): Handle it.
24181 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24182 * file-prefix-map.h: New file.
24183 (remap_debug_filename, add_debug_prefix_map): ...here.
24184 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24185 * final.c (debug_prefix_map, add_debug_prefix_map
24186 remap_debug_filename): Move to...
24187 * file-prefix-map.c: New file.
24188 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24189 generalize, get rid of alloca(), use strrchr() instead of strchr().
24190 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24191 Implement in terms of add_prefix_map().
24192 (remap_macro_filename, remap_debug_filename): Implement in term of
24194 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24195 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24196 * dbxout.c: Include file-prefix-map.h.
24197 * varasm.c: Likewise.
24198 * vmsdbgout.c: Likewise.
24199 * xcoffout.c: Likewise.
24200 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24201 * doc/cppopts.texi (-fmacro-prefix-map): Document.
24202 * doc/invoke.texi (-ffile-prefix-map): Document.
24203 (-fdebug-prefix-map): Update description.
24205 2018-01-18 Martin Liska <mliska@suse.cz>
24207 * config/i386/i386.c (indirect_thunk_name): Document that also
24209 (output_indirect_thunk): Document why both instructions
24210 (pause and lfence) are generated.
24212 2018-01-18 Richard Biener <rguenther@suse.de>
24214 PR tree-optimization/83887
24215 * graphite-scop-detection.c
24216 (scop_detection::get_nearest_dom_with_single_entry): Remove.
24217 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
24218 (scop_detection::merge_sese): Re-implement with a flood-fill
24219 algorithm that properly finds a SESE region if it exists.
24221 2018-01-18 Jakub Jelinek <jakub@redhat.com>
24224 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
24225 pointer_diff optimizations use view_convert instead of convert.
24227 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24229 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24230 Generate different code for -mno-speculate-indirect-jumps.
24231 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24232 (*call_indirect_aix<mode>): Disable for
24233 -mno-speculate-indirect-jumps.
24234 (*call_indirect_aix<mode>_nospec): New define_insn.
24235 (*call_value_indirect_aix<mode>): Disable for
24236 -mno-speculate-indirect-jumps.
24237 (*call_value_indirect_aix<mode>_nospec): New define_insn.
24238 (*sibcall_nonlocal_sysv<mode>): Generate different code for
24239 -mno-speculate-indirect-jumps.
24240 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24242 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
24244 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
24245 long double type, set the flags for noting the default long double
24246 type, even if we don't pass or return a long double type.
24248 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
24251 * ipa-inline.c (flatten_function): Do not overwrite final inlining
24254 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
24256 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
24257 support for merge[hl].
24258 (fold_mergehl_helper): New helper function.
24259 (tree-vector-builder.h): New #include for tree_vector_builder usage.
24260 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
24261 (altivec_vmrglw_direct): Add xxmrglw insn.
24263 2018-01-17 Andrew Waterman <andrew@sifive.com>
24265 * config/riscv/riscv.c (riscv_conditional_register_usage): If
24266 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
24268 2018-01-17 David Malcolm <dmalcolm@redhat.com>
24271 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
24272 call the lto_location_cache before reading the
24273 DECL_SOURCE_LOCATION of the types.
24275 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
24276 Richard Sandiford <richard.sandiford@linaro.org>
24278 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
24279 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
24280 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
24281 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
24282 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
24284 * config/aarch64/constraints.md (aarch64_movti_operand):
24286 * config/aarch64/predicates.md (Uti): Add new constraint.
24288 2018-01-17 Carl Love <cel@us.ibm.com>
24290 * config/rs6000/vsx.md (define_expand xl_len_r,
24291 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
24292 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
24294 (define_expand, define_insn): Move the shift left from the
24295 define_insn to the define_expand for lxvl and stxvl instructions.
24296 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
24297 and XL_LEN_R definitions to PURE.
24299 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
24301 * config/i386/i386.c (indirect_thunk_name): Declare regno
24302 as unsigned int. Compare regno with INVALID_REGNUM.
24303 (output_indirect_thunk): Ditto.
24304 (output_indirect_thunk_function): Ditto.
24305 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
24306 in the call to output_indirect_thunk_function.
24308 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
24310 PR middle-end/83884
24311 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
24312 rather than the size of inner_type to determine the stack slot size
24313 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
24315 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
24318 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
24321 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
24323 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
24324 endian Linux systems to optionally enable multilibs for selecting
24325 the long double type if the user configured an explicit type.
24326 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
24327 have no long double multilibs if not defined.
24328 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
24329 warn if the user used -mabi={ieee,ibm}longdouble and we built
24330 multilibs for long double.
24331 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
24332 appropriate multilib option.
24333 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
24335 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
24336 for building long double multilibs.
24337 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
24339 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
24341 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
24344 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
24346 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
24349 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
24352 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
24355 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
24357 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
24358 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
24360 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
24362 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
24363 different rtl trees depending on TARGET_64BIT.
24364 (rs6000_gen_lvx): Likewise.
24366 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
24368 * config/visium/visium.md (nop): Tweak comment.
24369 (hazard_nop): Likewise.
24371 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24373 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
24374 -mspeculate-indirect-jumps.
24375 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
24376 for -mno-speculate-indirect-jumps.
24377 (*call_indirect_elfv2<mode>_nospec): New define_insn.
24378 (*call_value_indirect_elfv2<mode>): Disable for
24379 -mno-speculate-indirect-jumps.
24380 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
24381 (indirect_jump): Emit different RTL for
24382 -mno-speculate-indirect-jumps.
24383 (*indirect_jump<mode>): Disable for
24384 -mno-speculate-indirect-jumps.
24385 (*indirect_jump<mode>_nospec): New define_insn.
24386 (tablejump): Emit different RTL for
24387 -mno-speculate-indirect-jumps.
24388 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
24389 (tablejumpsi_nospec): New define_expand.
24390 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
24391 (tablejumpdi_nospec): New define_expand.
24392 (*tablejump<mode>_internal1): Disable for
24393 -mno-speculate-indirect-jumps.
24394 (*tablejump<mode>_internal1_nospec): New define_insn.
24395 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
24398 2018-01-16 Artyom Skrobov tyomitch@gmail.com
24400 * caller-save.c (insert_save): Drop unnecessary parameter. All
24403 2018-01-16 Jakub Jelinek <jakub@redhat.com>
24404 Richard Biener <rguenth@suse.de>
24407 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
24408 return early, inline manually is_gimple_sizepos. Make sure if we
24409 call gimplify_expr we don't end up with a gimple constant.
24410 * tree.c (variably_modified_type_p): Don't return true for
24411 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
24412 * gimplify.h (is_gimple_sizepos): Remove.
24414 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24416 PR tree-optimization/83857
24417 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
24418 vectorizable_live_operation for pure SLP statements.
24419 (vectorizable_live_operation): Handle PHIs.
24421 2018-01-16 Richard Biener <rguenther@suse.de>
24423 PR tree-optimization/83867
24424 * tree-vect-stmts.c (vect_transform_stmt): Precompute
24425 nested_in_vect_loop_p since the scalar stmt may get invalidated.
24427 2018-01-16 Jakub Jelinek <jakub@redhat.com>
24430 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
24431 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
24432 If off is not INTEGER_CST, issue a may not be aligned warning
24433 rather than isn't aligned. Use isn%'t rather than isn't.
24434 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
24436 <case MULT_EXPR>: Improve the case when bottom and one of the
24437 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
24438 operand, in that case check if the other operand is multiple of
24439 bottom divided by the INTEGER_CST operand.
24441 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24444 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
24445 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
24446 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
24447 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
24448 * config/pa/pa.c (pa_function_arg_advance): Likewise.
24449 (pa_function_arg, pa_arg_partial_bytes): Likewise.
24450 (pa_function_arg_size): New function.
24452 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24454 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
24455 in a separate statement.
24457 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24459 PR tree-optimization/83847
24460 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
24461 group gathers and scatters.
24463 2018-01-16 Jakub Jelinek <jakub@redhat.com>
24465 PR rtl-optimization/86620
24466 * params.def (max-sched-ready-insns): Bump minimum value to 1.
24468 PR rtl-optimization/83213
24469 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
24470 to last if both are JUMP_INSNs.
24472 PR tree-optimization/83843
24473 * gimple-ssa-store-merging.c
24474 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
24475 store_immediate_info for bswap/nop orig_stores.
24477 2018-01-15 Andrew Waterman <andrew@sifive.com>
24479 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
24481 <UDIV>: Increase cost if !TARGET_DIV.
24483 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
24485 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
24486 (define_attr "cr_logical_3op"): New.
24487 (cceq_ior_compare): Adjust.
24488 (cceq_ior_compare_complement): Adjust.
24489 (*cceq_rev_compare): Adjust.
24490 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
24491 (is_cracked_insn): Adjust.
24492 (insn_must_be_first_in_group): Adjust.
24493 * config/rs6000/40x.md: Adjust.
24494 * config/rs6000/440.md: Adjust.
24495 * config/rs6000/476.md: Adjust.
24496 * config/rs6000/601.md: Adjust.
24497 * config/rs6000/603.md: Adjust.
24498 * config/rs6000/6xx.md: Adjust.
24499 * config/rs6000/7450.md: Adjust.
24500 * config/rs6000/7xx.md: Adjust.
24501 * config/rs6000/8540.md: Adjust.
24502 * config/rs6000/cell.md: Adjust.
24503 * config/rs6000/e300c2c3.md: Adjust.
24504 * config/rs6000/e500mc.md: Adjust.
24505 * config/rs6000/e500mc64.md: Adjust.
24506 * config/rs6000/e5500.md: Adjust.
24507 * config/rs6000/e6500.md: Adjust.
24508 * config/rs6000/mpc.md: Adjust.
24509 * config/rs6000/power4.md: Adjust.
24510 * config/rs6000/power5.md: Adjust.
24511 * config/rs6000/power6.md: Adjust.
24512 * config/rs6000/power7.md: Adjust.
24513 * config/rs6000/power8.md: Adjust.
24514 * config/rs6000/power9.md: Adjust.
24515 * config/rs6000/rs64.md: Adjust.
24516 * config/rs6000/titan.md: Adjust.
24518 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24520 * config/i386/predicates.md (indirect_branch_operand): Rewrite
24521 ix86_indirect_branch_register logic.
24523 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24525 * config/i386/constraints.md (Bs): Update
24526 ix86_indirect_branch_register check. Don't check
24527 ix86_indirect_branch_register with GOT_memory_operand.
24529 * config/i386/predicates.md (GOT_memory_operand): Don't check
24530 ix86_indirect_branch_register here.
24531 (GOT32_symbol_operand): Likewise.
24533 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24535 * config/i386/predicates.md (constant_call_address_operand):
24536 Rewrite ix86_indirect_branch_register logic.
24537 (sibcall_insn_operand): Likewise.
24539 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24541 * config/i386/constraints.md (Bs): Replace
24542 ix86_indirect_branch_thunk_register with
24543 ix86_indirect_branch_register.
24545 * config/i386/i386.md (indirect_jump): Likewise.
24546 (tablejump): Likewise.
24547 (*sibcall_memory): Likewise.
24548 (*sibcall_value_memory): Likewise.
24549 Peepholes of indirect call and jump via memory: Likewise.
24550 * config/i386/i386.opt: Likewise.
24551 * config/i386/predicates.md (indirect_branch_operand): Likewise.
24552 (GOT_memory_operand): Likewise.
24553 (call_insn_operand): Likewise.
24554 (sibcall_insn_operand): Likewise.
24555 (GOT32_symbol_operand): Likewise.
24557 2018-01-15 Jakub Jelinek <jakub@redhat.com>
24559 PR middle-end/83837
24560 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
24561 type rather than type addr's type points to.
24562 (expand_omp_atomic_mutex): Likewise.
24563 (expand_omp_atomic): Likewise.
24565 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24568 * config/i386/i386.c (output_indirect_thunk_function): Use
24569 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
24570 for __x86_return_thunk.
24572 2018-01-15 Richard Biener <rguenther@suse.de>
24574 PR middle-end/83850
24575 * expmed.c (extract_bit_field_1): Fix typo.
24577 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24580 * config/arm/iterators.md (VF): New mode iterator.
24581 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
24582 Remove integer-related logic from pattern.
24583 (neon_vabd<mode>_3): Likewise.
24585 2018-01-15 Jakub Jelinek <jakub@redhat.com>
24587 PR middle-end/82694
24588 * common.opt (fstrict-overflow): No longer an alias.
24589 (fwrapv-pointer): New option.
24590 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
24591 also for pointer types based on flag_wrapv_pointer.
24592 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
24593 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
24594 opts->x_flag_wrapv got set.
24595 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
24596 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
24597 POINTER_TYPE_OVERFLOW_UNDEFINED.
24598 * match.pd: Likewise in address comparison pattern.
24599 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
24601 2018-01-15 Richard Biener <rguenther@suse.de>
24604 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
24605 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
24606 Reset type names to their identifier if their TYPE_DECL doesn't
24607 have linkage (and thus is used for ODR and devirt).
24608 (save_debug_info_for_decl): Remove.
24609 (save_debug_info_for_type): Likewise.
24610 (add_tree_to_fld_list): Adjust.
24611 * tree-pretty-print.c (dump_generic_node): Make dumping of
24612 type names more robust.
24614 2018-01-15 Richard Biener <rguenther@suse.de>
24616 * BASE-VER: Bump to 8.0.1.
24618 2018-01-14 Martin Sebor <msebor@redhat.com>
24621 * builtins.c (check_access): Avoid warning when the no-warning bit
24624 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
24626 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
24627 * ira-color (allocno_hard_regs_compare): Likewise.
24629 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
24632 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
24633 Use .pushsection/.popsection.
24635 2018-01-14 Martin Sebor <msebor@redhat.com>
24638 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
24640 2018-01-14 Jakub Jelinek <jakub@redhat.com>
24642 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
24643 entry from extra_headers.
24644 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
24645 extra_headers, make the list bitwise identical to the i?86-*-* one.
24647 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24649 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
24650 -mcmodel=large with -mindirect-branch=thunk,
24651 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
24652 -mfunction-return=thunk-extern.
24653 * doc/invoke.texi: Document -mcmodel=large is incompatible with
24654 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
24655 -mfunction-return=thunk and -mfunction-return=thunk-extern.
24657 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24659 * config/i386/i386.c (print_reg): Print the name of the full
24660 integer register without '%'.
24661 (ix86_print_operand): Handle 'V'.
24662 * doc/extend.texi: Document 'V' modifier.
24664 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24666 * config/i386/constraints.md (Bs): Disallow memory operand for
24667 -mindirect-branch-register.
24669 * config/i386/predicates.md (indirect_branch_operand): Likewise.
24670 (GOT_memory_operand): Likewise.
24671 (call_insn_operand): Likewise.
24672 (sibcall_insn_operand): Likewise.
24673 (GOT32_symbol_operand): Likewise.
24674 * config/i386/i386.md (indirect_jump): Call convert_memory_address
24675 for -mindirect-branch-register.
24676 (tablejump): Likewise.
24677 (*sibcall_memory): Likewise.
24678 (*sibcall_value_memory): Likewise.
24679 Disallow peepholes of indirect call and jump via memory for
24680 -mindirect-branch-register.
24681 (*call_pop): Replace m with Bw.
24682 (*call_value_pop): Likewise.
24683 (*sibcall_pop_memory): Replace m with Bs.
24684 * config/i386/i386.opt (mindirect-branch-register): New option.
24685 * doc/invoke.texi: Document -mindirect-branch-register option.
24687 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24689 * config/i386/i386-protos.h (ix86_output_function_return): New.
24690 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
24691 set function_return_type.
24692 (indirect_thunk_name): Add ret_p to indicate thunk for function
24694 (output_indirect_thunk_function): Pass false to
24695 indirect_thunk_name.
24696 (ix86_output_indirect_branch_via_reg): Likewise.
24697 (ix86_output_indirect_branch_via_push): Likewise.
24698 (output_indirect_thunk_function): Create alias for function
24699 return thunk if regno < 0.
24700 (ix86_output_function_return): New function.
24701 (ix86_handle_fndecl_attribute): Handle function_return.
24702 (ix86_attribute_table): Add function_return.
24703 * config/i386/i386.h (machine_function): Add
24704 function_return_type.
24705 * config/i386/i386.md (simple_return_internal): Use
24706 ix86_output_function_return.
24707 (simple_return_internal_long): Likewise.
24708 * config/i386/i386.opt (mfunction-return=): New option.
24709 (indirect_branch): Mention -mfunction-return=.
24710 * doc/extend.texi: Document function_return function attribute.
24711 * doc/invoke.texi: Document -mfunction-return= option.
24713 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24715 * config/i386/i386-opts.h (indirect_branch): New.
24716 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
24717 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
24718 with local indirect jump when converting indirect call and jump.
24719 (ix86_set_indirect_branch_type): New.
24720 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
24721 (indirectlabelno): New.
24722 (indirect_thunk_needed): Likewise.
24723 (indirect_thunk_bnd_needed): Likewise.
24724 (indirect_thunks_used): Likewise.
24725 (indirect_thunks_bnd_used): Likewise.
24726 (INDIRECT_LABEL): Likewise.
24727 (indirect_thunk_name): Likewise.
24728 (output_indirect_thunk): Likewise.
24729 (output_indirect_thunk_function): Likewise.
24730 (ix86_output_indirect_branch_via_reg): Likewise.
24731 (ix86_output_indirect_branch_via_push): Likewise.
24732 (ix86_output_indirect_branch): Likewise.
24733 (ix86_output_indirect_jmp): Likewise.
24734 (ix86_code_end): Call output_indirect_thunk_function if needed.
24735 (ix86_output_call_insn): Call ix86_output_indirect_branch if
24737 (ix86_handle_fndecl_attribute): Handle indirect_branch.
24738 (ix86_attribute_table): Add indirect_branch.
24739 * config/i386/i386.h (machine_function): Add indirect_branch_type
24740 and has_local_indirect_jump.
24741 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
24743 (tablejump): Likewise.
24744 (*indirect_jump): Use ix86_output_indirect_jmp.
24745 (*tablejump_1): Likewise.
24746 (simple_return_indirect_internal): Likewise.
24747 * config/i386/i386.opt (mindirect-branch=): New option.
24748 (indirect_branch): New.
24751 (thunk-inline): Likewise.
24752 (thunk-extern): Likewise.
24753 * doc/extend.texi: Document indirect_branch function attribute.
24754 * doc/invoke.texi: Document -mindirect-branch= option.
24756 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
24759 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
24761 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
24763 * ipa-inline.c (want_inline_small_function_p): Return false if
24764 inlining has already failed with CIF_FINAL_ERROR.
24765 (update_caller_keys): Call want_inline_small_function_p before
24767 (update_callee_keys): Likewise.
24769 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
24771 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
24773 (rs6000_quadword_masked_address_p): Likewise.
24774 (quad_aligned_load_p): Likewise.
24775 (quad_aligned_store_p): Likewise.
24776 (const_load_sequence_p): Add comment to describe the outer-most loop.
24777 (mimic_memory_attributes_and_flags): New function.
24778 (rs6000_gen_stvx): Likewise.
24779 (replace_swapped_aligned_store): Likewise.
24780 (rs6000_gen_lvx): Likewise.
24781 (replace_swapped_aligned_load): Likewise.
24782 (replace_swapped_load_constant): Capitalize argument name in
24783 comment describing this function.
24784 (rs6000_analyze_swaps): Add a third pass to search for vector loads
24785 and stores that access quad-word aligned addresses and replace
24786 with stvx or lvx instructions when appropriate.
24787 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
24788 New function prototype.
24789 (rs6000_quadword_masked_address_p): Likewise.
24790 (rs6000_gen_lvx): Likewise.
24791 (rs6000_gen_stvx): Likewise.
24792 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
24793 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
24794 when memory address is aligned.
24795 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
24796 this split to select lvx instruction when memory address is aligned.
24797 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
24798 instruction when memory address is aligned.
24799 (*vsx_le_perm_load_v16qi): Likewise.
24800 (four unnamed splitters): Modify to select the stvx instruction
24801 when memory is aligned.
24803 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
24805 * predict.c (determine_unlikely_bbs): Handle correctly BBs
24806 which appears in the queue multiple times.
24808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24809 Alan Hayward <alan.hayward@arm.com>
24810 David Sherwood <david.sherwood@arm.com>
24812 * tree-vectorizer.h (vec_lower_bound): New structure.
24813 (_loop_vec_info): Add check_nonzero and lower_bounds.
24814 (LOOP_VINFO_CHECK_NONZERO): New macro.
24815 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
24816 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
24817 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
24818 fields. Make seg_len the distance travelled, not including the
24820 (dr_direction_indicator): Declare.
24821 (dr_zero_step_indicator): Likewise.
24822 (dr_known_forward_stride_p): Likewise.
24823 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
24825 (runtime_alias_check_p): Allow runtime alias checks with
24827 (operator ==): Compare access_size and align.
24828 (prune_runtime_alias_test_list): Rework for new distinction between
24829 the access_size and seg_len.
24830 (create_intersect_range_checks_index): Likewise. Cope with polynomial
24832 (get_segment_min_max): New function.
24833 (create_intersect_range_checks): Use it.
24834 (dr_step_indicator): New function.
24835 (dr_direction_indicator): Likewise.
24836 (dr_zero_step_indicator): Likewise.
24837 (dr_known_forward_stride_p): Likewise.
24838 * tree-loop-distribution.c (data_ref_segment_size): Return
24839 DR_STEP * (niters - 1).
24840 (compute_alias_check_pairs): Update call to the dr_with_seg_len
24842 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
24843 (vect_preserves_scalar_order_p): New function, split out from...
24844 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
24845 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
24846 (vect_vfa_access_size): New function.
24847 (vect_vfa_align): Likewise.
24848 (vect_compile_time_alias): Take access_size_a and access_b arguments.
24849 (dump_lower_bound): New function.
24850 (vect_check_lower_bound): Likewise.
24851 (vect_small_gap_p): Likewise.
24852 (vectorizable_with_step_bound_p): Likewise.
24853 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
24854 depencies if the vectorization factor is 1. Convert the checks
24855 for nonzero steps into checks on the bounds of DR_STEP. Try using
24856 a bunds check for variable steps if the minimum required step is
24857 relatively small. Update calls to the dr_with_seg_len
24858 constructor and to vect_compile_time_alias.
24859 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
24861 (vect_loop_versioning): Call it.
24862 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
24864 (vect_estimate_min_profitable_iters): Account for any bounds checks.
24866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24867 Alan Hayward <alan.hayward@arm.com>
24868 David Sherwood <david.sherwood@arm.com>
24870 * doc/sourcebuild.texi (vect_scatter_store): Document.
24871 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
24873 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
24875 * genopinit.c (main): Add supports_vec_scatter_store and
24876 supports_vec_scatter_store_cached to target_optabs.
24877 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
24878 IFN_MASK_SCATTER_STORE.
24879 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
24881 * internal-fn.h (internal_store_fn_p): Declare.
24882 (internal_fn_stored_value_index): Likewise.
24883 * internal-fn.c (scatter_store_direct): New macro.
24884 (expand_scatter_store_optab_fn): New function.
24885 (direct_scatter_store_optab_supported_p): New macro.
24886 (internal_store_fn_p): New function.
24887 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
24888 IFN_MASK_SCATTER_STORE.
24889 (internal_fn_mask_index): Likewise.
24890 (internal_fn_stored_value_index): New function.
24891 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
24892 for scatter stores.
24893 * optabs-query.h (supports_vec_scatter_store_p): Declare.
24894 * optabs-query.c (supports_vec_scatter_store_p): New function.
24895 * tree-vectorizer.h (vect_get_store_rhs): Declare.
24896 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
24897 true for scatter stores.
24898 (vect_gather_scatter_fn_p): Handle scatter stores too.
24899 (vect_check_gather_scatter): Consider using scatter stores if
24900 supports_vec_scatter_store_p.
24901 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
24902 scatter stores too.
24903 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24904 internal_fn_stored_value_index.
24905 (check_load_store_masking): Handle scatter stores too.
24906 (vect_get_store_rhs): Make public.
24907 (vectorizable_call): Use internal_store_fn_p.
24908 (vectorizable_store): Handle scatter store internal functions.
24909 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
24910 when deciding whether the end of the group has been reached.
24911 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
24912 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24913 (mask_scatter_store<mode>): New insns.
24915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24916 Alan Hayward <alan.hayward@arm.com>
24917 David Sherwood <david.sherwood@arm.com>
24919 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24920 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24921 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24923 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24924 Use vect_truncate_gather_scatter_offset if we can't treat the
24925 operation as a normal gather load or scatter store.
24926 (get_group_load_store_type): Take the gather_scatter_info
24927 as argument. Try using a gather load or scatter store for
24928 single-element groups.
24929 (get_load_store_type): Update calls to get_group_load_store_type
24930 and vect_use_strided_gather_scatters_p.
24932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24933 Alan Hayward <alan.hayward@arm.com>
24934 David Sherwood <david.sherwood@arm.com>
24936 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24937 optional tree argument.
24938 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24940 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24941 but continue to use the current value as a fallback.
24942 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24943 to compare the updates.
24944 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24945 (get_load_store_type): Use it when handling a strided access.
24946 (vect_get_strided_load_store_ops): New function.
24947 (vect_get_data_ptr_increment): Likewise.
24948 (vectorizable_load): Handle strided gather loads. Always pass
24949 a step to vect_create_data_ref_ptr and bump_vector_ptr.
24951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24952 Alan Hayward <alan.hayward@arm.com>
24953 David Sherwood <david.sherwood@arm.com>
24955 * doc/md.texi (gather_load@var{m}): Document.
24956 (mask_gather_load@var{m}): Likewise.
24957 * genopinit.c (main): Add supports_vec_gather_load and
24958 supports_vec_gather_load_cached to target_optabs.
24959 * optabs-tree.c (init_tree_optimization_optabs): Use
24960 ggc_cleared_alloc to allocate target_optabs.
24961 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24962 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24964 * internal-fn.h (internal_load_fn_p): Declare.
24965 (internal_gather_scatter_fn_p): Likewise.
24966 (internal_fn_mask_index): Likewise.
24967 (internal_gather_scatter_fn_supported_p): Likewise.
24968 * internal-fn.c (gather_load_direct): New macro.
24969 (expand_gather_load_optab_fn): New function.
24970 (direct_gather_load_optab_supported_p): New macro.
24971 (direct_internal_fn_optab): New function.
24972 (internal_load_fn_p): Likewise.
24973 (internal_gather_scatter_fn_p): Likewise.
24974 (internal_fn_mask_index): Likewise.
24975 (internal_gather_scatter_fn_supported_p): Likewise.
24976 * optabs-query.c (supports_at_least_one_mode_p): New function.
24977 (supports_vec_gather_load_p): Likewise.
24978 * optabs-query.h (supports_vec_gather_load_p): Declare.
24979 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24980 and memory_type field.
24981 (NUM_PATTERNS): Bump to 15.
24982 * tree-vect-data-refs.c: Include internal-fn.h.
24983 (vect_gather_scatter_fn_p): New function.
24984 (vect_describe_gather_scatter_call): Likewise.
24985 (vect_check_gather_scatter): Try using internal functions for
24986 gather loads. Recognize existing calls to a gather load function.
24987 (vect_analyze_data_refs): Consider using gather loads if
24988 supports_vec_gather_load_p.
24989 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24990 (vect_get_gather_scatter_offset_type): Likewise.
24991 (vect_convert_mask_for_vectype): Likewise.
24992 (vect_add_conversion_to_patterm): Likewise.
24993 (vect_try_gather_scatter_pattern): Likewise.
24994 (vect_recog_gather_scatter_pattern): New pattern recognizer.
24995 (vect_vect_recog_func_ptrs): Add it.
24996 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24997 internal_fn_mask_index and internal_gather_scatter_fn_p.
24998 (check_load_store_masking): Take the gather_scatter_info as an
24999 argument and handle gather loads.
25000 (vect_get_gather_scatter_ops): New function.
25001 (vectorizable_call): Check internal_load_fn_p.
25002 (vectorizable_load): Likewise. Handle gather load internal
25004 (vectorizable_store): Update call to check_load_store_masking.
25005 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
25006 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
25007 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
25008 (aarch64_gather_scale_operand_d): New predicates.
25009 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
25010 (mask_gather_load<mode>): New insns.
25012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25013 Alan Hayward <alan.hayward@arm.com>
25014 David Sherwood <david.sherwood@arm.com>
25016 * optabs.def (fold_left_plus_optab): New optab.
25017 * doc/md.texi (fold_left_plus_@var{m}): Document.
25018 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
25019 * internal-fn.c (fold_left_direct): Define.
25020 (expand_fold_left_optab_fn): Likewise.
25021 (direct_fold_left_optab_supported_p): Likewise.
25022 * fold-const-call.c (fold_const_fold_left): New function.
25023 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
25024 * tree-parloops.c (valid_reduction_p): New function.
25025 (gather_scalar_reductions): Use it.
25026 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
25027 (vect_finish_replace_stmt): Declare.
25028 * tree-vect-loop.c (fold_left_reduction_fn): New function.
25029 (needs_fold_left_reduction_p): New function, split out from...
25030 (vect_is_simple_reduction): ...here. Accept reductions that
25031 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
25032 (vect_force_simple_reduction): Also store the reduction type in
25033 the assignment's STMT_VINFO_REDUC_TYPE.
25034 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
25035 (merge_with_identity): New function.
25036 (vect_expand_fold_left): Likewise.
25037 (vectorize_fold_left_reduction): Likewise.
25038 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
25039 scalar phi in place for it. Check for target support and reject
25040 cases that would reassociate the operation. Defer the transform
25041 phase to vectorize_fold_left_reduction.
25042 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
25043 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
25044 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
25046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25048 * tree-if-conv.c (predicate_mem_writes): Remove redundant
25049 call to ifc_temp_var.
25051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25052 Alan Hayward <alan.hayward@arm.com>
25053 David Sherwood <david.sherwood@arm.com>
25055 * target.def (legitimize_address_displacement): Take the original
25056 offset as a poly_int.
25057 * targhooks.h (default_legitimize_address_displacement): Update
25059 * targhooks.c (default_legitimize_address_displacement): Likewise.
25060 * doc/tm.texi: Regenerate.
25061 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
25062 as an argument, moving assert of ad->disp == ad->disp_term to...
25063 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
25064 Try calling targetm.legitimize_address_displacement before expanding
25065 the address rather than afterwards, and adjust for the new interface.
25066 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
25067 Match the new hook interface. Handle SVE addresses.
25068 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
25069 new hook interface.
25071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25073 * Makefile.in (OBJS): Add early-remat.o.
25074 * target.def (select_early_remat_modes): New hook.
25075 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
25076 * doc/tm.texi: Regenerate.
25077 * targhooks.h (default_select_early_remat_modes): Declare.
25078 * targhooks.c (default_select_early_remat_modes): New function.
25079 * timevar.def (TV_EARLY_REMAT): New timevar.
25080 * passes.def (pass_early_remat): New pass.
25081 * tree-pass.h (make_pass_early_remat): Declare.
25082 * early-remat.c: New file.
25083 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
25085 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
25087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25088 Alan Hayward <alan.hayward@arm.com>
25089 David Sherwood <david.sherwood@arm.com>
25091 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
25092 vfm1 with a bound_epilog parameter.
25093 (vect_do_peeling): Update calls accordingly, and move the prologue
25094 call earlier in the function. Treat the base bound_epilog as 0 for
25095 fully-masked loops and retain vf - 1 for other loops. Add 1 to
25096 this base when peeling for gaps.
25097 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
25098 with fully-masked loops.
25099 (vect_estimate_min_profitable_iters): Handle the single peeled
25100 iteration in that case.
25102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25103 Alan Hayward <alan.hayward@arm.com>
25104 David Sherwood <david.sherwood@arm.com>
25106 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25107 single-element interleaving even if the size is not a power of 2.
25108 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25109 accesses for single-element interleaving if the group size is
25112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25113 Alan Hayward <alan.hayward@arm.com>
25114 David Sherwood <david.sherwood@arm.com>
25116 * doc/md.texi (fold_extract_last_@var{m}): Document.
25117 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25118 * optabs.def (fold_extract_last_optab): New optab.
25119 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25120 * internal-fn.c (fold_extract_direct): New macro.
25121 (expand_fold_extract_optab_fn): Likewise.
25122 (direct_fold_extract_optab_supported_p): Likewise.
25123 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25124 * tree-vect-loop.c (vect_model_reduction_cost): Handle
25125 EXTRACT_LAST_REDUCTION.
25126 (get_initial_def_for_reduction): Do not create an initial vector
25127 for EXTRACT_LAST_REDUCTION reductions.
25128 (vectorizable_reduction): Leave the scalar phi in place for
25129 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
25130 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
25131 epilogue code for EXTRACT_LAST_REDUCTION and defer the
25132 transform phase to vectorizable_condition.
25133 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25135 (vect_finish_stmt_generation): ...here.
25136 (vect_finish_replace_stmt): New function.
25137 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25138 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25140 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25143 Alan Hayward <alan.hayward@arm.com>
25144 David Sherwood <david.sherwood@arm.com>
25146 * doc/md.texi (extract_last_@var{m}): Document.
25147 * optabs.def (extract_last_optab): New optab.
25148 * internal-fn.def (EXTRACT_LAST): New internal function.
25149 * internal-fn.c (cond_unary_direct): New macro.
25150 (expand_cond_unary_optab_fn): Likewise.
25151 (direct_cond_unary_optab_supported_p): Likewise.
25152 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25153 loops using EXTRACT_LAST.
25154 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
25155 (extract_last_<mode>): ...this optab.
25156 (vec_extract<mode><Vel>): Update accordingly.
25158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25159 Alan Hayward <alan.hayward@arm.com>
25160 David Sherwood <david.sherwood@arm.com>
25162 * target.def (empty_mask_is_expensive): New hook.
25163 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
25164 * doc/tm.texi: Regenerate.
25165 * targhooks.h (default_empty_mask_is_expensive): Declare.
25166 * targhooks.c (default_empty_mask_is_expensive): New function.
25167 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25168 if the target says that empty masks are expensive.
25169 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25171 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25174 Alan Hayward <alan.hayward@arm.com>
25175 David Sherwood <david.sherwood@arm.com>
25177 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25178 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25179 (vect_use_loop_mask_for_alignment_p): New function.
25180 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25181 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25182 niters_skip argument. Make sure that the first niters_skip elements
25183 of the first iteration are inactive.
25184 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25185 Update call to vect_set_loop_masks_directly.
25186 (get_misalign_in_elems): New function, split out from...
25187 (vect_gen_prolog_loop_niters): ...here.
25188 (vect_update_init_of_dr): Take a code argument that specifies whether
25189 the adjustment should be added or subtracted.
25190 (vect_update_init_of_drs): Likewise.
25191 (vect_prepare_for_masked_peels): New function.
25192 (vect_do_peeling): Skip prologue peeling if we're using a mask
25193 instead. Update call to vect_update_inits_of_drs.
25194 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25196 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25197 alignment. Do not include the number of peeled iterations in
25198 the minimum threshold in that case.
25199 (vectorizable_induction): Adjust the start value down by
25200 LOOP_VINFO_MASK_SKIP_NITERS iterations.
25201 (vect_transform_loop): Call vect_prepare_for_masked_peels.
25202 Take the number of skipped iterations into account when calculating
25204 * tree-vect-stmts.c (vect_gen_while_not): New function.
25206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25207 Alan Hayward <alan.hayward@arm.com>
25208 David Sherwood <david.sherwood@arm.com>
25210 * doc/sourcebuild.texi (vect_fully_masked): Document.
25211 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
25212 default value to 0.
25213 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
25215 (vect_analyze_loop_2): ...here. Don't check the vectorization
25216 factor against the number of loop iterations if the loop is
25219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25220 Alan Hayward <alan.hayward@arm.com>
25221 David Sherwood <david.sherwood@arm.com>
25223 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
25224 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
25225 (dump_groups): Update accordingly.
25226 (iv_use::mem_type): New member variable.
25227 (address_p): New function.
25228 (record_use): Add a mem_type argument and initialize the new
25230 (record_group_use): Add a mem_type argument. Use address_p.
25231 Remove obsolete null checks of base_object. Update call to record_use.
25232 (find_interesting_uses_op): Update call to record_group_use.
25233 (find_interesting_uses_cond): Likewise.
25234 (find_interesting_uses_address): Likewise.
25235 (get_mem_type_for_internal_fn): New function.
25236 (find_address_like_use): Likewise.
25237 (find_interesting_uses_stmt): Try find_address_like_use before
25238 calling find_interesting_uses_op.
25239 (addr_offset_valid_p): Use the iv mem_type field as the type
25240 of the addressed memory.
25241 (add_autoinc_candidates): Likewise.
25242 (get_address_cost): Likewise.
25243 (split_small_address_groups_p): Use address_p.
25244 (split_address_groups): Likewise.
25245 (add_iv_candidate_for_use): Likewise.
25246 (autoinc_possible_for_pair): Likewise.
25247 (rewrite_groups): Likewise.
25248 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
25249 (determine_group_iv_cost): Update after split of USE_ADDRESS.
25250 (get_alias_ptr_type_for_ptr_address): New function.
25251 (rewrite_use_address): Rewrite address uses in calls that were
25252 identified by find_address_like_use.
25254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25255 Alan Hayward <alan.hayward@arm.com>
25256 David Sherwood <david.sherwood@arm.com>
25258 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
25260 * gimple-expr.h (is_gimple_addressable: Likewise.
25261 * gimple-expr.c (is_gimple_address): Likewise.
25262 * internal-fn.c (expand_call_mem_ref): New function.
25263 (expand_mask_load_optab_fn): Use it.
25264 (expand_mask_store_optab_fn): Likewise.
25266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25267 Alan Hayward <alan.hayward@arm.com>
25268 David Sherwood <david.sherwood@arm.com>
25270 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
25271 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
25272 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
25273 (cond_umax@var{mode}): Document.
25274 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
25275 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
25276 (cond_umin_optab, cond_umax_optab): New optabs.
25277 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
25278 (COND_IOR, COND_XOR): New internal functions.
25279 * internal-fn.h (get_conditional_internal_fn): Declare.
25280 * internal-fn.c (cond_binary_direct): New macro.
25281 (expand_cond_binary_optab_fn): Likewise.
25282 (direct_cond_binary_optab_supported_p): Likewise.
25283 (get_conditional_internal_fn): New function.
25284 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
25285 Cope with reduction statements that are vectorized as calls rather
25287 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
25288 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
25289 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
25290 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
25291 (UNSPEC_COND_EOR): New unspecs.
25292 (optab): Add mappings for them.
25293 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
25294 (sve_int_op, sve_fp_op): New int attributes.
25296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25297 Alan Hayward <alan.hayward@arm.com>
25298 David Sherwood <david.sherwood@arm.com>
25300 * optabs.def (while_ult_optab): New optab.
25301 * doc/md.texi (while_ult@var{m}@var{n}): Document.
25302 * internal-fn.def (WHILE_ULT): New internal function.
25303 * internal-fn.h (direct_internal_fn_supported_p): New override
25304 that takes two types as argument.
25305 * internal-fn.c (while_direct): New macro.
25306 (expand_while_optab_fn): New function.
25307 (convert_optab_supported_p): Likewise.
25308 (direct_while_optab_supported_p): New macro.
25309 * wide-int.h (wi::udiv_ceil): New function.
25310 * tree-vectorizer.h (rgroup_masks): New structure.
25311 (vec_loop_masks): New typedef.
25312 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
25313 and fully_masked_p.
25314 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
25315 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
25316 (vect_max_vf): New function.
25317 (slpeel_make_loop_iterate_ntimes): Delete.
25318 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
25319 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
25320 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
25321 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
25322 internal-fn.h, stor-layout.h and optabs-query.h.
25323 (vect_set_loop_mask): New function.
25324 (add_preheader_seq): Likewise.
25325 (add_header_seq): Likewise.
25326 (interleave_supported_p): Likewise.
25327 (vect_maybe_permute_loop_masks): Likewise.
25328 (vect_set_loop_masks_directly): Likewise.
25329 (vect_set_loop_condition_masked): Likewise.
25330 (vect_set_loop_condition_unmasked): New function, split out from
25331 slpeel_make_loop_iterate_ntimes.
25332 (slpeel_make_loop_iterate_ntimes): Rename to..
25333 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
25334 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
25335 (vect_do_peeling): Update call accordingly.
25336 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
25338 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25339 mask_compare_type, can_fully_mask_p and fully_masked_p.
25340 (release_vec_loop_masks): New function.
25341 (_loop_vec_info): Use it to free the loop masks.
25342 (can_produce_all_loop_masks_p): New function.
25343 (vect_get_max_nscalars_per_iter): Likewise.
25344 (vect_verify_full_masking): Likewise.
25345 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
25346 retries, and free the mask rgroups before retrying. Check loop-wide
25347 reasons for disallowing fully-masked loops. Make the final decision
25348 about whether use a fully-masked loop or not.
25349 (vect_estimate_min_profitable_iters): Do not assume that peeling
25350 for the number of iterations will be needed for fully-masked loops.
25351 (vectorizable_reduction): Disable fully-masked loops.
25352 (vectorizable_live_operation): Likewise.
25353 (vect_halve_mask_nunits): New function.
25354 (vect_double_mask_nunits): Likewise.
25355 (vect_record_loop_mask): Likewise.
25356 (vect_get_loop_mask): Likewise.
25357 (vect_transform_loop): Handle the case in which the final loop
25358 iteration might handle a partial vector. Call vect_set_loop_condition
25359 instead of slpeel_make_loop_iterate_ntimes.
25360 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
25361 (check_load_store_masking): New function.
25362 (prepare_load_store_mask): Likewise.
25363 (vectorizable_store): Handle fully-masked loops.
25364 (vectorizable_load): Likewise.
25365 (supportable_widening_operation): Use vect_halve_mask_nunits for
25367 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
25368 (vect_gen_while): New function.
25369 * config/aarch64/aarch64.md (umax<mode>3): New expander.
25370 (aarch64_uqdec<mode>): New insn.
25372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25373 Alan Hayward <alan.hayward@arm.com>
25374 David Sherwood <david.sherwood@arm.com>
25376 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
25377 (reduc_xor_scal_optab): New optabs.
25378 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
25379 (reduc_xor_scal_@var{m}): Document.
25380 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
25381 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
25382 internal functions.
25383 * fold-const-call.c (fold_const_call): Handle them.
25384 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
25385 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
25386 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
25387 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
25388 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
25389 (UNSPEC_XORV): New unspecs.
25390 (optab): Add entries for them.
25391 (BITWISEV): New int iterator.
25392 (bit_reduc_op): New int attributes.
25394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25395 Alan Hayward <alan.hayward@arm.com>
25396 David Sherwood <david.sherwood@arm.com>
25398 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
25399 * internal-fn.def (VEC_SHL_INSERT): New internal function.
25400 * optabs.def (vec_shl_insert_optab): New optab.
25401 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
25402 (duplicate_and_interleave): Likewise.
25403 * tree-vect-loop.c: Include internal-fn.h.
25404 (neutral_op_for_slp_reduction): New function, split out from
25405 get_initial_defs_for_reduction.
25406 (get_initial_def_for_reduction): Handle option 2 for variable-length
25407 vectors by loading the neutral value into a vector and then shifting
25408 the initial value into element 0.
25409 (get_initial_defs_for_reduction): Replace the code argument with
25410 the neutral value calculated by neutral_op_for_slp_reduction.
25411 Use gimple_build_vector for constant-length vectors.
25412 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
25413 but the first group_size elements have a neutral value.
25414 Use duplicate_and_interleave otherwise.
25415 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
25416 Update call to get_initial_defs_for_reduction. Handle SLP
25417 reductions for variable-length vectors by creating one vector
25418 result for each scalar result, with the elements associated
25419 with other scalar results stubbed out with the neutral value.
25420 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
25421 Require IFN_VEC_SHL_INSERT for double reductions on
25422 variable-length vectors, or SLP reductions that have
25423 a neutral value. Require can_duplicate_and_interleave_p
25424 support for variable-length unchained SLP reductions if there
25425 is no neutral value, such as for MIN/MAX reductions. Also require
25426 the number of vector elements to be a multiple of the number of
25427 SLP statements when doing variable-length unchained SLP reductions.
25428 Update call to vect_create_epilog_for_reduction.
25429 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
25430 and remove initial values.
25431 (duplicate_and_interleave): Make public.
25432 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
25433 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
25435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25436 Alan Hayward <alan.hayward@arm.com>
25437 David Sherwood <david.sherwood@arm.com>
25439 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
25440 (can_duplicate_and_interleave_p): New function.
25441 (vect_get_and_check_slp_defs): Take the vector of statements
25442 rather than just the current one. Remove excess parentheses.
25443 Restriction rejectinon of vect_constant_def and vect_external_def
25444 for variable-length vectors to boolean types, or types for which
25445 can_duplicate_and_interleave_p is false.
25446 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
25447 (duplicate_and_interleave): New function.
25448 (vect_get_constant_vectors): Use gimple_build_vector for
25449 constant-length vectors and suitable variable-length constant
25450 vectors. Use duplicate_and_interleave for other variable-length
25451 vectors. Don't defer the update when inserting new statements.
25453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25454 Alan Hayward <alan.hayward@arm.com>
25455 David Sherwood <david.sherwood@arm.com>
25457 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
25458 min_profitable_iters doesn't go negative.
25460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25461 Alan Hayward <alan.hayward@arm.com>
25462 David Sherwood <david.sherwood@arm.com>
25464 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
25465 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
25466 * optabs.def (vec_mask_load_lanes_optab): New optab.
25467 (vec_mask_store_lanes_optab): Likewise.
25468 * internal-fn.def (MASK_LOAD_LANES): New internal function.
25469 (MASK_STORE_LANES): Likewise.
25470 * internal-fn.c (mask_load_lanes_direct): New macro.
25471 (mask_store_lanes_direct): Likewise.
25472 (expand_mask_load_optab_fn): Handle masked operations.
25473 (expand_mask_load_lanes_optab_fn): New macro.
25474 (expand_mask_store_optab_fn): Handle masked operations.
25475 (expand_mask_store_lanes_optab_fn): New macro.
25476 (direct_mask_load_lanes_optab_supported_p): Likewise.
25477 (direct_mask_store_lanes_optab_supported_p): Likewise.
25478 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
25480 (vect_load_lanes_supported): Likewise.
25481 * tree-vect-data-refs.c (strip_conversion): New function.
25482 (can_group_stmts_p): Likewise.
25483 (vect_analyze_data_ref_accesses): Use it instead of checking
25484 for a pair of assignments.
25485 (vect_store_lanes_supported): Take a masked_p parameter.
25486 (vect_load_lanes_supported): Likewise.
25487 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
25488 vect_store_lanes_supported and vect_load_lanes_supported.
25489 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
25490 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
25491 parameter. Don't allow gaps for masked accesses.
25492 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
25493 and vect_load_lanes_supported.
25494 (get_load_store_type): Take a masked_p parameter and update
25495 call to get_group_load_store_type.
25496 (vectorizable_store): Update call to get_load_store_type.
25497 Handle IFN_MASK_STORE_LANES.
25498 (vectorizable_load): Update call to get_load_store_type.
25499 Handle IFN_MASK_LOAD_LANES.
25501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25502 Alan Hayward <alan.hayward@arm.com>
25503 David Sherwood <david.sherwood@arm.com>
25505 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
25507 * config/aarch64/aarch64-protos.h
25508 (aarch64_sve_struct_memory_operand_p): Declare.
25509 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
25510 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
25511 (VPRED, vpred): Handle SVE structure modes.
25512 * config/aarch64/constraints.md (Utx): New constraint.
25513 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
25514 (aarch64_sve_struct_nonimmediate_operand): New predicates.
25515 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
25516 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
25517 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
25518 structure modes. Split into pieces after RA.
25519 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
25520 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
25522 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
25523 SVE structure modes.
25524 (aarch64_classify_address): Likewise.
25525 (sizetochar): Move earlier in file.
25526 (aarch64_print_operand): Handle SVE register lists.
25527 (aarch64_array_mode): New function.
25528 (aarch64_sve_struct_memory_operand_p): Likewise.
25529 (TARGET_ARRAY_MODE): Redefine.
25531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25532 Alan Hayward <alan.hayward@arm.com>
25533 David Sherwood <david.sherwood@arm.com>
25535 * target.def (array_mode): New target hook.
25536 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
25537 * doc/tm.texi: Regenerate.
25538 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
25539 * hooks.c (hook_optmode_mode_uhwi_none): New function.
25540 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
25541 targetm.array_mode.
25542 * stor-layout.c (mode_for_array): Likewise. Support polynomial
25545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25546 Alan Hayward <alan.hayward@arm.com>
25547 David Sherwood <david.sherwood@arm.com>
25549 * fold-const.c (fold_binary_loc): Check the argument types
25550 rather than the result type when testing for a vector operation.
25552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25554 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
25555 * doc/tm.texi: Regenerate.
25557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25558 Alan Hayward <alan.hayward@arm.com>
25559 David Sherwood <david.sherwood@arm.com>
25561 * doc/invoke.texi (-msve-vector-bits=): Document new option.
25562 (sve): Document new AArch64 extension.
25563 * doc/md.texi (w): Extend the description of the AArch64
25564 constraint to include SVE vectors.
25565 (Upl, Upa): Document new AArch64 predicate constraints.
25566 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
25568 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
25569 (msve-vector-bits=): New option.
25570 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
25571 SVE when these are disabled.
25572 (sve): New extension.
25573 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
25574 modes. Adjust their number of units based on aarch64_sve_vg.
25575 (MAX_BITSIZE_MODE_ANY_MODE): Define.
25576 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
25577 aarch64_addr_query_type.
25578 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
25579 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
25580 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
25581 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
25582 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
25583 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
25584 (aarch64_simd_imm_zero_p): Delete.
25585 (aarch64_check_zero_based_sve_index_immediate): Declare.
25586 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25587 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25588 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25589 (aarch64_sve_float_mul_immediate_p): Likewise.
25590 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25591 rather than an rtx.
25592 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
25593 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
25594 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
25595 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
25596 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
25597 (aarch64_regmode_natural_size): Likewise.
25598 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
25599 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
25601 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
25602 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
25603 for VG and the SVE predicate registers.
25604 (V_ALIASES): Add a "z"-prefixed alias.
25605 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
25606 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
25607 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
25608 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
25609 (REG_CLASS_NAMES): Add entries for them.
25610 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
25611 and the predicate registers.
25612 (aarch64_sve_vg): Declare.
25613 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
25614 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
25615 (REGMODE_NATURAL_SIZE): Define.
25616 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
25618 * config/aarch64/aarch64.c: Include cfgrtl.h.
25619 (simd_immediate_info): Add a constructor for series vectors,
25620 and an associated step field.
25621 (aarch64_sve_vg): New variable.
25622 (aarch64_dbx_register_number): Handle VG and the predicate registers.
25623 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
25624 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
25625 (VEC_ANY_DATA, VEC_STRUCT): New constants.
25626 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
25627 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
25628 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
25629 (aarch64_get_mask_mode): New functions.
25630 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
25631 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
25632 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
25633 predicate modes and predicate registers. Explicitly restrict
25634 GPRs to modes of 16 bytes or smaller. Only allow FP registers
25635 to store a vector mode if it is recognized by
25636 aarch64_classify_vector_mode.
25637 (aarch64_regmode_natural_size): New function.
25638 (aarch64_hard_regno_caller_save_mode): Return the original mode
25640 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
25641 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
25642 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
25643 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
25645 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
25646 does not overlap dest if the function is frame-related. Handle
25648 (aarch64_split_add_offset): New function.
25649 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
25650 them aarch64_add_offset.
25651 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
25652 and update call to aarch64_sub_sp.
25653 (aarch64_add_cfa_expression): New function.
25654 (aarch64_expand_prologue): Pass extra temporary registers to the
25655 functions above. Handle the case in which we need to emit new
25656 DW_CFA_expressions for registers that were originally saved
25657 relative to the stack pointer, but now have to be expressed
25658 relative to the frame pointer.
25659 (aarch64_output_mi_thunk): Pass extra temporary registers to the
25661 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
25662 IP0 and IP1 values for SVE frames.
25663 (aarch64_expand_vec_series): New function.
25664 (aarch64_expand_sve_widened_duplicate): Likewise.
25665 (aarch64_expand_sve_const_vector): Likewise.
25666 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
25667 Handle SVE constants. Use emit_move_insn to move a force_const_mem
25668 into the register, rather than emitting a SET directly.
25669 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
25670 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
25671 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
25672 (offset_9bit_signed_scaled_p): New functions.
25673 (aarch64_replicate_bitmask_imm): New function.
25674 (aarch64_bitmask_imm): Use it.
25675 (aarch64_cannot_force_const_mem): Reject expressions involving
25676 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
25677 (aarch64_classify_index): Handle SVE indices, by requiring
25678 a plain register index with a scale that matches the element size.
25679 (aarch64_classify_address): Handle SVE addresses. Assert that
25680 the mode of the address is VOIDmode or an integer mode.
25681 Update call to aarch64_classify_symbol.
25682 (aarch64_classify_symbolic_expression): Update call to
25683 aarch64_classify_symbol.
25684 (aarch64_const_vec_all_in_range_p): New function.
25685 (aarch64_print_vector_float_operand): Likewise.
25686 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
25687 "vN" for FP registers with SVE modes. Handle (const ...) vectors
25688 and the FP immediates 1.0 and 0.5.
25689 (aarch64_print_address_internal): Handle SVE addresses.
25690 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
25691 (aarch64_regno_regclass): Handle predicate registers.
25692 (aarch64_secondary_reload): Handle big-endian reloads of SVE
25694 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
25695 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
25696 (aarch64_convert_sve_vector_bits): New function.
25697 (aarch64_override_options): Use it to handle -msve-vector-bits=.
25698 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25699 rather than an rtx.
25700 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
25701 Handle SVE vector and predicate modes. Accept VL-based constants
25702 that need only one temporary register, and VL offsets that require
25703 no temporary registers.
25704 (aarch64_conditional_register_usage): Mark the predicate registers
25705 as fixed if SVE isn't available.
25706 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
25707 Return true for SVE vector and predicate modes.
25708 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
25709 rather than an unsigned int. Handle SVE modes.
25710 (aarch64_preferred_simd_mode): Update call accordingly. Handle
25712 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
25714 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25715 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25716 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25717 (aarch64_sve_float_mul_immediate_p): New functions.
25718 (aarch64_sve_valid_immediate): New function.
25719 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
25720 Explicitly reject structure modes. Check for INDEX constants.
25721 Handle PTRUE and PFALSE constants.
25722 (aarch64_check_zero_based_sve_index_immediate): New function.
25723 (aarch64_simd_imm_zero_p): Delete.
25724 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
25725 vector modes. Accept constants in the range of CNT[BHWD].
25726 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
25727 ask for an Advanced SIMD mode.
25728 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
25729 (aarch64_simd_vector_alignment): Handle SVE predicates.
25730 (aarch64_vectorize_preferred_vector_alignment): New function.
25731 (aarch64_simd_vector_alignment_reachable): Use it instead of
25733 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
25734 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
25736 (MAX_VECT_LEN): Delete.
25737 (expand_vec_perm_d): Add a vec_flags field.
25738 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
25739 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
25740 (aarch64_evpc_ext): Don't apply a big-endian lane correction
25742 (aarch64_evpc_rev): Rename to...
25743 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
25744 (aarch64_evpc_rev_global): New function.
25745 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
25746 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
25748 (aarch64_evpc_sve_tbl): New function.
25749 (aarch64_expand_vec_perm_const_1): Update after rename of
25750 aarch64_evpc_rev. Handle SVE permutes too, trying
25751 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
25752 than aarch64_evpc_tbl.
25753 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
25754 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
25755 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
25756 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
25757 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
25758 (aarch64_expand_sve_vcond): New functions.
25759 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
25760 of aarch64_vector_mode_p.
25761 (aarch64_dwarf_poly_indeterminate_value): New function.
25762 (aarch64_compute_pressure_classes): Likewise.
25763 (aarch64_can_change_mode_class): Likewise.
25764 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
25765 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
25766 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
25767 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
25768 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
25769 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
25770 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
25771 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
25773 (Dn, Dl, Dr): Accept const as well as const_vector.
25774 (Dz): Likewise. Compare against CONST0_RTX.
25775 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
25776 of "vector" where appropriate.
25777 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
25778 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
25779 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
25780 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
25781 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
25782 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
25783 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
25784 (v_int_equiv): Extend to SVE modes.
25785 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
25787 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
25788 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
25789 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
25790 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
25791 (SVE_COND_FP_CMP): New int iterators.
25792 (perm_hilo): Handle the new unpack unspecs.
25793 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
25795 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
25796 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
25797 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
25798 (aarch64_equality_operator, aarch64_constant_vector_operand)
25799 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
25800 (aarch64_sve_nonimmediate_operand): Likewise.
25801 (aarch64_sve_general_operand): Likewise.
25802 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
25803 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
25804 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
25805 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
25806 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
25807 (aarch64_sve_float_arith_immediate): Likewise.
25808 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
25809 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
25810 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
25811 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
25812 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
25813 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
25814 (aarch64_sve_float_arith_operand): Likewise.
25815 (aarch64_sve_float_arith_with_sub_operand): Likewise.
25816 (aarch64_sve_float_mul_operand): Likewise.
25817 (aarch64_sve_vec_perm_operand): Likewise.
25818 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
25819 (aarch64_mov_operand): Accept const_poly_int and const_vector.
25820 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
25821 as well as const_vector.
25822 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
25823 in file. Use CONST0_RTX and CONSTM1_RTX.
25824 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
25825 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
25826 Use aarch64_simd_imm_zero.
25827 * config/aarch64/aarch64-sve.md: New file.
25828 * config/aarch64/aarch64.md: Include it.
25829 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
25830 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
25831 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
25832 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
25833 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
25834 (sve): New attribute.
25835 (enabled): Disable instructions with the sve attribute unless
25837 (movqi, movhi): Pass CONST_POLY_INT operaneds through
25838 aarch64_expand_mov_immediate.
25839 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
25840 CNT[BHSD] immediates.
25841 (movti): Split CONST_POLY_INT moves into two halves.
25842 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
25843 Split additions that need a temporary here if the destination
25844 is the stack pointer.
25845 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
25846 (*add<mode>3_poly_1): New instruction.
25847 (set_clobber_cc): New expander.
25849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25851 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
25852 parameter and use it instead of GET_MODE_SIZE (innermode). Use
25853 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
25854 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
25855 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
25856 Change innermode from fixed_mode_size to machine_mode.
25857 (simplify_subreg): Update call accordingly. Handle a constant-sized
25858 subreg of a variable-length CONST_VECTOR.
25860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25861 Alan Hayward <alan.hayward@arm.com>
25862 David Sherwood <david.sherwood@arm.com>
25864 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
25865 (add_offset_to_base): New function, split out from...
25866 (create_mem_ref): ...here. When handling a scale other than 1,
25867 check first whether the address is valid without the offset.
25868 Add it into the base if so, leaving the index and scale as-is.
25870 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25873 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
25874 fold_for_warn before checking if arg2 is INTEGER_CST.
25876 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
25878 * config/rs6000/predicates.md (load_multiple_operation): Delete.
25879 (store_multiple_operation): Delete.
25880 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
25881 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
25882 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
25883 guarded by TARGET_STRING.
25884 (rs6000_output_load_multiple): Delete.
25885 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
25886 OPTION_MASK_STRING / TARGET_STRING handling.
25887 (print_operand) <'N', 'O'>: Add comment that these are unused now.
25888 (const rs6000_opt_masks) <"string">: Change mask to 0.
25889 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
25890 (MASK_STRING): Delete.
25891 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
25893 (load_multiple): Delete.
25900 (store_multiple): Delete.
25907 (movmemsi_8reg): Delete.
25908 (corresponding unnamed define_insn): Delete.
25909 (movmemsi_6reg): Delete.
25910 (corresponding unnamed define_insn): Delete.
25911 (movmemsi_4reg): Delete.
25912 (corresponding unnamed define_insn): Delete.
25913 (movmemsi_2reg): Delete.
25914 (corresponding unnamed define_insn): Delete.
25915 (movmemsi_1reg): Delete.
25916 (corresponding unnamed define_insn): Delete.
25917 * config/rs6000/rs6000.opt (mno-string): New.
25918 (mstring): Replace by deprecation warning stub.
25919 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25921 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25923 * regrename.c (regrename_do_replace): If replacing the same
25924 reg multiple times, try to reuse last created gen_raw_REG.
25927 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25928 main to workaround a bug in GDB.
25930 2018-01-12 Tom de Vries <tom@codesourcery.com>
25933 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25935 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
25937 PR rtl-optimization/80481
25938 * ira-color.c (get_cap_member): New function.
25939 (allocnos_conflict_by_live_ranges_p): Use it.
25940 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25941 (setup_slot_coalesced_allocno_live_ranges): Ditto.
25943 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
25946 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25947 (*saddl_se_1): Ditto.
25948 (*ssubsi_1): Ditto.
25949 (*ssubl_se_1): Ditto.
25951 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25953 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25954 rather than wi::to_widest for DR_INITs.
25955 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25956 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25957 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25959 (vect_analyze_group_access_1): Note that here.
25961 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25963 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25964 polynomial type sizes.
25966 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25968 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25969 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25970 (gimple_add_tmp_var): Likewise.
25972 2018-01-12 Martin Liska <mliska@suse.cz>
25974 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25975 (gimple_alloc_sizes): Likewise.
25976 (dump_gimple_statistics): Use PRIu64 in printf format.
25977 * gimple.h: Change uint64_t to int.
25979 2018-01-12 Martin Liska <mliska@suse.cz>
25981 * tree-core.h: Use uint64_t instead of int.
25982 * tree.c (tree_node_counts): Likewise.
25983 (tree_node_sizes): Likewise.
25984 (dump_tree_statistics): Use PRIu64 in printf format.
25986 2018-01-12 Martin Liska <mliska@suse.cz>
25988 * Makefile.in: As qsort_chk is implemented in vec.c, add
25989 vec.o to linkage of gencfn-macros.
25990 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25991 passing the info to record_node_allocation_statistics.
25992 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25994 * ggc-common.c (struct ggc_usage): Add operator== and use
25995 it in operator< and compare function.
25996 * mem-stats.h (struct mem_usage): Likewise.
25997 * vec.c (struct vec_usage): Remove operator< and compare
25998 function. Can be simply inherited.
26000 2018-01-12 Martin Jambor <mjambor@suse.cz>
26003 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
26004 * tree-ssa-math-opts.c: Include domwalk.h.
26005 (convert_mult_to_fma_1): New function.
26006 (fma_transformation_info): New type.
26007 (fma_deferring_state): Likewise.
26008 (cancel_fma_deferring): New function.
26009 (result_of_phi): Likewise.
26010 (last_fma_candidate_feeds_initial_phi): Likewise.
26011 (convert_mult_to_fma): Added deferring logic, split actual
26012 transformation to convert_mult_to_fma_1.
26013 (math_opts_dom_walker): New type.
26014 (math_opts_dom_walker::after_dom_children): New method, body moved
26015 here from pass_optimize_widening_mul::execute, added deferring logic
26017 (pass_optimize_widening_mul::execute): Moved most of code to
26018 math_opts_dom_walker::after_dom_children.
26019 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
26020 * config/i386/i386.c (ix86_option_override_internal): Added
26021 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
26023 2018-01-12 Richard Biener <rguenther@suse.de>
26026 * dwarf2out.c (gen_variable_die): Do not reset old_die for
26027 inline instance vars.
26029 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
26032 * config/rx/rx.c (rx_is_restricted_memory_address):
26033 Handle SUBREG case.
26035 2018-01-12 Richard Biener <rguenther@suse.de>
26037 PR tree-optimization/80846
26038 * target.def (split_reduction): New target hook.
26039 * targhooks.c (default_split_reduction): New function.
26040 * targhooks.h (default_split_reduction): Declare.
26041 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
26042 target requests first reduce vectors by combining low and high
26044 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
26045 (get_vectype_for_scalar_type_and_size): Export.
26046 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
26047 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
26048 * doc/tm.texi: Regenerate.
26049 * config/i386/i386.c (ix86_split_reduction): Implement
26050 TARGET_VECTORIZE_SPLIT_REDUCTION.
26052 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
26055 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
26056 in PIC mode except for TARGET_VXWORKS_RTP.
26057 * config/sparc/sparc.c: Include cfgrtl.h.
26058 (TARGET_INIT_PIC_REG): Define.
26059 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
26060 (sparc_pic_register_p): New predicate.
26061 (sparc_legitimate_address_p): Use it.
26062 (sparc_legitimize_pic_address): Likewise.
26063 (sparc_delegitimize_address): Likewise.
26064 (sparc_mode_dependent_address_p): Likewise.
26065 (gen_load_pcrel_sym): Remove 4th parameter.
26066 (load_got_register): Adjust call to above. Remove obsolete stuff.
26067 (sparc_expand_prologue): Do not call load_got_register here.
26068 (sparc_flat_expand_prologue): Likewise.
26069 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
26070 (sparc_use_pseudo_pic_reg): New function.
26071 (sparc_init_pic_reg): Likewise.
26072 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26073 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
26075 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
26077 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
26078 Add item for branch_cost.
26080 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
26082 PR rtl-optimization/83565
26083 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
26084 not extend the result to a larger mode for rotate operations.
26085 (num_sign_bit_copies1): Likewise.
26087 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26090 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
26092 Use values-Xc.o for -pedantic.
26093 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
26095 2018-01-12 Martin Liska <mliska@suse.cz>
26098 * ipa-devirt.c (final_warning_record::grow_type_warnings):
26100 (possible_polymorphic_call_targets): Use it.
26101 (ipa_devirt): Likewise.
26103 2018-01-12 Martin Liska <mliska@suse.cz>
26105 * profile-count.h (enum profile_quality): Use 0 as invalid
26106 enum value of profile_quality.
26108 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
26110 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26111 -mext-string options.
26113 2018-01-12 Richard Biener <rguenther@suse.de>
26115 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26116 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26117 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26119 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26121 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
26123 * configure.ac (--with-long-double-format): Add support for the
26124 configuration option to change the default long double format on
26126 * config.gcc (powerpc*-linux*-*): Likewise.
26127 * configure: Regenerate.
26128 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26129 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26130 used without modification.
26132 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26134 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26135 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26136 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26137 MISC_BUILTIN_SPEC_BARRIER.
26138 (rs6000_init_builtins): Likewise.
26139 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26141 (speculation_barrier): New define_insn.
26142 * doc/extend.texi: Document __builtin_speculation_barrier.
26144 2018-01-11 Jakub Jelinek <jakub@redhat.com>
26147 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26148 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26149 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26151 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
26152 integral modes instead of "ss" and "sd".
26153 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
26154 vectors with 32-bit and 64-bit elements.
26155 (vecdupssescalarmodesuffix): New mode attribute.
26156 (vec_dup<mode>): Use it.
26158 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
26161 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
26162 frame if argument is passed on stack.
26164 2018-01-11 Jakub Jelinek <jakub@redhat.com>
26167 * ree.c (combine_reaching_defs): Optimize also
26168 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26169 reg2=any_extend(exp); reg1=reg2;, formatting fix.
26171 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
26173 PR middle-end/83189
26174 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26176 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
26178 PR middle-end/83718
26179 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26180 after they are computed.
26182 2018-01-11 Bin Cheng <bin.cheng@arm.com>
26184 PR tree-optimization/83695
26185 * gimple-loop-linterchange.cc
26186 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26187 reset cached scev information after interchange.
26188 (pass_linterchange::execute): Remove call to scev_reset_htab.
26190 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26192 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26193 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26194 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26195 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26196 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26197 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26198 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26199 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26200 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26201 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26202 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26203 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26204 (V_lane_reg): Likewise.
26205 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26207 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26208 (vfmal_lane_low<mode>_intrinsic,
26209 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26210 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
26211 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
26212 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
26213 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
26214 vfmsl_lane_high<mode>_intrinsic): New define_insns.
26216 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26218 * config/arm/arm-cpus.in (fp16fml): New feature.
26219 (ALL_SIMD): Add fp16fml.
26220 (armv8.2-a): Add fp16fml as an option.
26221 (armv8.3-a): Likewise.
26222 (armv8.4-a): Add fp16fml as part of fp16.
26223 * config/arm/arm.h (TARGET_FP16FML): Define.
26224 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
26226 * config/arm/arm-modes.def (V2HF): Define.
26227 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
26228 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
26229 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
26230 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
26231 vfmsl_low, vfmsl_high): New set of builtins.
26232 * config/arm/iterators.md (PLUSMINUS): New code iterator.
26233 (vfml_op): New code attribute.
26234 (VFMLHALVES): New int iterator.
26235 (VFML, VFMLSEL): New mode attributes.
26236 (V_reg): Define mapping for V2HF.
26237 (V_hi, V_lo): New mode attributes.
26238 (VF_constraint): Likewise.
26239 (vfml_half, vfml_half_selector): New int attributes.
26240 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
26242 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
26243 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
26245 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
26246 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
26247 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
26248 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
26250 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
26251 Document new effective target and option set.
26253 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26255 * config/arm/arm-cpus.in (armv8_4): New feature.
26256 (ARMv8_4a): New fgroup.
26257 (armv8.4-a): New arch.
26258 * config/arm/arm-tables.opt: Regenerate.
26259 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
26260 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
26261 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
26262 Add matching rules for -march=armv8.4-a and extensions.
26263 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
26265 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
26268 * config/rx/rx.md (BW): New mode attribute.
26269 (sync_lock_test_and_setsi): Add mode suffix to insn output.
26271 2018-01-11 Richard Biener <rguenther@suse.de>
26273 PR tree-optimization/83435
26274 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
26275 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
26276 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
26278 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26279 Alan Hayward <alan.hayward@arm.com>
26280 David Sherwood <david.sherwood@arm.com>
26282 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
26284 (aarch64_classify_address): Initialize it. Track polynomial offsets.
26285 (aarch64_print_address_internal): Use it to check for a zero offset.
26287 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26288 Alan Hayward <alan.hayward@arm.com>
26289 David Sherwood <david.sherwood@arm.com>
26291 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
26292 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
26293 Return a poly_int64 rather than a HOST_WIDE_INT.
26294 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
26295 rather than a HOST_WIDE_INT.
26296 * config/aarch64/aarch64.h (aarch64_frame): Protect with
26297 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
26298 hard_fp_offset, frame_size, initial_adjust, callee_offset and
26299 final_offset from HOST_WIDE_INT to poly_int64.
26300 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
26301 to_constant when getting the number of units in an Advanced SIMD
26303 (aarch64_builtin_vectorized_function): Check for a constant number
26305 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
26307 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
26308 attribute instead of GET_MODE_NUNITS.
26309 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
26310 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
26311 GET_MODE_SIZE for fixed-size registers.
26312 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
26313 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
26314 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
26315 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
26316 (aarch64_print_operand, aarch64_print_address_internal)
26317 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
26318 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
26319 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
26320 Handle polynomial GET_MODE_SIZE.
26321 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
26322 wider than SImode without modification.
26323 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
26324 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
26325 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
26326 passing and returning SVE modes.
26327 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
26328 rather than GEN_INT.
26329 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
26330 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
26331 (aarch64_allocate_and_probe_stack_space): Likewise.
26332 (aarch64_layout_frame): Cope with polynomial offsets.
26333 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
26334 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
26335 polynomial offsets.
26336 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
26337 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
26338 poly_int64 rather than a HOST_WIDE_INT.
26339 (aarch64_get_separate_components, aarch64_process_components)
26340 (aarch64_expand_prologue, aarch64_expand_epilogue)
26341 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
26342 (aarch64_anchor_offset): New function, split out from...
26343 (aarch64_legitimize_address): ...here.
26344 (aarch64_builtin_vectorization_cost): Handle polynomial
26345 TYPE_VECTOR_SUBPARTS.
26346 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
26348 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
26349 number of elements from the PARALLEL rather than the mode.
26350 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
26351 rather than GET_MODE_BITSIZE.
26352 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
26353 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
26354 (aarch64_expand_vec_perm_const_1): Handle polynomial
26355 d->perm.length () and d->perm elements.
26356 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
26357 Apply to_constant to d->perm elements.
26358 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
26359 polynomial CONST_VECTOR_NUNITS.
26360 (aarch64_move_pointer): Take amount as a poly_int64 rather
26362 (aarch64_progress_pointer): Avoid temporary variable.
26363 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
26364 the mode attribute instead of GET_MODE.
26366 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26367 Alan Hayward <alan.hayward@arm.com>
26368 David Sherwood <david.sherwood@arm.com>
26370 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
26371 x exists before using it.
26372 (aarch64_add_constant_internal): Rename to...
26373 (aarch64_add_offset_1): ...this. Replace regnum with separate
26374 src and dest rtxes. Handle the case in which they're different,
26375 including when the offset is zero. Replace scratchreg with an rtx.
26376 Use 2 additions if there is no spare register into which we can
26377 move a 16-bit constant.
26378 (aarch64_add_constant): Delete.
26379 (aarch64_add_offset): Replace reg with separate src and dest
26380 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
26381 Use aarch64_add_offset_1.
26382 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
26383 an rtx rather than an int. Take the delta as a poly_int64
26384 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
26385 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
26386 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
26387 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
26388 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
26389 and aarch64_add_sp.
26390 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
26391 aarch64_add_constant.
26393 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26395 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
26396 Use scalar_float_mode.
26398 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26400 * config/aarch64/aarch64-simd.md
26401 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
26402 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
26403 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
26404 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
26405 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
26406 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
26407 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
26408 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
26409 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
26410 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
26412 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26415 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
26416 targ_options->x_arm_arch_string is non NULL.
26418 2018-01-11 Tamar Christina <tamar.christina@arm.com>
26420 * config/aarch64/aarch64.h
26421 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
26423 2018-01-11 Sudakshina Das <sudi.das@arm.com>
26426 * expmed.c (emit_store_flag_force): Swap if const op0
26427 and change VOIDmode to mode of op0.
26429 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26431 PR rtl-optimization/83761
26432 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
26433 than bytes to mode_for_size.
26435 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
26437 PR middle-end/83189
26438 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
26439 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
26442 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
26444 PR middle-end/83575
26445 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
26446 when in layout mode.
26447 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
26448 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
26451 2018-01-10 Michael Collison <michael.collison@arm.com>
26453 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
26454 * config/aarch64/aarch64-option-extension.def: Add
26455 AARCH64_OPT_EXTENSION of 'fp16fml'.
26456 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26457 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
26458 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
26459 * config/aarch64/constraints.md (Ui7): New constraint.
26460 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
26461 (VFMLA_SEL_W): Ditto.
26464 (VFMLA16_LOW): New int iterator.
26465 (VFMLA16_HIGH): Ditto.
26466 (UNSPEC_FMLAL): New unspec.
26467 (UNSPEC_FMLSL): Ditto.
26468 (UNSPEC_FMLAL2): Ditto.
26469 (UNSPEC_FMLSL2): Ditto.
26470 (f16mac): New code attribute.
26471 * config/aarch64/aarch64-simd-builtins.def
26472 (aarch64_fmlal_lowv2sf): Ditto.
26473 (aarch64_fmlsl_lowv2sf): Ditto.
26474 (aarch64_fmlalq_lowv4sf): Ditto.
26475 (aarch64_fmlslq_lowv4sf): Ditto.
26476 (aarch64_fmlal_highv2sf): Ditto.
26477 (aarch64_fmlsl_highv2sf): Ditto.
26478 (aarch64_fmlalq_highv4sf): Ditto.
26479 (aarch64_fmlslq_highv4sf): Ditto.
26480 (aarch64_fmlal_lane_lowv2sf): Ditto.
26481 (aarch64_fmlsl_lane_lowv2sf): Ditto.
26482 (aarch64_fmlal_laneq_lowv2sf): Ditto.
26483 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
26484 (aarch64_fmlalq_lane_lowv4sf): Ditto.
26485 (aarch64_fmlsl_lane_lowv4sf): Ditto.
26486 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
26487 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
26488 (aarch64_fmlal_lane_highv2sf): Ditto.
26489 (aarch64_fmlsl_lane_highv2sf): Ditto.
26490 (aarch64_fmlal_laneq_highv2sf): Ditto.
26491 (aarch64_fmlsl_laneq_highv2sf): Ditto.
26492 (aarch64_fmlalq_lane_highv4sf): Ditto.
26493 (aarch64_fmlsl_lane_highv4sf): Ditto.
26494 (aarch64_fmlalq_laneq_highv4sf): Ditto.
26495 (aarch64_fmlsl_laneq_highv4sf): Ditto.
26496 * config/aarch64/aarch64-simd.md:
26497 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
26498 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26499 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
26500 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26501 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
26502 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
26503 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
26504 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
26505 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
26506 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
26507 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
26508 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
26509 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
26510 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
26511 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
26512 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
26513 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
26514 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
26515 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
26516 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
26517 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
26518 (vfmlsl_low_u32): Ditto.
26519 (vfmlalq_low_u32): Ditto.
26520 (vfmlslq_low_u32): Ditto.
26521 (vfmlal_high_u32): Ditto.
26522 (vfmlsl_high_u32): Ditto.
26523 (vfmlalq_high_u32): Ditto.
26524 (vfmlslq_high_u32): Ditto.
26525 (vfmlal_lane_low_u32): Ditto.
26526 (vfmlsl_lane_low_u32): Ditto.
26527 (vfmlal_laneq_low_u32): Ditto.
26528 (vfmlsl_laneq_low_u32): Ditto.
26529 (vfmlalq_lane_low_u32): Ditto.
26530 (vfmlslq_lane_low_u32): Ditto.
26531 (vfmlalq_laneq_low_u32): Ditto.
26532 (vfmlslq_laneq_low_u32): Ditto.
26533 (vfmlal_lane_high_u32): Ditto.
26534 (vfmlsl_lane_high_u32): Ditto.
26535 (vfmlal_laneq_high_u32): Ditto.
26536 (vfmlsl_laneq_high_u32): Ditto.
26537 (vfmlalq_lane_high_u32): Ditto.
26538 (vfmlslq_lane_high_u32): Ditto.
26539 (vfmlalq_laneq_high_u32): Ditto.
26540 (vfmlslq_laneq_high_u32): Ditto.
26541 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
26542 (AARCH64_FL_FOR_ARCH8_4): New.
26543 (AARCH64_ISA_F16FML): New ISA flag.
26544 (TARGET_F16FML): New feature flag for fp16fml.
26545 (doc/invoke.texi): Document new fp16fml option.
26547 2018-01-10 Michael Collison <michael.collison@arm.com>
26549 * config/aarch64/aarch64-builtins.c:
26550 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
26551 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26552 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
26553 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
26554 (AARCH64_ISA_SHA3): New ISA flag.
26555 (TARGET_SHA3): New feature flag for sha3.
26556 * config/aarch64/iterators.md (sha512_op): New int attribute.
26557 (CRYPTO_SHA512): New int iterator.
26558 (UNSPEC_SHA512H): New unspec.
26559 (UNSPEC_SHA512H2): Ditto.
26560 (UNSPEC_SHA512SU0): Ditto.
26561 (UNSPEC_SHA512SU1): Ditto.
26562 * config/aarch64/aarch64-simd-builtins.def
26563 (aarch64_crypto_sha512hqv2di): New builtin.
26564 (aarch64_crypto_sha512h2qv2di): Ditto.
26565 (aarch64_crypto_sha512su0qv2di): Ditto.
26566 (aarch64_crypto_sha512su1qv2di): Ditto.
26567 (aarch64_eor3qv8hi): Ditto.
26568 (aarch64_rax1qv2di): Ditto.
26569 (aarch64_xarqv2di): Ditto.
26570 (aarch64_bcaxqv8hi): Ditto.
26571 * config/aarch64/aarch64-simd.md:
26572 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
26573 (aarch64_crypto_sha512su0qv2di): Ditto.
26574 (aarch64_crypto_sha512su1qv2di): Ditto.
26575 (aarch64_eor3qv8hi): Ditto.
26576 (aarch64_rax1qv2di): Ditto.
26577 (aarch64_xarqv2di): Ditto.
26578 (aarch64_bcaxqv8hi): Ditto.
26579 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
26580 (vsha512h2q_u64): Ditto.
26581 (vsha512su0q_u64): Ditto.
26582 (vsha512su1q_u64): Ditto.
26583 (veor3q_u16): Ditto.
26584 (vrax1q_u64): Ditto.
26585 (vxarq_u64): Ditto.
26586 (vbcaxq_u16): Ditto.
26587 * config/arm/types.md (crypto_sha512): New type attribute.
26588 (crypto_sha3): Ditto.
26589 (doc/invoke.texi): Document new sha3 option.
26591 2018-01-10 Michael Collison <michael.collison@arm.com>
26593 * config/aarch64/aarch64-builtins.c:
26594 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
26595 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26596 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
26597 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
26598 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
26599 (AARCH64_ISA_SM4): New ISA flag.
26600 (TARGET_SM4): New feature flag for sm4.
26601 * config/aarch64/aarch64-simd-builtins.def
26602 (aarch64_sm3ss1qv4si): Ditto.
26603 (aarch64_sm3tt1aq4si): Ditto.
26604 (aarch64_sm3tt1bq4si): Ditto.
26605 (aarch64_sm3tt2aq4si): Ditto.
26606 (aarch64_sm3tt2bq4si): Ditto.
26607 (aarch64_sm3partw1qv4si): Ditto.
26608 (aarch64_sm3partw2qv4si): Ditto.
26609 (aarch64_sm4eqv4si): Ditto.
26610 (aarch64_sm4ekeyqv4si): Ditto.
26611 * config/aarch64/aarch64-simd.md:
26612 (aarch64_sm3ss1qv4si): Ditto.
26613 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
26614 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
26615 (aarch64_sm4eqv4si): Ditto.
26616 (aarch64_sm4ekeyqv4si): Ditto.
26617 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
26618 (sm3part_op): Ditto.
26619 (CRYPTO_SM3TT): Ditto.
26620 (CRYPTO_SM3PART): Ditto.
26621 (UNSPEC_SM3SS1): New unspec.
26622 (UNSPEC_SM3TT1A): Ditto.
26623 (UNSPEC_SM3TT1B): Ditto.
26624 (UNSPEC_SM3TT2A): Ditto.
26625 (UNSPEC_SM3TT2B): Ditto.
26626 (UNSPEC_SM3PARTW1): Ditto.
26627 (UNSPEC_SM3PARTW2): Ditto.
26628 (UNSPEC_SM4E): Ditto.
26629 (UNSPEC_SM4EKEY): Ditto.
26630 * config/aarch64/constraints.md (Ui2): New constraint.
26631 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
26632 * config/arm/types.md (crypto_sm3): New type attribute.
26633 (crypto_sm4): Ditto.
26634 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
26635 (vsm3tt1aq_u32): Ditto.
26636 (vsm3tt1bq_u32): Ditto.
26637 (vsm3tt2aq_u32): Ditto.
26638 (vsm3tt2bq_u32): Ditto.
26639 (vsm3partw1q_u32): Ditto.
26640 (vsm3partw2q_u32): Ditto.
26641 (vsm4eq_u32): Ditto.
26642 (vsm4ekeyq_u32): Ditto.
26643 (doc/invoke.texi): Document new sm4 option.
26645 2018-01-10 Michael Collison <michael.collison@arm.com>
26647 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
26648 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
26649 (AARCH64_FL_FOR_ARCH8_4): New.
26650 (AARCH64_FL_V8_4): New flag.
26651 (doc/invoke.texi): Document new armv8.4-a option.
26653 2018-01-10 Michael Collison <michael.collison@arm.com>
26655 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26656 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
26657 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
26658 * config/aarch64/aarch64-option-extension.def: Add
26659 AARCH64_OPT_EXTENSION of 'sha2'.
26660 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
26661 (crypto): Disable sha2 and aes if crypto disabled.
26662 (crypto): Enable aes and sha2 if enabled.
26663 (simd): Disable sha2 and aes if simd disabled.
26664 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
26666 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
26667 (TARGET_SHA2): New feature flag for sha2.
26668 (TARGET_AES): New feature flag for aes.
26669 * config/aarch64/aarch64-simd.md:
26670 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
26671 conditional on TARGET_AES.
26672 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
26673 (aarch64_crypto_sha1hsi): Make pattern conditional
26675 (aarch64_crypto_sha1hv4si): Ditto.
26676 (aarch64_be_crypto_sha1hv4si): Ditto.
26677 (aarch64_crypto_sha1su1v4si): Ditto.
26678 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
26679 (aarch64_crypto_sha1su0v4si): Ditto.
26680 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
26681 (aarch64_crypto_sha256su0v4si): Ditto.
26682 (aarch64_crypto_sha256su1v4si): Ditto.
26683 (doc/invoke.texi): Document new aes and sha2 options.
26685 2018-01-10 Martin Sebor <msebor@redhat.com>
26687 PR tree-optimization/83781
26688 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
26691 2018-01-11 Martin Sebor <msebor@gmail.com>
26692 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26694 PR tree-optimization/83501
26695 PR tree-optimization/81703
26697 * tree-ssa-strlen.c (get_string_cst): Rename...
26698 (get_string_len): ...to this. Handle global constants.
26699 (handle_char_store): Adjust.
26701 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
26702 Jim Wilson <jimw@sifive.com>
26704 * config/riscv/riscv-protos.h (riscv_output_return): New.
26705 * config/riscv/riscv.c (struct machine_function): New naked_p field.
26706 (riscv_attribute_table, riscv_output_return),
26707 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
26708 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
26709 (riscv_compute_frame_info): Only compute frame->mask if not a naked
26711 (riscv_expand_prologue): Add early return for naked function.
26712 (riscv_expand_epilogue): Likewise.
26713 (riscv_function_ok_for_sibcall): Return false for naked function.
26714 (riscv_set_current_function): New.
26715 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
26716 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
26717 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
26718 * doc/extend.texi (RISC-V Function Attributes): New.
26720 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
26722 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
26723 check for 128-bit long double before checking TCmode.
26724 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
26725 128-bit long doubles before checking TFmode or TCmode.
26726 (FLOAT128_IBM_P): Likewise.
26728 2018-01-10 Martin Sebor <msebor@redhat.com>
26730 PR tree-optimization/83671
26731 * builtins.c (c_strlen): Unconditionally return zero for the empty
26733 Use -Warray-bounds for warnings.
26734 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
26735 for non-constant array indices with COMPONENT_REF, arrays of
26736 arrays, and pointers to arrays.
26737 (gimple_fold_builtin_strlen): Determine and set length range for
26738 non-constant character arrays.
26740 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
26742 PR middle-end/81897
26743 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
26746 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
26748 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
26750 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
26753 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
26754 VECTOR_MEM_ALTIVEC_OR_VSX_P.
26755 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
26756 indexed_or_indirect_operand predicate.
26757 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
26758 (*vsx_le_perm_load_v8hi): Likewise.
26759 (*vsx_le_perm_load_v16qi): Likewise.
26760 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
26761 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
26762 (*vsx_le_perm_store_v8hi): Likewise.
26763 (*vsx_le_perm_store_v16qi): Likewise.
26764 (eight unnamed splitters): Likewise.
26766 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
26768 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
26769 * config/rs6000/emmintrin.h: Likewise.
26770 * config/rs6000/mmintrin.h: Likewise.
26771 * config/rs6000/xmmintrin.h: Likewise.
26773 2018-01-10 David Malcolm <dmalcolm@redhat.com>
26776 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
26778 * tree.c (tree_nop_conversion): Return true for location wrapper
26780 (maybe_wrap_with_location): New function.
26781 (selftest::check_strip_nops): New function.
26782 (selftest::test_location_wrappers): New function.
26783 (selftest::tree_c_tests): Call it.
26784 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
26785 (maybe_wrap_with_location): New decl.
26786 (EXPR_LOCATION_WRAPPER_P): New macro.
26787 (location_wrapper_p): New inline function.
26788 (tree_strip_any_location_wrapper): New inline function.
26790 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
26793 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
26794 stack_realign_offset for the largest alignment of stack slot
26796 (ix86_find_max_used_stack_alignment): New function.
26797 (ix86_finalize_stack_frame_flags): Use it. Set
26798 max_used_stack_alignment if we don't realign stack.
26799 * config/i386/i386.h (machine_function): Add
26800 max_used_stack_alignment.
26802 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
26804 * config/arm/arm.opt (-mbranch-cost): New option.
26805 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
26808 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
26811 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
26812 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
26814 2018-01-10 Richard Biener <rguenther@suse.de>
26817 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
26818 early out so it also covers the case where we have a non-NULL
26821 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
26823 PR tree-optimization/83753
26824 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
26825 for non-strided grouped accesses if the number of elements is 1.
26827 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
26830 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
26831 * i386.h (TARGET_USE_GATHER): Define.
26832 * x86-tune.def (X86_TUNE_USE_GATHER): New.
26834 2018-01-10 Martin Liska <mliska@suse.cz>
26837 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
26838 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
26840 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
26841 CLEANUP_NO_PARTITIONING is not set.
26843 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
26845 * doc/rtl.texi: Remove documentation of (const ...) wrappers
26846 for vectors, as a partial revert of r254296.
26847 * rtl.h (const_vec_p): Delete.
26848 (const_vec_duplicate_p): Don't test for vector CONSTs.
26849 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
26850 * expmed.c (make_tree): Likewise.
26853 * common.md (E, F): Use CONSTANT_P instead of checking for
26855 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
26856 checking for CONST_VECTOR.
26858 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26860 PR middle-end/83575
26861 * predict.c (force_edge_cold): Handle in more sane way edges
26862 with no prediction.
26864 2018-01-09 Carl Love <cel@us.ibm.com>
26866 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
26868 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
26869 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
26870 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
26871 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
26872 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
26873 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
26874 * config/rs6000/rs6000-protos.h: Add extern defition for
26875 rs6000_generate_float2_double_code.
26876 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
26878 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
26879 (float2_v2df): Add define_expand.
26881 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
26884 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
26885 op_mode in the force_to_mode call.
26887 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26889 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
26890 instead of checking each element individually.
26891 (aarch64_evpc_uzp): Likewise.
26892 (aarch64_evpc_zip): Likewise.
26893 (aarch64_evpc_ext): Likewise.
26894 (aarch64_evpc_rev): Likewise.
26895 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
26896 instead of checking each element individually. Return true without
26898 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
26899 whether all selected elements come from the same input, instead of
26900 checking each element individually. Remove calls to gen_rtx_REG,
26901 start_sequence and end_sequence and instead assert that no rtl is
26904 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26906 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
26907 order of HIGH and CONST checks.
26909 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26911 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
26912 if the destination isn't an SSA_NAME.
26914 2018-01-09 Richard Biener <rguenther@suse.de>
26916 PR tree-optimization/83668
26917 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26919 (canonicalize_loop_form): ... here, renamed from ...
26920 (canonicalize_loop_closed_ssa_form): ... this and amended to
26921 swap successor edges for loop exit blocks to make us use
26922 the RPO order we need for initial schedule generation.
26924 2018-01-09 Joseph Myers <joseph@codesourcery.com>
26926 PR tree-optimization/64811
26927 * match.pd: When optimizing comparisons with Inf, avoid
26928 introducing or losing exceptions from comparisons with NaN.
26930 2018-01-09 Martin Liska <mliska@suse.cz>
26933 * asan.c (shadow_mem_size): Add gcc_assert.
26935 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
26937 Don't save registers in main().
26940 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26941 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26942 * config/avr/avr.c (avr_set_current_function): Don't error if
26943 naked, OS_task or OS_main are specified at the same time.
26944 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26946 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26948 * common/config/avr/avr-common.c (avr_option_optimization_table):
26949 Switch on -mmain-is-OS_task for optimizing compilations.
26951 2018-01-09 Richard Biener <rguenther@suse.de>
26953 PR tree-optimization/83572
26954 * graphite.c: Include cfganal.h.
26955 (graphite_transform_loops): Connect infinite loops to exit
26956 and remove fake edges at the end.
26958 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26960 * ipa-inline.c (edge_badness): Revert accidental checkin.
26962 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26965 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26966 symbols; not inline clones.
26968 2018-01-09 Jakub Jelinek <jakub@redhat.com>
26971 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26972 hard registers. Formatting fixes.
26974 PR preprocessor/83722
26975 * gcc.c (try_generate_repro): Pass
26976 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26977 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26980 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
26981 Kito Cheng <kito.cheng@gmail.com>
26983 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26984 (riscv_leaf_function_p): Delete.
26985 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26987 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26989 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26991 (do_ifelse): New function.
26992 (do_isel): New function.
26993 (do_sub3): New function.
26994 (do_add3): New function.
26995 (do_load_mask_compare): New function.
26996 (do_overlap_load_compare): New function.
26997 (expand_compare_loop): New function.
26998 (expand_block_compare): Call expand_compare_loop() when appropriate.
26999 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
27000 option description.
27001 (-mblock-compare-inline-loop-limit): New option.
27003 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27006 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
27007 Reverse order of second and third operands in first alternative.
27008 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
27009 of first and second elements in UNSPEC_VPERMR vector.
27010 (altivec_expand_vec_perm_le): Likewise.
27012 2018-01-08 Jeff Law <law@redhat.com>
27014 PR rtl-optimizatin/81308
27015 * tree-switch-conversion.c (cfg_altered): New file scoped static.
27016 (process_switch): If group_case_labels makes a change, then set
27018 (pass_convert_switch::execute): If a switch is converted, then
27019 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
27021 PR rtl-optimization/81308
27022 * recog.c (split_all_insns): Conditionally cleanup the CFG after
27025 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
27027 PR target/83663 - Revert r255946
27028 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
27029 generation for cases where splatting a value is not useful.
27030 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
27031 across a vec_duplicate and a paradoxical subreg forming a vector
27032 mode to a vec_concat.
27034 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27036 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
27037 -march=armv8.3-a variants.
27038 * config/arm/t-multilib: Likewise.
27039 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
27041 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27043 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
27045 (cceq_ior_compare_complement): Give it a name so I can use it, and
27046 change boolean_or_operator predicate to boolean_operator so it can
27047 be used to generate a crand.
27048 (eqne): New code iterator.
27049 (bd/bd_neg): New code_attrs.
27050 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
27051 a single define_insn.
27052 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
27053 decrement (bdnzt/bdnzf/bdzt/bdzf).
27054 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
27055 with the new names of the branch decrement patterns, and added the
27056 names of the branch decrement conditional patterns.
27058 2018-01-08 Richard Biener <rguenther@suse.de>
27060 PR tree-optimization/83563
27061 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
27064 2018-01-08 Richard Biener <rguenther@suse.de>
27066 PR middle-end/83713
27067 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
27069 2018-01-08 Richard Biener <rguenther@suse.de>
27071 PR tree-optimization/83685
27072 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
27073 references to abnormals.
27075 2018-01-08 Richard Biener <rguenther@suse.de>
27078 * dwarf2out.c (output_indirect_strings): Handle empty
27079 skeleton_debug_str_hash.
27080 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
27082 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
27084 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
27085 (emit_store_direct): Likewise.
27086 (arc_trampoline_adjust_address): Likewise.
27087 (arc_asm_trampoline_template): New function.
27088 (arc_initialize_trampoline): Use asm_trampoline_template.
27089 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
27090 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
27091 * config/arc/arc.md (flush_icache): Delete pattern.
27093 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
27095 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
27096 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
27099 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
27102 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
27103 by not USED_FOR_TARGET.
27104 (make_pass_resolve_sw_modes): Likewise.
27106 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
27108 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27111 2018-01-08 Richard Biener <rguenther@suse.de>
27113 PR middle-end/83580
27114 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27116 2018-01-08 Richard Biener <rguenther@suse.de>
27118 PR middle-end/83517
27119 * match.pd ((t * 2) / 2) -> t): Add missing :c.
27121 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
27123 PR middle-end/81897
27124 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27125 basic blocks with a small number of successors.
27126 (convert_control_dep_chain_into_preds): Improve handling of
27128 (dump_predicates): Split apart into...
27129 (dump_pred_chain): ...here...
27130 (dump_pred_info): ...and here.
27131 (can_one_predicate_be_invalidated_p): Add debugging printfs.
27132 (can_chain_union_be_invalidated_p): Improve check for invalidation
27134 (uninit_uses_cannot_happen): Avoid unnecessary if
27135 convert_control_dep_chain_into_preds yielded nothing.
27137 2018-01-06 Martin Sebor <msebor@redhat.com>
27139 PR tree-optimization/83640
27140 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27141 subtracting negative offset from size.
27142 (builtin_access::overlap): Adjust offset bounds of the access to fall
27143 within the size of the object if possible.
27145 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
27147 PR rtl-optimization/83699
27148 * expmed.c (extract_bit_field_1): Restrict the vector usage of
27149 extract_bit_field_as_subreg to cases in which the extracted
27150 value is also a vector.
27152 * lra-constraints.c (process_alt_operands): Test for the equivalence
27153 substitutions when detecting a possible reload cycle.
27155 2018-01-06 Jakub Jelinek <jakub@redhat.com>
27158 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
27159 by default if flag_selective_schedling{,2}. Formatting fixes.
27161 PR rtl-optimization/83682
27162 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
27163 if it has non-VECTOR_MODE element mode.
27164 (vec_duplicate_p): Likewise.
27166 PR middle-end/83694
27167 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27168 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27170 2018-01-05 Jakub Jelinek <jakub@redhat.com>
27173 * config/i386/i386-builtin.def
27174 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27175 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27176 Require also OPTION_MASK_ISA_AVX512F in addition to
27177 OPTION_MASK_ISA_GFNI.
27178 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27179 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27180 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27181 to OPTION_MASK_ISA_GFNI.
27182 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27183 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27184 OPTION_MASK_ISA_AVX512BW.
27185 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27186 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27187 addition to OPTION_MASK_ISA_GFNI.
27188 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27189 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27190 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27191 to OPTION_MASK_ISA_GFNI.
27192 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27193 a requirement for all ISAs rather than any of them with a few
27195 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27197 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27198 bitmasks to be enabled with 3 exceptions, instead of requiring any
27199 enabled ISA with lots of exceptions.
27200 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27201 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27202 Change avx512bw in isa attribute to avx512f.
27203 * config/i386/sgxintrin.h: Add license boilerplate.
27204 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
27205 to __AVX512F__ and __AVX512VL to __AVX512VL__.
27206 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27207 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27209 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27210 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
27211 temporarily sse2 rather than sse if not enabled already.
27214 * config/i386/sse.md (VI248_VLBW): Rename to ...
27215 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
27216 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
27217 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
27218 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
27219 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
27220 mode iterator instead of VI248_VLBW.
27222 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
27224 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
27225 (record_modified): Skip clobbers; add debug output.
27226 (param_change_prob): Use sreal frequencies.
27228 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
27230 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
27231 punt for user-aligned variables.
27233 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
27235 * tree-chrec.c (chrec_contains_symbols): Return true for
27238 2018-01-05 Sudakshina Das <sudi.das@arm.com>
27241 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
27242 of (x|y) == x for BICS pattern.
27244 2018-01-05 Jakub Jelinek <jakub@redhat.com>
27246 PR tree-optimization/83605
27247 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
27248 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
27251 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
27253 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
27254 * config/epiphany/rtems.h: New file.
27256 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27257 Uros Bizjak <ubizjak@gmail.com>
27260 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
27261 QIreg_operand instead of register_operand predicate.
27262 * config/i386/i386.c (ix86_rop_should_change_byte_p,
27263 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
27264 comments instead of -fmitigate[-_]rop.
27266 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27269 * cgraphunit.c (symbol_table::compile): Switch to text_section
27270 before calling assembly_start debug hook.
27271 * run-rtl-passes.c (run_rtl_passes): Likewise.
27274 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27276 * tree-vrp.c (extract_range_from_binary_expr_1): Check
27277 range_int_cst_p rather than !symbolic_range_p before calling
27278 extract_range_from_multiplicative_op_1.
27280 2018-01-04 Jeff Law <law@redhat.com>
27282 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
27283 redundant test in assertion.
27285 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27287 * doc/rtl.texi: Document machine_mode wrapper classes.
27289 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27291 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
27292 using tree_to_uhwi.
27294 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27296 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
27297 the VEC_PERM_EXPR fold to fail.
27299 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27302 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
27303 to switched_sections.
27305 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27308 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
27309 test for d.testing.
27311 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
27314 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
27315 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
27317 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27320 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
27321 is BLKmode and bitpos not zero or mode change is needed.
27323 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27326 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
27329 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
27332 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
27333 instead of MULT rtx. Update all corresponding splitters.
27334 (*saddl_se): Ditto.
27335 (*ssub<modesuffix>): Ditto.
27336 (*ssubl_se): Ditto.
27337 (*cmp_sadd_di): Update split patterns.
27338 (*cmp_sadd_si): Ditto.
27339 (*cmp_sadd_sidi): Ditto.
27340 (*cmp_ssub_di): Ditto.
27341 (*cmp_ssub_si): Ditto.
27342 (*cmp_ssub_sidi): Ditto.
27343 * config/alpha/predicates.md (const23_operand): New predicate.
27344 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
27345 Look for ASHIFT, not MULT inner operand.
27346 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
27348 2018-01-04 Martin Liska <mliska@suse.cz>
27350 PR gcov-profile/83669
27351 * gcov.c (output_intermediate_file): Add version to intermediate
27353 * doc/gcov.texi: Document new field 'version' in intermediate
27354 file format. Fix location of '-k' option of gcov command.
27356 2018-01-04 Martin Liska <mliska@suse.cz>
27359 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
27361 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27363 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
27365 2018-01-03 Martin Sebor <msebor@redhat.com>
27367 PR tree-optimization/83655
27368 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
27369 checking calls with invalid arguments.
27371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27373 * tree-vect-stmts.c (vect_get_store_rhs): New function.
27374 (vectorizable_mask_load_store): Delete.
27375 (vectorizable_call): Return false for masked loads and stores.
27376 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
27377 instead of gimple_assign_rhs1.
27378 (vectorizable_load): Handle IFN_MASK_LOAD.
27379 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
27381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27383 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
27385 (vectorizable_mask_load_store): ...here.
27386 (vectorizable_load): ...and here.
27388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27390 * tree-vect-stmts.c (vect_build_all_ones_mask)
27391 (vect_build_zero_merge_argument): New functions, split out from...
27392 (vectorizable_load): ...here.
27394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27396 * tree-vect-stmts.c (vect_check_store_rhs): New function,
27398 (vectorizable_mask_load_store): ...here.
27399 (vectorizable_store): ...and here.
27401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27403 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
27405 (vectorizable_mask_load_store): ...here.
27407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27409 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
27410 (vect_model_store_cost): Take a vec_load_store_type instead of a
27412 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
27413 (vect_model_store_cost): Take a vec_load_store_type instead of a
27415 (vectorizable_mask_load_store): Update accordingly.
27416 (vectorizable_store): Likewise.
27417 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
27419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27421 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
27422 IFN_MASK_LOAD calls here rather than...
27423 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
27425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27426 Alan Hayward <alan.hayward@arm.com>
27427 David Sherwood <david.sherwood@arm.com>
27429 * expmed.c (extract_bit_field_1): For vector extracts,
27430 fall back to extract_bit_field_as_subreg if vec_extract
27433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27434 Alan Hayward <alan.hayward@arm.com>
27435 David Sherwood <david.sherwood@arm.com>
27437 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
27438 they are variable or constant sized.
27439 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
27440 slots for constant-sized data.
27442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27443 Alan Hayward <alan.hayward@arm.com>
27444 David Sherwood <david.sherwood@arm.com>
27446 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
27447 handling COND_EXPRs with boolean comparisons, try to find a better
27448 basis for the mask type than the boolean itself.
27450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27452 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
27453 is calculated and how it can be overridden.
27454 * genmodes.c (max_bitsize_mode_any_mode): New variable.
27455 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
27457 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
27460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27461 Alan Hayward <alan.hayward@arm.com>
27462 David Sherwood <david.sherwood@arm.com>
27464 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
27465 Remove the mode argument.
27466 (aarch64_simd_valid_immediate): Remove the mode and inverse
27468 * config/aarch64/iterators.md (bitsize): New iterator.
27469 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
27470 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
27471 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
27472 aarch64_simd_valid_immediate.
27473 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
27474 (aarch64_reg_or_bic_imm): Likewise.
27475 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
27476 with an insn_type enum and msl with a modifier_type enum.
27477 Replace element_width with a scalar_mode. Change the shift
27478 to unsigned int. Add constructors for scalar_float_mode and
27479 scalar_int_mode elements.
27480 (aarch64_vect_float_const_representable_p): Delete.
27481 (aarch64_can_const_movi_rtx_p)
27482 (aarch64_simd_scalar_immediate_valid_for_move)
27483 (aarch64_simd_make_constant): Update call to
27484 aarch64_simd_valid_immediate.
27485 (aarch64_advsimd_valid_immediate_hs): New function.
27486 (aarch64_advsimd_valid_immediate): Likewise.
27487 (aarch64_simd_valid_immediate): Remove mode and inverse
27488 arguments. Rewrite to use the above. Use const_vec_duplicate_p
27489 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
27490 and aarch64_float_const_representable_p on the result.
27491 (aarch64_output_simd_mov_immediate): Remove mode argument.
27492 Update call to aarch64_simd_valid_immediate and use of
27493 simd_immediate_info.
27494 (aarch64_output_scalar_simd_mov_immediate): Update call
27497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27498 Alan Hayward <alan.hayward@arm.com>
27499 David Sherwood <david.sherwood@arm.com>
27501 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
27502 (mode_nunits): Likewise CONST_MODE_NUNITS.
27503 * machmode.def (ADJUST_NUNITS): Document.
27504 * genmodes.c (mode_data::need_nunits_adj): New field.
27505 (blank_mode): Update accordingly.
27506 (adj_nunits): New variable.
27507 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
27509 (emit_mode_size_inline): Set need_bytesize_adj for all modes
27510 listed in adj_nunits.
27511 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
27512 listed in adj_nunits. Don't emit case statements for such modes.
27513 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
27514 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
27515 nothing if adj_nunits is nonnull.
27516 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
27517 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
27518 (emit_mode_fbit): Update use of print_maybe_const_decl.
27519 (emit_move_size): Likewise. Treat the array as non-const
27521 (emit_mode_adjustments): Handle adj_nunits.
27523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27525 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
27526 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
27527 (VECTOR_MODES): Use it.
27528 (make_vector_modes): Take the prefix as an argument.
27530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27531 Alan Hayward <alan.hayward@arm.com>
27532 David Sherwood <david.sherwood@arm.com>
27534 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
27535 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
27536 for MODE_VECTOR_BOOL.
27537 * machmode.def (VECTOR_BOOL_MODE): Document.
27538 * genmodes.c (VECTOR_BOOL_MODE): New macro.
27539 (make_vector_bool_mode): New function.
27540 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
27542 * lto-streamer-in.c (lto_input_mode_table): Likewise.
27543 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
27545 * stor-layout.c (int_mode_for_mode): Likewise.
27546 * tree.c (build_vector_type_for_mode): Likewise.
27547 * varasm.c (output_constant_pool_2): Likewise.
27548 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
27549 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
27550 for MODE_VECTOR_BOOL.
27551 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
27552 of mode class checks.
27553 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
27554 instead of a list of mode class checks.
27555 (expand_vector_scalar_condition): Likewise.
27556 (type_for_widest_vector_mode): Handle BImode as an inner mode.
27558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27559 Alan Hayward <alan.hayward@arm.com>
27560 David Sherwood <david.sherwood@arm.com>
27562 * machmode.h (mode_size): Change from unsigned short to
27564 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
27565 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27566 or if measurement_type is not polynomial.
27567 (fixed_size_mode::includes_p): Check for constant-sized modes.
27568 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
27569 return a poly_uint16 rather than an unsigned short.
27570 (emit_mode_size): Change the type of mode_size from unsigned short
27571 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
27572 (emit_mode_adjustments): Cope with polynomial vector sizes.
27573 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27575 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27577 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
27578 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
27579 * caller-save.c (setup_save_areas): Likewise.
27580 (replace_reg_with_saved_mem): Likewise.
27581 * calls.c (emit_library_call_value_1): Likewise.
27582 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
27583 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
27584 (gen_lowpart_for_combine): Likewise.
27585 * convert.c (convert_to_integer_1): Likewise.
27586 * cse.c (equiv_constant, cse_insn): Likewise.
27587 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
27588 (cselib_subst_to_values): Likewise.
27589 * dce.c (word_dce_process_block): Likewise.
27590 * df-problems.c (df_word_lr_mark_ref): Likewise.
27591 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
27592 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
27593 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
27594 (rtl_for_decl_location): Likewise.
27595 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
27596 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
27597 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
27598 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
27599 (expand_expr_real_1): Likewise.
27600 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
27601 (pad_below): Likewise.
27602 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27603 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
27604 * ira.c (get_subreg_tracking_sizes): Likewise.
27605 * ira-build.c (ira_create_allocno_objects): Likewise.
27606 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
27607 (ira_sort_regnos_for_alter_reg): Likewise.
27608 * ira-costs.c (record_operand_costs): Likewise.
27609 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
27610 (resolve_simple_move): Likewise.
27611 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
27612 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
27613 (lra_constraints): Likewise.
27614 (CONST_POOL_OK_P): Reject variable-sized modes.
27615 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
27616 (add_pseudo_to_slot, lra_spill): Likewise.
27617 * omp-low.c (omp_clause_aligned_alignment): Likewise.
27618 * optabs-query.c (get_best_extraction_insn): Likewise.
27619 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27620 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
27621 (expand_mult_highpart, valid_multiword_target_p): Likewise.
27622 * recog.c (offsettable_address_addr_space_p): Likewise.
27623 * regcprop.c (maybe_mode_change): Likewise.
27624 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
27625 * regrename.c (build_def_use): Likewise.
27626 * regstat.c (dump_reg_info): Likewise.
27627 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
27628 (find_reloads, find_reloads_subreg_address): Likewise.
27629 * reload1.c (eliminate_regs_1): Likewise.
27630 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
27631 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
27632 (simplify_binary_operation_1, simplify_subreg): Likewise.
27633 * targhooks.c (default_function_arg_padding): Likewise.
27634 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
27635 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
27636 (verify_gimple_assign_ternary): Likewise.
27637 * tree-inline.c (estimate_move_cost): Likewise.
27638 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27639 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
27640 (get_address_cost_ainc): Likewise.
27641 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
27642 (vect_supportable_dr_alignment): Likewise.
27643 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27644 (vectorizable_reduction): Likewise.
27645 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
27646 (vectorizable_operation, vectorizable_load): Likewise.
27647 * tree.c (build_same_sized_truth_vector_type): Likewise.
27648 * valtrack.c (cleanup_auto_inc_dec): Likewise.
27649 * var-tracking.c (emit_note_insn_var_location): Likewise.
27650 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
27651 (ADDR_VEC_ALIGN): Likewise.
27653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27654 Alan Hayward <alan.hayward@arm.com>
27655 David Sherwood <david.sherwood@arm.com>
27657 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
27659 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27660 or if measurement_type is polynomial.
27661 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
27662 * combine.c (make_extraction): Likewise.
27663 * dse.c (find_shift_sequence): Likewise.
27664 * dwarf2out.c (mem_loc_descriptor): Likewise.
27665 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
27666 (extract_bit_field, extract_low_bits): Likewise.
27667 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
27668 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
27669 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
27670 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
27671 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27672 * reload.c (find_reloads): Likewise.
27673 * reload1.c (alter_reg): Likewise.
27674 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
27675 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
27676 * tree-if-conv.c (predicate_mem_writes): Likewise.
27677 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
27678 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
27679 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
27680 * valtrack.c (dead_debug_insert_temp): Likewise.
27681 * varasm.c (mergeable_constant_section): Likewise.
27682 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
27684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27685 Alan Hayward <alan.hayward@arm.com>
27686 David Sherwood <david.sherwood@arm.com>
27688 * expr.c (expand_assignment): Cope with polynomial mode sizes
27689 when assigning to a CONCAT.
27691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27692 Alan Hayward <alan.hayward@arm.com>
27693 David Sherwood <david.sherwood@arm.com>
27695 * machmode.h (mode_precision): Change from unsigned short to
27697 (mode_to_precision): Return a poly_uint16 rather than an unsigned
27699 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
27700 or if measurement_type is not polynomial.
27701 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
27702 in which the mode is already known to be a scalar_int_mode.
27703 * genmodes.c (emit_mode_precision): Change the type of mode_precision
27704 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
27706 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27707 for GET_MODE_PRECISION.
27708 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27709 for GET_MODE_PRECISION.
27710 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
27712 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
27713 (expand_field_assignment, make_extraction): Likewise.
27714 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
27715 (get_last_value): Likewise.
27716 * convert.c (convert_to_integer_1): Likewise.
27717 * cse.c (cse_insn): Likewise.
27718 * expr.c (expand_expr_real_1): Likewise.
27719 * lra-constraints.c (simplify_operand_subreg): Likewise.
27720 * optabs-query.c (can_atomic_load_p): Likewise.
27721 * optabs.c (expand_atomic_load): Likewise.
27722 (expand_atomic_store): Likewise.
27723 * ree.c (combine_reaching_defs): Likewise.
27724 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
27725 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
27726 * tree.h (type_has_mode_precision_p): Likewise.
27727 * ubsan.c (instrument_si_overflow): Likewise.
27729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27730 Alan Hayward <alan.hayward@arm.com>
27731 David Sherwood <david.sherwood@arm.com>
27733 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
27734 polynomial numbers of units.
27735 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
27736 (valid_vector_subparts_p): New function.
27737 (build_vector_type): Remove temporary shim and take the number
27738 of units as a poly_uint64 rather than an int.
27739 (build_opaque_vector_type): Take the number of units as a
27740 poly_uint64 rather than an int.
27741 * tree.c (build_vector_from_ctor): Handle polynomial
27742 TYPE_VECTOR_SUBPARTS.
27743 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
27744 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
27745 (build_vector_from_val): If the number of units is variable,
27746 use build_vec_duplicate_cst for constant operands and
27747 VEC_DUPLICATE_EXPR otherwise.
27748 (make_vector_type): Remove temporary is_constant ().
27749 (build_vector_type, build_opaque_vector_type): Take the number of
27750 units as a poly_uint64 rather than an int.
27751 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
27753 * cfgexpand.c (expand_debug_expr): Likewise.
27754 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
27755 (store_constructor, expand_expr_real_1): Likewise.
27756 (const_scalar_mask_from_tree): Likewise.
27757 * fold-const-call.c (fold_const_reduction): Likewise.
27758 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
27759 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
27760 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
27761 (fold_relational_const): Likewise.
27762 (native_interpret_vector): Likewise. Change the size from an
27763 int to an unsigned int.
27764 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
27765 TYPE_VECTOR_SUBPARTS.
27766 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
27767 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
27768 duplicating a non-constant operand into a variable-length vector.
27769 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
27770 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
27771 * ipa-icf.c (sem_variable::equals): Likewise.
27772 * match.pd: Likewise.
27773 * omp-simd-clone.c (simd_clone_subparts): Likewise.
27774 * print-tree.c (print_node): Likewise.
27775 * stor-layout.c (layout_type): Likewise.
27776 * targhooks.c (default_builtin_vectorization_cost): Likewise.
27777 * tree-cfg.c (verify_gimple_comparison): Likewise.
27778 (verify_gimple_assign_binary): Likewise.
27779 (verify_gimple_assign_ternary): Likewise.
27780 (verify_gimple_assign_single): Likewise.
27781 * tree-pretty-print.c (dump_generic_node): Likewise.
27782 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27783 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
27784 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
27785 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
27786 (vect_shift_permute_load_chain): Likewise.
27787 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
27788 (expand_vector_condition, optimize_vector_constructor): Likewise.
27789 (lower_vec_perm, get_compute_type): Likewise.
27790 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27791 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
27792 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
27793 (vect_recog_mask_conversion_pattern): Likewise.
27794 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
27795 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
27796 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27797 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
27798 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
27799 (vectorizable_shift, vectorizable_operation, vectorizable_store)
27800 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
27801 (supportable_widening_operation): Likewise.
27802 (supportable_narrowing_operation): Likewise.
27803 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
27805 * varasm.c (output_constant): Likewise.
27807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27808 Alan Hayward <alan.hayward@arm.com>
27809 David Sherwood <david.sherwood@arm.com>
27811 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
27812 so that both the length == 3 and length != 3 cases set up their
27813 own permute vectors. Add comments explaining why we know the
27814 number of elements is constant.
27815 (vect_permute_load_chain): Likewise.
27817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27818 Alan Hayward <alan.hayward@arm.com>
27819 David Sherwood <david.sherwood@arm.com>
27821 * machmode.h (mode_nunits): Change from unsigned char to
27823 (ONLY_FIXED_SIZE_MODES): New macro.
27824 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
27825 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
27826 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
27828 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
27829 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
27830 or if measurement_type is not polynomial.
27831 * genmodes.c (ZERO_COEFFS): New macro.
27832 (emit_mode_nunits_inline): Make mode_nunits_inline return a
27834 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
27835 Use ZERO_COEFFS when emitting initializers.
27836 * data-streamer.h (bp_pack_poly_value): New function.
27837 (bp_unpack_poly_value): Likewise.
27838 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27839 for GET_MODE_NUNITS.
27840 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27841 for GET_MODE_NUNITS.
27842 * tree.c (make_vector_type): Remove temporary shim and make
27843 the real function take the number of units as a poly_uint64
27844 rather than an int.
27845 (build_vector_type_for_mode): Handle polynomial nunits.
27846 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
27847 * emit-rtl.c (const_vec_series_p_1): Likewise.
27848 (gen_rtx_CONST_VECTOR): Likewise.
27849 * fold-const.c (test_vec_duplicate_folding): Likewise.
27850 * genrecog.c (validate_pattern): Likewise.
27851 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
27852 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27853 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
27854 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
27855 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
27856 * rtlanal.c (subreg_get_info): Likewise.
27857 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27858 (vect_grouped_load_supported): Likewise.
27859 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
27860 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
27861 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
27862 (simplify_const_unary_operation, simplify_binary_operation_1)
27863 (simplify_const_binary_operation, simplify_ternary_operation)
27864 (test_vector_ops_duplicate, test_vector_ops): Likewise.
27865 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
27866 instead of CONST_VECTOR_NUNITS.
27867 * varasm.c (output_constant_pool_2): Likewise.
27868 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
27869 explicit-encoded elements in the XVEC for variable-length vectors.
27871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27873 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
27875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27876 Alan Hayward <alan.hayward@arm.com>
27877 David Sherwood <david.sherwood@arm.com>
27879 * coretypes.h (fixed_size_mode): Declare.
27880 (fixed_size_mode_pod): New typedef.
27881 * builtins.h (target_builtins::x_apply_args_mode)
27882 (target_builtins::x_apply_result_mode): Change type to
27883 fixed_size_mode_pod.
27884 * builtins.c (apply_args_size, apply_result_size, result_vector)
27885 (expand_builtin_apply_args_1, expand_builtin_apply)
27886 (expand_builtin_return): Update accordingly.
27888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27890 * cse.c (hash_rtx_cb): Hash only the encoded elements.
27891 * cselib.c (cselib_hash_rtx): Likewise.
27892 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
27893 CONST_VECTOR encoding.
27895 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27896 Jeff Law <law@redhat.com>
27899 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
27900 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
27901 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
27902 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
27905 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
27906 explicitly probe *sp in a noreturn function if there were any callee
27907 register saves or frame pointer is needed.
27909 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27912 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27913 BLKmode for ternary, binary or unary expressions.
27916 * var-tracking.c (delete_vta_debug_insn): New inline function.
27917 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27918 insns from get_insns () to NULL instead of each bb separately.
27919 Use delete_vta_debug_insn. No longer static.
27920 (vt_debug_insns_local, variable_tracking_main_1): Adjust
27921 delete_vta_debug_insns callers.
27922 * rtl.h (delete_vta_debug_insns): Declare.
27923 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27924 instead of variable_tracking_main.
27926 2018-01-03 Martin Sebor <msebor@redhat.com>
27928 PR tree-optimization/83603
27929 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27930 arguments past the endof the argument list in functions declared
27931 without a prototype.
27932 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27933 Avoid checking when arguments are null.
27935 2018-01-03 Martin Sebor <msebor@redhat.com>
27938 * doc/extend.texi (attribute const): Fix a typo.
27939 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27940 issuing -Wsuggest-attribute for void functions.
27942 2018-01-03 Martin Sebor <msebor@redhat.com>
27944 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27945 offset_int::from instead of wide_int::to_shwi.
27946 (maybe_diag_overlap): Remove assertion.
27947 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27948 * gimple-ssa-sprintf.c (format_directive): Same.
27949 (parse_directive): Same.
27950 (sprintf_dom_walker::compute_format_length): Same.
27951 (try_substitute_return_value): Same.
27953 2018-01-03 Jeff Law <law@redhat.com>
27955 PR middle-end/83654
27956 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27957 non-constant residual for zero at runtime and avoid probing in
27958 that case. Reorganize code for trailing problem to mirror handling
27961 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27963 PR tree-optimization/83501
27964 * tree-ssa-strlen.c (get_string_cst): New.
27965 (handle_char_store): Call get_string_cst.
27967 2018-01-03 Martin Liska <mliska@suse.cz>
27969 PR tree-optimization/83593
27970 * tree-ssa-strlen.c: Include tree-cfg.h.
27971 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27972 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27973 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27975 (strlen_dom_walker::before_dom_children): Call
27976 gimple_purge_dead_eh_edges. Dump tranformation with details
27978 (strlen_dom_walker::before_dom_children): Update call by adding
27979 new argument cleanup_eh.
27980 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27982 2018-01-03 Martin Liska <mliska@suse.cz>
27985 * cif-code.def (VARIADIC_THUNK): New enum value.
27986 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27989 2018-01-03 Jan Beulich <jbeulich@suse.com>
27991 * sse.md (mov<mode>_internal): Tighten condition for when to use
27992 vmovdqu<ssescalarsize> for TI and OI modes.
27994 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27996 Update copyright years.
27998 2018-01-03 Martin Liska <mliska@suse.cz>
28001 * ipa-visibility.c (function_and_variable_visibility): Skip
28002 functions with noipa attribure.
28004 2018-01-03 Jakub Jelinek <jakub@redhat.com>
28006 * gcc.c (process_command): Update copyright notice dates.
28007 * gcov-dump.c (print_version): Ditto.
28008 * gcov.c (print_version): Ditto.
28009 * gcov-tool.c (print_version): Ditto.
28010 * gengtype.c (create_file): Ditto.
28011 * doc/cpp.texi: Bump @copying's copyright year.
28012 * doc/cppinternals.texi: Ditto.
28013 * doc/gcc.texi: Ditto.
28014 * doc/gccint.texi: Ditto.
28015 * doc/gcov.texi: Ditto.
28016 * doc/install.texi: Ditto.
28017 * doc/invoke.texi: Ditto.
28019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28021 * vector-builder.h (vector_builder::m_full_nelts): Change from
28022 unsigned int to poly_uint64.
28023 (vector_builder::full_nelts): Update prototype accordingly.
28024 (vector_builder::new_vector): Likewise.
28025 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
28026 (vector_builder::operator ==): Likewise.
28027 (vector_builder::finalize): Likewise.
28028 * int-vector-builder.h (int_vector_builder::int_vector_builder):
28029 Take the number of elements as a poly_uint64 rather than an
28031 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
28032 from unsigned int to poly_uint64.
28033 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
28034 (vec_perm_indices::new_vector): Likewise.
28035 (vec_perm_indices::length): Likewise.
28036 (vec_perm_indices::nelts_per_input): Likewise.
28037 (vec_perm_indices::input_nelts): Likewise.
28038 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
28039 number of elements per input as a poly_uint64 rather than an
28040 unsigned int. Use the original encoding for variable-length
28041 vectors, rather than clamping each individual element.
28042 For the second and subsequent elements in each pattern,
28043 clamp the step and base before clamping their sum.
28044 (vec_perm_indices::series_p): Handle polynomial element counts.
28045 (vec_perm_indices::all_in_range_p): Likewise.
28046 (vec_perm_indices_to_tree): Likewise.
28047 (vec_perm_indices_to_rtx): Likewise.
28048 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
28049 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
28050 (tree_vector_builder::new_binary_operation): Handle polynomial
28051 element counts. Return false if we need to know the number
28052 of elements at compile time.
28053 * fold-const.c (fold_vec_perm): Punt if the number of elements
28054 isn't known at compile time.
28056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28058 * vec-perm-indices.h (vec_perm_builder): Change element type
28059 from HOST_WIDE_INT to poly_int64.
28060 (vec_perm_indices::element_type): Update accordingly.
28061 (vec_perm_indices::clamp): Handle polynomial element_types.
28062 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28063 (vec_perm_indices::all_in_range_p): Likewise.
28064 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
28066 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
28067 polynomial vec_perm_indices element types.
28068 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
28069 * fold-const.c (fold_vec_perm): Likewise.
28070 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
28071 * tree-vect-generic.c (lower_vec_perm): Likewise.
28072 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28073 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
28074 element type to HOST_WIDE_INT.
28076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28077 Alan Hayward <alan.hayward@arm.com>
28078 David Sherwood <david.sherwood@arm.com>
28080 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
28081 rather than an int. Use plus_constant.
28082 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
28083 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
28085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28086 Alan Hayward <alan.hayward@arm.com>
28087 David Sherwood <david.sherwood@arm.com>
28089 * calls.c (emit_call_1, expand_call): Change struct_value_size from
28090 a HOST_WIDE_INT to a poly_int64.
28092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28093 Alan Hayward <alan.hayward@arm.com>
28094 David Sherwood <david.sherwood@arm.com>
28096 * calls.c (load_register_parameters): Cope with polynomial
28097 mode sizes. Require a constant size for BLKmode parameters
28098 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
28099 forces a parameter to be padded at the lsb end in order to
28100 fill a complete number of words, require the parameter size
28101 to be ordered wrt UNITS_PER_WORD.
28103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28104 Alan Hayward <alan.hayward@arm.com>
28105 David Sherwood <david.sherwood@arm.com>
28107 * reload1.c (spill_stack_slot_width): Change element type
28108 from unsigned int to poly_uint64_pod.
28109 (alter_reg): Treat mode sizes as polynomial.
28111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28112 Alan Hayward <alan.hayward@arm.com>
28113 David Sherwood <david.sherwood@arm.com>
28115 * reload.c (complex_word_subreg_p): New function.
28116 (reload_inner_reg_of_subreg, push_reload): Use it.
28118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28119 Alan Hayward <alan.hayward@arm.com>
28120 David Sherwood <david.sherwood@arm.com>
28122 * lra-constraints.c (process_alt_operands): Reject matched
28123 operands whose sizes aren't ordered.
28124 (match_reload): Refer to this check here.
28126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28127 Alan Hayward <alan.hayward@arm.com>
28128 David Sherwood <david.sherwood@arm.com>
28130 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28131 that the mode size is in the set {1, 2, 4, 8, 16}.
28133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28134 Alan Hayward <alan.hayward@arm.com>
28135 David Sherwood <david.sherwood@arm.com>
28137 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28138 Use plus_constant instead of gen_rtx_PLUS.
28140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28141 Alan Hayward <alan.hayward@arm.com>
28142 David Sherwood <david.sherwood@arm.com>
28144 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28145 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28146 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28147 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28148 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28149 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28150 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28151 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28152 * config/i386/i386.c (ix86_push_rounding): ...this new function.
28153 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
28155 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
28156 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
28157 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
28158 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
28159 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
28160 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
28161 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
28162 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
28163 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
28164 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28166 * expr.c (emit_move_resolve_push): Treat the input and result
28167 of PUSH_ROUNDING as a poly_int64.
28168 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28169 (emit_push_insn): Likewise.
28170 * lra-eliminations.c (mark_not_eliminable): Likewise.
28171 * recog.c (push_operand): Likewise.
28172 * reload1.c (elimination_effects): Likewise.
28173 * rtlanal.c (nonzero_bits1): Likewise.
28174 * calls.c (store_one_arg): Likewise. Require the padding to be
28175 known at compile time.
28177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28178 Alan Hayward <alan.hayward@arm.com>
28179 David Sherwood <david.sherwood@arm.com>
28181 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28182 Use plus_constant instead of gen_rtx_PLUS.
28184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28185 Alan Hayward <alan.hayward@arm.com>
28186 David Sherwood <david.sherwood@arm.com>
28188 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28189 rather than an int.
28191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28192 Alan Hayward <alan.hayward@arm.com>
28193 David Sherwood <david.sherwood@arm.com>
28195 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28196 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28197 via stack temporaries. Treat the mode size as polynomial too.
28199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28200 Alan Hayward <alan.hayward@arm.com>
28201 David Sherwood <david.sherwood@arm.com>
28203 * expr.c (expand_expr_real_2): When handling conversions involving
28204 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28205 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
28206 as a poly_uint64 too.
28208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28209 Alan Hayward <alan.hayward@arm.com>
28210 David Sherwood <david.sherwood@arm.com>
28212 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
28214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28215 Alan Hayward <alan.hayward@arm.com>
28216 David Sherwood <david.sherwood@arm.com>
28218 * combine.c (can_change_dest_mode): Handle polynomial
28219 REGMODE_NATURAL_SIZE.
28220 * expmed.c (store_bit_field_1): Likewise.
28221 * expr.c (store_constructor): Likewise.
28222 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
28223 and polynomial REGMODE_NATURAL_SIZE.
28224 (gen_lowpart_common): Likewise.
28225 * reginfo.c (record_subregs_of_mode): Likewise.
28226 * rtlanal.c (read_modify_subreg_p): Likewise.
28228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28229 Alan Hayward <alan.hayward@arm.com>
28230 David Sherwood <david.sherwood@arm.com>
28232 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
28233 numbers of elements.
28235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28236 Alan Hayward <alan.hayward@arm.com>
28237 David Sherwood <david.sherwood@arm.com>
28239 * match.pd: Cope with polynomial numbers of vector elements.
28241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28242 Alan Hayward <alan.hayward@arm.com>
28243 David Sherwood <david.sherwood@arm.com>
28245 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
28246 in a POINTER_PLUS_EXPR.
28248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28249 Alan Hayward <alan.hayward@arm.com>
28250 David Sherwood <david.sherwood@arm.com>
28252 * omp-simd-clone.c (simd_clone_subparts): New function.
28253 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
28254 (ipa_simd_modify_function_body): Likewise.
28256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28257 Alan Hayward <alan.hayward@arm.com>
28258 David Sherwood <david.sherwood@arm.com>
28260 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
28261 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
28262 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
28263 (expand_vector_condition, vector_element): Likewise.
28264 (subparts_gt): New function.
28265 (get_compute_type): Use subparts_gt.
28266 (count_type_subparts): Delete.
28267 (expand_vector_operations_1): Use subparts_gt instead of
28268 count_type_subparts.
28270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28271 Alan Hayward <alan.hayward@arm.com>
28272 David Sherwood <david.sherwood@arm.com>
28274 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
28275 (vect_compile_time_alias): ...this new function. Do the calculation
28276 on poly_ints rather than trees.
28277 (vect_prune_runtime_alias_test_list): Update call accordingly.
28279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28280 Alan Hayward <alan.hayward@arm.com>
28281 David Sherwood <david.sherwood@arm.com>
28283 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
28285 (vect_schedule_slp_instance): Likewise.
28287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28288 Alan Hayward <alan.hayward@arm.com>
28289 David Sherwood <david.sherwood@arm.com>
28291 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
28292 constant and extern definitions for variable-length vectors.
28293 (vect_get_constant_vectors): Note that the number of units
28294 is known to be constant.
28296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28297 Alan Hayward <alan.hayward@arm.com>
28298 David Sherwood <david.sherwood@arm.com>
28300 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
28301 of units as polynomial. Choose between WIDE and NARROW based
28304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28305 Alan Hayward <alan.hayward@arm.com>
28306 David Sherwood <david.sherwood@arm.com>
28308 * tree-vect-stmts.c (simd_clone_subparts): New function.
28309 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
28311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28312 Alan Hayward <alan.hayward@arm.com>
28313 David Sherwood <david.sherwood@arm.com>
28315 * tree-vect-stmts.c (vectorizable_call): Treat the number of
28316 vectors as polynomial. Use build_index_vector for
28317 IFN_GOMP_SIMD_LANE.
28319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28320 Alan Hayward <alan.hayward@arm.com>
28321 David Sherwood <david.sherwood@arm.com>
28323 * tree-vect-stmts.c (get_load_store_type): Treat the number of
28324 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
28325 for variable-length vectors.
28326 (vectorizable_mask_load_store): Treat the number of units as
28327 polynomial, asserting that it is constant if the condition has
28328 already been enforced.
28329 (vectorizable_store, vectorizable_load): Likewise.
28331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28332 Alan Hayward <alan.hayward@arm.com>
28333 David Sherwood <david.sherwood@arm.com>
28335 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
28336 of units as polynomial. Punt if we can't tell at compile time
28337 which vector contains the final result.
28339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28340 Alan Hayward <alan.hayward@arm.com>
28341 David Sherwood <david.sherwood@arm.com>
28343 * tree-vect-loop.c (vectorizable_induction): Treat the number
28344 of units as polynomial. Punt on SLP inductions. Use an integer
28345 VEC_SERIES_EXPR for variable-length integer reductions. Use a
28346 cast of such a series for variable-length floating-point
28349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28350 Alan Hayward <alan.hayward@arm.com>
28351 David Sherwood <david.sherwood@arm.com>
28353 * tree.h (build_index_vector): Declare.
28354 * tree.c (build_index_vector): New function.
28355 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
28356 of units as polynomial, forcibly converting it to a constant if
28357 vectorizable_reduction has already enforced the condition.
28358 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
28359 to create a {1,2,3,...} vector.
28360 (vectorizable_reduction): Treat the number of units as polynomial.
28361 Choose vectype_in based on the largest scalar element size rather
28362 than the smallest number of units. Enforce the restrictions
28365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28366 Alan Hayward <alan.hayward@arm.com>
28367 David Sherwood <david.sherwood@arm.com>
28369 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
28370 number of units as polynomial.
28372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28373 Alan Hayward <alan.hayward@arm.com>
28374 David Sherwood <david.sherwood@arm.com>
28376 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
28377 * target.def (autovectorize_vector_sizes): Return the vector sizes
28378 by pointer, using vector_sizes rather than a bitmask.
28379 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
28380 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
28381 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
28383 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
28384 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
28385 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
28386 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
28387 * omp-general.c (omp_max_vf): Likewise.
28388 * omp-low.c (omp_clause_aligned_alignment): Likewise.
28389 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
28390 * tree-vect-loop.c (vect_analyze_loop): Likewise.
28391 * tree-vect-slp.c (vect_slp_bb): Likewise.
28392 * doc/tm.texi: Regenerate.
28393 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
28395 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
28396 the vector size as a poly_uint64 rather than an unsigned int.
28397 (current_vector_size): Change from an unsigned int to a poly_uint64.
28398 (get_vectype_for_scalar_type): Update accordingly.
28399 * tree.h (build_truth_vector_type): Take the size and number of
28400 units as a poly_uint64 rather than an unsigned int.
28401 (build_vector_type): Add a temporary overload that takes
28402 the number of units as a poly_uint64 rather than an unsigned int.
28403 * tree.c (make_vector_type): Likewise.
28404 (build_truth_vector_type): Take the number of units as a poly_uint64
28405 rather than an unsigned int.
28407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28408 Alan Hayward <alan.hayward@arm.com>
28409 David Sherwood <david.sherwood@arm.com>
28411 * target.def (get_mask_mode): Take the number of units and length
28412 as poly_uint64s rather than unsigned ints.
28413 * targhooks.h (default_get_mask_mode): Update accordingly.
28414 * targhooks.c (default_get_mask_mode): Likewise.
28415 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
28416 * doc/tm.texi: Regenerate.
28418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28419 Alan Hayward <alan.hayward@arm.com>
28420 David Sherwood <david.sherwood@arm.com>
28422 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
28423 * omp-general.c (omp_max_vf): Likewise.
28424 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
28425 (expand_omp_simd): Handle polynomial safelen.
28426 * omp-low.c (omplow_simd_context): Add a default constructor.
28427 (omplow_simd_context::max_vf): Change from int to poly_uint64.
28428 (lower_rec_simd_input_clauses): Update accordingly.
28429 (lower_rec_input_clauses): Likewise.
28431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28432 Alan Hayward <alan.hayward@arm.com>
28433 David Sherwood <david.sherwood@arm.com>
28435 * tree-vectorizer.h (vect_nunits_for_cost): New function.
28436 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
28437 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
28438 (vect_analyze_slp_cost): Likewise.
28439 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
28440 (vect_model_load_cost): Likewise.
28442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28443 Alan Hayward <alan.hayward@arm.com>
28444 David Sherwood <david.sherwood@arm.com>
28446 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
28447 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
28448 from an unsigned int * to a poly_uint64_pod *.
28449 (calculate_unrolling_factor): New function.
28450 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
28452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28453 Alan Hayward <alan.hayward@arm.com>
28454 David Sherwood <david.sherwood@arm.com>
28456 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
28457 from an unsigned int to a poly_uint64.
28458 (_loop_vec_info::slp_unrolling_factor): Likewise.
28459 (_loop_vec_info::vectorization_factor): Change from an int
28461 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
28462 (vect_get_num_vectors): New function.
28463 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
28464 (vect_get_num_copies): Use vect_get_num_vectors.
28465 (vect_analyze_data_ref_dependences): Change max_vf from an int *
28466 to an unsigned int *.
28467 (vect_analyze_data_refs): Change min_vf from an int * to a
28469 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28470 than an unsigned HOST_WIDE_INT.
28471 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
28472 (vect_analyze_data_ref_dependence): Change max_vf from an int *
28473 to an unsigned int *.
28474 (vect_analyze_data_ref_dependences): Likewise.
28475 (vect_compute_data_ref_alignment): Handle polynomial vf.
28476 (vect_enhance_data_refs_alignment): Likewise.
28477 (vect_prune_runtime_alias_test_list): Likewise.
28478 (vect_shift_permute_load_chain): Likewise.
28479 (vect_supportable_dr_alignment): Likewise.
28480 (dependence_distance_ge_vf): Take the vectorization factor as a
28481 poly_uint64 rather than an unsigned HOST_WIDE_INT.
28482 (vect_analyze_data_refs): Change min_vf from an int * to a
28484 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
28485 vfm1 as a poly_uint64 rather than an int. Make the same change
28486 for the returned bound_scalar.
28487 (vect_gen_vector_loop_niters): Handle polynomial vf.
28488 (vect_do_peeling): Likewise. Update call to
28489 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
28490 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
28492 * tree-vect-loop.c (vect_determine_vectorization_factor)
28493 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
28494 (vect_get_known_peeling_cost): Likewise.
28495 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
28496 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
28497 (vect_transform_loop): Likewise. Use the lowest possible VF when
28498 updating the upper bounds of the loop.
28499 (vect_min_worthwhile_factor): Make static. Return an unsigned int
28500 rather than an int.
28501 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
28502 polynomial unroll factors.
28503 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
28504 (vect_make_slp_decision): Likewise.
28505 (vect_supported_load_permutation_p): Likewise, and polynomial
28507 (vect_analyze_slp_cost): Handle polynomial vf.
28508 (vect_slp_analyze_node_operations): Likewise.
28509 (vect_slp_analyze_bb_1): Likewise.
28510 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28511 than an unsigned HOST_WIDE_INT.
28512 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
28513 (vectorizable_load): Handle polynomial vf.
28514 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
28516 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
28518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28519 Alan Hayward <alan.hayward@arm.com>
28520 David Sherwood <david.sherwood@arm.com>
28522 * match.pd: Handle bit operations involving three constants
28523 and try to fold one pair.
28525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28527 * tree-vect-loop-manip.c: Include gimple-fold.h.
28528 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
28529 niters_maybe_zero parameters. Handle other cases besides a step of 1.
28530 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
28531 Add a path that uses a step of VF instead of 1, but disable it
28533 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
28534 and niters_no_overflow parameters. Update calls to
28535 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
28536 Create a new SSA name if the latter choses to use a ste other
28537 than zero, and return it via niters_vector_mult_vf_var.
28538 * tree-vect-loop.c (vect_transform_loop): Update calls to
28539 vect_do_peeling, vect_gen_vector_loop_niters and
28540 slpeel_make_loop_iterate_ntimes.
28541 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
28542 (vect_gen_vector_loop_niters): Update declarations after above changes.
28544 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
28546 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
28547 128-bit round to integer instructions.
28548 (ceil<mode>2): Likewise.
28549 (btrunc<mode>2): Likewise.
28550 (round<mode>2): Likewise.
28552 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28554 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
28555 unaligned VSX load/store on P8/P9.
28556 (expand_block_clear): Allow the use of unaligned VSX
28557 load/store on P8/P9.
28559 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28561 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
28563 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
28564 swap associated with both a load and a store.
28566 2018-01-02 Andrew Waterman <andrew@sifive.com>
28568 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
28569 * config/riscv/riscv.md (clear_cache): Use it.
28571 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
28573 * web.c: Remove out-of-date comment.
28575 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28577 * expr.c (fixup_args_size_notes): Check that any existing
28578 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
28579 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
28580 (emit_single_push_insn): ...here.
28582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28584 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
28585 (const_vector_encoded_nelts): New function.
28586 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
28587 (const_vector_int_elt, const_vector_elt): Declare.
28588 * emit-rtl.c (const_vector_int_elt_1): New function.
28589 (const_vector_elt): Likewise.
28590 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
28591 of CONST_VECTOR_ELT.
28593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28595 * expr.c: Include rtx-vector-builder.h.
28596 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
28597 directly on the tree encoding.
28598 (const_vector_from_tree): Likewise.
28599 * optabs.c: Include rtx-vector-builder.h.
28600 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
28601 sequence of "u" values.
28602 * vec-perm-indices.c: Include rtx-vector-builder.h.
28603 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
28604 directly on the vec_perm_indices encoding.
28606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28608 * doc/rtl.texi (const_vector): Describe new encoding scheme.
28609 * Makefile.in (OBJS): Add rtx-vector-builder.o.
28610 * rtx-vector-builder.h: New file.
28611 * rtx-vector-builder.c: Likewise.
28612 * rtl.h (rtx_def::u2): Add a const_vector field.
28613 (CONST_VECTOR_NPATTERNS): New macro.
28614 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
28615 (CONST_VECTOR_DUPLICATE_P): Likewise.
28616 (CONST_VECTOR_STEPPED_P): Likewise.
28617 (CONST_VECTOR_ENCODED_ELT): Likewise.
28618 (const_vec_duplicate_p): Check for a duplicated vector encoding.
28619 (unwrap_const_vec_duplicate): Likewise.
28620 (const_vec_series_p): Check for a non-duplicated vector encoding.
28621 Say that the function only returns true for integer vectors.
28622 * emit-rtl.c: Include rtx-vector-builder.h.
28623 (gen_const_vec_duplicate_1): Delete.
28624 (gen_const_vector): Call gen_const_vec_duplicate instead of
28625 gen_const_vec_duplicate_1.
28626 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
28627 (gen_const_vec_duplicate): Use rtx_vector_builder.
28628 (gen_const_vec_series): Likewise.
28629 (gen_rtx_CONST_VECTOR): Likewise.
28630 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
28631 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28632 Build a new vector rather than modifying a CONST_VECTOR in-place.
28633 (handle_special_swappables): Update call accordingly.
28634 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
28635 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28636 Build a new vector rather than modifying a CONST_VECTOR in-place.
28637 (handle_special_swappables): Update call accordingly.
28639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28641 * simplify-rtx.c (simplify_const_binary_operation): Use
28642 CONST_VECTOR_ELT instead of XVECEXP.
28644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28646 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
28647 the selector elements to be different from the data elements
28648 if the selector is a VECTOR_CST.
28649 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
28650 ssizetype for the selector.
28652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28654 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
28655 before testing each element individually.
28656 * tree-vect-generic.c (lower_vec_perm): Likewise.
28658 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28660 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
28661 * selftest-run-tests.c (selftest::run_tests): Call it.
28662 * vector-builder.h (vector_builder::operator ==): New function.
28663 (vector_builder::operator !=): Likewise.
28664 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
28665 (vec_perm_indices::all_from_input_p): New function.
28666 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28667 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
28668 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
28669 instead of reading the VECTOR_CST directly. Detect whether both
28670 vector inputs are the same before constructing the vec_perm_indices,
28671 and update the number of inputs argument accordingly. Use the
28672 utility functions added above. Only construct sel2 if we need to.
28674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28676 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
28677 the broadcast of the low byte.
28678 (expand_mult_highpart): Use an explicit encoding for the permutes.
28679 * optabs-query.c (can_mult_highpart_p): Likewise.
28680 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
28681 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28682 (vectorizable_bswap): Likewise.
28683 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
28684 explicit encoding for the power-of-2 permutes.
28685 (vect_permute_store_chain): Likewise.
28686 (vect_grouped_load_supported): Likewise.
28687 (vect_permute_load_chain): Likewise.
28689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28691 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
28692 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
28693 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
28694 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28695 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
28696 (vect_gen_perm_mask_any): Likewise.
28698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28700 * int-vector-builder.h: New file.
28701 * vec-perm-indices.h: Include int-vector-builder.h.
28702 (vec_perm_indices): Redefine as an int_vector_builder.
28703 (auto_vec_perm_indices): Delete.
28704 (vec_perm_builder): Redefine as a stand-alone class.
28705 (vec_perm_indices::vec_perm_indices): New function.
28706 (vec_perm_indices::clamp): Likewise.
28707 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
28708 (vec_perm_indices::new_vector): New function.
28709 (vec_perm_indices::new_expanded_vector): Update for new
28710 vec_perm_indices class.
28711 (vec_perm_indices::rotate_inputs): New function.
28712 (vec_perm_indices::all_in_range_p): Operate directly on the
28713 encoded form, without computing elided elements.
28714 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
28715 encoding. Update for new vec_perm_indices class.
28716 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
28717 the given vec_perm_builder.
28718 (expand_vec_perm_var): Update vec_perm_builder constructor.
28719 (expand_mult_highpart): Use vec_perm_builder instead of
28720 auto_vec_perm_indices.
28721 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
28722 vec_perm_indices instead of auto_vec_perm_indices. Use a single
28723 or double series encoding as appropriate.
28724 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
28725 vec_perm_indices instead of auto_vec_perm_indices.
28726 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28727 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28728 (vect_permute_store_chain): Likewise.
28729 (vect_grouped_load_supported): Likewise.
28730 (vect_permute_load_chain): Likewise.
28731 (vect_shift_permute_load_chain): Likewise.
28732 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28733 (vect_transform_slp_perm_load): Likewise.
28734 (vect_schedule_slp_instance): Likewise.
28735 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28736 (vectorizable_mask_load_store): Likewise.
28737 (vectorizable_bswap): Likewise.
28738 (vectorizable_store): Likewise.
28739 (vectorizable_load): Likewise.
28740 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
28741 vec_perm_indices instead of auto_vec_perm_indices. Use
28742 tree_to_vec_perm_builder to read the vector from a tree.
28743 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
28744 vec_perm_builder instead of a vec_perm_indices.
28745 (have_whole_vector_shift): Use vec_perm_builder and
28746 vec_perm_indices instead of auto_vec_perm_indices. Leave the
28747 truncation to calc_vec_perm_mask_for_shift.
28748 (vect_create_epilog_for_reduction): Likewise.
28749 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
28750 from auto_vec_perm_indices to vec_perm_indices.
28751 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28752 instead of changing individual elements.
28753 (aarch64_vectorize_vec_perm_const): Use new_vector to install
28754 the vector in d.perm.
28755 * config/arm/arm.c (expand_vec_perm_d::perm): Change
28756 from auto_vec_perm_indices to vec_perm_indices.
28757 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28758 instead of changing individual elements.
28759 (arm_vectorize_vec_perm_const): Use new_vector to install
28760 the vector in d.perm.
28761 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
28762 Update vec_perm_builder constructor.
28763 (rs6000_expand_interleave): Likewise.
28764 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
28765 (rs6000_expand_interleave): Likewise.
28767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28769 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
28770 to qimode could truncate the indices.
28771 * optabs.c (expand_vec_perm_var): Likewise.
28773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28775 * Makefile.in (OBJS): Add vec-perm-indices.o.
28776 * vec-perm-indices.h: New file.
28777 * vec-perm-indices.c: Likewise.
28778 * target.h (vec_perm_indices): Replace with a forward class
28780 (auto_vec_perm_indices): Move to vec-perm-indices.h.
28781 * optabs.h: Include vec-perm-indices.h.
28782 (expand_vec_perm): Delete.
28783 (selector_fits_mode_p, expand_vec_perm_var): Declare.
28784 (expand_vec_perm_const): Declare.
28785 * target.def (vec_perm_const_ok): Replace with...
28786 (vec_perm_const): ...this new hook.
28787 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
28788 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
28789 * doc/tm.texi: Regenerate.
28790 * optabs.def (vec_perm_const): Delete.
28791 * doc/md.texi (vec_perm_const): Likewise.
28792 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
28793 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
28794 expand_vec_perm for constant permutation vectors. Assert that
28795 the mode of variable permutation vectors is the integer equivalent
28796 of the mode that is being permuted.
28797 * optabs-query.h (selector_fits_mode_p): Declare.
28798 * optabs-query.c: Include vec-perm-indices.h.
28799 (selector_fits_mode_p): New function.
28800 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
28801 is defined, instead of checking whether the vec_perm_const_optab
28802 exists. Use targetm.vectorize.vec_perm_const instead of
28803 targetm.vectorize.vec_perm_const_ok. Check whether the indices
28804 fit in the vector mode before using a variable permute.
28805 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
28806 vec_perm_indices instead of an rtx.
28807 (expand_vec_perm): Replace with...
28808 (expand_vec_perm_const): ...this new function. Take the selector
28809 as a vec_perm_indices rather than an rtx. Also take the mode of
28810 the selector. Update call to shift_amt_for_vec_perm_mask.
28811 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
28812 Use vec_perm_indices::new_expanded_vector to expand the original
28813 selector into bytes. Check whether the indices fit in the vector
28814 mode before using a variable permute.
28815 (expand_vec_perm_var): Make global.
28816 (expand_mult_highpart): Use expand_vec_perm_const.
28817 * fold-const.c: Includes vec-perm-indices.h.
28818 * tree-ssa-forwprop.c: Likewise.
28819 * tree-vect-data-refs.c: Likewise.
28820 * tree-vect-generic.c: Likewise.
28821 * tree-vect-loop.c: Likewise.
28822 * tree-vect-slp.c: Likewise.
28823 * tree-vect-stmts.c: Likewise.
28824 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
28826 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
28827 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
28828 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
28829 (aarch64_vectorize_vec_perm_const): ...this new function.
28830 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28831 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28832 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
28833 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
28834 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28835 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28836 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
28838 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
28839 check for NEON modes.
28840 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
28841 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
28842 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
28843 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
28845 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
28846 the old VEC_PERM_CONST conditions.
28847 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
28848 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
28849 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
28850 (ia64_vectorize_vec_perm_const_ok): Merge into...
28851 (ia64_vectorize_vec_perm_const): ...this new function.
28852 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
28853 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
28854 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
28855 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
28856 * config/mips/mips.c (mips_expand_vec_perm_const)
28857 (mips_vectorize_vec_perm_const_ok): Merge into...
28858 (mips_vectorize_vec_perm_const): ...this new function.
28859 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
28860 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
28861 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
28862 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
28863 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
28864 (rs6000_expand_vec_perm_const): Delete.
28865 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
28867 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28868 (altivec_expand_vec_perm_const_le): Take each operand individually.
28869 Operate on constant selectors rather than rtxes.
28870 (altivec_expand_vec_perm_const): Likewise. Update call to
28871 altivec_expand_vec_perm_const_le.
28872 (rs6000_expand_vec_perm_const): Delete.
28873 (rs6000_vectorize_vec_perm_const_ok): Delete.
28874 (rs6000_vectorize_vec_perm_const): New function.
28875 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28876 an element count and rtx array.
28877 (rs6000_expand_extract_even): Update call accordingly.
28878 (rs6000_expand_interleave): Likewise.
28879 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
28880 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
28881 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
28882 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
28883 (rs6000_expand_vec_perm_const): Delete.
28884 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28885 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28886 (altivec_expand_vec_perm_const_le): Take each operand individually.
28887 Operate on constant selectors rather than rtxes.
28888 (altivec_expand_vec_perm_const): Likewise. Update call to
28889 altivec_expand_vec_perm_const_le.
28890 (rs6000_expand_vec_perm_const): Delete.
28891 (rs6000_vectorize_vec_perm_const_ok): Delete.
28892 (rs6000_vectorize_vec_perm_const): New function. Remove stray
28893 reference to the SPE evmerge intructions.
28894 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28895 an element count and rtx array.
28896 (rs6000_expand_extract_even): Update call accordingly.
28897 (rs6000_expand_interleave): Likewise.
28898 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
28899 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
28901 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28905 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
28906 vector mode and that that mode matches the mode of the data
28908 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
28909 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
28910 directly using expand_vec_perm_1 when forcing selectors into
28912 (expand_vec_perm_var): New function, split out from expand_vec_perm.
28914 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28916 * optabs-query.h (can_vec_perm_p): Delete.
28917 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28918 * optabs-query.c (can_vec_perm_p): Split into...
28919 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28920 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28921 particular selector is valid.
28922 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28923 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28924 (vect_grouped_load_supported): Likewise.
28925 (vect_shift_permute_load_chain): Likewise.
28926 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28927 (vect_transform_slp_perm_load): Likewise.
28928 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28929 (vectorizable_bswap): Likewise.
28930 (vect_gen_perm_mask_checked): Likewise.
28931 * fold-const.c (fold_ternary_loc): Likewise. Don't take
28932 implementations of variable permutation vectors into account
28933 when deciding which selector to use.
28934 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28935 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28936 with a false third argument.
28937 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28938 to test whether the constant selector is valid and can_vec_perm_var_p
28939 to test whether a variable selector is valid.
28941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28943 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28944 * optabs-query.c (can_vec_perm_p): Likewise.
28945 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28946 instead of vec_perm_indices.
28947 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28948 (vect_gen_perm_mask_checked): Likewise,
28949 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28950 (vect_gen_perm_mask_checked): Likewise,
28952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28954 * optabs-query.h (qimode_for_vec_perm): Declare.
28955 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28956 (qimode_for_vec_perm): ...this new function.
28957 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28959 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28961 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28962 does not have a conditional at the top.
28964 2018-01-02 Richard Biener <rguenther@suse.de>
28966 * ipa-inline.c (big_speedup_p): Fix expression.
28968 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28971 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28974 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28978 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28979 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28980 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28981 cond_taken_branch_cost 3->4.
28983 2018-01-01 Jakub Jelinek <jakub@redhat.com>
28985 PR tree-optimization/83581
28986 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28987 TODO_cleanup_cfg if any changes have been made.
28989 PR middle-end/83608
28990 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28991 convert_modes if target mode has the right side, but different mode
28994 PR middle-end/83609
28995 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28996 last argument when extracting from CONCAT. If either from_real or
28997 from_imag is NULL, use expansion through memory. If result is not
28998 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28999 the parts directly to inner mode, if even that fails, use expansion
29002 PR middle-end/83623
29003 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
29004 check for bswap in mode rather than HImode and use that in expand_unop
29007 Copyright (C) 2018 Free Software Foundation, Inc.
29009 Copying and distribution of this file, with or without modification,
29010 are permitted in any medium without royalty provided the copyright
29011 notice and this notice are preserved.