1 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
4 * params.opt (modref-max-depth): Add range.
5 (modref-max-adjustments): Fix range.
7 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
9 * tree-pass.h (make_pass_vrp_threader): Remove.
10 * tree-ssa-threadbackward.c
11 (back_threader_profitability::profitable_path_p): Remove
12 ASSERT_EXPR references.
13 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
14 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
16 (class hybrid_threader): Remove.
17 (hybrid_threader::hybrid_threader): Remove.
18 (hybrid_threader::~hybrid_threader): Remove.
19 (hybrid_threader::before_dom_children): Remove.
20 (hybrid_threader::after_dom_children): Remove.
21 (execute_vrp_threader): Remove.
22 (class pass_vrp_threader): Remove.
23 (make_pass_vrp_threader): Remove.
25 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
27 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
28 (gimple_call_retslot_flags): Likewise.
29 (gimple_call_static_chain_flags): Likewise.
30 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
32 (modref_summary::useful_p): Likewise.
33 (modref_summary_lto::useful_p): Likewise.
34 (analyze_parms): Do not give up on NOVOPS.
35 (analyze_function): When dumping report chnages in EAF flags
36 between IPA and local pass.
37 (modref_merge_call_site_flags): Compute implicit eaf flags
38 based on callee ecf_flags and fnspec; if the function does not
39 bind to current defs use interposable_eaf_flags.
40 (modref_propagate_flags_in_scc): Update.
41 * ipa-modref.h (interposable_eaf_flags): New function.
43 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
45 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
47 (rs6000_invalid_new_builtin): New function.
48 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
49 (rs6000_expand_ldst_mask): New function.
50 (new_cpu_expand_builtin): Likewise.
51 (elemrev_icode): Likewise.
52 (ldv_expand_builtin): Likewise.
53 (lxvrse_expand_builtin): Likewise.
54 (lxvrze_expand_builtin): Likewise.
55 (stv_expand_builtin): Likewise.
56 (new_mma_expand_builtin): Likewise.
57 (new_htm_spr_num): Likewise.
58 (new_htm_expand_builtin): Likewise.
59 (rs6000_expand_new_builtin): Likewise.
60 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
62 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
64 * ipa-modref.c (modref_lattice): Add do_dataflow,
65 changed and propagate_to fields.
66 (modref_lattice::release): Free propagate_to
67 (modref_lattice::merge): Do not give up early on unknown
69 (modref_lattice::merge_deref): Likewise.
70 (modref_eaf_analysis): Update toplevel comment.
71 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
72 do optimistic dataflow initialization.
73 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
74 (modref_eaf_analysis::propagate): New member function.
75 (analyze_parms): Update to new API of modref_eaf_analysis.
77 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
79 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
80 return true on functions from other partition.
82 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
84 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
85 (vsx_splat_v4si_di): Change constraint to "wa".
87 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
89 PR tree-optimization/103061
90 * value-relation.cc (path_oracle::path_oracle): Initialize
92 (path_oracle::killing_def): Set m_killed_defs.
93 (path_oracle::query_relation): Do not look at the root oracle for
95 * value-relation.h (class path_oracle): Add m_killed_defs.
97 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
99 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
100 Remove gotos and other cleanups.
102 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
105 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
106 paradoxical and zero sized accesses.
108 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
111 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
114 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
116 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
117 to the list of symbol prefixes that must be made linker-
120 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
122 * config.host: Add support for aarch64-*-darwin.
123 * config/aarch64/host-aarch64-darwin.c: New file.
124 * config/aarch64/x-darwin: New file.
126 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
128 PR tree-optimization/103093
129 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
131 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
133 PR tree-optimization/102943
134 * gimple-range-cache.cc (class update_list): New.
135 (update_list::add): Replace add_to_update.
136 (update_list::pop): New.
137 (ranger_cache::ranger_cache): Adjust.
138 (ranger_cache::~ranger_cache): Adjust.
139 (ranger_cache::add_to_update): Delete.
140 (ranger_cache::propagate_cache): Adjust to new class.
141 (ranger_cache::propagate_updated_value): Ditto.
142 (ranger_cache::fill_block_cache): Ditto.
143 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
145 2021-11-05 Richard Biener <rguenther@suse.de>
147 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
148 comment and expand on another one. Combine nested if.
150 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
153 * config/pa/pa-modes.def: Add OImode integer type.
154 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
156 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
157 to UNITS_PER_WORD if IN_LIBGCC2.
158 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
159 negvti2, ashlti3, shrpd_internal): New patterns.
160 Change some multi instruction types to multi.
162 2021-11-05 Jakub Jelinek <jakub@redhat.com>
166 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
169 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
172 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
174 * config/aarch64/constraints.md: Support const in Usw.
176 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
178 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
179 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
181 2021-11-05 Richard Biener <rguenther@suse.de>
183 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
184 (LOOP_VINFO_N_STMTS): Likewise.
185 (vec_info_for_bb): Remove unused function.
186 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
187 Initialize n_stmts member.
188 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
189 (vect_create_loop_vinfo): Do not set loop->aux.
190 (vect_analyze_loop_2): Do not get n_stmts as argument,
191 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
193 (vect_analyze_loop_1): Remove callback, get the mode iterator
194 and autodetected_vector_mode as argument, advancing the
195 iterator and initializing autodetected_vector_mode here.
196 (vect_analyze_loop): Split analysis loop into two, first
197 processing main loops only and then epilogues.
199 2021-11-05 Martin Jambor <mjambor@suse.cz>
201 * ipa-prop.c (compute_complex_assign_jump_func): Remove
202 unnecessary check for RECORD_TYPE.
204 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
206 * doc/install.texi2html: Do not generate old.html any longer.
208 2021-11-05 Martin Liska <mliska@suse.cz>
211 * opts.c (finish_options): Reset flag_gtoggle when it is used.
213 2021-11-05 Jakub Jelinek <jakub@redhat.com>
216 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
217 in CONST_WIDE_INT handling. Adjust recursive calls.
218 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
219 new add_const_value_attribute argument.
220 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
221 add_const_value_attribute argument.
223 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
225 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
228 2021-11-05 Richard Biener <rguenther@suse.de>
230 * tree-vectorizer.h (struct vect_loop_form_info): New.
231 (vect_analyze_loop_form): Adjust.
232 (vect_create_loop_vinfo): New.
233 * tree-parloops.c (gather_scalar_reductions): Adjust for
234 vect_analyze_loop_form API change.
235 * tree-vect-loop.c: Include <functional>.
236 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
237 take struct vect_loop_form_info as output parameter and adjust.
238 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
239 split out call to the original vect_analyze_loop_form_1.
240 (vect_reanalyze_as_main_loop): Rename to...
241 (vect_analyze_loop_1): ... this, factor out the call to
242 vect_analyze_loop_form and generalize to be able to use it twice ...
243 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
246 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
249 * config/rs6000/fusion.md: Regenerate.
250 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
252 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
254 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
257 2021-11-04 Tamar Christina <tamar.christina@arm.com>
259 * config/aarch64/aarch64-builtins.c
260 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
261 ashr_simd, lshr, lshr_simd.
262 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
263 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
264 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
266 2021-11-04 Tamar Christina <tamar.christina@arm.com>
268 * match.pd: New negate+shift pattern.
270 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
272 PR tree-optimization/103079
273 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
275 (gimple_range_calc_op2): Ditto.
277 2021-11-04 Martin Jambor <mjambor@suse.cz>
280 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
281 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
282 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
283 parameter to mark_dead_statements.
284 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
285 (ipa_param_body_adjustments::mark_dead_statements): New parameter
286 debugstack, push into it all SSA names used in debug statements,
287 produce m_dead_ssa_debug_equiv mapping for the removed param.
288 (replace_with_mapped_expr): New function.
289 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
290 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
291 (ipa_param_body_adjustments::common_initialization): Gather and
292 procecc SSA which will be removed but are in debug statements. Simplify.
293 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
295 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
296 when avoiding a copy of an unnecessary statement. Remap removed SSA
297 names in existing debug statements.
298 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
299 parameters if we have already done so.
301 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
304 * gimple.c (gimple_call_static_chain_flags): Handle case when
305 nested function does not bind locally.
307 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
309 * config/aarch64/aarch64.c (aarch64_function_value): Generate
310 a register rtx for Neon vector-tuple modes.
311 (aarch64_layout_arg): Likewise.
313 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
315 * lower-subreg.c (simple_move): Prevent decomposition if
316 modes are not tieable.
318 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
319 Richard Sandiford <richard.sandiford@arm.com>
321 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
322 (v2x4hi_UP): Likewise.
323 (v2x4hf_UP): Likewise.
324 (v2x4bf_UP): Likewise.
325 (v2x2si_UP): Likewise.
326 (v2x2sf_UP): Likewise.
327 (v2x1di_UP): Likewise.
328 (v2x1df_UP): Likewise.
329 (v2x16qi_UP): Likewise.
330 (v2x8hi_UP): Likewise.
331 (v2x8hf_UP): Likewise.
332 (v2x8bf_UP): Likewise.
333 (v2x4si_UP): Likewise.
334 (v2x4sf_UP): Likewise.
335 (v2x2di_UP): Likewise.
336 (v2x2df_UP): Likewise.
337 (v3x8qi_UP): Likewise.
338 (v3x4hi_UP): Likewise.
339 (v3x4hf_UP): Likewise.
340 (v3x4bf_UP): Likewise.
341 (v3x2si_UP): Likewise.
342 (v3x2sf_UP): Likewise.
343 (v3x1di_UP): Likewise.
344 (v3x1df_UP): Likewise.
345 (v3x16qi_UP): Likewise.
346 (v3x8hi_UP): Likewise.
347 (v3x8hf_UP): Likewise.
348 (v3x8bf_UP): Likewise.
349 (v3x4si_UP): Likewise.
350 (v3x4sf_UP): Likewise.
351 (v3x2di_UP): Likewise.
352 (v3x2df_UP): Likewise.
353 (v4x8qi_UP): Likewise.
354 (v4x4hi_UP): Likewise.
355 (v4x4hf_UP): Likewise.
356 (v4x4bf_UP): Likewise.
357 (v4x2si_UP): Likewise.
358 (v4x2sf_UP): Likewise.
359 (v4x1di_UP): Likewise.
360 (v4x1df_UP): Likewise.
361 (v4x16qi_UP): Likewise.
362 (v4x8hi_UP): Likewise.
363 (v4x8hf_UP): Likewise.
364 (v4x8bf_UP): Likewise.
365 (v4x4si_UP): Likewise.
366 (v4x4sf_UP): Likewise.
367 (v4x2di_UP): Likewise.
368 (v4x2df_UP): Likewise.
369 (TYPES_GETREGP): Delete.
370 (TYPES_SETREGP): Likewise.
371 (TYPES_LOADSTRUCT_U): Define.
372 (TYPES_LOADSTRUCT_P): Likewise.
373 (TYPES_LOADSTRUCT_LANE_U): Likewise.
374 (TYPES_LOADSTRUCT_LANE_P): Likewise.
375 (TYPES_STORE1P): Move for consistency.
376 (TYPES_STORESTRUCT_U): Define.
377 (TYPES_STORESTRUCT_P): Likewise.
378 (TYPES_STORESTRUCT_LANE_U): Likewise.
379 (TYPES_STORESTRUCT_LANE_P): Likewise.
380 (aarch64_simd_tuple_types): Define.
381 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
382 (aarch64_init_simd_builtin_functions): Update frontend lookup
383 for builtin functions after handling arm_neon.h pragma.
384 (register_tuple_type): Manually set modes of single-integer
385 tuple types. Record tuple types.
386 * config/aarch64/aarch64-modes.def
387 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
388 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
389 (SVE_MODES): Give single-vector modes priority over vector-
391 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
392 be after all single-vector modes.
393 * config/aarch64/aarch64-simd-builtins.def: Update builtin
394 generator macros to reflect modifications to the backend
396 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
397 Use vector-tuple mode iterator and rename to...
398 (aarch64_simd_ld2<vstruct_elt>): This.
399 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
401 (aarch64_simd_ld2r<vstruct_elt>): This.
402 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
403 iterator and rename to...
404 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
405 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
407 (vec_load_lanes<mode><vstruct_elt>): This.
408 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
410 (aarch64_simd_st2<vstruct_elt>): This.
411 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
412 iterator and rename to...
413 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
414 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
416 (vec_store_lanes<mode><vstruct_elt>): This.
417 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
419 (aarch64_simd_ld3<vstruct_elt>): This.
420 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
422 (aarch64_simd_ld3r<vstruct_elt>): This.
423 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
424 iterator and rename to...
425 (vec_load_lanesci<mode>): This.
426 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
428 (aarch64_simd_st3<vstruct_elt>): This.
429 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
430 iterator and rename to...
431 (vec_store_lanesci<mode>): This.
432 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
434 (aarch64_simd_ld4<vstruct_elt>): This.
435 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
437 (aarch64_simd_ld4r<vstruct_elt>): This.
438 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
439 iterator and rename to...
440 (vec_load_lanesxi<mode>): This.
441 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
443 (aarch64_simd_st4<vstruct_elt>): This.
444 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
445 iterator and rename to...
446 (vec_store_lanesxi<mode>): This.
447 (mov<mode>): Define for Neon vector-tuple modes.
448 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
450 (aarch64_ld1x3<vstruct_elt>): This.
451 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
453 (aarch64_ld1_x3_<vstruct_elt>): This.
454 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
456 (aarch64_ld1x4<vstruct_elt>): This.
457 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
459 (aarch64_ld1_x4_<vstruct_elt>): This.
460 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
462 (aarch64_st1x2<vstruct_elt>): This.
463 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
465 (aarch64_st1_x2_<vstruct_elt>): This.
466 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
468 (aarch64_st1x3<vstruct_elt>): This.
469 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
471 (aarch64_st1_x3_<vstruct_elt>): This.
472 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
474 (aarch64_st1x4<vstruct_elt>): This.
475 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
477 (aarch64_st1_x4_<vstruct_elt>): This.
478 (*aarch64_mov<mode>): Define for vector-tuple modes.
479 (*aarch64_be_mov<mode>): Likewise.
480 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
481 mode iterator and rename to...
482 (aarch64_ld<nregs>r<vstruct_elt>): This.
483 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
485 (aarch64_ld2<vstruct_elt>_dreg): This.
486 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
488 (aarch64_ld3<vstruct_elt>_dreg): This.
489 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
491 (aarch64_ld4<vstruct_elt>_dreg): This.
492 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
493 iterator and rename to...
494 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
495 iterator and rename to...
496 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
497 (aarch64_ld1x2<VQ:mode>): Delete.
498 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
500 (aarch64_ld1x2<vstruct_elt>): This.
501 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
502 tuple mode iterator and rename to...
503 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
504 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
505 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
506 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
508 (aarch64_st2<vstruct_elt>_dreg): This.
509 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
511 (aarch64_st3<vstruct_elt>_dreg): This.
512 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
514 (aarch64_st4<vstruct_elt>_dreg): This.
515 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
516 iterator and rename to...
517 (aarch64_st<nregs><vstruct_elt>): This.
518 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
519 iterator and rename to aarch64_st<nregs><vstruct_elt>.
520 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
521 tuple mode iterator and rename to...
522 (aarch64_st<nregs>_lane<vstruct_elt>): This.
523 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
524 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
526 (aarch64_simd_ld1<vstruct_elt>_x2): This.
527 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
528 Refactor to include new vector-tuple modes.
529 (aarch64_classify_vector_mode): Add cases for new vector-
531 (aarch64_advsimd_partial_struct_mode_p): Define.
532 (aarch64_advsimd_full_struct_mode_p): Likewise.
533 (aarch64_advsimd_vector_array_mode): Likewise.
534 (aarch64_sve_data_mode): Change location in file.
535 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
536 (aarch64_hard_regno_nregs): Handle case of partial Neon
538 (aarch64_classify_address): Refactor to include handling of
539 Neon vector-tuple modes.
540 (aarch64_print_operand): Print "d" for "%R" for a partial
541 Neon vector structure.
542 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
543 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
544 modes with scalar machines modes larger than 8 bytes.
545 (aarch64_can_change_mode_class): Don't allow changes between
546 partial and full Neon vector-structure modes.
547 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
548 builtin and remove boiler-plate code for opaque mode.
549 (vst2_lane_f32): Likewise.
550 (vst2_lane_f64): Likewise.
551 (vst2_lane_p8): Likewise.
552 (vst2_lane_p16): Likewise.
553 (vst2_lane_p64): Likewise.
554 (vst2_lane_s8): Likewise.
555 (vst2_lane_s16): Likewise.
556 (vst2_lane_s32): Likewise.
557 (vst2_lane_s64): Likewise.
558 (vst2_lane_u8): Likewise.
559 (vst2_lane_u16): Likewise.
560 (vst2_lane_u32): Likewise.
561 (vst2_lane_u64): Likewise.
562 (vst2q_lane_f16): Likewise.
563 (vst2q_lane_f32): Likewise.
564 (vst2q_lane_f64): Likewise.
565 (vst2q_lane_p8): Likewise.
566 (vst2q_lane_p16): Likewise.
567 (vst2q_lane_p64): Likewise.
568 (vst2q_lane_s8): Likewise.
569 (vst2q_lane_s16): Likewise.
570 (vst2q_lane_s32): Likewise.
571 (vst2q_lane_s64): Likewise.
572 (vst2q_lane_u8): Likewise.
573 (vst2q_lane_u16): Likewise.
574 (vst2q_lane_u32): Likewise.
575 (vst2q_lane_u64): Likewise.
576 (vst3_lane_f16): Likewise.
577 (vst3_lane_f32): Likewise.
578 (vst3_lane_f64): Likewise.
579 (vst3_lane_p8): Likewise.
580 (vst3_lane_p16): Likewise.
581 (vst3_lane_p64): Likewise.
582 (vst3_lane_s8): Likewise.
583 (vst3_lane_s16): Likewise.
584 (vst3_lane_s32): Likewise.
585 (vst3_lane_s64): Likewise.
586 (vst3_lane_u8): Likewise.
587 (vst3_lane_u16): Likewise.
588 (vst3_lane_u32): Likewise.
589 (vst3_lane_u64): Likewise.
590 (vst3q_lane_f16): Likewise.
591 (vst3q_lane_f32): Likewise.
592 (vst3q_lane_f64): Likewise.
593 (vst3q_lane_p8): Likewise.
594 (vst3q_lane_p16): Likewise.
595 (vst3q_lane_p64): Likewise.
596 (vst3q_lane_s8): Likewise.
597 (vst3q_lane_s16): Likewise.
598 (vst3q_lane_s32): Likewise.
599 (vst3q_lane_s64): Likewise.
600 (vst3q_lane_u8): Likewise.
601 (vst3q_lane_u16): Likewise.
602 (vst3q_lane_u32): Likewise.
603 (vst3q_lane_u64): Likewise.
604 (vst4_lane_f16): Likewise.
605 (vst4_lane_f32): Likewise.
606 (vst4_lane_f64): Likewise.
607 (vst4_lane_p8): Likewise.
608 (vst4_lane_p16): Likewise.
609 (vst4_lane_p64): Likewise.
610 (vst4_lane_s8): Likewise.
611 (vst4_lane_s16): Likewise.
612 (vst4_lane_s32): Likewise.
613 (vst4_lane_s64): Likewise.
614 (vst4_lane_u8): Likewise.
615 (vst4_lane_u16): Likewise.
616 (vst4_lane_u32): Likewise.
617 (vst4_lane_u64): Likewise.
618 (vst4q_lane_f16): Likewise.
619 (vst4q_lane_f32): Likewise.
620 (vst4q_lane_f64): Likewise.
621 (vst4q_lane_p8): Likewise.
622 (vst4q_lane_p16): Likewise.
623 (vst4q_lane_p64): Likewise.
624 (vst4q_lane_s8): Likewise.
625 (vst4q_lane_s16): Likewise.
626 (vst4q_lane_s32): Likewise.
627 (vst4q_lane_s64): Likewise.
628 (vst4q_lane_u8): Likewise.
629 (vst4q_lane_u16): Likewise.
630 (vst4q_lane_u32): Likewise.
631 (vst4q_lane_u64): Likewise.
632 (vtbl3_s8): Likewise.
633 (vtbl3_u8): Likewise.
634 (vtbl3_p8): Likewise.
635 (vtbl4_s8): Likewise.
636 (vtbl4_u8): Likewise.
637 (vtbl4_p8): Likewise.
638 (vld1_u8_x3): Likewise.
639 (vld1_s8_x3): Likewise.
640 (vld1_u16_x3): Likewise.
641 (vld1_s16_x3): Likewise.
642 (vld1_u32_x3): Likewise.
643 (vld1_s32_x3): Likewise.
644 (vld1_u64_x3): Likewise.
645 (vld1_s64_x3): Likewise.
646 (vld1_f16_x3): Likewise.
647 (vld1_f32_x3): Likewise.
648 (vld1_f64_x3): Likewise.
649 (vld1_p8_x3): Likewise.
650 (vld1_p16_x3): Likewise.
651 (vld1_p64_x3): Likewise.
652 (vld1q_u8_x3): Likewise.
653 (vld1q_s8_x3): Likewise.
654 (vld1q_u16_x3): Likewise.
655 (vld1q_s16_x3): Likewise.
656 (vld1q_u32_x3): Likewise.
657 (vld1q_s32_x3): Likewise.
658 (vld1q_u64_x3): Likewise.
659 (vld1q_s64_x3): Likewise.
660 (vld1q_f16_x3): Likewise.
661 (vld1q_f32_x3): Likewise.
662 (vld1q_f64_x3): Likewise.
663 (vld1q_p8_x3): Likewise.
664 (vld1q_p16_x3): Likewise.
665 (vld1q_p64_x3): Likewise.
666 (vld1_u8_x2): Likewise.
667 (vld1_s8_x2): Likewise.
668 (vld1_u16_x2): Likewise.
669 (vld1_s16_x2): Likewise.
670 (vld1_u32_x2): Likewise.
671 (vld1_s32_x2): Likewise.
672 (vld1_u64_x2): Likewise.
673 (vld1_s64_x2): Likewise.
674 (vld1_f16_x2): Likewise.
675 (vld1_f32_x2): Likewise.
676 (vld1_f64_x2): Likewise.
677 (vld1_p8_x2): Likewise.
678 (vld1_p16_x2): Likewise.
679 (vld1_p64_x2): Likewise.
680 (vld1q_u8_x2): Likewise.
681 (vld1q_s8_x2): Likewise.
682 (vld1q_u16_x2): Likewise.
683 (vld1q_s16_x2): Likewise.
684 (vld1q_u32_x2): Likewise.
685 (vld1q_s32_x2): Likewise.
686 (vld1q_u64_x2): Likewise.
687 (vld1q_s64_x2): Likewise.
688 (vld1q_f16_x2): Likewise.
689 (vld1q_f32_x2): Likewise.
690 (vld1q_f64_x2): Likewise.
691 (vld1q_p8_x2): Likewise.
692 (vld1q_p16_x2): Likewise.
693 (vld1q_p64_x2): Likewise.
694 (vld1_s8_x4): Likewise.
695 (vld1q_s8_x4): Likewise.
696 (vld1_s16_x4): Likewise.
697 (vld1q_s16_x4): Likewise.
698 (vld1_s32_x4): Likewise.
699 (vld1q_s32_x4): Likewise.
700 (vld1_u8_x4): Likewise.
701 (vld1q_u8_x4): Likewise.
702 (vld1_u16_x4): Likewise.
703 (vld1q_u16_x4): Likewise.
704 (vld1_u32_x4): Likewise.
705 (vld1q_u32_x4): Likewise.
706 (vld1_f16_x4): Likewise.
707 (vld1q_f16_x4): Likewise.
708 (vld1_f32_x4): Likewise.
709 (vld1q_f32_x4): Likewise.
710 (vld1_p8_x4): Likewise.
711 (vld1q_p8_x4): Likewise.
712 (vld1_p16_x4): Likewise.
713 (vld1q_p16_x4): Likewise.
714 (vld1_s64_x4): Likewise.
715 (vld1_u64_x4): Likewise.
716 (vld1_p64_x4): Likewise.
717 (vld1q_s64_x4): Likewise.
718 (vld1q_u64_x4): Likewise.
719 (vld1q_p64_x4): Likewise.
720 (vld1_f64_x4): Likewise.
721 (vld1q_f64_x4): Likewise.
722 (vld2_s64): Likewise.
723 (vld2_u64): Likewise.
724 (vld2_f64): Likewise.
727 (vld2_p64): Likewise.
728 (vld2_s16): Likewise.
729 (vld2_p16): Likewise.
730 (vld2_s32): Likewise.
732 (vld2_u16): Likewise.
733 (vld2_u32): Likewise.
734 (vld2_f16): Likewise.
735 (vld2_f32): Likewise.
736 (vld2q_s8): Likewise.
737 (vld2q_p8): Likewise.
738 (vld2q_s16): Likewise.
739 (vld2q_p16): Likewise.
740 (vld2q_p64): Likewise.
741 (vld2q_s32): Likewise.
742 (vld2q_s64): Likewise.
743 (vld2q_u8): Likewise.
744 (vld2q_u16): Likewise.
745 (vld2q_u32): Likewise.
746 (vld2q_u64): Likewise.
747 (vld2q_f16): Likewise.
748 (vld2q_f32): Likewise.
749 (vld2q_f64): Likewise.
750 (vld3_s64): Likewise.
751 (vld3_u64): Likewise.
752 (vld3_f64): Likewise.
755 (vld3_s16): Likewise.
756 (vld3_p16): Likewise.
757 (vld3_s32): Likewise.
759 (vld3_u16): Likewise.
760 (vld3_u32): Likewise.
761 (vld3_f16): Likewise.
762 (vld3_f32): Likewise.
763 (vld3_p64): Likewise.
764 (vld3q_s8): Likewise.
765 (vld3q_p8): Likewise.
766 (vld3q_s16): Likewise.
767 (vld3q_p16): Likewise.
768 (vld3q_s32): Likewise.
769 (vld3q_s64): Likewise.
770 (vld3q_u8): Likewise.
771 (vld3q_u16): Likewise.
772 (vld3q_u32): Likewise.
773 (vld3q_u64): Likewise.
774 (vld3q_f16): Likewise.
775 (vld3q_f32): Likewise.
776 (vld3q_f64): Likewise.
777 (vld3q_p64): Likewise.
778 (vld4_s64): Likewise.
779 (vld4_u64): Likewise.
780 (vld4_f64): Likewise.
783 (vld4_s16): Likewise.
784 (vld4_p16): Likewise.
785 (vld4_s32): Likewise.
787 (vld4_u16): Likewise.
788 (vld4_u32): Likewise.
789 (vld4_f16): Likewise.
790 (vld4_f32): Likewise.
791 (vld4_p64): Likewise.
792 (vld4q_s8): Likewise.
793 (vld4q_p8): Likewise.
794 (vld4q_s16): Likewise.
795 (vld4q_p16): Likewise.
796 (vld4q_s32): Likewise.
797 (vld4q_s64): Likewise.
798 (vld4q_u8): Likewise.
799 (vld4q_u16): Likewise.
800 (vld4q_u32): Likewise.
801 (vld4q_u64): Likewise.
802 (vld4q_f16): Likewise.
803 (vld4q_f32): Likewise.
804 (vld4q_f64): Likewise.
805 (vld4q_p64): Likewise.
806 (vld2_dup_s8): Likewise.
807 (vld2_dup_s16): Likewise.
808 (vld2_dup_s32): Likewise.
809 (vld2_dup_f16): Likewise.
810 (vld2_dup_f32): Likewise.
811 (vld2_dup_f64): Likewise.
812 (vld2_dup_u8): Likewise.
813 (vld2_dup_u16): Likewise.
814 (vld2_dup_u32): Likewise.
815 (vld2_dup_p8): Likewise.
816 (vld2_dup_p16): Likewise.
817 (vld2_dup_p64): Likewise.
818 (vld2_dup_s64): Likewise.
819 (vld2_dup_u64): Likewise.
820 (vld2q_dup_s8): Likewise.
821 (vld2q_dup_p8): Likewise.
822 (vld2q_dup_s16): Likewise.
823 (vld2q_dup_p16): Likewise.
824 (vld2q_dup_s32): Likewise.
825 (vld2q_dup_s64): Likewise.
826 (vld2q_dup_u8): Likewise.
827 (vld2q_dup_u16): Likewise.
828 (vld2q_dup_u32): Likewise.
829 (vld2q_dup_u64): Likewise.
830 (vld2q_dup_f16): Likewise.
831 (vld2q_dup_f32): Likewise.
832 (vld2q_dup_f64): Likewise.
833 (vld2q_dup_p64): Likewise.
834 (vld3_dup_s64): Likewise.
835 (vld3_dup_u64): Likewise.
836 (vld3_dup_f64): Likewise.
837 (vld3_dup_s8): Likewise.
838 (vld3_dup_p8): Likewise.
839 (vld3_dup_s16): Likewise.
840 (vld3_dup_p16): Likewise.
841 (vld3_dup_s32): Likewise.
842 (vld3_dup_u8): Likewise.
843 (vld3_dup_u16): Likewise.
844 (vld3_dup_u32): Likewise.
845 (vld3_dup_f16): Likewise.
846 (vld3_dup_f32): Likewise.
847 (vld3_dup_p64): Likewise.
848 (vld3q_dup_s8): Likewise.
849 (vld3q_dup_p8): Likewise.
850 (vld3q_dup_s16): Likewise.
851 (vld3q_dup_p16): Likewise.
852 (vld3q_dup_s32): Likewise.
853 (vld3q_dup_s64): Likewise.
854 (vld3q_dup_u8): Likewise.
855 (vld3q_dup_u16): Likewise.
856 (vld3q_dup_u32): Likewise.
857 (vld3q_dup_u64): Likewise.
858 (vld3q_dup_f16): Likewise.
859 (vld3q_dup_f32): Likewise.
860 (vld3q_dup_f64): Likewise.
861 (vld3q_dup_p64): Likewise.
862 (vld4_dup_s64): Likewise.
863 (vld4_dup_u64): Likewise.
864 (vld4_dup_f64): Likewise.
865 (vld4_dup_s8): Likewise.
866 (vld4_dup_p8): Likewise.
867 (vld4_dup_s16): Likewise.
868 (vld4_dup_p16): Likewise.
869 (vld4_dup_s32): Likewise.
870 (vld4_dup_u8): Likewise.
871 (vld4_dup_u16): Likewise.
872 (vld4_dup_u32): Likewise.
873 (vld4_dup_f16): Likewise.
874 (vld4_dup_f32): Likewise.
875 (vld4_dup_p64): Likewise.
876 (vld4q_dup_s8): Likewise.
877 (vld4q_dup_p8): Likewise.
878 (vld4q_dup_s16): Likewise.
879 (vld4q_dup_p16): Likewise.
880 (vld4q_dup_s32): Likewise.
881 (vld4q_dup_s64): Likewise.
882 (vld4q_dup_u8): Likewise.
883 (vld4q_dup_u16): Likewise.
884 (vld4q_dup_u32): Likewise.
885 (vld4q_dup_u64): Likewise.
886 (vld4q_dup_f16): Likewise.
887 (vld4q_dup_f32): Likewise.
888 (vld4q_dup_f64): Likewise.
889 (vld4q_dup_p64): Likewise.
890 (vld2_lane_u8): Likewise.
891 (vld2_lane_u16): Likewise.
892 (vld2_lane_u32): Likewise.
893 (vld2_lane_u64): Likewise.
894 (vld2_lane_s8): Likewise.
895 (vld2_lane_s16): Likewise.
896 (vld2_lane_s32): Likewise.
897 (vld2_lane_s64): Likewise.
898 (vld2_lane_f16): Likewise.
899 (vld2_lane_f32): Likewise.
900 (vld2_lane_f64): Likewise.
901 (vld2_lane_p8): Likewise.
902 (vld2_lane_p16): Likewise.
903 (vld2_lane_p64): Likewise.
904 (vld2q_lane_u8): Likewise.
905 (vld2q_lane_u16): Likewise.
906 (vld2q_lane_u32): Likewise.
907 (vld2q_lane_u64): Likewise.
908 (vld2q_lane_s8): Likewise.
909 (vld2q_lane_s16): Likewise.
910 (vld2q_lane_s32): Likewise.
911 (vld2q_lane_s64): Likewise.
912 (vld2q_lane_f16): Likewise.
913 (vld2q_lane_f32): Likewise.
914 (vld2q_lane_f64): Likewise.
915 (vld2q_lane_p8): Likewise.
916 (vld2q_lane_p16): Likewise.
917 (vld2q_lane_p64): Likewise.
918 (vld3_lane_u8): Likewise.
919 (vld3_lane_u16): Likewise.
920 (vld3_lane_u32): Likewise.
921 (vld3_lane_u64): Likewise.
922 (vld3_lane_s8): Likewise.
923 (vld3_lane_s16): Likewise.
924 (vld3_lane_s32): Likewise.
925 (vld3_lane_s64): Likewise.
926 (vld3_lane_f16): Likewise.
927 (vld3_lane_f32): Likewise.
928 (vld3_lane_f64): Likewise.
929 (vld3_lane_p8): Likewise.
930 (vld3_lane_p16): Likewise.
931 (vld3_lane_p64): Likewise.
932 (vld3q_lane_u8): Likewise.
933 (vld3q_lane_u16): Likewise.
934 (vld3q_lane_u32): Likewise.
935 (vld3q_lane_u64): Likewise.
936 (vld3q_lane_s8): Likewise.
937 (vld3q_lane_s16): Likewise.
938 (vld3q_lane_s32): Likewise.
939 (vld3q_lane_s64): Likewise.
940 (vld3q_lane_f16): Likewise.
941 (vld3q_lane_f32): Likewise.
942 (vld3q_lane_f64): Likewise.
943 (vld3q_lane_p8): Likewise.
944 (vld3q_lane_p16): Likewise.
945 (vld3q_lane_p64): Likewise.
946 (vld4_lane_u8): Likewise.
947 (vld4_lane_u16): Likewise.
948 (vld4_lane_u32): Likewise.
949 (vld4_lane_u64): Likewise.
950 (vld4_lane_s8): Likewise.
951 (vld4_lane_s16): Likewise.
952 (vld4_lane_s32): Likewise.
953 (vld4_lane_s64): Likewise.
954 (vld4_lane_f16): Likewise.
955 (vld4_lane_f32): Likewise.
956 (vld4_lane_f64): Likewise.
957 (vld4_lane_p8): Likewise.
958 (vld4_lane_p16): Likewise.
959 (vld4_lane_p64): Likewise.
960 (vld4q_lane_u8): Likewise.
961 (vld4q_lane_u16): Likewise.
962 (vld4q_lane_u32): Likewise.
963 (vld4q_lane_u64): Likewise.
964 (vld4q_lane_s8): Likewise.
965 (vld4q_lane_s16): Likewise.
966 (vld4q_lane_s32): Likewise.
967 (vld4q_lane_s64): Likewise.
968 (vld4q_lane_f16): Likewise.
969 (vld4q_lane_f32): Likewise.
970 (vld4q_lane_f64): Likewise.
971 (vld4q_lane_p8): Likewise.
972 (vld4q_lane_p16): Likewise.
973 (vld4q_lane_p64): Likewise.
974 (vqtbl2_s8): Likewise.
975 (vqtbl2_u8): Likewise.
976 (vqtbl2_p8): Likewise.
977 (vqtbl2q_s8): Likewise.
978 (vqtbl2q_u8): Likewise.
979 (vqtbl2q_p8): Likewise.
980 (vqtbl3_s8): Likewise.
981 (vqtbl3_u8): Likewise.
982 (vqtbl3_p8): Likewise.
983 (vqtbl3q_s8): Likewise.
984 (vqtbl3q_u8): Likewise.
985 (vqtbl3q_p8): Likewise.
986 (vqtbl4_s8): Likewise.
987 (vqtbl4_u8): Likewise.
988 (vqtbl4_p8): Likewise.
989 (vqtbl4q_s8): Likewise.
990 (vqtbl4q_u8): Likewise.
991 (vqtbl4q_p8): Likewise.
992 (vqtbx2_s8): Likewise.
993 (vqtbx2_u8): Likewise.
994 (vqtbx2_p8): Likewise.
995 (vqtbx2q_s8): Likewise.
996 (vqtbx2q_u8): Likewise.
997 (vqtbx2q_p8): Likewise.
998 (vqtbx3_s8): Likewise.
999 (vqtbx3_u8): Likewise.
1000 (vqtbx3_p8): Likewise.
1001 (vqtbx3q_s8): Likewise.
1002 (vqtbx3q_u8): Likewise.
1003 (vqtbx3q_p8): Likewise.
1004 (vqtbx4_s8): Likewise.
1005 (vqtbx4_u8): Likewise.
1006 (vqtbx4_p8): Likewise.
1007 (vqtbx4q_s8): Likewise.
1008 (vqtbx4q_u8): Likewise.
1009 (vqtbx4q_p8): Likewise.
1010 (vst1_s64_x2): Likewise.
1011 (vst1_u64_x2): Likewise.
1012 (vst1_f64_x2): Likewise.
1013 (vst1_s8_x2): Likewise.
1014 (vst1_p8_x2): Likewise.
1015 (vst1_s16_x2): Likewise.
1016 (vst1_p16_x2): Likewise.
1017 (vst1_s32_x2): Likewise.
1018 (vst1_u8_x2): Likewise.
1019 (vst1_u16_x2): Likewise.
1020 (vst1_u32_x2): Likewise.
1021 (vst1_f16_x2): Likewise.
1022 (vst1_f32_x2): Likewise.
1023 (vst1_p64_x2): Likewise.
1024 (vst1q_s8_x2): Likewise.
1025 (vst1q_p8_x2): Likewise.
1026 (vst1q_s16_x2): Likewise.
1027 (vst1q_p16_x2): Likewise.
1028 (vst1q_s32_x2): Likewise.
1029 (vst1q_s64_x2): Likewise.
1030 (vst1q_u8_x2): Likewise.
1031 (vst1q_u16_x2): Likewise.
1032 (vst1q_u32_x2): Likewise.
1033 (vst1q_u64_x2): Likewise.
1034 (vst1q_f16_x2): Likewise.
1035 (vst1q_f32_x2): Likewise.
1036 (vst1q_f64_x2): Likewise.
1037 (vst1q_p64_x2): Likewise.
1038 (vst1_s64_x3): Likewise.
1039 (vst1_u64_x3): Likewise.
1040 (vst1_f64_x3): Likewise.
1041 (vst1_s8_x3): Likewise.
1042 (vst1_p8_x3): Likewise.
1043 (vst1_s16_x3): Likewise.
1044 (vst1_p16_x3): Likewise.
1045 (vst1_s32_x3): Likewise.
1046 (vst1_u8_x3): Likewise.
1047 (vst1_u16_x3): Likewise.
1048 (vst1_u32_x3): Likewise.
1049 (vst1_f16_x3): Likewise.
1050 (vst1_f32_x3): Likewise.
1051 (vst1_p64_x3): Likewise.
1052 (vst1q_s8_x3): Likewise.
1053 (vst1q_p8_x3): Likewise.
1054 (vst1q_s16_x3): Likewise.
1055 (vst1q_p16_x3): Likewise.
1056 (vst1q_s32_x3): Likewise.
1057 (vst1q_s64_x3): Likewise.
1058 (vst1q_u8_x3): Likewise.
1059 (vst1q_u16_x3): Likewise.
1060 (vst1q_u32_x3): Likewise.
1061 (vst1q_u64_x3): Likewise.
1062 (vst1q_f16_x3): Likewise.
1063 (vst1q_f32_x3): Likewise.
1064 (vst1q_f64_x3): Likewise.
1065 (vst1q_p64_x3): Likewise.
1066 (vst1_s8_x4): Likewise.
1067 (vst1q_s8_x4): Likewise.
1068 (vst1_s16_x4): Likewise.
1069 (vst1q_s16_x4): Likewise.
1070 (vst1_s32_x4): Likewise.
1071 (vst1q_s32_x4): Likewise.
1072 (vst1_u8_x4): Likewise.
1073 (vst1q_u8_x4): Likewise.
1074 (vst1_u16_x4): Likewise.
1075 (vst1q_u16_x4): Likewise.
1076 (vst1_u32_x4): Likewise.
1077 (vst1q_u32_x4): Likewise.
1078 (vst1_f16_x4): Likewise.
1079 (vst1q_f16_x4): Likewise.
1080 (vst1_f32_x4): Likewise.
1081 (vst1q_f32_x4): Likewise.
1082 (vst1_p8_x4): Likewise.
1083 (vst1q_p8_x4): Likewise.
1084 (vst1_p16_x4): Likewise.
1085 (vst1q_p16_x4): Likewise.
1086 (vst1_s64_x4): Likewise.
1087 (vst1_u64_x4): Likewise.
1088 (vst1_p64_x4): Likewise.
1089 (vst1q_s64_x4): Likewise.
1090 (vst1q_u64_x4): Likewise.
1091 (vst1q_p64_x4): Likewise.
1092 (vst1_f64_x4): Likewise.
1093 (vst1q_f64_x4): Likewise.
1094 (vst2_s64): Likewise.
1095 (vst2_u64): Likewise.
1096 (vst2_f64): Likewise.
1097 (vst2_s8): Likewise.
1098 (vst2_p8): Likewise.
1099 (vst2_s16): Likewise.
1100 (vst2_p16): Likewise.
1101 (vst2_s32): Likewise.
1102 (vst2_u8): Likewise.
1103 (vst2_u16): Likewise.
1104 (vst2_u32): Likewise.
1105 (vst2_f16): Likewise.
1106 (vst2_f32): Likewise.
1107 (vst2_p64): Likewise.
1108 (vst2q_s8): Likewise.
1109 (vst2q_p8): Likewise.
1110 (vst2q_s16): Likewise.
1111 (vst2q_p16): Likewise.
1112 (vst2q_s32): Likewise.
1113 (vst2q_s64): Likewise.
1114 (vst2q_u8): Likewise.
1115 (vst2q_u16): Likewise.
1116 (vst2q_u32): Likewise.
1117 (vst2q_u64): Likewise.
1118 (vst2q_f16): Likewise.
1119 (vst2q_f32): Likewise.
1120 (vst2q_f64): Likewise.
1121 (vst2q_p64): Likewise.
1122 (vst3_s64): Likewise.
1123 (vst3_u64): Likewise.
1124 (vst3_f64): Likewise.
1125 (vst3_s8): Likewise.
1126 (vst3_p8): Likewise.
1127 (vst3_s16): Likewise.
1128 (vst3_p16): Likewise.
1129 (vst3_s32): Likewise.
1130 (vst3_u8): Likewise.
1131 (vst3_u16): Likewise.
1132 (vst3_u32): Likewise.
1133 (vst3_f16): Likewise.
1134 (vst3_f32): Likewise.
1135 (vst3_p64): Likewise.
1136 (vst3q_s8): Likewise.
1137 (vst3q_p8): Likewise.
1138 (vst3q_s16): Likewise.
1139 (vst3q_p16): Likewise.
1140 (vst3q_s32): Likewise.
1141 (vst3q_s64): Likewise.
1142 (vst3q_u8): Likewise.
1143 (vst3q_u16): Likewise.
1144 (vst3q_u32): Likewise.
1145 (vst3q_u64): Likewise.
1146 (vst3q_f16): Likewise.
1147 (vst3q_f32): Likewise.
1148 (vst3q_f64): Likewise.
1149 (vst3q_p64): Likewise.
1150 (vst4_s64): Likewise.
1151 (vst4_u64): Likewise.
1152 (vst4_f64): Likewise.
1153 (vst4_s8): Likewise.
1154 (vst4_p8): Likewise.
1155 (vst4_s16): Likewise.
1156 (vst4_p16): Likewise.
1157 (vst4_s32): Likewise.
1158 (vst4_u8): Likewise.
1159 (vst4_u16): Likewise.
1160 (vst4_u32): Likewise.
1161 (vst4_f16): Likewise.
1162 (vst4_f32): Likewise.
1163 (vst4_p64): Likewise.
1164 (vst4q_s8): Likewise.
1165 (vst4q_p8): Likewise.
1166 (vst4q_s16): Likewise.
1167 (vst4q_p16): Likewise.
1168 (vst4q_s32): Likewise.
1169 (vst4q_s64): Likewise.
1170 (vst4q_u8): Likewise.
1171 (vst4q_u16): Likewise.
1172 (vst4q_u32): Likewise.
1173 (vst4q_u64): Likewise.
1174 (vst4q_f16): Likewise.
1175 (vst4q_f32): Likewise.
1176 (vst4q_f64): Likewise.
1177 (vst4q_p64): Likewise.
1178 (vtbx4_s8): Likewise.
1179 (vtbx4_u8): Likewise.
1180 (vtbx4_p8): Likewise.
1181 (vld1_bf16_x2): Likewise.
1182 (vld1q_bf16_x2): Likewise.
1183 (vld1_bf16_x3): Likewise.
1184 (vld1q_bf16_x3): Likewise.
1185 (vld1_bf16_x4): Likewise.
1186 (vld1q_bf16_x4): Likewise.
1187 (vld2_bf16): Likewise.
1188 (vld2q_bf16): Likewise.
1189 (vld2_dup_bf16): Likewise.
1190 (vld2q_dup_bf16): Likewise.
1191 (vld3_bf16): Likewise.
1192 (vld3q_bf16): Likewise.
1193 (vld3_dup_bf16): Likewise.
1194 (vld3q_dup_bf16): Likewise.
1195 (vld4_bf16): Likewise.
1196 (vld4q_bf16): Likewise.
1197 (vld4_dup_bf16): Likewise.
1198 (vld4q_dup_bf16): Likewise.
1199 (vst1_bf16_x2): Likewise.
1200 (vst1q_bf16_x2): Likewise.
1201 (vst1_bf16_x3): Likewise.
1202 (vst1q_bf16_x3): Likewise.
1203 (vst1_bf16_x4): Likewise.
1204 (vst1q_bf16_x4): Likewise.
1205 (vst2_bf16): Likewise.
1206 (vst2q_bf16): Likewise.
1207 (vst3_bf16): Likewise.
1208 (vst3q_bf16): Likewise.
1209 (vst4_bf16): Likewise.
1210 (vst4q_bf16): Likewise.
1211 (vld2_lane_bf16): Likewise.
1212 (vld2q_lane_bf16): Likewise.
1213 (vld3_lane_bf16): Likewise.
1214 (vld3q_lane_bf16): Likewise.
1215 (vld4_lane_bf16): Likewise.
1216 (vld4q_lane_bf16): Likewise.
1217 (vst2_lane_bf16): Likewise.
1218 (vst2q_lane_bf16): Likewise.
1219 (vst3_lane_bf16): Likewise.
1220 (vst3q_lane_bf16): Likewise.
1221 (vst4_lane_bf16): Likewise.
1222 (vst4q_lane_bf16): Likewise.
1223 * config/aarch64/geniterators.sh: Modify iterator regex to
1224 match new vector-tuple modes.
1225 * config/aarch64/iterators.md (insn_count): Extend mode
1226 attribute with vector-tuple type information.
1228 (Vendreg): Likewise.
1231 (VSTRUCT_2D): New mode iterator.
1232 (VSTRUCT_2DNX): Likewise.
1233 (VSTRUCT_2DX): Likewise.
1234 (VSTRUCT_2Q): Likewise.
1235 (VSTRUCT_2QD): Likewise.
1236 (VSTRUCT_3D): Likewise.
1237 (VSTRUCT_3DNX): Likewise.
1238 (VSTRUCT_3DX): Likewise.
1239 (VSTRUCT_3Q): Likewise.
1240 (VSTRUCT_3QD): Likewise.
1241 (VSTRUCT_4D): Likewise.
1242 (VSTRUCT_4DNX): Likewise.
1243 (VSTRUCT_4DX): Likewise.
1244 (VSTRUCT_4Q): Likewise.
1245 (VSTRUCT_4QD): Likewise.
1246 (VSTRUCT_D): Likewise.
1247 (VSTRUCT_Q): Likewise.
1248 (VSTRUCT_QD): Likewise.
1249 (VSTRUCT_ELT): New mode attribute.
1250 (vstruct_elt): Likewise.
1251 * genmodes.c (VECTOR_MODE): Add default prefix and order
1253 (VECTOR_MODE_WITH_PREFIX): Define.
1254 (make_vector_mode): Add mode prefix and order parameters.
1256 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1258 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
1260 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1262 * expr.c (emit_group_load_1): Remove historic workaround.
1264 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1266 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
1267 Factor out main loop to...
1268 (aarch64_init_simd_builtin_functions): This new function.
1269 (register_tuple_type): Define.
1270 (aarch64_scalar_builtin_type_p): Define.
1271 (handle_arm_neon_h): Define.
1272 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
1273 pragma for arm_neon.h.
1274 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
1276 (handle_arm_neon_h): Likewise.
1277 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
1278 Remove static modifier.
1279 * config/aarch64/arm_neon.h (target): Remove Neon vector
1280 structure type definitions.
1282 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1284 PR tree-optimization/102943
1285 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1286 Prefer range_of_expr unless there are no statements in the BB.
1288 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1290 PR tree-optimization/102943
1291 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
1292 Avoid duplicate calculation of paths.
1294 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1296 PR tree-optimization/102943
1297 * gimple-range-path.cc (path_range_query::compute_phi_relations):
1298 Only compute relations for SSA names in the import list.
1299 (path_range_query::compute_outgoing_relations): Same.
1300 * gimple-range-path.h (path_range_query::import_p): New.
1302 2021-11-04 Richard Biener <rguenther@suse.de>
1304 PR rtl-optimization/103075
1305 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
1306 false for a VOIDmode operand.
1308 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1310 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
1311 variables private and add "m_" to their names. Remove is_loop.
1312 (aarch64_record_potential_advsimd_unrolling): Replace with...
1313 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
1314 (aarch64_analyze_loop_vinfo): Replace with...
1315 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
1316 Move initialization of (m_)vec_flags to add_stmt_cost.
1317 (aarch64_analyze_bb_vinfo): Delete.
1318 (aarch64_count_ops): Replace with...
1319 (aarch64_vector_costs::count_ops): ...this.
1320 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
1321 using m_costing_for_scalar to test whether we're costing
1322 scalar or vector code.
1323 (aarch64_adjust_body_cost_sve): Replace with...
1324 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
1325 (aarch64_adjust_body_cost): Replace with...
1326 (aarch64_vector_costs::adjust_body_cost): ...this.
1327 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
1329 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1331 * target.def (targetm.vectorize.init_cost): Replace with...
1332 (targetm.vectorize.create_costs): ...this.
1333 (targetm.vectorize.add_stmt_cost): Delete.
1334 (targetm.vectorize.finish_cost): Likewise.
1335 (targetm.vectorize.destroy_cost_data): Likewise.
1336 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
1337 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1338 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1339 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1340 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1341 * doc/tm.texi: Regenerate.
1342 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
1344 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
1345 (vector_costs): New class.
1346 (init_cost): Take a vec_info and return a vector_costs.
1347 (dump_stmt_cost): Remove data parameter.
1348 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
1349 (add_stmt_costs): Likewise.
1350 (finish_cost): Replace data parameter with a vector_costs.
1351 (destroy_cost_data): Delete.
1352 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
1354 (vec_info::vec_info): Remove the target_cost_data parameter and
1355 initialize the member variable to null instead.
1356 (vec_info::~vec_info): Delete target_cost_data instead of calling
1358 (vector_costs::add_stmt_cost): New function.
1359 (vector_costs::finish_cost): Likewise.
1360 (vector_costs::record_stmt_cost): Likewise.
1361 (vector_costs::adjust_cost_for_freq): Likewise.
1362 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
1363 call to vec_info::vec_info.
1364 (vect_compute_single_scalar_iteration_cost): Update after above
1365 changes to costing interface.
1366 (vect_analyze_loop_operations): Likewise.
1367 (vect_estimate_min_profitable_iters): Likewise.
1368 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
1369 at the start_over point, where it needs to be recreated after
1370 trying without slp. Update retry code accordingly.
1371 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
1372 to vec_info::vec_info.
1373 (vect_slp_analyze_operation): Update after above changes to costing
1375 (vect_bb_vectorization_profitable_p): Likewise.
1376 * targhooks.h (default_init_cost): Replace with...
1377 (default_vectorize_create_costs): ...this.
1378 (default_add_stmt_cost): Delete.
1379 (default_finish_cost, default_destroy_cost_data): Likewise.
1380 * targhooks.c (default_init_cost): Replace with...
1381 (default_vectorize_create_costs): ...this.
1382 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
1383 (default_finish_cost, default_destroy_cost_data): Delete.
1384 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
1385 vector_costs. Add a constructor.
1386 (aarch64_init_cost): Replace with...
1387 (aarch64_vectorize_create_costs): ...this.
1388 (aarch64_add_stmt_cost): Replace with...
1389 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
1390 to adjust the cost for inner loops.
1391 (aarch64_finish_cost): Replace with...
1392 (aarch64_vector_costs::finish_cost): ...this.
1393 (aarch64_destroy_cost_data): Delete.
1394 (TARGET_VECTORIZE_INIT_COST): Replace with...
1395 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1396 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1397 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1398 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1399 * config/i386/i386.c (ix86_vector_costs): New structure.
1400 (ix86_init_cost): Replace with...
1401 (ix86_vectorize_create_costs): ...this.
1402 (ix86_add_stmt_cost): Replace with...
1403 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
1404 to adjust the cost for inner loops.
1405 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
1406 (TARGET_VECTORIZE_INIT_COST): Replace with...
1407 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1408 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1409 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1410 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1411 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
1412 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1413 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1414 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1415 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1416 (rs6000_cost_data): Inherit from vector_costs.
1417 Add a constructor. Drop loop_info, cost and costing_for_scalar
1418 in favor of the corresponding vector_costs member variables.
1419 Add "m_" to the names of the remaining member variables and
1421 (rs6000_density_test): Replace with...
1422 (rs6000_cost_data::density_test): ...this.
1423 (rs6000_init_cost): Replace with...
1424 (rs6000_vectorize_create_costs): ...this.
1425 (rs6000_update_target_cost_per_stmt): Replace with...
1426 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
1427 (rs6000_add_stmt_cost): Replace with...
1428 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
1429 to adjust the cost for inner loops.
1430 (rs6000_adjust_vect_cost_per_loop): Replace with...
1431 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
1432 (rs6000_finish_cost): Replace with...
1433 (rs6000_cost_data::finish_cost): ...this. Group loop code
1434 into a single if statement and pass the loop_vinfo down to
1436 (rs6000_destroy_cost_data): Delete.
1438 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1440 PR tree-optimization/103062
1441 PR tree-optimization/103062
1442 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
1443 Increase size of allocation by 1.
1444 (ssa_equiv_stack::push_replacement): Grow as needed.
1445 (ssa_equiv_stack::get_replacement): Same.
1446 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
1447 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
1448 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
1449 (pointer_equiv_analyzer::get_equiv): Same.
1450 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
1451 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
1452 const markers. Use auto_vec instead of tree *.
1454 2021-11-04 Richard Biener <rguenther@suse.de>
1456 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
1457 true parameter and inline valueization.
1458 (vn_nary_op_lookup_1): Inline valueization from ...
1459 (vn_nary_op_compute_hash): ... here and remove it here.
1460 * tree-ssa-pre.c (phi_translate_1): Do not valueize
1461 before vn_nary_lookup_pieces.
1462 (get_representative_for): Mark created SSA representatives
1465 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1467 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
1468 that the operand has a vector mode. Use subreg_lowpart_offset
1469 to test whether an index corresponds to the low part.
1471 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1473 * read-rtl.c: Remove dead !GENERATOR_FILE block.
1474 * read-rtl-function.c (function_reader::consolidate_singletons):
1475 Generate canonical CONST_VECTORs.
1477 2021-11-04 liuhongt <hongtao.liu@intel.com>
1480 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
1481 (regmem_or_bitnot_regmem_operand): .. and extend to handle
1483 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
1484 the operands which are required to be register_operand.
1485 (*<avx512>_vpternlog<mode>_2): Ditto.
1486 (*<avx512>_vpternlog<mode>_3): Ditto.
1487 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
1488 vector HFmodes since it's not a real AVX512FP16 instruction.
1490 2021-11-04 liuhongt <hongtao.liu@intel.com>
1493 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
1494 .COPYSIGN (a,b) when a and b are same type as the truncation
1495 type and has less precision than extend type.
1497 2021-11-04 Richard Biener <rguenther@suse.de>
1499 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
1502 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1504 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
1505 (VI1_AVX512VNNI): Likewise.
1506 (SDOT_VPDP_SUF): New mode_attr.
1509 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
1510 vpdpwssd when VNNI targets available.
1511 (usdot_prod<mode>): New expander for vector QImode.
1513 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1515 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
1516 parentheses to base and stride.
1517 (_tile_stream_loadd_internal): Likewise.
1518 (_tile_stored_internal): Likewise.
1520 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
1522 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
1523 arguments to `reg_class_subset_p'.
1525 2021-11-03 Joseph Myers <joseph@codesourcery.com>
1528 * fold-const.c (fold_init): New function.
1529 * fold-const.h (fold_init): New prototype.
1531 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1533 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
1535 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1537 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
1538 a range from edge if arg == phidef.
1540 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1542 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
1543 for constant before any other processing.
1545 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1547 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
1549 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1551 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
1552 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
1553 no explcit context and current_bb is set.
1554 (gimple_ranger::fold_stmt): New.
1555 * gimple-range.h (current_bb, fold_stmt): New.
1556 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
1558 2021-11-03 Richard Biener <rguenther@suse.de>
1560 PR tree-optimization/102970
1561 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
1562 when translating a MEM_REF over a backedge.
1564 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
1566 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
1567 * config/aarch64/aarch64-tune.md: Regenerate.
1568 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
1569 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
1570 * doc/invoke.texi: Add documentation for Ampere-1 core.
1572 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
1574 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
1576 (ldr_got_small_<mode>): Remove pattern.
1577 (ldr_got_small_sidi): Likewise.
1578 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
1579 GOT accesses as moves.
1580 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
1581 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
1582 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
1584 2021-11-03 Martin Liska <mliska@suse.cz>
1586 * gcov.c (read_line): Remove dead variable.
1588 2021-11-03 Martin Liska <mliska@suse.cz>
1591 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
1592 (ipa_fn_summary::account_size_time): Likewise.
1593 (edge_set_predicate): Likewise.
1594 (set_hint_predicate): Likewise.
1595 (add_freqcounting_predicate): Likewise.
1596 (evaluate_conditions_for_known_args): Likewise.
1597 (evaluate_properties_for_edge): Likewise.
1598 (remap_freqcounting_preds_after_dup): Likewise.
1599 (ipa_fn_summary_t::duplicate): Likewise.
1600 (set_cond_stmt_execution_predicate): Likewise.
1601 (set_switch_stmt_execution_predicate): Likewise.
1602 (compute_bb_predicates): Likewise.
1603 (will_be_nonconstant_expr_predicate): Likewise.
1604 (will_be_nonconstant_predicate): Likewise.
1605 (phi_result_unknown_predicate): Likewise.
1606 (predicate_for_phi_result): Likewise.
1607 (analyze_function_body): Likewise.
1608 (compute_fn_summary): Likewise.
1609 (summarize_calls_size_and_time): Likewise.
1610 (estimate_calls_size_and_time): Likewise.
1611 (ipa_call_context::estimate_size_and_time): Likewise.
1612 (remap_edge_summaries): Likewise.
1613 (remap_freqcounting_predicate): Likewise.
1614 (ipa_merge_fn_summary_after_inlining): Likewise.
1615 (ipa_update_overall_fn_summary): Likewise.
1616 (read_ipa_call_summary): Likewise.
1617 (inline_read_section): Likewise.
1618 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
1619 * ipa-predicate.c (predicate::add_clause): Likewise.
1620 (ipa_predicate::add_clause): Likewise.
1621 (predicate::or_with): Likewise.
1622 (ipa_predicate::or_with): Likewise.
1623 (predicate::evaluate): Likewise.
1624 (ipa_predicate::evaluate): Likewise.
1625 (predicate::probability): Likewise.
1626 (ipa_predicate::probability): Likewise.
1627 (dump_condition): Likewise.
1628 (dump_clause): Likewise.
1629 (predicate::dump): Likewise.
1630 (ipa_predicate::dump): Likewise.
1631 (predicate::debug): Likewise.
1632 (ipa_predicate::debug): Likewise.
1633 (predicate::remap_after_duplication): Likewise.
1634 (ipa_predicate::remap_after_duplication): Likewise.
1635 (predicate::remap_after_inlining): Likewise.
1636 (ipa_predicate::remap_after_inlining): Likewise.
1637 (predicate::stream_in): Likewise.
1638 (ipa_predicate::stream_in): Likewise.
1639 (predicate::stream_out): Likewise.
1640 (ipa_predicate::stream_out): Likewise.
1641 (add_condition): Likewise.
1642 * ipa-predicate.h (class predicate): Likewise.
1643 (class ipa_predicate): Likewise.
1644 (add_condition): Likewise.
1646 2021-11-03 Richard Biener <rguenther@suse.de>
1648 * bitmap.h (bitmap_bit_p): Change the return type to bool.
1649 * bitmap.c (bitmap_bit_p): Likewise.
1650 * sbitmap.h (bitmap_bit_p): Likewise.
1651 (bitmap_set_bit): Return whether the bit changed.
1652 (bitmap_clear_bit): Likewise.
1653 * tree-ssa.c (verify_vssa): Make use of the changed state
1654 from bitmap_set_bit.
1656 2021-11-03 Richard Biener <rguenther@suse.de>
1658 PR middle-end/103033
1659 * internal-fn.c (expand_DEFERRED_INIT): Elide the
1660 native_interpret_expr path in favor of folding the
1661 VIEW_CONVERT_EXPR generated when punning the RHS.
1663 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1665 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
1668 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
1671 * ipa-modref.c (callee_to_caller_flags): New function.
1672 (modref_eaf_analysis::analyze_ssa_name): Use it.
1673 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
1675 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1677 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
1680 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1681 Uroš Bizjak <ubizjak@gmail.com>
1683 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
1684 rotations by non-constant amounts.
1686 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1688 * ipa-modref.c: Fix anonymous namespace placement.
1689 (class modref_eaf_analysis): New class.
1690 (analyze_ssa_name_flags): Turn to ...
1691 (modref_eaf_analysis::analyze_ssa_name): ... this one.
1692 (merge_call_lhs_flags): Turn to ...
1693 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
1694 (modref_eaf_analysis::merge_with_ssa_name): New member function.
1695 (record_escape_points): Turn to ...
1696 (modref_eaf_analysis::record_escape_points): ... this one.
1697 (analyze_parms): Updat
1698 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
1700 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1702 * gimple.c (gimple_call_static_chain_flags): New function.
1703 * gimple.h (gimple_call_static_chain_flags): Declare
1704 * ipa-modref.c (modref_summary::modref_summary): Initialize
1706 (modref_summary_lto::modref_summary_lto): Likewise.
1707 (modref_summary::useful_p): Test static_chain_flags.
1708 (modref_summary_lto::useful_p): Likewise.
1709 (struct modref_summary_lto): Add static_chain_flags.
1710 (modref_summary::dump): Dump static_chain_flags.
1711 (modref_summary_lto::dump): Likewise.
1712 (struct escape_point): Add static_cahin_arg.
1713 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
1714 (analyze_parms): Handle static chains.
1715 (modref_summaries::duplicate): Duplicate static_chain_flags.
1716 (modref_summaries_lto::duplicate): Likewise.
1717 (modref_write): Stream static_chain_flags.
1718 (read_section): Likewise.
1719 (modref_merge_call_site_flags): Handle static_chain_flags.
1720 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
1721 * tree-ssa-structalias.c (handle_rhs_call): Use
1722 gimple_static_chain_flags.
1724 2021-11-02 Richard Biener <rguenther@suse.de>
1726 PR tree-optimization/103029
1727 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
1728 the PHI nodes in the loop maintain their original operand
1731 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1733 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
1734 (EAF_NOREAD): Renumber.
1735 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
1736 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
1737 (deref_flags): Likewise.
1738 (modref_lattice::init): Likewise.
1739 (modref_lattice::merge): Likewise.
1740 (merge_call_lhs_flags): Likewise.
1741 (analyze_ssa_name_flags): Likewise.
1742 (modref_merge_call_site_flags): Likewise.
1743 * tree-ssa-structalias.c (handle_call_arg): Likewise.
1745 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
1747 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
1748 check and remove an unused local variable with shNadd/shNadd.uw
1751 2021-11-02 Tamar Christina <tamar.christina@arm.com>
1753 PR tree-optimization/103007
1754 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
1757 2021-11-02 Richard Biener <rguenther@suse.de>
1759 PR middle-end/103038
1760 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
1761 (can_native_interpret_type_p): Likewise.
1762 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
1763 RHS if the LHS is an SSA name.
1765 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
1767 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
1768 * langhooks-def.h (lhd_simulate_record_decl): Declare.
1769 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
1770 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
1771 * langhooks.c (lhd_simulate_record_decl): New function.
1773 2021-11-02 Jakub Jelinek <jakub@redhat.com>
1776 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
1777 modes with TImode inner mode if 32-bit.
1779 2021-11-02 liuhongt <hongtao.liu@intel.com>
1781 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
1782 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
1783 vect_slp_v4si_store_unalign): Document efficient target.
1784 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
1785 vect_slp_v16qi_store_unalign_1): Ditto.
1786 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
1787 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
1788 (struct_4char_block_move, struct_8char_block_move,
1789 struct_16char_block_move): Ditto.
1791 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1792 Jakub Jelinek <jakub@redhat.com>
1795 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
1796 ix86_expand_ti_to_v1ti): New helper functions.
1797 (ix86_expand_v1ti_shift): Check if the amount operand is an
1798 integer constant, and expand as a TImode shift if it isn't.
1799 (ix86_expand_v1ti_rotate): Check if the amount operand is an
1800 integer constant, and expand as a TImode rotate if it isn't.
1801 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
1802 right shifts of V1TImode quantities.
1803 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
1804 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
1805 to QImode general_operand, and let the helper functions lower
1806 shifts by non-constant operands, as TImode shifts. Make
1807 conditional on TARGET_64BIT.
1808 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
1809 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
1810 Make conditional on TARGET_64BIT.
1812 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1814 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
1815 "*tabort_1_plus"): Remove operands non-null check.
1817 2021-11-01 Martin Liska <mliska@suse.cz>
1819 * opt-functions.awk: Add new sanity checking.
1820 * optc-gen.awk: Add new argument to integer_range_info.
1821 * params.opt: Update 2 params which have negative IntegerRange.
1823 2021-11-01 qing zhao <qing.zhao@oracle.com>
1825 * gimplify.c (gimplify_decl_expr): Do not add call to
1826 __builtin_clear_padding when a variable is a gimple register
1827 or it might not have padding.
1828 (gimplify_init_constructor): Likewise.
1830 2021-11-01 Tamar Christina <tamar.christina@arm.com>
1832 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
1833 movi, dup and extract costing fields.
1834 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
1835 thunderx_extra_costs, thunderx2t99_extra_costs,
1836 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
1838 * config/arm/aarch-cost-tables.h (generic_extra_costs,
1839 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
1840 exynosm1_extra_costs, xgene1_extra_costs): Likewise
1841 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
1842 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
1844 2021-11-01 Tamar Christina <tamar.christina@arm.com>
1846 * cse.c (add_to_set): New.
1847 (find_sets_in_insn): Register constants in sets.
1848 (canonicalize_insn): Use auto_vec instead.
1849 (cse_insn): Try materializing using vec_dup.
1850 * rtl.h (simplify_context::simplify_gen_vec_select,
1851 simplify_gen_vec_select): New.
1852 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
1854 2021-11-01 David Malcolm <dmalcolm@redhat.com>
1856 * common.opt (fdiagnostics-escape-format=): New.
1857 (diagnostics_escape_format): New enum.
1858 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
1859 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
1860 * diagnostic-format-json.cc (json_end_diagnostic): Add
1861 "escape-source" attribute.
1862 * diagnostic-show-locus.c
1863 (exploc_with_display_col::exploc_with_display_col): Replace
1864 "tabstop" param with a cpp_char_column_policy and add an "aspect"
1865 param. Use these to compute m_display_col accordingly.
1866 (struct char_display_policy): New struct.
1867 (layout::m_policy): New field.
1868 (layout::m_escape_on_output): New field.
1869 (def_policy): New function.
1870 (make_range): Update for changes to exploc_with_display_col ctor.
1871 (default_print_decoded_ch): New.
1872 (width_per_escaped_byte): New.
1873 (escape_as_bytes_width): New.
1874 (escape_as_bytes_print): New.
1875 (escape_as_unicode_width): New.
1876 (escape_as_unicode_print): New.
1878 (layout::layout): Initialize new fields. Update m_exploc ctor
1879 call for above change to ctor.
1880 (layout::maybe_add_location_range): Update for changes to
1881 exploc_with_display_col ctor.
1882 (layout::calculate_x_offset_display): Update for change to
1884 (layout::print_source_line): Pass policy
1885 to cpp_display_width_computation. Capture cpp_decoded_char when
1886 calling process_next_codepoint. Move printing of source code to
1887 m_policy.m_print_cb.
1888 (line_label::line_label): Pass in policy rather than context.
1889 (layout::print_any_labels): Update for change to line_label ctor.
1890 (get_affected_range): Pass in policy rather than context, updating
1891 calls to location_compute_display_column accordingly.
1892 (get_printed_columns): Likewise, also for cpp_display_width.
1893 (correction::correction): Pass in policy rather than tabstop.
1894 (correction::compute_display_cols): Pass m_policy rather than
1895 m_tabstop to cpp_display_width.
1896 (correction::m_tabstop): Replace with...
1897 (correction::m_policy): ...this.
1898 (line_corrections::line_corrections): Pass in policy rather than
1900 (line_corrections::m_context): Replace with...
1901 (line_corrections::m_policy): ...this.
1902 (line_corrections::add_hint): Update to use m_policy rather than
1904 (line_corrections::add_hint): Likewise.
1905 (layout::print_trailing_fixits): Likewise.
1906 (selftest::test_display_widths): New.
1907 (selftest::test_layout_x_offset_display_utf8): Update to use
1908 policy rather than tabstop.
1909 (selftest::test_one_liner_labels_utf8): Add test of escaping
1911 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
1912 use policy rather than tabstop.
1913 (selftest::test_overlapped_fixit_printing): Likewise.
1914 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
1915 (selftest::test_overlapped_fixit_printing_2): Likewise.
1916 (selftest::test_tab_expansion): Likewise.
1917 (selftest::test_escaping_bytes_1): New.
1918 (selftest::test_escaping_bytes_2): New.
1919 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
1920 * diagnostic.c (diagnostic_initialize): Initialize
1921 context->escape_format.
1922 (convert_column_unit): Update to use default character width policy.
1923 (selftest::test_diagnostic_get_location_text): Likewise.
1924 * diagnostic.h (enum diagnostics_escape_format): New enum.
1925 (diagnostic_context::escape_format): New field.
1926 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
1927 (-fdiagnostics-format=): Add "escape-source" attribute to examples
1928 of JSON output, and document it.
1929 * input.c (location_compute_display_column): Pass in "policy"
1930 rather than "tabstop", passing to
1931 cpp_byte_column_to_display_column.
1932 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
1933 * input.h (class cpp_char_column_policy): New forward decl.
1934 (location_compute_display_column): Pass in "policy" rather than
1936 * opts.c (common_handle_option): Handle
1937 OPT_fdiagnostics_escape_format_.
1938 * selftest.c (temp_source_file::temp_source_file): New ctor
1939 overload taking a size_t.
1940 * selftest.h (temp_source_file::temp_source_file): Likewise.
1942 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
1944 * dbgcnt.def: Add debug counter for back_thread[12] and
1945 back_threadfull[12].
1946 * passes.def: Pass "first" argument to each back threading pass.
1947 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
1949 (back_threader::debug_counter): New.
1950 (back_threader::maybe_register_path): Call debug_counter.
1952 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
1954 * tree-ssa-threadbackward.c (BT_NONE): New.
1957 (back_threader::back_threader): Add flags.
1958 Move loop initialization here.
1959 (back_threader::~back_threader): New.
1960 (back_threader::find_taken_edge_switch): Change solver and ranger
1962 (back_threader::find_taken_edge_cond): Same.
1963 (back_threader::find_paths_to_names): Same.
1964 (back_threader::find_paths): Same.
1965 (back_threader::dump): Same.
1966 (try_thread_blocks): Merge into thread_blocks.
1967 (back_threader::thread_blocks): New.
1968 (do_early_thread_jumps): Merge into thread_blocks.
1969 (do_thread_jumps): Merge into thread_blocks.
1970 (back_threader::thread_through_all_blocks): Remove.
1972 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
1974 PR tree-optimization/103003
1975 * value-relation.cc (dom_oracle::register_relation): If the 2
1976 ssa names are the same, don't register any relation.
1978 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
1980 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
1981 redundant check for calls_eh_return.
1982 * config/aarch64/aarch64.md (*do_return): Likewise.
1984 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
1986 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
1987 duplicate_loop_to_header_edge to
1988 duplicate_loop_body_to_header_edge.
1989 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
1990 * cfghooks.h (struct cfg_hooks): Likewise.
1991 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
1992 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
1993 (clone_loop_to_header_edge): Likewise.
1994 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
1995 * cfgrtl.c (struct cfg_hooks): Likewise.
1996 * doc/loop.texi: Likewise.
1997 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
1998 (unroll_loop_runtime_iterations): Likewise.
1999 (unroll_loop_stupid): Likewise.
2000 (apply_opt_in_copies): Likewise.
2001 * tree-cfg.c (struct cfg_hooks): Likewise.
2002 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
2003 (try_peel_loop): Likewise.
2004 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
2005 (gimple_duplicate_loop_body_to_header_edge): Likewise.
2006 (tree_transform_and_unroll_loop): Likewise.
2007 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
2010 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
2012 * cfgloopmanip.c (loop_version): Refactor loopify to
2013 loop_version. Move condition generation after loopify.
2015 * cfgloopmanip.h (loopify): Delete.
2017 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
2019 * ipa-fnsummary.c: Include tree-dfa.h.
2020 (points_to_local_or_readonly_memory_p): Return true on return
2022 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
2025 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
2027 PR middle-end/102972
2028 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
2029 internal Fortran name; new permit_num_teams arg to permit
2030 omp_get_num_teams and omp_get_team_num.
2031 (scan_omp_1_stmt): Update call to it, add missing call for
2032 reverse offload, and check for strictly nested API calls in teams.
2034 2021-10-30 Jakub Jelinek <jakub@redhat.com>
2036 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
2038 2021-10-29 David Malcolm <dmalcolm@redhat.com>
2040 * selftest.c (assert_streq): Add newlines when emitting non-equal
2043 2021-10-29 David Malcolm <dmalcolm@redhat.com>
2045 * Makefile.in: Fix syntax for reference to LIBDEPS in
2048 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
2050 * doc/install.texi: Bump required minimum DejaGnu version.
2052 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
2053 Andrew MacLeod <amacleod@redhat.com>
2055 * value-relation.cc (path_oracle::killing_def): Add a
2056 self-equivalence so we don't look to the root oracle.
2058 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
2060 * passes.def: Replace the pass_thread_jumps before VRP* with
2061 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
2062 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
2063 Remove hyphen from "thread-full" name.
2065 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
2067 PR tree-optimization/102983
2068 * gimple-range-cache.h (propagate_updated_value): Make public.
2069 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
2070 when processing gcond stmts.
2072 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
2074 * gimple.c (gimple_call_retslot_flags): New function.
2075 * gimple.h (gimple_call_retslot_flags): Declare.
2076 * ipa-modref.c: Include tree-cfg.h.
2077 (struct escape_entry): Turn parm_index to signed.
2078 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
2079 (modref_summary::modref_summary): Initialize retslot_flags.
2080 (struct modref_summary_lto): Likewise.
2081 (modref_summary::useful_p): Check retslot_flags.
2082 (modref_summary_lto::useful_p): Likewise.
2083 (modref_summary::dump): Dump retslot_flags.
2084 (modref_summary_lto::dump): Likewise.
2085 (struct escape_point): Add hidden_args enum.
2086 (analyze_ssa_name_flags): Ignore return slot return;
2087 use gimple_call_retslot_flags.
2088 (record_escape_points): Break out from ...
2089 (analyze_parms): ... here; handle retslot_flags.
2090 (modref_summaries::duplicate): Duplicate retslot_flags.
2091 (modref_summaries_lto::duplicate): Likewise.
2092 (modref_write_escape_summary): Stream parm_index as signed.
2093 (modref_read_escape_summary): Likewise.
2094 (modref_write): Stream retslot_flags.
2095 (read_section): Likewise.
2096 (struct escape_map): Fix typo in comment.
2097 (update_escape_summary_1): Fix whitespace.
2098 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
2099 (modref_merge_call_site_flags): Merge retslot_flags.
2100 * ipa-modref.h (struct modref_summary): Add retslot_flags.
2101 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
2103 2021-10-29 Tamar Christina <tamar.christina@arm.com>
2105 PR tree-optimization/102977
2106 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
2107 (vect_detect_pair_op): Add crosslane check.
2108 (vect_match_call_complex_mla): Remove.
2109 (class complex_mul_pattern): Update comment.
2110 (complex_mul_pattern::matches): Update detection.
2111 (class complex_fma_pattern): Remove.
2112 (complex_fma_pattern::matches): Remove.
2113 (complex_fma_pattern::recognize): Remove.
2114 (complex_fma_pattern::build): Remove.
2115 (class complex_fms_pattern): Update comment.
2116 (complex_fms_pattern::matches): Remove.
2117 (complex_operations_pattern::recognize): Remove complex_fma_pattern
2119 2021-10-29 Jakub Jelinek <jakub@redhat.com>
2121 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
2124 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
2127 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
2128 patterns match and emit for VSX xxpermdi.
2130 2021-10-29 liuhongt <hongtao.liu@intel.com>
2133 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
2135 (V16HF_FTYPE_V16HF): Ditto.
2136 (V32HF_FTYPE_V32HF): Ditto.
2137 (V8HF_FTYPE_V8HF_ROUND): Ditto.
2138 (V16HF_FTYPE_V16HF_ROUND): Ditto.
2139 (V32HF_FTYPE_V32HF_ROUND): Ditto.
2140 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
2141 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
2142 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
2143 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
2144 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
2145 * config/i386/i386-builtins.c
2146 (ix86_builtin_vectorized_function): Enable vectorization for
2147 HFmode FLOOR/CEIL/TRUNC operation.
2148 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
2150 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
2153 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2154 Andrew MacLeod <amacleod@redhat.com>
2156 * value-relation.cc (path_oracle::killing_def): Walk the
2157 equivalency list and remove SSA from any equivalencies.
2159 2021-10-28 Stafford Horne <shorne@gmail.com>
2161 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
2164 2021-10-28 Jakub Jelinek <jakub@redhat.com>
2166 PR tree-optimization/102951
2167 * fold-const.h (address_compare): Declare.
2168 * fold-const.c (address_compare): New function.
2169 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
2170 address_compare helper.
2171 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
2174 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
2176 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
2178 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
2180 PR tree-optimization/102940
2181 * tree-vrp.c (execute_ranger_vrp): Reset scev.
2183 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
2185 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
2187 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
2189 * configure: Regenerate.
2190 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
2192 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
2194 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
2196 2021-10-28 Richard Biener <rguenther@suse.de>
2199 * fold-const.c (fold_convert_const): Avoid int to float
2200 constant folding with -frounding-math and inexact result.
2201 * simplify-rtx.c (simplify_const_unary_operation): Likewise
2202 for both float and unsigned_float.
2204 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2206 * tree-ssa-threadbackward.c
2207 (back_threader::find_taken_edge_switch): Use find_case_label_range
2208 instead of find_taken_edge.
2210 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2212 * tree-ssa-threadbackward.c (class back_threader_registry):
2213 Inherit from back_jt_path_registry.
2214 (back_threader_registry::thread_through_all_blocks): Remove.
2215 (back_threader_registry::register_path): Remove
2216 m_lowlevel_registry prefix.
2218 2021-10-28 Richard Biener <rguenther@suse.de>
2221 * fold-const.c (fold_convert_const_real_from_real): Honor
2222 -frounding-math if the conversion is not exact.
2223 * simplify-rtx.c (simplify_const_unary_operation): Do not
2224 simplify FLOAT_TRUNCATE with sign dependent rounding.
2226 2021-10-28 Richard Biener <rguenther@suse.de>
2228 PR tree-optimization/102949
2229 * tree-vect-stmts.c (ensure_base_align): Look at the
2230 dr_info of a group leader and assert we are looking at
2231 one with analyzed alignment.
2233 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
2236 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
2237 V1T1 mode for unaligned load and store.
2239 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
2241 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
2244 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
2246 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
2249 2021-10-28 Alexandre Oliva <oliva@adacore.com>
2251 * common.opt (fharden-compares): New.
2252 (fharden-conditional-branches): New.
2253 * doc/invoke.texi: Document new options.
2254 * gimple-harden-conditionals.cc: New.
2255 * Makefile.in (OBJS): Build it.
2256 * passes.def: Add new passes.
2257 * tree-pass.h (make_pass_harden_compares): Declare.
2258 (make_pass_harden_conditional_branches): Declare.
2260 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2263 * config/rs6000/altivec.md: Add vsx register constraints.
2264 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
2265 (vsx_xxsel<mode>2): Likewise.
2266 (vsx_xxsel<mode>3): Likewise.
2267 (vsx_xxsel<mode>4): Likewise.
2269 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2272 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
2273 (altivec_vsel<mode>): ... this and update define.
2274 (*altivec_vsel<mode>_uns): Delete.
2275 (altivec_vsel<mode>2): New define_insn.
2276 (altivec_vsel<mode>3): Likewise.
2277 (altivec_vsel<mode>4): Likewise.
2278 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
2279 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
2281 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
2282 selection instead of per element.
2283 * config/rs6000/vector.md:
2284 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
2285 (vsx_xxsel<mode>): ... this and update define.
2286 (*vsx_xxsel<mode>_uns): Delete.
2287 (vsx_xxsel<mode>2): New define_insn.
2288 (vsx_xxsel<mode>3): Likewise.
2289 (vsx_xxsel<mode>4): Likewise.
2291 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
2293 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
2294 HFmode rsqrt without TARGET_SSE_MATH.
2295 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
2296 btrunc condition to be restricted by -ftrapping-math, adjust
2297 use_rsqrt_p function call.
2298 * config/i386/i386.md (rcphf2): New define_insn.
2299 (rsqrthf2): Likewise.
2300 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
2301 (div<mode>3): New expander for HF mode.
2302 (rsqrt<mode>2): Likewise.
2303 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
2304 (*avx512fp16_vmrsqrtv8hf2): Likewise.
2306 2021-10-27 Saagar Jha <saagar@saagarjha.com>
2308 * config.gcc: Adjust for Darwin21.
2309 * config/darwin-c.c (macosx_version_as_macro): Likewise.
2310 * config/darwin-driver.c (validate_macosx_version_min):
2312 (darwin_find_version_from_kernel): Likewise.
2314 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2316 * gimple-range-path.cc
2317 (path_range_query::range_defined_in_block): Call killing_def.
2319 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2321 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
2322 operands as well as relation.
2323 * gimple-range-path.cc
2324 (path_range_query::compute_ranges_in_block): Compute PHI relations
2325 first. Compute outgoing relations at the end.
2326 (path_range_query::compute_ranges): Remove call to compute_relations.
2327 (path_range_query::compute_relations): Remove.
2328 (path_range_query::maybe_register_phi_relation): New.
2329 (path_range_query::compute_phi_relations): Abstract out
2330 registering one PHI relation to...
2331 (path_range_query::compute_outgoing_relations): ...here.
2332 * gimple-range-path.h (class path_range_query): Remove
2334 Add maybe_register_phi_relation.
2336 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2337 Andrew MacLeod <amacleod@redhat.com>
2339 * value-relation.cc (path_oracle::killing_def): Kill second
2342 2021-10-27 Martin Jambor <mjambor@suse.cz>
2344 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
2345 profile feedback depending on their local availability.
2347 2021-10-27 Martin Jambor <mjambor@suse.cz>
2349 * params.opt (param_ipa_cp_profile_count_base): New parameter.
2350 * doc/invoke.texi (Optimize Options): Add entry for
2351 ipa-cp-profile-count-base.
2352 * ipa-cp.c (max_count): Replace with base_count, replace all
2353 occurrences too, unless otherwise stated.
2354 (ipcp_cloning_candidate_p): identify mostly-directly called
2355 functions based on their counts, not max_count.
2356 (compare_edge_profile_counts): New function.
2357 (ipcp_propagate_stage): Instead of setting max_count, find the
2358 appropriate edge count in a sorted vector of counts of eligible
2359 edges and make it the base_count.
2361 2021-10-27 Martin Jambor <mjambor@suse.cz>
2363 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
2364 n_nonrec_calls and itself, document all fields.
2365 (init_caller_stats): Initialize the above new fields.
2366 (gather_caller_stats): Gather self-recursive counts and calls number.
2367 (get_info_about_necessary_edges): Gather counts of self-recursive and
2368 other edges bringing in the requested value separately.
2369 (dump_profile_updates): Rework to dump info about a single node only.
2370 (lenient_count_portion_handling): New function.
2371 (struct gather_other_count_struct): New type.
2372 (gather_count_of_non_rec_edges): New function.
2373 (struct desc_incoming_count_struct): New type.
2374 (analyze_clone_icoming_counts): New function.
2375 (adjust_clone_incoming_counts): Likewise.
2376 (update_counts_for_self_gen_clones): Likewise.
2377 (update_profiling_info): Rewritten.
2378 (update_specialized_profile): Adjust call to dump_profile_updates.
2379 (create_specialized_node): Do not update profiling info.
2380 (decide_about_value): New parameter self_gen_clones, either push new
2381 clones into it or updat their profile counts. For self-recursively
2382 generated values, use a portion of the node count instead of count
2383 from self-recursive edges to estimate goodness.
2384 (decide_whether_version_node): Gather clones for self-generated values
2385 in a new vector, update their profiles at once at the end.
2387 2021-10-27 Richard Biener <rguenther@suse.de>
2389 * tree-vectorizer.c (vect_transform_loops): New function,
2391 (try_vectorize_loop_1): ... here. Simplify as epilogues
2392 are now fully handled in the split part.
2394 2021-10-27 Jakub Jelinek <jakub@redhat.com>
2396 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
2397 iterators with pointer types.
2398 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
2400 2021-10-26 Martin Sebor <msebor@redhat.com>
2402 PR tree-optimization/102238
2403 PR tree-optimization/102919
2404 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
2405 (array_elt_at_offset): Move to pointer-query.
2406 (set_aggregate_size_and_offset): New function.
2407 (field_at_offset): Move to pointer-query.
2408 (get_origin_and_offset): Rename...
2409 (get_origin_and_offset_r): this. Add an argument. Make aggregate
2410 handling more robust.
2411 (get_origin_and_offset): New.
2412 (alias_offset): Add an argument.
2413 (format_string): Use subobject size determined by get_origin_and_offset.
2414 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
2415 Improve/correct handling of aggregates.
2416 (array_elt_at_offset): Same.
2417 * pointer-query.h (field_at_offset): Declare.
2418 (array_elt_at_offset): Declare.
2420 2021-10-26 Martin Sebor <msebor@redhat.com>
2422 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
2424 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
2425 (expand_builtin_fork_or_exec): Same.
2426 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
2427 compute_objsize additional arguments.
2428 (inbounds_memaccess_p): Same.
2429 (array_bounds_checker::check_array_bounds): Add an assert. Stash
2430 statement in a member.
2431 (check_array_bounds_dom_walker::before_dom_children): Same.
2432 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
2433 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
2434 (handle_printf_call): Pass a new argument.
2435 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
2436 (check_access): Add an argument and pass it along to callees.
2437 (check_read_access): Make a member function.
2438 (pass_waccess::check_strcat): Pass access_data ctor additional
2440 (pass_waccess::check_strncat): Same.
2441 (pass_waccess::check_stxcpy): Same.
2442 (pass_waccess::check_stxncpy): Same.
2443 (pass_waccess::check_strncmp): Same.
2444 (pass_waccess::check_read_access): Same.
2445 (pass_waccess::check_builtin): Same.
2446 (pass_waccess::maybe_check_access_sizes): Same.
2447 (pass_waccess::maybe_check_dealloc_call): Same.
2448 * gimple-ssa-warn-access.h (check_read_access): Declare a new
2450 * pointer-query.cc (compute_objsize_r): Add an argument.
2451 (gimple_call_return_array): Same.
2452 (gimple_call_alloc_size): Same.
2453 (access_ref::access_ref): Same.
2454 (access_ref::get_ref): Same.
2455 (pointer_query::get_ref): Same.
2456 (handle_min_max_size): Pass an arguments to callees.
2457 (handle_array_ref): Add an argument.
2458 (handle_mem_ref): Same.
2459 (compute_objsize): Same.
2460 * pointer-query.h (struct access_ref): Adjust signatures.
2461 (struct access_data): Same.
2462 (gimple_call_alloc_size): Add an argument.
2463 (gimple_parm_array_size): Same.
2464 (compute_objsize): Same.
2465 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
2466 argument to compute_objsize.
2467 (strlen_pass::maybe_warn_overflow): Same.
2468 (maybe_diag_stxncpy_trunc): Same.
2470 2021-10-26 Martin Sebor <msebor@redhat.com>
2472 PR middle-end/102453
2473 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
2474 (pass_waccess::check_atomic_builtin): Call it.
2476 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
2478 PR rtl-optimization/102842
2479 * lra-constraints.c (match_reload): Ignore out in checking values
2481 (curr_insn_transform): Collect outputs before doing reloads of operands.
2483 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
2486 * config/rs6000/x86intrin.h: Move some included headers to new
2487 headers. Include new immintrin.h instead of those headers.
2488 * config/rs6000/immintrin.h: New.
2489 * config/rs6000/x86gprintrin.h: New.
2490 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
2492 2021-10-26 Richard Biener <rguenther@suse.de>
2494 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
2495 Remove byte_offset parameter.
2496 (vect_create_data_ref_ptr): Likewise.
2497 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2499 (vect_create_data_ref_ptr): Likewise.
2500 * tree-vect-stmts.c (vectorizable_store): Adjust.
2501 (vectorizable_load): Likewise.
2503 2021-10-26 Richard Biener <rguenther@suse.de>
2505 PR tree-optimization/96109
2506 * tree-vectorizer.h (dr_misalignment): Add optional offset
2508 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
2509 offset applied for negative stride accesses.
2510 (vect_enhance_data_refs_alignment): Compute negative stride
2511 access offset and pass it to dr_misalignment.
2512 * tree-vect-stmts.c (get_negative_load_store_type): Pass
2513 negative offset to dr_misalignment.
2514 (get_group_load_store_type): Likewise.
2515 (get_load_store_type): Likewise.
2516 (vectorizable_store): Remove asserts about alignment.
2517 (vectorizable_load): Likewise.
2519 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2521 PR tree-optimization/102897
2522 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
2524 2021-10-26 Richard Biener <rguenther@suse.de>
2526 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2527 Take offset in bytes.
2528 (vect_create_data_ref_ptr): Likewise.
2529 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
2530 offset by element size.
2531 (vect_create_cond_for_align_checks): Likewise.
2532 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
2533 (vectorizable_load): Remove duplicate leftover from merge
2536 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
2538 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
2539 function to expand V1TI mode logical shifts by integer constants.
2540 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
2541 mode rotations by integer constants.
2542 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
2543 ix86_expand_v1ti_rotate): Prototype new functions here.
2544 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
2545 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
2547 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2549 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2550 Avoid threading circular paths.
2552 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2554 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
2555 Attempt to resolve all incoming paths to a PHI.
2556 (back_threader::resolve_def): Always return true for PHIs.
2558 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2560 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
2561 Always try to resolve path without looking back.
2562 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
2563 edge is a back edge.
2565 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2567 PR tree-optimization/102789
2568 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2569 update inits of simd_lane_access.
2571 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2573 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
2574 move to vr-values.c.
2575 (simplify_casted_conds): Move to vrp_folder class.
2576 (execute_vrp): Call via vrp_folder now.
2577 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
2578 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
2579 * vr-values.h (simplify_casted_cond): Add prototype.
2581 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2583 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
2586 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
2588 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
2590 (_mm_testc_si128): Likewise.
2591 (_mm_testnzc_si128): Likewise.
2592 (_mm_blend_ps): Likewise.
2593 (_mm_blendv_ps): Likewise.
2594 (_mm_blend_pd): Likewise.
2595 (_mm_blendv_pd): Likewise.
2596 (_mm_ceil_pd): Likewise.
2597 (_mm_ceil_sd): Likewise.
2598 (_mm_ceil_ps): Likewise.
2599 (_mm_ceil_ss): Likewise.
2600 (_mm_floor_pd): Likewise.
2601 (_mm_floor_sd): Likewise.
2602 (_mm_floor_ps): Likewise.
2603 (_mm_floor_ss): Likewise.
2604 (_mm_minpos_epu16): Likewise.
2605 (_mm_mul_epi32): Likewise.
2606 (_mm_cvtepi8_epi16): Likewise.
2607 (_mm_packus_epi32): Likewise.
2608 (_mm_cmpgt_epi64): Likewise.
2610 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
2612 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
2613 shifts of the mode's smin_value and smax_value when the bit count
2614 operand doesn't have side-effects.
2615 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
2616 when the bit count operand doesn't have side-effects.
2617 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
2618 compile-time evaluation of saturating left shifts with constant
2621 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2623 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
2626 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2628 * flag-types.h (enum ranger_debug): Adjust values.
2629 * params.opt (ranger_debug): Ditto.
2631 2021-10-25 Martin Jambor <mjambor@suse.cz>
2633 PR tree-optimization/102886
2634 * tree-sra.c (totally_scalarize_subtree): Fix the out of
2637 2021-10-25 Andrew Pinski <apinski@marvell.com>
2639 * tree-ssa-dce.c (simple_dce_from_worklist):
2640 Check stmt_unremovable_because_of_non_call_eh_p also
2641 before removing the statement.
2643 2021-10-25 Richard Biener <rguenther@suse.de>
2645 PR tree-optimization/102905
2646 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2647 Use vect_supportable_dr_alignment again to determine whether
2648 an access is supported when not aligned.
2650 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2652 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
2655 2021-10-25 Jim Wilson <jimw@sifive.com>
2656 Kito Cheng <kito.cheng@sifive.com>
2657 Jia-Wei Chen <jiawei@iscas.ac.cn>
2658 Shi-Hua Liao <shihua@iscas.ac.cn>
2660 * config/riscv/bitmanip.md (shiftm1): New.
2661 (*bset<mode>): Ditto.
2662 (*bset<mode>_mask): Ditto.
2663 (*bset<mode>_1): Ditto.
2664 (*bset<mode>_1_mask): Ditto.
2665 (*bseti<mode>): Ditto.
2666 (*bclr<mode>): Ditto.
2667 (*bclri<mode>): Ditto.
2668 (*binv<mode>): Ditto.
2669 (*binvi<mode>): Ditto.
2670 (*bext<mode>): Ditto.
2672 * config/riscv/predicates.md (splittable_const_int_operand):
2674 (single_bit_mask_operand): New.
2675 (not_single_bit_mask_operand): Ditto.
2676 (const31_operand): Ditto.
2677 (const63_operand): Ditto.
2678 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
2679 (riscv_output_move): Ditto.
2680 (riscv_print_operand): Handle new operand type: T and S.
2681 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
2683 2021-10-25 Jim Wilson <jimw@sifive.com>
2685 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
2688 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2690 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
2692 (riscv_rtx_costs): Ditto.
2694 2021-10-25 Jim Wilson <jimw@sifive.com>
2695 Kito Cheng <kito.cheng@sifive.com>
2696 Jia-Wei Chen <jiawei@iscas.ac.cn>
2698 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
2699 (bitmanip_minmax): New.
2700 (clz_ctz_pcnt): New.
2701 (bitmanip_optab): New.
2702 (bitmanip_insn): New.
2703 (*<optab>_not<mode>): New.
2704 (*xor_not<mode>): New.
2705 (<bitmanip_optab>si2): New.
2706 (*<bitmanip_optab>disi2): New.
2707 (<bitmanip_optab>di2): New.
2708 (*zero_extendhi<GPR:mode>2_bitmanip): New.
2709 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
2710 (*zero_extendhi<GPR:mode>2_zbb): New.
2713 (rotrsi3_sext): New.
2716 (rotlsi3_sext): New.
2717 (bswap<mode>2): New.
2718 (<bitmanip_optab><mode>3): New.
2719 * config/riscv/riscv.md (type): Add rotate.
2720 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
2721 (*zero_extendhi<GPR:mode>2): New.
2722 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
2723 (*extend<SHORT:mode><SUPERQI:mode>2): New.
2725 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2727 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
2729 (riscv_rtx_costs): Ditto.
2731 2021-10-25 Jim Wilson <jimw@sifive.com>
2732 Kito Cheng <kito.cheng@sifive.com>
2733 Jia-Wei Chen <jiawei@iscas.ac.cn>
2735 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
2740 (riscv_rtx_costs): Ditto.
2741 * config/riscv/riscv.md: Include bitmanip.md
2742 (type): Add bitmanip bype.
2743 (zero_extendsidi2): Change to define_expand pattern.
2744 (*zero_extendsidi2_internal): New.
2745 (zero_extendsidi2_shifted): Disable for ZBA.
2747 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2749 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
2750 Add zba, zbb, zbc and zbs.
2751 (riscv_ext_flag_table): Ditto.
2752 * config/riscv/riscv-opts.h (MASK_ZBA): New.
2756 (TARGET_ZBA): Ditto.
2757 (TARGET_ZBB): Ditto.
2758 (TARGET_ZBC): Ditto.
2759 (TARGET_ZBS): Ditto.
2760 * config/riscv/riscv.opt (riscv_zb_subext): New.
2762 2021-10-25 liuhongt <hongtao.liu@intel.com>
2765 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
2766 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
2768 2021-10-25 Richard Biener <rguenther@suse.de>
2770 PR tree-optimization/102920
2771 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
2772 controlling VN_TOP matching behavior.
2773 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
2774 (vn_phi_eq): Do not optimistically match VN_TOP.
2776 2021-10-25 konglin1 <lingling.kong@intel.com>
2778 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
2779 define_insn_and_split.
2780 (fma_<mode>_fadd_fcmul):Likewise
2781 (fma_<complexopname>_<mode>_fma_zero):Likewise
2783 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
2785 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
2786 TARGET_DISABLE_FPREGS.
2787 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
2788 MASK_DISABLE_FPREGS.
2789 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
2790 cost of hardware integer multiplication.
2791 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
2792 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
2793 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
2794 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
2795 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
2796 SOFT_FLOAT. Revise comment.
2797 (msoft-float): New option.
2799 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
2801 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
2803 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
2805 * doc/install.texi (Specific): Remove obsolete details
2806 around GNU/Linux on Itanium.
2807 (Specific): Remove reference to Windows for Itanium.
2809 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
2811 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
2814 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
2816 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
2818 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
2820 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
2821 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
2822 (one_cmplv1ti2): New define expand.
2824 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
2827 * Makefile.in: Handle dvidir and install-dvi target.
2828 * configure: Regenerate.
2829 * configure.ac: Add install-dvi to target_list.
2831 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
2833 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
2836 2021-10-22 Richard Biener <rguenther@suse.de>
2838 PR tree-optimization/102893
2839 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
2840 test for an exit edge.
2842 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
2843 Andrew MacLeod <amacleod@redhat.com>
2845 * gimple-range-path.cc (path_range_query::compute_phi_relations):
2846 Kill any global relations we may know before registering a new
2848 * value-relation.cc (path_oracle::killing_def): New.
2849 * value-relation.h (path_oracle::killing_def): New.
2851 2021-10-22 Richard Biener <rguenther@suse.de>
2854 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
2856 (vn_phi_lookup): Likewise.
2858 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
2861 * doc/invoke.texi: Document -fcf-protection requires i686 or
2864 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
2866 PR middle-end/102764
2867 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
2869 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
2871 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
2872 macro and all invocations.
2874 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2876 * doc/invoke.texi (ranger-debug): Document.
2877 * flag-types.h (enum ranger_debug): New.
2878 (enum evrp_mode): Remove debug values.
2879 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
2880 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
2881 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
2882 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
2883 (execute_early_vrp): Use evrp-mode directly.
2884 * params.opt (enum evrp_mode): Remove debug values.
2885 (ranger-debug): New.
2886 (ranger-logical-depth): Relocate to be in alphabetical order.
2888 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2890 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
2891 * flag-types.h: (enum vrp_mode): New.
2892 * params.opt: (vrp1-mode, vrp2-mode): New.
2893 * tree-vrp.c (vrp_pass_num): New.
2894 (pass_vrp::pass_vrp): Set pass number.
2895 (pass_vrp::execute): Choose which VRP mode to execute.
2897 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2899 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
2900 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
2901 * tree-vrp.c (class rvrp_folder): Relocate here.
2902 (execute_ranger_vrp): New.
2903 * tree-vrp.h (execute_ranger_vrp): Export.
2905 2021-10-21 Martin Liska <mliska@suse.cz>
2909 * opts.c (finish_options): Process flag_var_tracking* options
2910 here as they can be adjusted by optimize attribute.
2911 Process also flag_syntax_only and flag_gtoggle.
2912 * toplev.c (process_options): Remove it here.
2913 * common.opt: Make debug_nonbind_markers_p as PerFunction
2914 attribute as it depends on optimization level.
2916 2021-10-21 Martin Jambor <mjambor@suse.cz>
2918 PR tree-optimization/102505
2919 * tree-sra.c (totally_scalarize_subtree): Check that the
2920 encountered field fits within the acces we would like to put it
2923 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
2925 * tree-ssa-threadbackward.c
2926 (back_threader::maybe_register_path): Remove circular paths check.
2928 2021-10-21 Richard Biener <rguenther@suse.de>
2930 * toplev.c (process_options): Move the initial debug_hooks
2932 (toplev::main): ... before the call of the post_options
2935 2021-10-21 Richard Biener <rguenther@suse.de>
2937 PR tree-optimization/102847
2938 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
2939 load cost in the prologue for VMAT_INVARIANT.
2941 2021-10-21 Richard Biener <rguenther@suse.de>
2943 PR tree-optimization/102847
2944 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
2945 handle VMAT_INVARIANT as a splat in the prologue.
2947 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
2950 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
2951 move to use the same logic as HImode.
2953 2021-10-21 Richard Biener <rguenther@suse.de>
2955 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
2956 superfluous gimple_call_nothrow_p check.
2958 2021-10-21 Andrew Pinski <apinski@marvell.com>
2960 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
2961 Mark the ssa-name of the rhs as one to be removed.
2962 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
2963 Call simple_dce_from_worklist at the end to a simple dce.
2965 2021-10-21 Andrew Pinski <apinski@marvell.com>
2967 * tree-cfg.c (maybe_remove_writeonly_store): New function
2968 factored out from ...
2969 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
2971 2021-10-21 Andrew Pinski <apinski@marvell.com>
2973 * tree-cfg.c (execute_fixup_cfg): Remove comment
2974 about standalone pass.
2976 2021-10-21 Andrew Pinski <apinski@marvell.com>
2978 * tree-cfg.c (execute_fixup_cfg): Output when the statement
2979 is removed when it is a write only var.
2981 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
2983 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2984 Avoid threading circular paths.
2986 2021-10-20 Alex Coplan <alex.coplan@arm.com>
2988 * calls.c (initialize_argument_information): Remove some dead
2989 code, remove handling for function_arg returning const_int.
2990 * doc/tm.texi: Delete documentation for unused target hooks.
2991 * doc/tm.texi.in: Likewise.
2992 * target.def (load_bounds_for_arg): Delete.
2993 (store_bounds_for_arg): Delete.
2994 (load_returned_bounds): Delete.
2995 (store_returned_bounds): Delete.
2996 * targhooks.c (default_load_bounds_for_arg): Delete.
2997 (default_store_bounds_for_arg): Delete.
2998 (default_load_returned_bounds): Delete.
2999 (default_store_returned_bounds): Delete.
3000 * targhooks.h (default_load_bounds_for_arg): Delete.
3001 (default_store_bounds_for_arg): Delete.
3002 (default_load_returned_bounds): Delete.
3003 (default_store_returned_bounds): Delete.
3005 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
3007 * doc/extend.texi (Basic Asm): Clarify that asm is not an
3009 * doc/invoke.texi (-fno-asm): Fix description for C++.
3011 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
3013 * doc/install.texi: Remove link to old.html
3015 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3017 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
3019 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3021 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
3023 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3025 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
3027 * config/aarch64/constraints.md (D1): New.
3029 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3031 * config/aarch64/aarch64-simd.md
3032 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
3033 (*aarch64_topbits_shuffle<mode>_le): New.
3034 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
3035 (*aarch64_topbits_shuffle<mode>_be): New.
3036 * config/aarch64/predicates.md
3037 (aarch64_simd_shift_imm_vec_exact_top): New.
3039 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3041 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
3042 *aarch64_<srn_op>shrn<mode>2_vect_le,
3043 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
3044 * config/aarch64/iterators.md (srn_op): New.
3046 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
3048 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
3050 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
3051 ctx->outer instead of ctx.
3053 2021-10-20 Martin Liska <mliska@suse.cz>
3056 2021-10-19 Martin Liska <mliska@suse.cz>
3059 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
3060 * system.h (strip_whilespaces): New function.
3062 2021-10-20 Martin Liska <mliska@suse.cz>
3065 2021-10-19 Martin Liska <mliska@suse.cz>
3068 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
3071 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
3073 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
3074 Change to 8 with -Os, 11 otherwise.
3076 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
3078 * config/aarch64/aarch64.c (neoversev1_tunings):
3079 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
3080 (neoversen2_tunings): Likewise.
3082 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
3085 * config/aarch64/aarch64.md (frint_pattern): Update comment.
3086 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
3087 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
3088 * config/aarch64/arm_neon.h: Likewise.
3089 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
3091 2021-10-20 Martin Liska <mliska@suse.cz>
3093 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
3094 already declared global variable.
3095 (arm_unwind_emit_set): Use out_file as function argument.
3096 (arm_unwind_emit): Likewise.
3097 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
3098 (machopic_output_stub_indirection): Likewise.
3099 (machopic_output_indirection): Likewise.
3100 (machopic_finish): Likewise.
3101 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
3102 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
3103 * config/ia64/ia64.c (process_epilogue): Likewise.
3104 (process_cfa_adjust_cfa): Likewise.
3105 (process_cfa_register): Likewise.
3106 (process_cfa_offset): Likewise.
3107 (ia64_asm_unwind_emit): Likewise.
3108 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
3110 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
3112 * config/aarch64/aarch64-builtins.c
3113 (aarch64_general_gimple_fold_builtin):
3114 lower vld1 and vst1 variants of the neon builtins
3115 * config/aarch64/aarch64-protos.h:
3116 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
3117 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
3120 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
3122 * match.pd: Generate IFN_TRUNC.
3124 2021-10-20 Richard Biener <rguenther@suse.de>
3126 PR tree-optimization/102853
3127 * tree-data-ref.c (split_constant_offset_1): Bail out
3128 immediately if the expression traps on overflow.
3130 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3132 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
3134 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3136 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
3137 Remove superflous debugging message.
3138 (back_jt_path_registry::duplicate_thread_path): Same.
3140 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3142 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
3144 (back_threader_registry::register_path): Remove m_threaded_paths.
3146 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3148 PR tree-optimization/102814
3149 * doc/invoke.texi: Document --param=max-fsm-thread-length.
3150 * params.opt: Add --param=max-fsm-thread-length.
3151 * tree-ssa-threadbackward.c
3152 (back_threader_profitability::profitable_path_p): Fail on paths
3153 longer than max-fsm-thread-length.
3155 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
3157 PR middle-end/102764
3158 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
3159 statement to reset the current location for the outgoing edges.
3161 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3162 Richard Biener <rguenther@suse.de>
3164 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
3165 on the same line as the threading cancellation.
3166 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
3167 Avoid threading through loop headers where the path remains in the
3170 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
3172 * tree-object-size.c (unknown): Make into a function. Adjust
3174 (unknown_object_size): Simplify implementation.
3176 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
3177 Kewen Lin <linkw@linux.ibm.com>
3179 * doc/sourcebuild.texi (Effective-Target Keywords): Document
3180 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
3181 vect_slp_v16qi_store, vect_slp_v2hi_store,
3182 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
3184 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
3186 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
3188 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
3192 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
3193 * config/rs6000/pmmintrin.h: Same.
3194 * config/rs6000/smmintrin.h: Same.
3195 * config/rs6000/tmmintrin.h: Same.
3197 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
3199 * config.gcc (extra_headers): Add nmmintrin.h.
3201 2021-10-19 Richard Biener <rguenther@suse.de>
3203 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
3204 misalignment parameter.
3205 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
3206 Do not change DR_MISALIGNMENT in place, instead pass the
3207 adjusted misalignment to vect_supportable_dr_alignment.
3208 (vect_peeling_supportable): Likewise.
3209 (vect_peeling_hash_get_lowest_cost): Adjust.
3210 (vect_enhance_data_refs_alignment): Likewise.
3211 (vect_vfa_access_size): Likewise.
3212 (vect_supportable_dr_alignment): Add misalignment
3213 parameter and simplify.
3214 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
3215 (get_group_load_store_type): Likewise.
3216 (get_load_store_type): Likewise.
3218 2021-10-19 Clément Chigot <clement.chigot@atos.net>
3220 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
3221 __tls_get_addr reference to .text csect.
3223 2021-10-19 Martin Liska <mliska@suse.cz>
3226 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
3229 2021-10-19 Richard Biener <rguenther@suse.de>
3231 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
3232 (vect_get_load_cost): Likewise.
3233 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3234 alignment support scheme and misalignment as arguments
3236 (vect_get_peeling_costs_all_drs): Compute that info here
3237 and note that we shouldn't need to.
3238 * tree-vect-stmts.c (vect_model_store_cost): Get
3239 alignment support scheme and misalignment as arguments.
3240 (vect_get_store_cost): Likewise.
3241 (vect_model_load_cost): Likewise.
3242 (vect_get_load_cost): Likewise.
3243 (vectorizable_store): Pass down alignment support scheme
3244 and misalignment to costing.
3245 (vectorizable_load): Likewise.
3247 2021-10-19 Richard Biener <rguenther@suse.de>
3249 * tree-vect-stmts.c (get_negative_load_store_type): Add
3250 offset output parameter and initialize it.
3251 (get_group_load_store_type): Likewise.
3252 (get_load_store_type): Likewise.
3253 (vectorizable_store): Use offset as computed by
3254 get_load_store_type.
3255 (vectorizable_load): Likewise.
3257 2021-10-19 Richard Biener <rguenther@suse.de>
3259 PR tree-optimization/102827
3260 * tree-if-conv.c (predicate_statements): Add pe parameter
3261 and use that edge to insert invariant stmts on.
3262 (combine_blocks): Pass through pe.
3263 (tree_if_conversion): Compute the edge to insert invariant
3264 stmts on and pass it along.
3266 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
3269 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
3270 sssubaddv2hi3): Swap the order of operators in vec_concat.
3272 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
3274 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
3275 (altivec_vmrghb_direct): New.
3276 (*altivec_vmrghh_internal): Delete.
3277 (altivec_vmrghh_direct): New.
3278 (*altivec_vmrghw_internal): Delete.
3279 (altivec_vmrghw_direct_<mode>): New.
3280 (altivec_vmrghw_direct): Delete.
3281 (*altivec_vmrglb_internal): Delete.
3282 (altivec_vmrglb_direct): New.
3283 (*altivec_vmrglh_internal): Delete.
3284 (altivec_vmrglh_direct): New.
3285 (*altivec_vmrglw_internal): Delete.
3286 (altivec_vmrglw_direct_<mode>): New.
3287 (altivec_vmrglw_direct): Delete.
3288 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3289 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
3291 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
3292 (vsx_xxmrglw_<mode>): Adjust.
3294 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
3296 * passes.def: Change threading comment before pass_ccp pass.
3298 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
3300 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
3301 Modify the expansion for sign extension. All extensions are done
3302 within VSX registers.
3304 2021-10-19 Richard Biener <rguenther@suse.de>
3306 * tree-vect-stmts.c (get_group_load_store_type): Add
3307 misalignment output parameter and initialize it.
3308 (get_group_load_store_type): Likewise.
3309 (vectorizable_store): Remove now redundant queries.
3310 (vectorizable_load): Likewise.
3312 2021-10-19 Richard Biener <rguenther@suse.de>
3314 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
3315 check_aligned argument.
3316 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
3318 (vect_peeling_hash_insert): Add supportable_if_not_aligned
3319 argument and do not call vect_supportable_dr_alignment here.
3320 (vect_peeling_supportable): Adjust.
3321 (vect_enhance_data_refs_alignment): Compute whether the
3322 access is supported with different alignment here and
3323 pass that down to vect_peeling_hash_insert.
3324 (vect_vfa_access_size): Adjust.
3325 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
3326 (vect_get_load_cost): Likewise.
3327 (get_negative_load_store_type): Likewise.
3328 (get_group_load_store_type): Likewise.
3329 (get_load_store_type): Likewise.
3331 2021-10-19 Martin Liska <mliska@suse.cz>
3334 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
3335 * system.h (strip_whilespaces): New function.
3337 2021-10-19 dianhong xu <dianhong.xu@intel.com>
3339 * config/i386/avx512fp16intrin.h:
3340 (_mm512_set1_pch): New intrinsic.
3341 * config/i386/avx512fp16vlintrin.h:
3342 (_mm256_set1_pch): New intrinsic.
3343 (_mm_set1_pch): Ditto.
3345 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
3347 PR tree-optimization/102796
3348 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
3349 normally. Return get_tree_range for non gimple_range_ssa_p names.
3350 (gimple_ranger::range_of_stmt): Use get_tree_range for non
3351 gimple_range_ssa_p names.
3353 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
3356 * config/i386/i386.c (ix86_print_operand_address):
3357 Error out for non-address_operand asm operands.
3359 2021-10-18 Richard Biener <rguenther@suse.de>
3361 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
3362 not auto-convert dr_alignment_support to bool.
3363 (vect_peeling_supportable): Likewise.
3364 (vect_enhance_data_refs_alignment): Likewise.
3365 (vect_supportable_dr_alignment): Commonize read/write case.
3366 * tree-vect-stmts.c (vect_get_store_cost): Use
3367 dr_alignment_support, not int, for the vect_supportable_dr_alignment
3369 (vect_get_load_cost): Likewise.
3371 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
3373 * tree-object-size.c (object_sizes_execute): Consolidate LHS
3374 null check and do it early.
3376 2021-10-18 Richard Biener <rguenther@suse.de>
3378 * tree-vect-stmts.c (vectorizable_store): Use the
3379 computed alignment scheme instead of querying
3382 2021-10-18 Richard Biener <rguenther@suse.de>
3384 * tree-vect-stmts.c (vectorizable_store): Do not recompute
3385 alignment scheme already determined by get_load_store_type.
3387 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3389 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
3390 Clone corresponding pass.
3392 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3394 * combine.c (recog_for_combine): For an unrecognized move/set of
3395 a constant, try force_const_mem to place it in the constant pool.
3396 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
3397 Delete global variables (that are no longer assigned a cost value).
3398 (cse_insn): Simplify logic for deciding whether to place a folded
3399 constant in the constant pool using force_const_mem.
3400 (cse_main): Remove zero initialization of constant_pool_entries_cost
3401 and constant_pool_entries_regcost.
3402 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
3403 fractionally more expensive, when optimizing for speed.
3405 2021-10-18 Martin Liska <mliska@suse.cz>
3407 PR gcov-profile/102746
3408 PR gcov-profile/102747
3409 * gcov.c (main): Return return_code.
3410 (output_gcov_file): Mark return_code when error happens.
3411 (generate_results): Likewise.
3412 (read_graph_file): Likewise.
3413 (read_count_file): Likewise.
3415 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3417 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
3418 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
3419 a popcount, allowing compile-time evaluation/simplification.
3420 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
3422 2021-10-18 Richard Biener <rguenther@suse.de>
3424 PR tree-optimization/102788
3425 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
3427 (vect_pattern_recog_1): Likewise.
3428 (vect_recog_bool_pattern): Continue matching the pattern
3429 even if we do not have a vector type for a conversion
3432 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3434 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
3435 Evalute SS_NEG and SS_ABS of a constant argument.
3437 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
3440 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
3441 * internal-fn.def: Add entry for COND_NEG internal function.
3442 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
3443 NEGATE_EXPR, COND_NEG mapping.
3444 * optabs.def: Add entry for cond_neg_optab.
3445 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
3446 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
3447 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
3449 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3451 * tree-ssa-strlen.c (class strlen_pass): Rename from
3453 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
3454 (handle_assign): Same.
3455 (adjust_last_stmt): Same.
3456 (maybe_warn_overflow): Same.
3457 (handle_builtin_strlen): Same.
3458 (handle_builtin_strchr): Same.
3459 (handle_builtin_strcpy): Same.
3460 (handle_builtin_strncat): Same.
3461 (handle_builtin_stxncpy_strncat): Same.
3462 (handle_builtin_memcpy): Same.
3463 (handle_builtin_strcat): Same.
3464 (handle_alloc_call): Same.
3465 (handle_builtin_memset): Same.
3466 (handle_builtin_memcmp): Same.
3467 (get_len_or_size): Same.
3468 (strxcmp_eqz_result): Same.
3469 (handle_builtin_string_cmp): Same.
3470 (handle_pointer_plus): Same.
3471 (count_nonzero_bytes_addr): Same.
3472 (count_nonzero_bytes): Same.
3473 (handle_store): Same.
3474 (strlen_check_and_optimize_call): Same.
3475 (handle_integral_assign): Same.
3476 (check_and_optimize_stmt): Same.
3477 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
3479 2021-10-18 Richard Biener <rguenther@suse.de>
3481 PR tree-optimization/102798
3482 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
3483 Only copy points-to info to newly generated SSA names.
3485 2021-10-18 Martin Liska <mliska@suse.cz>
3487 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
3488 * gcov.c (get_cycles_count): Likewise.
3489 * lto-compress.c (lto_compression_zlib): Likewise.
3490 (lto_uncompression_zlib): Likewise.
3491 * targhooks.c (default_pch_valid_p): Likewise.
3493 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
3495 * tree-pass.h (make_pass_thread_jumps_full): New.
3496 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
3497 (try_thread_blocks): Add resolve and speed arguments.
3498 (pass_thread_jumps::execute): Inline.
3499 (do_early_thread_jumps): New.
3500 (do_thread_jumps): New.
3501 (make_pass_thread_jumps): Move.
3502 (pass_early_thread_jumps::gate): Inline.
3503 (pass_early_thread_jumps::execute): Inline.
3504 (class pass_thread_jumps_full): New.
3506 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
3508 * configure.ac: Treat powerpc64*-*-freebsd* the same as
3509 powerpc64-*-freebsd*.
3510 * configure: Regenerate.
3512 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
3514 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
3515 set_ptr_nonull to set_ptr_nonnull in comments.
3517 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
3519 PR tree-optimization/102720
3520 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
3521 of call used and clobbered sets.
3523 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3525 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
3526 skip gomp and itm when r or nodefaultlibs is given.
3527 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
3528 Update ordering of exclusions, remove duplicate 'v' addition
3529 (collect2 will add this from the main command line).
3531 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3533 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
3534 filelist and framework options in specs instead of code.
3535 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
3536 specs that are really driver ones.
3537 (DARWIN_CC1_SPEC): Likewise.
3538 (CPP_SPEC): Likewise.
3539 (SYSROOT_SPEC): Append space.
3540 (LINK_SYSROOT_SPEC): Remove most driver link specs.
3541 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
3542 (STARTFILE_SPEC): Likewise.
3543 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
3544 (ASM_SPEC): Update driver-related specs.
3545 (ASM_FINAL_SPEC): Likewise.
3546 * config/darwin.opt: Remove now unused option aliases.
3547 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
3548 (ASM_SPEC): Update driver-related specs.
3550 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
3552 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
3553 allow standard_sse_constant_p immediate constants.
3555 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3557 * config.gcc: Add tm-dwarf2.h to tm_d-file.
3559 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
3561 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
3562 that occur in abnormal phis.
3563 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
3564 abnormal and EH edges.
3565 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
3566 (rvrp_folder::value_on_edge): Ditto.
3567 (rvrp_folder::value_of_stmt): Ditto.
3568 (hybrid_folder::value_of_expr): Ditto for ranger queries.
3569 (hybrid_folder::value_on_edge): Ditto.
3570 (hybrid_folder::value_of_stmt): Ditto.
3571 * value-query.cc (gimple_range_global): Always return a range if
3572 the type is supported.
3574 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
3576 * config/pa/pa.md: Consistently use "rG" constraint for copy
3577 instruction in move patterns.
3579 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
3581 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
3584 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
3585 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
3586 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
3587 (ELFABIVERSION_AMDGPU_HSA_V4): New.
3588 (SET_SRAM_ECC_UNSUPPORTED): New.
3589 (copy_early_debug_info): Create elf flags to match the other objects.
3590 (main): Just let the attribute flags pass through.
3592 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3594 * tree-loop-distribution.c (reduction_var_overflows_first):
3595 Pass the type of reduction_var as first argument as it is also
3596 done for the load type.
3597 (loop_distribution::transform_reduction_loop): Add missing
3598 TREE_TYPE while determining precission of reduction_var.
3600 2021-10-15 Richard Biener <rguenther@suse.de>
3602 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
3604 * toplev.c (process_options): Warn when STABS debugging is
3605 enabled but not the preferred format.
3606 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
3608 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
3610 2021-10-15 Richard Biener <rguenther@suse.de>
3613 * tree-inline.c (copy_bb): Avoid underflowing nargs.
3615 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3617 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
3618 HFmode input operand to HImode.
3619 (ix86_vectorize_vec_perm_const): Likewise.
3620 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
3622 (*avx512f_permvar_truncv8siv8hi_1_hf):
3625 2021-10-15 Richard Biener <rguenther@suse.de>
3627 PR middle-end/102682
3628 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
3629 not create a paradoxical subreg.
3631 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3633 * config/i386/i386-expand.c (ix86_expand_vector_init):
3634 For half_vector concat for HFmode, handle them like HImode.
3636 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3638 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
3639 (back_threader::back_threader): Same.
3640 (back_threader::resolve_phi): Try to solve without looking back if
3642 (back_threader::find_paths_to_names): Same.
3643 (try_thread_blocks): Pass resolve argument to back threader.
3644 (pass_early_thread_jumps::execute): Same.
3646 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3648 * doc/invoke.texi: Remove max-fsm-thread-length,
3649 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
3651 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
3653 (back_threader_registry::back_threader_registry): Same.
3654 (back_threader_profitability::profitable_path_p): Remove
3655 param_max_fsm_thread-length.
3656 (back_threader_registry::register_path): Remove
3657 m_max_allowable_paths.
3659 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3661 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
3663 (back_threader::~back_threader): Do not release m_path.
3665 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
3667 PR tree-optimization/102738
3668 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
3670 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
3672 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
3673 (omp_mark_declare_variant): Move from c-omp.c.
3674 (omp_context_name_list_prop): Update for Fortran strings.
3675 * omp-general.h (omp_check_context_selector): New prototype.
3676 (omp_mark_declare_variant): New prototype.
3678 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3680 * config/rs6000/rs6000.c (rs6000_density_test): Move early
3681 exit test further up the function.
3683 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3685 * config/arm/arm.c (arm_add_stmt_cost): Delete.
3686 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3688 2021-10-14 Martin Jambor <mjambor@suse.cz>
3690 * doc/invoke.texi (Optimize Options): Add entry for
3691 ipa-cp-recursive-freq-factor.
3693 2021-10-14 Tamar Christina <tamar.christina@arm.com>
3695 * match.pd: New rule.
3697 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
3700 * ipa-modref-tree.h (modref_access_node::update2):
3701 Also check that parm_offset is unchanged.
3702 (modref_ref_node::insert_access): Fix updating of
3705 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3707 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
3710 2021-10-14 Richard Biener <rguenther@suse.de>
3712 PR tree-optimization/102659
3713 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
3714 rewrite pointer typed undefined overflow operations.
3715 (predicate_statements): Likewise. Make sure to emit invariant
3716 conversions in the preheader.
3717 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
3718 non-empty preheaders.
3719 * tree-data-ref.c (dr_analyze_indices): Strip useless
3720 conversions to the MEM_REF base type.
3722 2021-10-14 Martin Liska <mliska@suse.cz>
3724 * common.opt: Stop using AUTODETECT_VALUE
3725 and use EnabledBy where possible.
3726 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
3727 * toplev.c (AUTODETECT_VALUE): Remove macro.
3728 (process_options): Simplify by using EnabledBy and
3729 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
3732 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
3734 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
3736 (vld1_s32_x3): Likewise.
3738 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3740 PR tree-optimization/102736
3741 PR tree-optimization/102736
3742 * gimple-range-path.cc (path_range_query::range_on_path_entry):
3743 Assert that the requested range is defined outside the path.
3744 (path_range_query::ssa_range_in_phi): Do not call
3745 range_on_path_entry for SSA names that are defined within the
3748 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
3750 * config/darwin-driver.c (darwin_find_version_from_kernel):
3751 Quote internal identifiers and avoid contractions in
3753 (darwin_default_min_version): Likewise.
3754 (darwin_driver_init): Likewise.
3756 2021-10-14 Martin Jambor <mjambor@suse.cz>
3758 * params.opt (ipa-cp-recursive-freq-factor): New.
3759 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
3760 scc_no, self_recursion_generated_level, same_scc and
3761 self_recursion_generated_p.
3762 (ipcp_lattice::add_value): Replaced parameter unlimited with
3763 same_lat_gen_level, usit it determine limit of values and store it to
3765 (ipcp_lattice<valtype>::print): Dump the new fileds.
3766 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
3767 parameter and store it to the new value.
3768 (self_recursively_generated_p): Removed.
3769 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
3770 instead of self_recursively_generated_p, store self generation level
3772 (value_topo_info<valtype>::add_val): Set scc_no.
3773 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
3774 recursively feeding values and self generated values by appropriate
3777 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3779 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
3780 redundant test for flag_vect_cost_model.
3782 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3784 * bitmap.c (debug): New overloaded function for auto_bitmaps.
3785 * bitmap.h (debug): Same.
3787 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
3789 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
3790 Adjust builtin call.
3791 (_mm512_mask3_fcmadd_pch): Likewise.
3792 (_mm512_mask_fmadd_pch): Likewise
3793 (_mm512_mask3_fmadd_pch): Likewise
3794 (_mm512_mask_fcmadd_round_pch): Likewise
3795 (_mm512_mask3_fcmadd_round_pch): Likewise
3796 (_mm512_mask_fmadd_round_pch): Likewise
3797 (_mm512_mask3_fmadd_round_pch): Likewise
3798 (_mm_mask_fcmadd_sch): Likewise
3799 (_mm_mask3_fcmadd_sch): Likewise
3800 (_mm_mask_fmadd_sch): Likewise
3801 (_mm_mask3_fmadd_sch): Likewise
3802 (_mm_mask_fcmadd_round_sch): Likewise
3803 (_mm_mask3_fcmadd_round_sch): Likewise
3804 (_mm_mask_fmadd_round_sch): Likewise
3805 (_mm_mask3_fmadd_round_sch): Likewise
3806 (_mm_fcmadd_round_sch): Likewise
3807 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
3808 Adjust builtin call.
3809 (_mm_mask3_fmadd_pch): Likewise
3810 (_mm256_mask_fmadd_pch): Likewise
3811 (_mm256_mask3_fmadd_pch): Likewise
3812 (_mm_mask_fcmadd_pch): Likewise
3813 (_mm_mask3_fcmadd_pch): Likewise
3814 (_mm256_mask_fcmadd_pch): Likewise
3815 (_mm256_mask3_fcmadd_pch): Likewise
3816 * config/i386/i386-builtin.def: Add mask3 builtin for complex
3817 fma, and adjust mask_builtin to corresponding expander.
3818 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3819 Skip eraseing embedded rounding for expanders that emits
3821 * config/i386/sse.md (complexmove): New mode_attr.
3822 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
3823 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
3824 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
3825 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
3826 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
3827 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
3828 * config/i386/subst.md (round_embedded_complex): New subst.
3830 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
3832 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
3833 (cbranchdf4): Likewise.
3834 Add missing move patterns for TARGET_SOFT_FLOAT.
3836 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
3838 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
3839 pseudo intermediate when moving a SUBREG into a hard register,
3840 by checking ix86_hardreg_mov_ok.
3841 (ix86_expand_vector_extract): Store zero-extended SImode
3842 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
3844 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
3845 complex (SUBREG) sets of (vector) hard registers before reload, by
3846 checking ix86_hardreg_mov_ok.
3848 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
3850 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
3852 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
3854 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
3856 2021-10-13 Martin Sebor <msebor@redhat.com>
3858 PR middle-end/102630
3859 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
3861 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
3863 * collect2.c (is_lto_object_file): Release simple-object
3864 resources, close files.
3866 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
3868 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
3869 when checking mve addressing modes.
3870 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
3872 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
3873 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
3874 there is a single Ux constraint.
3876 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
3878 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
3879 multiplication on 32-bit PA 1.1 and 2.0 targets.
3881 2021-10-13 Martin Liska <mliska@suse.cz>
3883 PR gcov-profile/90364
3884 * coverage.c (build_info): Emit checksum to the global variable.
3885 (build_info_type): Add new field for checksum.
3886 (coverage_obj_finish): Pass object_checksum.
3887 (coverage_init): Use 0 as checksum for .gcno files.
3888 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
3889 * gcov.c (read_graph_file): Read also checksum.
3890 * doc/invoke.texi: Document the behaviour change.
3892 2021-10-13 Richard Biener <rguenther@suse.de>
3894 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
3895 start at integer value 2.
3896 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
3897 the iterator for GSI_LAST_NEW_STMT.
3898 (gsi_insert_seq_nodes_after): Likewise.
3899 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
3900 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
3901 arguments to gsi_replace.
3903 2021-10-13 Martin Liska <mliska@suse.cz>
3906 * common.opt: Use EnabledBy instead of detection in
3907 finish_options and process_options.
3908 * opts.c (finish_options): Remove handling of
3909 x_flag_unroll_all_loops.
3910 * toplev.c (process_options): Likewise for flag_web and
3911 flag_rename_registers.
3913 2021-10-13 Richard Biener <rguenther@suse.de>
3915 PR tree-optimization/102659
3916 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
3917 (if_convertible_gimple_assign_stmt_p): Mark the loop for
3918 rewrite when stmts with undefined behavior on integer
3920 (combine_blocks): Predicate also when we need to rewrite stmts.
3921 (predicate_statements): Rewrite affected stmts to something
3922 with well-defined behavior on overflow.
3923 (tree_if_conversion): Initialize need_to_rewrite_undefined.
3925 2021-10-13 Richard Biener <rguenther@suse.de>
3928 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
3931 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
3933 * dwarf2ctf.c: Fix typo in comment.
3935 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
3938 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
3939 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
3940 avoid generating an automatic #define of vec_cpsgn. Use the
3941 correct built-in for V4SFmode that doesn't depend on VSX.
3943 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
3947 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
3948 define_insn_and_split pattern. Add alternative 1 and split it
3949 post reload to insert operand 1 into the low part of operand 0.
3950 (*sub<mode>_1_slp): Ditto.
3951 (*and<mode>_1_slp): Ditto.
3952 (*<any_or:code><mode>_1_slp): Ditto.
3953 (*ashl<mode>3_1_slp): Ditto.
3954 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
3955 (*<any_rotate:insn><mode>3_1_slp): Ditto.
3956 (*neg<mode>_1_slp): New insn_and_split pattern.
3957 (*one_cmpl<mode>_1_slp): Ditto.
3959 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
3961 * doc/install.texi: Update MinGW and mingw-64 Binaries
3964 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
3966 * doc/extend.texi (Common Variable Attributes): Fix typos in
3967 alloc_size documentation.
3969 2021-10-12 Richard Biener <rguenther@suse.de>
3971 PR tree-optimization/102696
3972 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
3973 the tree fatally failed when we reject a BIT_FIELD_REF.
3975 2021-10-12 Richard Biener <rguenther@suse.de>
3977 PR tree-optimization/102572
3978 * tree-vect-stmts.c (vect_build_gather_load_calls): When
3979 gathering the vectorized defs for the mask pass in the
3980 desired mask vector type so invariants will be handled
3983 2021-10-12 Tamar Christina <tamar.christina@arm.com>
3985 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
3986 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
3987 *fcmuo<mode>_nor_combine): New.
3989 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
3992 * config/sparc/sparc-modes.def (OI): New integer mode.
3994 2021-10-12 Jakub Jelinek <jakub@redhat.com>
3996 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
3997 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
4000 2021-10-12 Jakub Jelinek <jakub@redhat.com>
4002 * tree-vectorizer.h (loop_cost_model): New function.
4003 (unlimited_cost_model): Use it.
4004 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
4005 call instead of flag_vect_cost_model.
4006 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
4007 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
4008 of flag_simd_cost_model.
4010 2021-10-12 liuhongt <hongtao.liu@intel.com>
4013 * config/i386/i386-expand.c (emit_reduc_half): Handle
4015 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
4016 (reduc_plus_scal_v4qi): Ditto.
4018 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4020 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
4021 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
4022 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
4024 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4026 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
4027 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
4028 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
4029 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
4030 _mm_cvtepu32_epi64): New.
4032 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4034 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
4035 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
4037 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4039 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
4040 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
4041 _mm_max_epi32, _mm_max_epu32): New.
4043 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
4045 * ipa-modref-tree.h (struct modref_access_node): Revert
4047 (struct modref_ref_node): Likewise.
4049 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
4051 * ipa-modref-tree.h (modref_tree::global_access_p): New member
4054 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
4055 ignore_stores_eaf_flags): Move to ipa-modref.h
4056 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
4057 (modref_summary::global_memory_read_p): New member function.
4058 (modref_summary::global_memory_written_p): New member function.
4059 * ipa-modref.h (modref_summary::global_memory_read_p,
4060 modref_summary::global_memory_written_p): Declare.
4061 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
4062 ignore_stores_eaf_flags): move here.
4063 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
4065 (handle_rhs_call): Rewrite.
4066 (handle_call_arg): New function.
4067 (determine_global_memory_access): New function.
4068 (handle_const_call): Remove
4069 (handle_pure_call): Remove
4070 (find_func_aliases_for_call): Update use of handle_rhs_call.
4071 (compute_points_to_sets): Handle global memory acccesses
4074 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
4076 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
4077 documentation, mention UBSAN_OPTIONS, similar to what is done
4078 for AddressSanitizer.
4080 2021-10-11 Richard Biener <rguenther@suse.de>
4082 PR middle-end/102683
4083 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
4084 availability before building an integer type for storage
4087 2021-10-11 Richard Biener <rguenther@suse.de>
4089 PR middle-end/101480
4090 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
4093 2021-10-11 Martin Liska <mliska@suse.cz>
4095 * common.opt: Remove Init(2) for some options.
4096 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
4097 use rather OPTION_SET_P.
4099 2021-10-11 Martin Liska <mliska@suse.cz>
4101 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
4102 * flag-types.h (enum ira_region): Likewise.
4103 * toplev.c (process_options): Use OPTION_SET_P instead of
4104 IRA_REGION_AUTODETECT.
4106 2021-10-11 Jakub Jelinek <jakub@redhat.com>
4108 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
4109 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
4111 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4113 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
4114 * config/s390/s390.c (s390_rawmemchr): New function.
4115 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
4116 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
4117 the pattern vfees<mode> from vx-builtins.md.
4118 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
4120 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4122 * builtins.c (get_memory_rtx): Change to external linkage.
4123 * builtins.h (get_memory_rtx): Add function prototype.
4124 * doc/md.texi (rawmemchr<mode>): Document.
4125 * internal-fn.c (expand_RAWMEMCHR): Define.
4126 * internal-fn.def (RAWMEMCHR): Add.
4127 * optabs.def (rawmemchr_optab): Add.
4128 * tree-loop-distribution.c (find_single_drs): Change return code
4129 behaviour by also returning true if no single store was found
4131 (loop_distribution::classify_partition): Respect the new return
4132 code behaviour of function find_single_drs.
4133 (loop_distribution::execute): Call new function
4134 transform_reduction_loop in order to replace rawmemchr or strlen
4135 like loops by calls into builtins.
4136 (generate_reduction_builtin_1): New function.
4137 (generate_rawmemchr_builtin): New function.
4138 (generate_strlen_builtin_1): New function.
4139 (generate_strlen_builtin): New function.
4140 (generate_strlen_builtin_using_rawmemchr): New function.
4141 (reduction_var_overflows_first): New function.
4142 (determine_reduction_stmt_1): New function.
4143 (determine_reduction_stmt): New function.
4144 (loop_distribution::transform_reduction_loop): New function.
4146 2021-10-11 Martin Liska <mliska@suse.cz>
4148 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
4149 and remove legacy hashing code.
4151 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
4154 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
4155 CONST_INT for BEGIN and END, and use gcc_assert rather than
4158 2021-10-10 Jakub Jelinek <jakub@redhat.com>
4161 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
4162 use MO_VAL_SET if loc is not sp.
4164 2021-10-10 Andrew Pinski <apinski@marvell.com>
4166 PR tree-optimization/102622
4167 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
4168 Swap the order of a?0:pow2cst and a?0:-1 transformations.
4170 2021-10-09 liuhongt <hongtao.liu@intel.com>
4173 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
4175 (ix86_use_mask_cmp_p): Ditto.
4176 (ix86_expand_sse_movcc): Ditto.
4177 * config/i386/i386.md (setcc_hf_mask): New define_insn.
4178 (movhf_mask): Ditto.
4179 (UNSPEC_MOVCC_MASK): New unspec.
4180 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
4182 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
4184 PR rtl-optimization/102627
4185 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
4187 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
4189 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
4192 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
4194 * value-range.cc (irange::debug): New.
4195 * value-range.h (irange::debug): New.
4197 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
4199 PR tree-optimization/102385
4200 * predict.h (change_edge_frequency): Declare.
4201 * predict.c (change_edge_frequency): New function.
4202 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
4204 (tree_unroll_loop): Likewise.
4205 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
4206 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
4208 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
4209 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
4210 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
4211 to retrieve the exit edges. Make all the old profile update code
4212 conditional on !single_loop_p -- the case it was written for --
4213 and use a different approach for the single-loop case.
4215 2021-10-08 Martin Liska <mliska@suse.cz>
4217 * config/alpha/alpha.c (alpha_option_override): Use new macro
4219 * config/arc/arc.c (arc_override_options): Likewise.
4220 * config/arm/arm.c (arm_option_override): Likewise.
4221 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
4222 * config/c6x/c6x.c (c6x_option_override): Likewise.
4223 * config/csky/csky.c: Likewise.
4224 * config/darwin.c (darwin_override_options): Likewise.
4225 * config/frv/frv.c (frv_option_override): Likewise.
4226 * config/i386/djgpp.h: Likewise.
4227 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
4228 (ix86_max_noce_ifcvt_seq_cost): Likewise.
4229 * config/ia64/ia64.c (ia64_option_override): Likewise.
4230 (ia64_override_options_after_change): Likewise.
4231 * config/m32c/m32c.c (m32c_option_override): Likewise.
4232 * config/m32r/m32r.c (m32r_init): Likewise.
4233 * config/m68k/m68k.c (m68k_option_override): Likewise.
4234 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
4235 * config/mips/mips.c (mips_option_override): Likewise.
4236 * config/nios2/nios2.c (nios2_option_override): Likewise.
4237 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
4238 * config/pa/pa.c (pa_option_override): Likewise.
4239 * config/riscv/riscv.c (riscv_option_override): Likewise.
4240 * config/rs6000/aix71.h: Likewise.
4241 * config/rs6000/aix72.h: Likewise.
4242 * config/rs6000/aix73.h: Likewise.
4243 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
4244 (rs6000_override_options_after_change): Likewise.
4245 (rs6000_linux64_override_options): Likewise.
4246 (glibc_supports_ieee_128bit): Likewise.
4247 (rs6000_option_override_internal): Likewise.
4248 (rs6000_file_start): Likewise.
4249 (rs6000_darwin_file_start): Likewise.
4250 * config/rs6000/rtems.h: Likewise.
4251 * config/rs6000/sysv4.h: Likewise.
4252 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
4253 * config/s390/s390.c (s390_option_override): Likewise.
4254 * config/sh/linux.h: Likewise.
4255 * config/sh/netbsd-elf.h (while): Likewise.
4256 * config/sh/sh.c (sh_option_override): Likewise.
4257 * config/sol2.c (solaris_override_options): Likewise.
4258 * config/sparc/sparc.c (sparc_option_override): Likewise.
4259 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
4260 * config/visium/visium.c (visium_option_override): Likewise.
4261 * config/vxworks.c (vxworks_override_options): Likewise.
4262 * lto-opts.c (lto_write_options): Likewise.
4263 * omp-expand.c (expand_omp_simd): Likewise.
4264 * omp-general.c (omp_max_vf): Likewise.
4265 * omp-offload.c (oacc_xform_loop): Likewise.
4266 * opts.h (OPTION_SET_P): Likewise.
4267 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
4268 * toplev.c (process_options): Likewise.
4269 * tree-predcom.c: Likewise.
4270 * tree-sra.c (analyze_all_variable_accesses): Likewise.
4272 2021-10-08 liuhongt <hongtao.liu@intel.com>
4275 * config/i386/i386.c (ix86_optab_supported_p):
4276 Return true for HFmode.
4277 * match.pd: Simplify (_Float16) ceil ((double) x) to
4278 __builtin_ceilf16 (a) when a is _Float16 type and
4279 direct_internal_fn_supported_p.
4281 2021-10-08 liuhongt <hongtao.liu@intel.com>
4284 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
4285 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
4286 (reduc_<code>_scal_v4hi): New.
4288 2021-10-08 liuhongt <hongtao.liu@intel.com>
4290 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
4291 * doc/invoke.texi (Options That Control Optimization): Update
4293 * opts.c (default_options_table): Enable auto-vectorization at
4294 O2 with very-cheap cost model.
4295 (finish_options): Use cheap cost model for
4296 explicit -ftree{,-loop}-vectorize.
4298 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4300 * ctfc.c (ctfc_delete_container): Free hash table contents.
4302 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4304 * toplev.c (process_options): Do not warn for GNU GIMPLE.
4306 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
4308 * tree-object-size.c (addr_object_size,
4309 compute_builtin_object_size): Drop PDECL and POFF arguments.
4310 (addr_object_size): Adjust calls.
4311 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
4314 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
4316 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
4317 representing signed and unsigned high-part multiplication resp.
4318 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
4319 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
4320 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
4321 SS_DIV, US_DIV]: Similar simplifications for saturating
4323 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
4324 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
4325 Implement compile-time evaluation for constant operands.
4326 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
4328 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
4329 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
4330 Mention the new smul_highpart and umul_highpart RTX codes.
4331 * doc/invoke.texi: Silence @xref "compilation" warnings.
4333 2021-10-07 Martin Jambor <mjambor@suse.cz>
4336 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
4337 case when the source reference description corresponds to a
4338 referance taken in a function src->caller is inlined to.
4340 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
4343 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
4345 (modref_access_node::try_merge_with): Add sanity check that there
4346 are no redundant entries in the list.
4348 2021-10-07 Richard Biener <rguenther@suse.de>
4350 PR tree-optimization/102608
4351 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
4354 2021-10-07 Martin Liska <mliska@suse.cz>
4356 * toplev.c (toplev::main): Make
4357 save_opt_decoded_options a pointer type
4358 * toplev.h: Likewise.
4360 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4362 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
4363 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
4364 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
4366 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4368 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
4369 Adjust for new -msram-ecc=any behaviour.
4370 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
4371 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
4372 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
4373 (EF_AMDGPU_XNACK_V3): ... this.
4374 (EF_AMDGPU_SRAM_ECC): Rename to ...
4375 (EF_AMDGPU_SRAM_ECC_V3): ... this.
4376 (EF_AMDGPU_FEATURE_XNACK_V4): New.
4377 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
4378 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
4379 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
4380 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
4381 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
4382 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
4383 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
4384 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
4385 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
4386 (SET_XNACK_ON): New.
4387 (SET_XNACK_OFF): New.
4389 (SET_SRAM_ECC_ON): New.
4390 (SET_SRAM_ECC_ANY): New.
4391 (SET_SRAM_ECC_OFF): New.
4392 (TEST_SRAM_ECC_ANY): New.
4393 (TEST_SRAM_ECC_ON): New.
4394 (main): Implement HSACOv4 and -msram-ecc=any.
4396 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4398 * config.in: Regenerate.
4399 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
4403 (A_FIJI): Rename to ...
4405 (A_900): Rename to ...
4407 (A_906): Rename to ...
4409 (A_908): Rename to ...
4411 (SRAMOPT): New macro.
4412 (ASM_SPEC): Adjust xnack option usage.
4413 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
4414 * configure: Regenerate.
4415 * configure.ac: Detect LLVM assembler dialect.
4417 2021-10-07 Richard Biener <rguenther@suse.de>
4419 * tree-pretty-print.c (dump_generic_node): Do not elide
4420 printing '&' when dumping with -gimple.
4422 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4424 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
4426 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
4427 (adjust_imagpart_expr): Ditto.
4428 * value-range.cc (irange::irange_intersect): Call new routine if
4429 RHS is a single pair.
4430 (irange::intersect): New wide_int version.
4431 * value-range.h (class irange): New prototype.
4433 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4435 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
4436 Add parameter to limit size when recognizing switches.
4437 (gimple_outgoing_range::edge_range_p): Check size limit.
4438 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
4439 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
4440 that exceed the size limit.
4441 (gori_compute::gori_compute): Add initializer.
4442 * params.opt (evrp-switch-limit): New.
4443 * doc/invoke.texi: Update docs.
4445 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4447 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
4448 TYPE_MAX_VALUE instead of creating new trees when possible.
4450 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4452 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
4453 zero and non-zero more efficently.
4455 2021-10-06 Richard Biener <rguenther@suse.de>
4458 * dumpfile.h (TDF_GIMPLE_VAL): New.
4459 (dump_flag): Re-order and adjust TDF_* flags. Make
4460 the enum uint32_t. Use std::underlying_type in the
4462 (optgroup_flag): Likewise for the operator overloads.
4463 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
4464 in _Literal if TDF_GIMPLE_VAL.
4465 * gimple-pretty-print.c (dump_gimple_assign): Add
4466 TDF_GIMPLE_VAL to flags when dumping operands where only
4467 is_gimple_val are allowed.
4468 (dump_gimple_cond): Likewise.
4470 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
4472 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
4475 2021-10-05 qing zhao <qing.zhao@oracle.com>
4477 PR middle-end/102359
4478 * gimplify.c (gimplify_decl_expr): Not add initialization for an
4479 auto variable when it has been initialized by frontend.
4481 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
4483 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
4486 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4488 * common/config/avr/avr-common.c (avr_handle_option): Mark
4489 argument as ATTRIBUTE_UNUSED.
4491 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4493 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
4494 Undefine before redefinition.
4496 2021-10-05 Richard Biener <rguenther@suse.de>
4498 * toplev.c (no_backend): Remove global var.
4499 (process_options): Pass in no_backend, move post_options
4500 langhook call to toplev::main.
4501 (do_compile): Pass in no_backend, move process_options call
4503 (toplev::run_self_tests): Check no_backend at the caller.
4504 (toplev::main): Call post_options and process_options
4505 split out from do_compile, do self-tests only if
4506 no_backend is initialized.
4508 2021-10-05 Richard Biener <rguenther@suse.de>
4510 * tree-cfg.c (dump_function_to_file): Dump the UID of the
4511 function as part of the name when requested.
4512 * tree-pretty-print.c (dump_function_name): Dump the UID when
4513 requested and the langhook produced the actual name.
4515 2021-10-05 Richard Biener <rguenther@suse.de>
4517 PR middle-end/102587
4518 PR middle-end/102285
4519 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
4520 zero-initialization as last resort, use the constant
4521 size as given by the DEFERRED_INIT argument to build
4524 2021-10-04 Marek Polacek <polacek@redhat.com>
4527 * doc/invoke.texi: Document -Warray-compare.
4529 2021-10-04 Richard Biener <rguenther@suse.de>
4531 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
4532 variables are not to be initialized.
4534 2021-10-04 Richard Biener <rguenther@suse.de>
4536 * expr.h (non_mem_decl_p): Declare.
4537 (mem_ref_refers_to_non_mem_p): Likewise.
4538 * expr.c (non_mem_decl_p): Export.
4539 (mem_ref_refers_to_non_mem_p): Likewise.
4540 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
4541 but check the base with mem_ref_refers_to_non_mem_p
4544 2021-10-04 Richard Biener <rguenther@suse.de>
4546 PR tree-optimization/102570
4547 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
4548 we are using clique for the internal function code.
4549 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
4550 internal function code.
4551 (print_vn_reference_ops): Print the internal function code.
4552 (vn_reference_op_compute_hash): Hash it.
4553 (copy_reference_ops_from_call): Record it.
4554 (visit_stmt): Remove the restriction around internal function
4556 (fully_constant_vn_reference_p): Use fold_const_call and handle
4558 (vn_reference_eq): Compare call return types.
4559 * tree-ssa-pre.c (create_expression_by_pieces): Handle
4560 generating calls to internal functions.
4561 (compute_avail): Remove the restriction around internal function
4564 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
4566 PR tree-optimization/102560
4567 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
4568 marker for invalid_range.
4570 2021-10-04 Richard Biener <rguenther@suse.de>
4572 PR middle-end/102587
4573 * internal-fn.c (expand_DEFERRED_INIT): Guard register
4574 initialization path an avoid initializing VLA registers
4577 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
4579 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
4581 2021-10-03 Martin Liska <mliska@suse.cz>
4583 * toplev.c (toplev::main): Check opt_index if it is a part
4586 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
4588 PR tree-optimization/102563
4589 * range-op.cc (operator_lshift::op1_range): Do not clobber
4592 2021-10-02 Martin Liska <mliska@suse.cz>
4594 * toplev.c (toplev::main): save_decoded_options[0] is program
4595 name and so it should be skipped.
4597 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4599 PR tree-optimization/102546
4600 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
4601 X << Y is non-zero implies X is also non-zero.
4603 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4605 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4607 * config/aarch64/aarch64-tune.md: Regenerate.
4608 * doc/invoke.texi: Update docs.
4610 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4612 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4614 * config/aarch64/aarch64-tune.md: Regenerate.
4615 * doc/invoke.texi: Update docs.
4617 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4619 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4621 * config/aarch64/aarch64-tune.md: Regenerate.
4622 * doc/invoke.texi: Update docs.
4624 2021-10-01 Martin Sebor <msebor@redhat.com>
4627 * doc/invoke.texi (-Waddress): Update.
4628 * gengtype.c (write_types): Avoid -Waddress.
4629 * poly-int.h (POLY_SET_COEFF): Avoid using null.
4631 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
4634 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
4637 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4639 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
4641 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4643 * gimple-range-path.cc (path_range_query::compute_ranges): Use
4645 * gimple-range-path.h (class path_range_query): Remove shadowed
4647 (path_range_query::get_path_oracle): New.
4649 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4650 Richard Biener <rguenther@suse.de>
4653 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
4654 INT_MIN / -1 division detection from here ...
4655 (-fsanitize=signed-integer-overflow): ... and add it here.
4657 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4659 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
4661 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
4662 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
4663 (AARCH64_ISA_V9): New ISA flag.
4664 * doc/invoke.texi: Update docs.
4666 2021-10-01 Martin Liska <mliska@suse.cz>
4668 * toplev.c (toplev::main): Save decoded optimization options.
4669 * toplev.h (save_opt_decoded_options): New.
4670 * doc/extend.texi: Be more clear about optimize and target
4673 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4675 * explow.c: Include langhooks.h.
4676 (set_stack_check_libfunc): Build a proper function type.
4678 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4681 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
4682 not return true for external weak function symbols in medium model.
4684 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4686 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
4687 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
4688 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
4689 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
4690 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
4691 OMP_CLAUSE_SCHEDULE_STATIC.
4693 2021-10-01 Richard Biener <rguenther@suse.de>
4695 PR middle-end/102518
4696 * tree-inline.c (setup_one_parameter): Avoid substituting
4697 an invariant into contexts where a GIMPLE register is not valid.
4699 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4701 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
4702 * config/arm/arm-tables.opt: Regenerate.
4703 * config/arm/arm-tune.md: Regenerate.
4704 * doc/invoke.texi: Update docs.
4706 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
4709 * config/i386/i386.md
4710 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
4713 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
4715 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
4716 omp_{c,re}alloc, fix omp_alloc/omp_free.
4718 2021-09-30 Martin Liska <mliska@suse.cz>
4720 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
4721 asm_out_file and stream directly to MYFILE.
4723 2021-09-30 Richard Biener <rguenther@suse.de>
4725 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4726 Restore and fix condition under which we apply npeel to
4727 the DRs misalignment value.
4729 2021-09-30 Richard Biener <rguenther@suse.de>
4731 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4732 Fix npeel check for variable amount of peeling.
4734 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4736 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
4738 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4740 * gimple-range.cc (gimple_ranger::debug): New.
4741 * gimple-range.h (class gimple_ranger): Add debug.
4743 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4745 PR middle-end/102519
4746 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
4748 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
4751 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
4752 (btf_finalize): Empty the hash_map btf_var_ids.
4754 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
4756 * tree-vrp.c (thread_through_all_blocks): Return bool.
4757 (execute_vrp_threader): Return TODO_* flags.
4758 (pass_data_vrp_threader): Set todo_flags_finish to 0.
4760 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
4762 * timevar.def (TV_TREE_VRP_THREADER): New.
4763 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
4765 2021-09-29 David Faust <david.faust@oracle.com>
4767 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
4769 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
4771 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
4772 new type qualifier enum.
4773 (TYPES_TERNOP_SSSU): Likewise.
4774 (TYPES_TERNOP_PPPU): Likewise.
4775 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
4776 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
4778 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
4779 builtin and remove casts.
4780 (vqtbl1_s8): Likewise.
4781 (vqtbl1q_p8): Likewise.
4782 (vqtbl1q_s8): Likewise.
4783 (vqtbx1_s8): Likewise.
4784 (vqtbx1_p8): Likewise.
4785 (vqtbx1q_s8): Likewise.
4786 (vqtbx1q_p8): Likewise.
4787 (vtbl1_p8): Likewise.
4788 (vtbl2_p8): Likewise.
4789 (vtbx2_p8): Likewise.
4791 2021-09-29 Richard Biener <rguenther@suse.de>
4793 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
4795 (vect_update_misalignment_for_peel): Use it to update
4796 misaligned to the value necessary for an aligned access.
4797 (vect_get_peeling_costs_all_drs): Likewise.
4798 (vect_enhance_data_refs_alignment): Likewise.
4800 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4802 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
4803 emitted operations and adjust heuristic for code size.
4805 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4807 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
4808 emitted operations and adjust heuristic for code size.
4810 2021-09-29 Jakub Jelinek <jakub@redhat.com>
4812 PR middle-end/102504
4813 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
4814 in OMP_SCOPE clauses, not just on worksharing construct clauses.
4816 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
4818 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
4819 (<u>mulditi3): Call <su>muldi3_highpart.
4820 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
4821 (<u>mulsidi3): Call <su>mulsi3_highpart.
4822 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
4824 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
4826 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
4828 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
4830 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
4831 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
4834 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4836 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
4837 (AARCH64_FL_V8_7): Likewise.
4838 (AARCH64_FL_FOR_ARCH8_7): Likewise.
4839 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
4840 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
4841 * doc/invoke.texi: Document the above.
4843 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4845 * dbgcnt.c (dbg_cnt_counter): New.
4846 * dbgcnt.h (dbg_cnt_counter): New.
4847 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
4848 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
4849 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
4850 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
4853 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4855 * cfgcleanup.c (pass_jump::execute): Check
4856 flag_expensive_optimizations.
4857 (pass_jump_after_combine::gate): Same.
4858 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
4859 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
4860 * tree-ssa-threadupdate.c
4861 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
4864 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
4866 * tree-ssa-reassoc.c (biased_names): New global.
4867 (propagate_bias_p): New function.
4868 (loop_carried_phi): Remove.
4869 (propagate_rank): Propagate bias along single uses.
4870 (get_rank): Update biased_names when needed.
4872 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
4874 * passes.def (pass_reassoc): Rename parameter to early_p.
4875 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
4877 (phi_rank): Don't bias loop-carried phi ranks
4878 before vectorization pass.
4879 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
4880 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
4882 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
4884 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
4886 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
4888 2021-09-28 Jakub Jelinek <jakub@redhat.com>
4891 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
4892 special 80387 instruction XFmode constants if flag_rounding_math.
4894 2021-09-28 Richard Biener <rguenther@suse.de>
4896 PR tree-optimization/100112
4897 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
4898 referece into the hashtable twice in case last_vuse is
4899 different from the original vuse on the stmt.
4901 2021-09-28 Jakub Jelinek <jakub@redhat.com>
4903 PR middle-end/102492
4904 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
4905 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
4906 clauses on SIMD constructs.
4908 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4910 PR tree-optimization/102511
4911 * gimple-range-path.cc (path_range_query::range_on_path_entry):
4912 Return VARYING when nothing found.
4914 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
4917 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
4919 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
4920 (VALID_MMX_REG_MODE): Add V4HFmode.
4921 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
4922 vector mode condition.
4923 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
4925 (function_arg_32): Add V4HFmode.
4926 (function_arg_advance_32): Likewise.
4927 * config/i386/i386.md (mode): Add V4HF/V2HF.
4928 (MODE_SIZE): Likewise.
4929 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
4930 (V_32): Add V2HF mode.
4931 (VHF_32_64): New mode iterator.
4932 (*mov<mode>_internal): Adjust sse alternatives to support
4934 (*mov<mode>_internal): Adjust sse alternatives to support
4936 (<insn><mode>3): New define_insn for add/sub/mul/div.
4938 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4940 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
4942 (pass_early_thread_jumps::gate): Same.
4943 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
4944 Return if !flag_thread_jumps.
4945 * tree-ssa-threadupdate.c
4946 (jt_path_registry::register_jump_thread): Assert that
4947 flag_thread_jumps is true.
4949 2021-09-28 liuhongt <hongtao.liu@intel.com>
4952 (simplify_context::simplify_binary_operation_1): Relax
4953 condition of simplifying (vec_concat:M (vec_select op0
4954 index0)(vec_select op1 index1)) to allow different modes
4955 between op0 and M, but have same inner mode.
4957 2021-09-28 liuhongt <hongtao.liu@intel.com>
4959 * config/i386/i386-expand.c (emit_reduc_half): Handle
4960 V8HF/V16HF/V32HFmode.
4961 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
4962 (REDUC_SSE_SMINMAX_MODE): Ditto.
4963 (REDUC_PLUS_MODE): Add V16HF and V32HF.
4964 (REDUC_SMINMAX_MODE): Ditto.
4966 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4968 * gimple-range-path.cc
4969 (path_range_query::precompute_ranges_in_block): Rename to...
4970 (path_range_query::compute_ranges_in_block): ...this.
4971 (path_range_query::precompute_ranges): Rename to...
4972 (path_range_query::compute_ranges): ...this.
4973 (path_range_query::precompute_relations): Rename to...
4974 (path_range_query::compute_relations): ...this.
4975 (path_range_query::precompute_phi_relations): Rename to...
4976 (path_range_query::compute_phi_relations): ...this.
4977 * gimple-range-path.h: Rename precompute* to compute*.
4978 * tree-ssa-threadbackward.c
4979 (back_threader::find_taken_edge_switch): Same.
4980 (back_threader::find_taken_edge_cond): Same.
4981 * tree-ssa-threadedge.c
4982 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
4983 (hybrid_jt_state::register_equivs_stmt): Inline...
4984 * tree-ssa-threadedge.h: ...here.
4986 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4988 * tree-vrp.c (lhs_of_dominating_assert): Remove.
4989 (class vrp_jt_state): Remove.
4990 (class vrp_jt_simplifier): Remove.
4991 (vrp_jt_simplifier::simplify): Remove.
4992 (class vrp_jump_threader): Remove.
4993 (vrp_jump_threader::vrp_jump_threader): Remove.
4994 (vrp_jump_threader::~vrp_jump_threader): Remove.
4995 (vrp_jump_threader::before_dom_children): Remove.
4996 (vrp_jump_threader::after_dom_children): Remove.
4998 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5000 * passes.def (pass_vrp_threader): New.
5001 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
5002 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
5003 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
5004 (hybrid_jt_simplifier::simplify): New.
5005 (hybrid_jt_simplifier::compute_ranges_from_state): New.
5006 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
5007 (class hybrid_jt_simplifier): New.
5008 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
5010 (class hybrid_threader): New.
5011 (hybrid_threader::hybrid_threader): New.
5012 (hybrid_threader::~hybrid_threader): New.
5013 (hybrid_threader::before_dom_children): New.
5014 (hybrid_threader::after_dom_children): New.
5015 (execute_vrp_threader): New.
5016 (class pass_vrp_threader): New.
5017 (make_pass_vrp_threader): New.
5019 2021-09-27 Martin Liska <mliska@suse.cz>
5021 * output.h (enum section_flag): New.
5022 (SECTION_FORGET): Remove.
5023 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
5024 (SECTION_STYLE_MASK): Define it based on other enum
5026 * varasm.c (switch_to_section): Remove unused handling of
5029 2021-09-27 Martin Liska <mliska@suse.cz>
5031 * common.opt: Add new variable flag_default_complex_method.
5032 * opts.c (finish_options): Handle flags related to
5033 x_flag_complex_method.
5034 * toplev.c (process_options): Remove option handling related
5035 to flag_complex_method.
5037 2021-09-27 Richard Biener <rguenther@suse.de>
5039 PR middle-end/102450
5040 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
5041 type_for_size, instead use int_mode_for_size.
5043 2021-09-27 Andrew Pinski <apinski@marvell.com>
5046 * gimplify.c (gimplify_save_expr): Return early
5047 if the type of val is error_mark_node.
5049 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5051 * tree-ssanames.c (ssa_name_has_boolean_range): Use
5054 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5056 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
5057 vrp_visit_cond_stmt.
5058 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
5059 (cprop_into_stmt): Same.
5060 (dom_opt_dom_walker::optimize_stmt): Same.
5062 2021-09-27 Richard Biener <rguenther@suse.de>
5064 PR tree-optimization/97351
5065 PR tree-optimization/97352
5066 PR tree-optimization/82426
5067 * tree-vectorizer.h (dr_misalignment): Add vector type
5069 (aligned_access_p): Likewise.
5070 (known_alignment_for_access_p): Likewise.
5071 (vect_supportable_dr_alignment): Likewise.
5072 (vect_known_alignment_in_bytes): Likewise. Refactor.
5073 (DR_MISALIGNMENT): Remove.
5074 (vect_update_shared_vectype): Likewise.
5075 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
5076 a vector type with larger alignment requirement and apply
5077 the negative step adjustment here.
5078 (vect_calculate_target_alignment): Remove.
5079 (vect_compute_data_ref_alignment): Get explicit vector type
5080 argument, do not apply a negative step alignment adjustment
5082 (vect_slp_analyze_node_alignment): Re-analyze alignment
5083 when we re-visit the DR with a bigger desired alignment but
5084 keep more precise results from smaller alignments.
5085 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
5086 (vect_slp_analyze_node_operations_1): Do not update the
5087 shared vector type on stmts.
5088 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
5089 vector type of an SLP node to the representative stmt-info.
5090 (vect_transform_stmt): Likewise.
5092 2021-09-27 liuhongt <hongtao.liu@intel.com>
5095 2021-09-09 liuhongt <hongtao.liu@intel.com>
5098 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
5099 (reduc_plus_scal_v4sf): .. this, New define_expand.
5100 (reduc_plus_scal_v2df): .. and this, New define_expand.
5102 2021-09-26 liuhongt <hongtao.liu@intel.com>
5104 * doc/extend.texi (Half-Precision): Remove storage only
5105 description for _Float16 w/o avx512fp16.
5107 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
5109 * config/pru/constraints.md (Rrio): New constraint.
5110 * config/pru/predicates.md (regio_operand): New predicate.
5111 * config/pru/pru-pragma.c (pru_register_pragmas): Register
5112 the __regio_symbol address space.
5113 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
5114 * config/pru/pru.c (pru_symref2ioregno): New helper function.
5115 (pru_legitimate_address_p): Remove.
5116 (pru_addr_space_legitimate_address_p): Use the address space
5118 (pru_nongeneric_pointer_addrspace): New helper function.
5119 (pru_insert_attributes): New function to validate __regio_symbol
5121 (TARGET_INSERT_ATTRIBUTES): New macro.
5122 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
5123 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
5124 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
5125 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
5127 (*regio_nozext_writesi): New pattern to write to I/O registers.
5128 (*regio_zext_write_r30<EQS0:mode>): Ditto.
5129 * doc/extend.texi: Document the new PRU Named Address Space.
5131 2021-09-24 Patrick Palka <ppalka@redhat.com>
5135 * real.c (encode_ieee_double): Avoid unwanted sign extension.
5136 (encode_ieee_quad): Likewise.
5138 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
5140 PR rtl-optimization/102147
5141 * ira-build.c (ira_conflict_vector_profitable_p): Make
5142 profitability calculation independent of host compiler pointer and
5145 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
5147 * gimple-range-path.cc (path_range_query::path_range_query):
5148 Move debugging header...
5149 (path_range_query::precompute_ranges): ...here.
5150 (path_range_query::internal_range_of_expr): Do not call
5151 range_on_path_entry if NAME is defined in the current block.
5153 2021-09-24 Richard Biener <rguenther@suse.de>
5155 * cfghooks.c (verify_flow_info): Verify unallocated BB and
5156 edge flags are not set.
5158 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
5160 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
5162 (jt_path_registry::register_jump_thread): Call
5163 cancel_invalid_paths.
5164 * tree-ssa-threadupdate.h (class jt_path_registry): Add
5165 cancel_invalid_paths.
5167 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
5169 PR tree-optimization/102400
5170 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
5171 result_vdef to zero value.
5173 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
5175 PR tree-optimization/102451
5176 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
5179 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
5181 * config/i386/sse.md (cond_<insn><mode>): Extend to support
5183 (cond_mul<mode>): Likewise.
5184 (cond_div<mode>): Likewise.
5185 (cond_<code><mode>): Likewise.
5186 (cond_fma<mode>): Likewise.
5187 (cond_fms<mode>): Likewise.
5188 (cond_fnma<mode>): Likewise.
5189 (cond_fnms<mode>): Likewise.
5191 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
5193 PR tree-optimization/102463
5194 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
5195 there is no range-ops handler, don't look for a relation.
5197 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
5199 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
5200 non-executable_edge flag as parameter.
5201 * gimple-range-cache.h (ranger_cache): Adjust prototype.
5202 * gimple-range-gori.cc (gori_compute::gori_compute): Take
5203 non-executable_edge flag as parameter.
5204 (gori_compute::outgoing_edge_range_p): Check new flag.
5205 * gimple-range-gori.h (gori_compute): Adjust prototype.
5206 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
5207 (gimple_ranger::range_on_edge): Check new flag.
5208 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
5209 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
5210 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
5211 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
5212 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
5213 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
5214 (simplify_using_ranges::simplify_switch_using_ranges): Clear
5215 EDGE_EXECUTABLE like it originally did.
5216 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
5217 NON_EXECUTABLE flags.
5218 (simplify_using_ranges::simplify_using_ranges): Adjust.
5219 * vr-values.h (class simplify_using_ranges): Adjust.
5220 (simplify_using_ranges::set_range_query): Add non-executable flag param.
5222 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
5225 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
5226 zero-width bit fields and return indicator.
5227 (rs6000_discover_homogeneous_aggregate): Diagnose when the
5228 presence of a zero-width bit field changes parameter passing in
5231 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5233 * gimple-range-fold.cc (fold_using_range::range_of_phi):
5234 Remove dominator check.
5236 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5238 * gimple-range-path.cc (path_range_query::precompute_relations):
5239 Hoist edge calculations before using EDGE_SUCC.
5241 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
5243 * configure.ac: Fix --with-multilib-list description.
5244 * configure: Regenerate.
5246 2021-09-23 Richard Biener <rguenther@suse.de>
5248 PR tree-optimization/102448
5249 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
5250 Clear alignment info copied from DR_PTR_INFO.
5252 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5254 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
5256 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
5257 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
5258 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
5259 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
5260 (vcond<mode><mode>): New expander.
5261 (vcond<mode><sseintvecmodelower>): Likewise.
5262 (vcond<sseintvecmodelower><mode>): Likewise.
5263 (vcondu<mode><sseintvecmodelower>): Likewise.
5265 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5267 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
5269 (extendv4hf<mode>2): Likewise.
5270 (extendv2hfv2df2): Likewise.
5271 (trunc<mode><ssePHmodelower>2): Likewise.
5272 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
5273 (trunc<mode>v4hf2): ... this, and drop constraints.
5274 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
5275 (truncv2dfv2hf2): ... this, and likewise.
5277 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5279 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
5281 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
5283 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
5284 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
5285 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
5287 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5289 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
5290 (fixuns_trunchfhi2): Likewise.
5291 (*fixuns_trunchfsi2zext): New define_insn.
5292 * config/i386/sse.md (ssePHmodelower): New mode_attr.
5293 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5294 New expander for same element vector fix_truncate.
5295 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5296 Likewise for V4HF to V4SI/V4DI fix_truncate.
5297 (fix<fixunssuffix>_truncv2hfv2di2):
5298 Likeise for V2HF to V2DI fix_truncate.
5300 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5302 * config/i386/i386.md (<code>hf3): New expander.
5304 2021-09-23 liuhongt <hongtao.liu@intel.com>
5306 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
5307 (VFH_SF_AVX512VL): Extend to handle HFmode.
5308 (VF_SF_AVX512VL): Deleted.
5310 2021-09-23 liuhongt <hongtao.liu@intel.com>
5312 * config/i386/i386.md (rinthf2): New expander.
5313 (nearbyinthf2): New expander.
5315 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5317 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
5318 (class dom_jt_state): ...this and provide virtual overrides.
5319 (dom_jt_state::register_equiv): New.
5320 (class dom_jt_simplifier): Rename from
5321 dom_jump_threader_simplifier.
5322 (dom_jump_threader_simplifier::simplify): Rename...
5323 (dom_jt_simplifier::simplify): ...to this.
5324 (pass_dominator::execute): Use dom_jt_simplifier and
5326 * tree-ssa-threadedge.c (jump_threader::jump_threader):
5328 (jt_state::register_equivs_stmt): Abstract out...
5329 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
5331 (jump_threader::thread_around_empty_blocks): Update state.
5332 (jump_threader::thread_through_normal_block): Same.
5333 (jt_state::jt_state): Remove.
5334 (jt_state::push): Remove pass specific bits. Keep block vector
5336 (jt_state::append_path): New.
5337 (jt_state::pop): Remove pass specific bits.
5338 (jt_state::register_equiv): Same.
5339 (jt_state::record_ranges_from_stmt): Same.
5340 (jt_state::register_equivs_on_edge): Same. Rename...
5341 (jt_state::register_equivs_edge): ...to this.
5342 (jt_state::dump): New.
5343 (jt_state::debug): New.
5344 (jump_threader_simplifier::simplify): Remove.
5345 (jt_state::get_path): New.
5346 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
5347 class. Expose common functionality as virtual methods.
5348 (class jump_threader_simplifier): Same. Rename...
5349 (class jt_simplifier): ...to this.
5350 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
5351 (class vrp_jt_simplifier): ...to this. Provide pass specific
5353 (class vrp_jt_state): New.
5354 (vrp_jump_threader_simplifier::simplify): Rename...
5355 (vrp_jt_simplifier::simplify): ...to this. Inline code from
5356 what used to be the base class.
5357 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
5360 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
5363 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
5366 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
5367 Richard Biener <rguenther@suse.de>
5369 * match.pd (negation simplifications): Implement some negation
5370 folding transformations from fold-const.c's fold_negate_expr.
5371 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
5372 argument, to control whether the op should be simplified prior
5373 to looking up/assigning a value number.
5374 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
5375 (vn_nary_simplify): Likewise.
5376 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
5377 now call vn_nary_build_or_lookup_1 disabling simplification.
5379 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
5381 PR tree-optimization/102087
5382 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5383 Update bound/cmp/control for niter.
5385 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5387 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5388 Move check for non-empty BB here.
5389 (fur_source::register_outgoing_edges): ...from here.
5391 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5393 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
5394 Remove call to improve_range_with_equivs.
5395 (path_range_query::improve_range_with_equivs): Remove
5396 * gimple-range-path.h: Remove improve_range_with_equivs.
5398 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5400 * config/i386/avx512fp16intrin.h:
5401 (_mm512_mask_blend_ph): New intrinsic.
5402 (_mm512_permutex2var_ph): Ditto.
5403 (_mm512_permutexvar_ph): Ditto.
5404 * config/i386/avx512fp16vlintrin.h:
5405 (_mm256_mask_blend_ph): New intrinsic.
5406 (_mm256_permutex2var_ph): Ditto.
5407 (_mm256_permutexvar_ph): Ditto.
5408 (_mm_mask_blend_ph): Ditto.
5409 (_mm_permutex2var_ph): Ditto.
5410 (_mm_permutexvar_ph): Ditto.
5412 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5414 * config/i386/avx512fp16intrin.h: Add new intrinsics.
5415 (_mm512_conj_pch): New intrinsic.
5416 (_mm512_mask_conj_pch): Ditto.
5417 (_mm512_maskz_conj_pch): Ditto.
5418 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
5419 (_mm256_conj_pch): New intrinsic.
5420 (_mm256_mask_conj_pch): Ditto.
5421 (_mm256_maskz_conj_pch): Ditto.
5422 (_mm_conj_pch): Ditto.
5423 (_mm_mask_conj_pch): Ditto.
5424 (_mm_maskz_conj_pch): Ditto.
5426 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5428 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
5429 (_mm512_reduce_add_ph): New intrinsic.
5430 (_mm512_reduce_mul_ph): Ditto.
5431 (_mm512_reduce_min_ph): Ditto.
5432 (_mm512_reduce_max_ph): Ditto.
5433 * config/i386/avx512fp16vlintrin.h
5434 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
5435 (_mm256_reduce_add_ph): New intrinsic.
5436 (_mm256_reduce_mul_ph): Ditto.
5437 (_mm256_reduce_min_ph): Ditto.
5438 (_mm256_reduce_max_ph): Ditto.
5439 (_mm_reduce_add_ph): Ditto.
5440 (_mm_reduce_mul_ph): Ditto.
5441 (_mm_reduce_min_ph): Ditto.
5442 (_mm_reduce_max_ph): Ditto.
5444 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5446 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
5447 __m128h_u): New typedef.
5448 (_mm512_load_ph): New intrinsic.
5449 (_mm256_load_ph): Ditto.
5450 (_mm_load_ph): Ditto.
5451 (_mm512_loadu_ph): Ditto.
5452 (_mm256_loadu_ph): Ditto.
5453 (_mm_loadu_ph): Ditto.
5454 (_mm512_store_ph): Ditto.
5455 (_mm256_store_ph): Ditto.
5456 (_mm_store_ph): Ditto.
5457 (_mm512_storeu_ph): Ditto.
5458 (_mm256_storeu_ph): Ditto.
5459 (_mm_storeu_ph): Ditto.
5460 (_mm512_abs_ph): Ditto.
5461 * config/i386/avx512fp16vlintrin.h
5462 (_mm_abs_ph): Ditto.
5463 (_mm256_abs_ph): Ditto.
5465 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5467 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
5469 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5472 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
5473 is actually a full copy of the source operand into the target.
5474 Don't emit a strict low part move if source and target mode match.
5476 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5478 PR middle-end/102415
5479 * omp-expand.c (expand_omp_single): If region->exit is NULL,
5480 assert region->entry is GIMPLE_OMP_SCOPE region and return.
5482 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5484 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
5485 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
5486 arguments from 2 to 3.
5487 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
5488 allocate clause allocator and print align if present.
5489 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
5490 for omp_default_mem_alloc if align modifier is present. If align
5491 modifier is present, use TREE_LIST to encode both allocator and
5493 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
5494 Handle align modifier on allocator clause if present.
5496 2021-09-22 liuhongt <hongtao.liu@intel.com>
5498 * config/i386/i386.md (define_attr "isa"): Add
5500 (define_attr "enabled"): Correspond fma_or_avx512vl to
5501 TARGET_FMA || TARGET_AVX512VL.
5502 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
5507 2021-09-22 liuhongt <hongtao.liu@intel.com>
5509 * config/i386/i386.md (cstorehf3): New define_expand.
5511 2021-09-22 liuhongt <hongtao.liu@intel.com>
5513 * config/i386/i386.md (<rounding_insn>hf2): New expander.
5514 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
5515 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
5516 Extend from VF_128 to VFH_128.
5518 2021-09-22 liuhongt <hongtao.liu@intel.com>
5520 * config/i386/i386-features.c (i386-features.c): Handle
5522 * config/i386/i386.md (sqrthf2): New expander.
5523 (*sqrthf2): New define_insn.
5524 * config/i386/sse.md
5525 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
5528 2021-09-22 liuhongt <hongtao.liu@intel.com>
5530 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
5532 (_mm_mask3_fcmadd_sch): Likewise.
5533 (_mm_maskz_fcmadd_sch): Likewise.
5534 (_mm_fcmadd_sch): Likewise.
5535 (_mm_mask_fmadd_sch): Likewise.
5536 (_mm_mask3_fmadd_sch): Likewise.
5537 (_mm_maskz_fmadd_sch): Likewise.
5538 (_mm_fmadd_sch): Likewise.
5539 (_mm_mask_fcmadd_round_sch): Likewise.
5540 (_mm_mask3_fcmadd_round_sch): Likewise.
5541 (_mm_maskz_fcmadd_round_sch): Likewise.
5542 (_mm_fcmadd_round_sch): Likewise.
5543 (_mm_mask_fmadd_round_sch): Likewise.
5544 (_mm_mask3_fmadd_round_sch): Likewise.
5545 (_mm_maskz_fmadd_round_sch): Likewise.
5546 (_mm_fmadd_round_sch): Likewise.
5547 (_mm_fcmul_sch): Likewise.
5548 (_mm_mask_fcmul_sch): Likewise.
5549 (_mm_maskz_fcmul_sch): Likewise.
5550 (_mm_fmul_sch): Likewise.
5551 (_mm_mask_fmul_sch): Likewise.
5552 (_mm_maskz_fmul_sch): Likewise.
5553 (_mm_fcmul_round_sch): Likewise.
5554 (_mm_mask_fcmul_round_sch): Likewise.
5555 (_mm_maskz_fcmul_round_sch): Likewise.
5556 (_mm_fmul_round_sch): Likewise.
5557 (_mm_mask_fmul_round_sch): Likewise.
5558 (_mm_maskz_fmul_round_sch): Likewise.
5559 * config/i386/i386-builtin.def: Add corresponding new builtins.
5560 * config/i386/sse.md
5561 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
5562 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
5563 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
5565 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
5566 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
5568 * config/i386/subst.md (mask_scalarcz_name): New.
5569 (mask_scalarc_name): Ditto.
5570 (mask_scalarc_operand3): Ditto.
5571 (mask_scalarcz_operand4): Ditto.
5572 (round_scalarcz_name): Ditto.
5573 (round_scalarc_mask_operand3): Ditto.
5574 (round_scalarcz_mask_operand4): Ditto.
5575 (round_scalarc_mask_op3): Ditto.
5576 (round_scalarcz_mask_op4): Ditto.
5577 (round_scalarcz_constraint): Ditto.
5578 (round_scalarcz_nimm_predicate): Ditto.
5579 (mask_scalarcz): Ditto.
5580 (mask_scalarc): Ditto.
5581 (round_scalarcz): Ditto.
5583 2021-09-22 liuhongt <hongtao.liu@intel.com>
5585 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
5587 (_mm512_mask_fcmadd_pch): Likewise.
5588 (_mm512_mask3_fcmadd_pch): Likewise.
5589 (_mm512_maskz_fcmadd_pch): Likewise.
5590 (_mm512_fmadd_pch): Likewise.
5591 (_mm512_mask_fmadd_pch): Likewise.
5592 (_mm512_mask3_fmadd_pch): Likewise.
5593 (_mm512_maskz_fmadd_pch): Likewise.
5594 (_mm512_fcmadd_round_pch): Likewise.
5595 (_mm512_mask_fcmadd_round_pch): Likewise.
5596 (_mm512_mask3_fcmadd_round_pch): Likewise.
5597 (_mm512_maskz_fcmadd_round_pch): Likewise.
5598 (_mm512_fmadd_round_pch): Likewise.
5599 (_mm512_mask_fmadd_round_pch): Likewise.
5600 (_mm512_mask3_fmadd_round_pch): Likewise.
5601 (_mm512_maskz_fmadd_round_pch): Likewise.
5602 (_mm512_fcmul_pch): Likewise.
5603 (_mm512_mask_fcmul_pch): Likewise.
5604 (_mm512_maskz_fcmul_pch): Likewise.
5605 (_mm512_fmul_pch): Likewise.
5606 (_mm512_mask_fmul_pch): Likewise.
5607 (_mm512_maskz_fmul_pch): Likewise.
5608 (_mm512_fcmul_round_pch): Likewise.
5609 (_mm512_mask_fcmul_round_pch): Likewise.
5610 (_mm512_maskz_fcmul_round_pch): Likewise.
5611 (_mm512_fmul_round_pch): Likewise.
5612 (_mm512_mask_fmul_round_pch): Likewise.
5613 (_mm512_maskz_fmul_round_pch): Likewise.
5614 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
5616 (_mm_mask_fmadd_pch): Likewise.
5617 (_mm_mask3_fmadd_pch): Likewise.
5618 (_mm_maskz_fmadd_pch): Likewise.
5619 (_mm256_fmadd_pch): Likewise.
5620 (_mm256_mask_fmadd_pch): Likewise.
5621 (_mm256_mask3_fmadd_pch): Likewise.
5622 (_mm256_maskz_fmadd_pch): Likewise.
5623 (_mm_fcmadd_pch): Likewise.
5624 (_mm_mask_fcmadd_pch): Likewise.
5625 (_mm_mask3_fcmadd_pch): Likewise.
5626 (_mm_maskz_fcmadd_pch): Likewise.
5627 (_mm256_fcmadd_pch): Likewise.
5628 (_mm256_mask_fcmadd_pch): Likewise.
5629 (_mm256_mask3_fcmadd_pch): Likewise.
5630 (_mm256_maskz_fcmadd_pch): Likewise.
5631 (_mm_fmul_pch): Likewise.
5632 (_mm_mask_fmul_pch): Likewise.
5633 (_mm_maskz_fmul_pch): Likewise.
5634 (_mm256_fmul_pch): Likewise.
5635 (_mm256_mask_fmul_pch): Likewise.
5636 (_mm256_maskz_fmul_pch): Likewise.
5637 (_mm_fcmul_pch): Likewise.
5638 (_mm_mask_fcmul_pch): Likewise.
5639 (_mm_maskz_fcmul_pch): Likewise.
5640 (_mm256_fcmul_pch): Likewise.
5641 (_mm256_mask_fcmul_pch): Likewise.
5642 (_mm256_maskz_fcmul_pch): Likewise.
5643 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
5644 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
5645 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
5646 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
5647 * config/i386/i386-builtin.def: Add new builtins.
5648 * config/i386/i386-expand.c: Handle new builtin types.
5649 * config/i386/subst.md (SUBST_CV): New.
5650 (maskc_name): Ditto.
5651 (maskc_operand3): Ditto.
5653 (sdc_maskz_name): Ditto.
5654 (sdc_mask_op4): Ditto.
5655 (sdc_mask_op5): Ditto.
5656 (sdc_mask_mode512bit_condition): Ditto.
5658 (round_maskc_operand3): Ditto.
5659 (round_sdc_mask_operand4): Ditto.
5660 (round_maskc_op3): Ditto.
5661 (round_sdc_mask_op4): Ditto.
5662 (round_saeonly_sdc_mask_operand5): Ditto.
5663 * config/i386/sse.md (unspec): Add complex fma unspecs.
5664 (avx512fmaskcmode): New.
5665 (UNSPEC_COMPLEX_F_C_MA): Ditto.
5666 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
5667 (complexopname): Ditto.
5668 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
5669 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
5670 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
5672 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
5673 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
5675 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
5677 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
5678 rs6000-density-size-threshold, rs6000-density-penalty,
5679 rs6000-density-load-pct-threshold,
5680 rs6000-density-load-num-threshold): New parameter.
5681 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
5682 corresponding parameters.
5684 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5686 * gimple-range-path.cc (path_range_query::defined_outside_path):
5688 (path_range_query::range_on_path_entry): New.
5689 (path_range_query::internal_range_of_expr): Resolve unknowns
5691 (path_range_query::improve_range_with_equivs): New.
5692 (path_range_query::ssa_range_in_phi): Resolve unknowns with
5694 * gimple-range-path.h (class path_range_query): Add
5695 defined_outside_path, range_on_path_entry, and
5696 improve_range_with_equivs.
5698 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5700 * gimple-range-path.cc (path_range_query::add_to_imports): New.
5701 (path_range_query::add_copies_to_imports): New.
5702 (path_range_query::precompute_ranges): Call
5703 add_copies_to_imports.
5704 * gimple-range-path.h (class path_range_query): Add prototypes
5705 for add_copies_to_imports and add_to_imports.
5707 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5709 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5710 Remove useless code.
5712 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5714 * gimple-range-fold.h (class fur_source): Make oracle protected.
5715 * gimple-range-path.cc (path_range_query::path_range_query): Add
5716 resolve argument. Initialize oracle.
5717 (path_range_query::~path_range_query): Delete oracle.
5718 (path_range_query::range_of_stmt): Adapt to use relations.
5719 (path_range_query::precompute_ranges): Pre-compute relations.
5720 (class jt_fur_source): New
5721 (jt_fur_source::jt_fur_source): New.
5722 (jt_fur_source::register_relation): New.
5723 (jt_fur_source::query_relation): New.
5724 (path_range_query::precompute_relations): New.
5725 (path_range_query::precompute_phi_relations): New.
5726 * gimple-range-path.h (path_range_query): Add resolve argument.
5727 Add oracle, precompute_relations, precompute_phi_relations.
5728 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
5729 resolve argument to solver.
5731 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5733 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5734 Rename postfold_gcond_edges to register_outgoing_edges and
5736 (fold_using_range::postfold_gcond_edges): Rename...
5737 (fur_source::register_outgoing_edges): ...to this.
5738 * gimple-range-fold.h (postfold_gcond_edges): Rename to
5739 register_outgoing_edges and move to fur_source.
5741 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5743 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
5744 dom_info_available_p.
5746 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5748 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
5749 and quick_grow_cleared instead of safe_grow_cleared.
5751 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
5754 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
5757 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
5759 * configure.ac: Detect when the assembler supports new-style
5760 architecture extensions.
5761 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
5762 the full CPU string if the assembler can grok it.
5763 (arm_rewrite_march): Likewise but for the architecture.
5764 * config.in: Regenerate.
5765 * configure: Regenerate.
5767 2021-09-21 Richard Biener <rguenther@suse.de>
5769 PR tree-optimization/102421
5770 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
5771 adjust alignment info.
5773 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
5775 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
5776 bitfield stream out.
5778 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
5780 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
5781 undefined edges, apply an equivalence if appropriate.
5782 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
5783 UNDEFINED if EDGE_EXECUTABLE is not set.
5784 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
5785 as EXECUTABLE upon startup.
5786 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
5787 EDGE_EXECUTABLE set.
5788 * vr-values.c (set_and_propagate_unexecutable): New.
5789 (simplify_using_ranges::fold_cond): Call set_and_propagate.
5790 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
5791 * vr-values.h: Add prototype.
5793 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
5795 * value-relation.cc (equiv_oracle::register_initial_def): New.
5796 (equiv_oracle::register_relation): Call register_initial_def.
5797 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
5798 (relation_oracle::register_stmt): Check def block of PHI arguments.
5799 * value-relation.h (equiv_oracle): Add new prototypes.
5801 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
5803 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5804 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
5805 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
5806 __ROUNDING_MATH__ according to their corresponding flags.
5807 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
5808 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
5809 and __ROUNDING_MATH__.
5811 2021-09-20 Richard Biener <rguenther@suse.de>
5813 * tree-vect-stmts.c (vectorizable_load): Use the vectype
5816 2021-09-20 Richard Biener <rguenther@suse.de>
5818 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
5819 Do not compute alignment of the vectorized access here.
5821 2021-09-20 Richard Biener <rguenther@suse.de>
5823 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
5824 Store -1 for runtime alias peeling iterations.
5826 2021-09-20 Richard Biener <rguenther@suse.de>
5828 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
5830 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
5832 * input.c (string_concat_db::record_string_concatenation)
5833 (string_concat_db::get_string_concatenation): Skip for
5834 'RESERVED_LOCATION_P'.
5836 2021-09-20 Richard Biener <rguenther@suse.de>
5838 PR tree-optimization/65206
5839 * tree-data-ref.h (struct data_reference): Add alt_indices,
5841 * tree-data-ref.c (free_data_ref): Release alt_indices.
5842 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
5843 (create_data_ref): Adjust.
5844 (initialize_data_dependence_relation): Split into head
5845 and tail. When the base objects fail to match up try
5846 again with pointer-based analysis of indices.
5847 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
5848 not compare the lazily computed alternate set of indices.
5850 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
5852 * gcc.c: Test for execute OK when we find the
5853 programs for assembler linker and dsymutil and those
5854 were specified at configure-time.
5856 2021-09-19 Martin Sebor <msebor@redhat.com>
5858 PR middle-end/102403
5859 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
5860 Correct a function pre/postcondition.
5862 2021-09-19 Martin Sebor <msebor@redhat.com>
5864 PR middle-end/102243
5865 * tree-ssa-strlen.c (get_range): Handle null cfun.
5867 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
5869 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
5870 unwinder shim as a convenience library.
5872 2021-09-19 Andrew Pinski <apinski@marvell.com>
5874 * doc/install.texi: Add note about
5875 binutils 2.35 is required for LTO usage.
5877 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
5879 * tree-ssa-threadbackward.c
5880 (back_threader_registry::register_path): Use push_edge.
5881 * tree-ssa-threadedge.c
5882 (jump_threader::thread_around_empty_blocks): Same.
5883 (jump_threader::thread_through_normal_block): Same.
5884 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
5886 * tree-ssa-threadupdate.c
5887 (jt_path_registry::allocate_thread_edge): Remove.
5888 (jt_path_registry::push_edge): New.
5889 (dump_jump_thread_path): Make static.
5890 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
5893 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
5895 * gimple-range-path.cc (path_range_query::path_range_query): Add
5897 (path_range_query::dump): Remove extern declaration of dump_ranger.
5898 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
5899 * gimple-range-trace.h (dump_ranger): Add prototype.
5901 2021-09-19 John Ericson <git@JohnEricson.me>
5903 * gcc.c (find_a_program): New function, factored out of...
5904 (find_a_file): Here.
5905 (execute): Use find_a_program when looking for programs rather
5908 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
5910 * config/avr/avr-mcus.def: Add atmega324pb.
5911 * doc/avr-mmcu.texi: Corresponding changes.
5913 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
5916 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
5917 the same transformations as to @1. For comparisons against NaN,
5918 don't check HONOR_SNANS but confirm that neither operand is a
5921 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
5923 * attribs.c (make_unique_name): Delete.
5924 * attribs.h (make_unique_name): Delete.
5926 2021-09-19 Andrew Pinski <apinski@marvell.com>
5928 * lra-constraints.c (check_and_process_move): Assert
5929 that dclass and sclass are greater than or equal to NO_REGS.
5931 2021-09-18 Jakub Jelinek <jakub@redhat.com>
5933 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
5934 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
5935 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
5937 2021-09-18 liuhongt <hongtao.liu@intel.com>
5939 * config/i386/i386-features.c (remove_partial_avx_dependency):
5940 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
5941 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
5943 2021-09-18 Jakub Jelinek <jakub@redhat.com>
5945 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
5946 if file/namespace scope variable doesn't have predetermined sharing,
5947 treat it as if there was default(none).
5949 2021-09-18 liuhongt <hongtao.liu@intel.com>
5951 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
5953 (_mm_mask_fmadd_sh): Likewise.
5954 (_mm_mask3_fmadd_sh): Likewise.
5955 (_mm_maskz_fmadd_sh): Likewise.
5956 (_mm_fmadd_round_sh): Likewise.
5957 (_mm_mask_fmadd_round_sh): Likewise.
5958 (_mm_mask3_fmadd_round_sh): Likewise.
5959 (_mm_maskz_fmadd_round_sh): Likewise.
5960 (_mm_fnmadd_sh): Likewise.
5961 (_mm_mask_fnmadd_sh): Likewise.
5962 (_mm_mask3_fnmadd_sh): Likewise.
5963 (_mm_maskz_fnmadd_sh): Likewise.
5964 (_mm_fnmadd_round_sh): Likewise.
5965 (_mm_mask_fnmadd_round_sh): Likewise.
5966 (_mm_mask3_fnmadd_round_sh): Likewise.
5967 (_mm_maskz_fnmadd_round_sh): Likewise.
5968 (_mm_fmsub_sh): Likewise.
5969 (_mm_mask_fmsub_sh): Likewise.
5970 (_mm_mask3_fmsub_sh): Likewise.
5971 (_mm_maskz_fmsub_sh): Likewise.
5972 (_mm_fmsub_round_sh): Likewise.
5973 (_mm_mask_fmsub_round_sh): Likewise.
5974 (_mm_mask3_fmsub_round_sh): Likewise.
5975 (_mm_maskz_fmsub_round_sh): Likewise.
5976 (_mm_fnmsub_sh): Likewise.
5977 (_mm_mask_fnmsub_sh): Likewise.
5978 (_mm_mask3_fnmsub_sh): Likewise.
5979 (_mm_maskz_fnmsub_sh): Likewise.
5980 (_mm_fnmsub_round_sh): Likewise.
5981 (_mm_mask_fnmsub_round_sh): Likewise.
5982 (_mm_mask3_fnmsub_round_sh): Likewise.
5983 (_mm_maskz_fnmsub_round_sh): Likewise.
5984 * config/i386/i386-builtin-types.def
5985 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
5986 * config/i386/i386-builtin.def: Add new builtins.
5987 * config/i386/i386-expand.c: Handle new builtin type.
5988 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
5989 Ajdust to support FP16.
5990 (fmai_vmfmsub_<mode><round_name>): Ditto.
5991 (fmai_vmfnmadd_<mode><round_name>): Ditto.
5992 (fmai_vmfnmsub_<mode><round_name>): Ditto.
5993 (*fmai_fmadd_<mode>): Ditto.
5994 (*fmai_fmsub_<mode>): Ditto.
5995 (*fmai_fnmadd_<mode><round_name>): Ditto.
5996 (*fmai_fnmsub_<mode><round_name>): Ditto.
5997 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
5998 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
5999 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
6000 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
6001 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
6002 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
6003 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
6004 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
6005 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
6006 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
6007 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
6008 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
6009 adjust to support FP16.
6010 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
6011 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
6012 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
6015 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
6017 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
6019 (maskload<mode><avx512fmaskmodelower>): Ditto.
6020 (maskstore<mode><avx512fmaskmodelower>): Ditto.
6022 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
6024 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
6026 (ix86_expand_copysign): Ditto.
6027 (ix86_expand_xorsign): Ditto.
6028 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
6030 (ix86_build_signbit_mask): Ditto.
6031 (ix86_can_change_mode_class): Ditto.
6032 * config/i386/i386.md
6033 (SSEMODEF): Add HFmode.
6034 (ssevecmodef): Ditto.
6035 (<code>hf2): New define_expand.
6036 (*<code>hf2_1): New define_insn_and_split.
6037 (copysign<mode>): Extend to support HFmode under AVX512FP16.
6038 (xorsign<mode>): Ditto.
6039 * config/i386/sse.md (VFB): New mode iterator.
6040 (VFB_128_256): Ditto.
6042 (sseintvecmode2): Support HF vector mode.
6043 (<code><mode>2): Use new mode iterator.
6044 (*<code><mode>2): Ditto.
6045 (copysign<mode>3): Ditto.
6046 (xorsign<mode>3): Ditto.
6047 (<code><mode>3<mask_name>): Ditto.
6048 (<code><mode>3<mask_name>): Ditto.
6049 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
6050 (<sse>_andnot<mode>3<mask_name>): Ditto.
6051 (*<code><mode>3<mask_name>): Ditto.
6052 (*<code><mode>3<mask_name>): Ditto.
6054 2021-09-18 liuhongt <hongtao.liu@intel.com>
6056 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
6058 (_mm512_mask3_fmadd_ph): Likewise.
6059 (_mm512_maskz_fmadd_ph): Likewise.
6060 (_mm512_fmadd_round_ph): Likewise.
6061 (_mm512_mask_fmadd_round_ph): Likewise.
6062 (_mm512_mask3_fmadd_round_ph): Likewise.
6063 (_mm512_maskz_fmadd_round_ph): Likewise.
6064 (_mm512_fnmadd_ph): Likewise.
6065 (_mm512_mask_fnmadd_ph): Likewise.
6066 (_mm512_mask3_fnmadd_ph): Likewise.
6067 (_mm512_maskz_fnmadd_ph): Likewise.
6068 (_mm512_fnmadd_round_ph): Likewise.
6069 (_mm512_mask_fnmadd_round_ph): Likewise.
6070 (_mm512_mask3_fnmadd_round_ph): Likewise.
6071 (_mm512_maskz_fnmadd_round_ph): Likewise.
6072 (_mm512_fmsub_ph): Likewise.
6073 (_mm512_mask_fmsub_ph): Likewise.
6074 (_mm512_mask3_fmsub_ph): Likewise.
6075 (_mm512_maskz_fmsub_ph): Likewise.
6076 (_mm512_fmsub_round_ph): Likewise.
6077 (_mm512_mask_fmsub_round_ph): Likewise.
6078 (_mm512_mask3_fmsub_round_ph): Likewise.
6079 (_mm512_maskz_fmsub_round_ph): Likewise.
6080 (_mm512_fnmsub_ph): Likewise.
6081 (_mm512_mask_fnmsub_ph): Likewise.
6082 (_mm512_mask3_fnmsub_ph): Likewise.
6083 (_mm512_maskz_fnmsub_ph): Likewise.
6084 (_mm512_fnmsub_round_ph): Likewise.
6085 (_mm512_mask_fnmsub_round_ph): Likewise.
6086 (_mm512_mask3_fnmsub_round_ph): Likewise.
6087 (_mm512_maskz_fnmsub_round_ph): Likewise.
6088 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
6090 (_mm256_mask_fmadd_ph): Likewise.
6091 (_mm256_mask3_fmadd_ph): Likewise.
6092 (_mm256_maskz_fmadd_ph): Likewise.
6093 (_mm_fmadd_ph): Likewise.
6094 (_mm_mask_fmadd_ph): Likewise.
6095 (_mm_mask3_fmadd_ph): Likewise.
6096 (_mm_maskz_fmadd_ph): Likewise.
6097 (_mm256_fnmadd_ph): Likewise.
6098 (_mm256_mask_fnmadd_ph): Likewise.
6099 (_mm256_mask3_fnmadd_ph): Likewise.
6100 (_mm256_maskz_fnmadd_ph): Likewise.
6101 (_mm_fnmadd_ph): Likewise.
6102 (_mm_mask_fnmadd_ph): Likewise.
6103 (_mm_mask3_fnmadd_ph): Likewise.
6104 (_mm_maskz_fnmadd_ph): Likewise.
6105 (_mm256_fmsub_ph): Likewise.
6106 (_mm256_mask_fmsub_ph): Likewise.
6107 (_mm256_mask3_fmsub_ph): Likewise.
6108 (_mm256_maskz_fmsub_ph): Likewise.
6109 (_mm_fmsub_ph): Likewise.
6110 (_mm_mask_fmsub_ph): Likewise.
6111 (_mm_mask3_fmsub_ph): Likewise.
6112 (_mm_maskz_fmsub_ph): Likewise.
6113 (_mm256_fnmsub_ph): Likewise.
6114 (_mm256_mask_fnmsub_ph): Likewise.
6115 (_mm256_mask3_fnmsub_ph): Likewise.
6116 (_mm256_maskz_fnmsub_ph): Likewise.
6117 (_mm_fnmsub_ph): Likewise.
6118 (_mm_mask_fnmsub_ph): Likewise.
6119 (_mm_mask3_fnmsub_ph): Likewise.
6120 (_mm_maskz_fnmsub_ph): Likewise.
6121 * config/i386/i386-builtin.def: Add corresponding new builtins.
6122 * config/i386/sse.md
6123 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
6124 support HF vector modes.
6125 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
6127 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
6128 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
6129 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
6130 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
6131 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
6132 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
6133 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
6135 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
6136 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
6137 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
6138 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
6139 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
6140 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
6142 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
6143 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
6144 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
6145 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
6146 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
6147 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
6148 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
6150 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
6151 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
6152 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
6153 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
6154 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
6156 2021-09-18 liuhongt <hongtao.liu@intel.com>
6158 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
6160 (_mm512_mask_fmaddsub_ph): Likewise.
6161 (_mm512_mask3_fmaddsub_ph): Likewise.
6162 (_mm512_maskz_fmaddsub_ph): Likewise.
6163 (_mm512_fmaddsub_round_ph): Likewise.
6164 (_mm512_mask_fmaddsub_round_ph): Likewise.
6165 (_mm512_mask3_fmaddsub_round_ph): Likewise.
6166 (_mm512_maskz_fmaddsub_round_ph): Likewise.
6167 (_mm512_mask_fmsubadd_ph): Likewise.
6168 (_mm512_mask3_fmsubadd_ph): Likewise.
6169 (_mm512_maskz_fmsubadd_ph): Likewise.
6170 (_mm512_fmsubadd_round_ph): Likewise.
6171 (_mm512_mask_fmsubadd_round_ph): Likewise.
6172 (_mm512_mask3_fmsubadd_round_ph): Likewise.
6173 (_mm512_maskz_fmsubadd_round_ph): Likewise.
6174 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
6176 (_mm256_mask_fmaddsub_ph): Likewise.
6177 (_mm256_mask3_fmaddsub_ph): Likewise.
6178 (_mm256_maskz_fmaddsub_ph): Likewise.
6179 (_mm_fmaddsub_ph): Likewise.
6180 (_mm_mask_fmaddsub_ph): Likewise.
6181 (_mm_mask3_fmaddsub_ph): Likewise.
6182 (_mm_maskz_fmaddsub_ph): Likewise.
6183 (_mm256_fmsubadd_ph): Likewise.
6184 (_mm256_mask_fmsubadd_ph): Likewise.
6185 (_mm256_mask3_fmsubadd_ph): Likewise.
6186 (_mm256_maskz_fmsubadd_ph): Likewise.
6187 (_mm_fmsubadd_ph): Likewise.
6188 (_mm_mask_fmsubadd_ph): Likewise.
6189 (_mm_mask3_fmsubadd_ph): Likewise.
6190 (_mm_maskz_fmsubadd_ph): Likewise.
6191 * config/i386/i386-builtin.def: Add corresponding new builtins.
6192 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
6193 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
6194 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
6196 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
6198 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
6199 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
6200 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
6202 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
6203 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
6205 2021-09-18 liuhongt <hongtao.liu@intel.com>
6208 * config/i386/i386.c (ix86_print_operand): Handle
6209 V8HF/V16HF/V32HFmode.
6210 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
6211 * config/i386/sse.md (avx512bcst): Remove.
6213 2021-09-17 Martin Sebor <msebor@redhat.com>
6215 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
6216 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
6217 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
6218 (check_defs): Add comment.
6219 (can_skip_redundant_opnd): Update comment.
6220 (compute_uninit_opnds_pos): Adjust to namespace change.
6221 (find_pdom): Move to gimple-predicate-analysis.cc.
6223 (struct uninit_undef_val_t): New.
6224 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
6225 (find_control_equiv_block): Same.
6226 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
6227 (MAX_SWITCH_CASES): Same.
6228 (compute_control_dep_chain): Same.
6229 (find_uninit_use): Use predicate analyzer.
6230 (struct pred_info): Move to gimple-predicate-analysis.
6231 (convert_control_dep_chain_into_preds): Same.
6232 (find_predicates): Same.
6233 (collect_phi_def_edges): Same.
6234 (warn_uninitialized_phi): Use predicate analyzer.
6235 (find_def_preds): Move to gimple-predicate-analysis.
6236 (dump_pred_info): Same.
6237 (dump_pred_chain): Same.
6238 (dump_predicates): Same.
6239 (destroy_predicate_vecs): Remove.
6240 (execute_late_warn_uninitialized): New.
6241 (get_cmp_code): Move to gimple-predicate-analysis.
6242 (is_value_included_in): Same.
6243 (value_sat_pred_p): Same.
6244 (find_matching_predicate_in_rest_chains): Same.
6245 (is_use_properly_guarded): Same.
6246 (prune_uninit_phi_opnds): Same.
6247 (find_var_cmp_const): Same.
6248 (use_pred_not_overlap_with_undef_path_pred): Same.
6249 (pred_equal_p): Same.
6250 (is_neq_relop_p): Same.
6251 (is_neq_zero_form_p): Same.
6252 (pred_expr_equal_p): Same.
6253 (is_pred_expr_subset_of): Same.
6254 (is_pred_chain_subset_of): Same.
6255 (is_included_in): Same.
6256 (is_superset_of): Same.
6258 (simplify_pred): Same.
6259 (simplify_preds_2): Same.
6260 (simplify_preds_3): Same.
6261 (simplify_preds_4): Same.
6262 (simplify_preds): Same.
6264 (push_to_worklist): Same.
6265 (get_pred_info_from_cmp): Same.
6266 (is_degenerated_phi): Same.
6267 (normalize_one_pred_1): Same.
6268 (normalize_one_pred): Same.
6269 (normalize_one_pred_chain): Same.
6270 (normalize_preds): Same.
6271 (can_one_predicate_be_invalidated_p): Same.
6272 (can_chain_union_be_invalidated_p): Same.
6273 (uninit_uses_cannot_happen): Same.
6274 (pass_late_warn_uninitialized::execute): Define.
6275 * gimple-predicate-analysis.cc: New file.
6276 * gimple-predicate-analysis.h: New file.
6278 2021-09-17 Julian Brown <julian@codesourcery.com>
6280 * config/gcn/gcn.c (gimple.h): Include.
6281 (gcn_fork_join): Emit barrier for worker-level joins.
6282 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
6283 writes_gang_private bitmap parameter. Set bit for blocks
6284 containing gang-private variable writes.
6285 (worker_single_simple): Don't emit barrier after predicated block.
6286 (worker_single_copy): Don't emit barrier if we're not broadcasting
6287 anything and the block contains no gang-private writes.
6288 (neuter_worker_single): Don't predicate blocks that only contain
6289 NOPs or internal marker functions. Pass has_gang_private_write
6290 argument to worker_single_copy.
6291 (oacc_do_neutering): Add writes_gang_private bitmap handling.
6293 2021-09-17 Julian Brown <julian@codesourcery.com>
6295 * config/gcn/gcn-protos.h
6296 (gcn_goacc_create_worker_broadcast_record): Update prototype.
6297 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
6298 preallocated block of LDS memory. Do not cache/share decls for
6299 reduction temporaries between invocations.
6300 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
6301 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
6302 and return temporary LDS space at that offset. Return pointer in
6304 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
6306 (ACC_LDS_SIZE): Define as acc_lds_size.
6307 (gcn_init_machine_status): Don't initialise lds_allocated,
6308 lds_allocs, reduc_decls fields of machine function struct.
6309 (gcn_option_override): Handle default size for gang-private
6310 variables and -mgang-private-size option.
6311 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
6312 initialising M0_REG.
6313 (gcn_shared_mem_layout): New function.
6314 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
6315 gang_private_hwm variable.
6316 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
6317 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
6318 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
6319 * config/gcn/gcn.opt (gang_private_size_opt): New global.
6320 (mgang-private-size=): New option.
6321 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
6323 * doc/tm.texi: Regenerate.
6324 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
6326 (build_sender_ref): Handle sender_decl being pointer.
6327 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
6328 parameters. Pass placement argument to
6329 create_worker_broadcast_record hook invocations. Handle
6330 sender_decl being pointer and isolate_broadcasts inserting extra
6332 (blk_offset_map_t): Add typedef.
6333 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
6334 preallocated range to worker_single_copy call.
6335 (dfs_broadcast_reachable_1): New function.
6336 (idx_decl_pair_t, used_range_vec_t): New typedefs.
6337 (sort_size_descending): New function.
6338 (addr_range): New class.
6339 (splay_tree_compare_addr_range, splay_tree_free_key)
6340 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
6341 (execute_omp_oacc_neuter_broadcast): Rename to...
6342 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
6343 parameters. Arrange layout of shared memory for broadcast
6345 (execute_omp_oacc_neuter_broadcast): New function.
6346 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
6347 handling from here. Enable pass for all OpenACC routines in order
6348 to call shared memory-layout hook.
6349 * target.def (create_worker_broadcast_record): Add OFFSET
6351 (shared_mem_layout): New hook.
6353 2021-09-17 Julian Brown <julian@codesourcery.com>
6354 Thomas Schwinge <thomas@codesourcery.com>
6356 * omp-oacc-neuter-broadcast.cc
6357 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
6359 (execute_omp_oacc_neuter_broadcast): Adjust.
6361 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6363 * value-relation.cc (class equiv_chain): Move to header file.
6364 (path_oracle::path_oracle): New.
6365 (path_oracle::~path_oracle): New.
6366 (path_oracle::register_relation): New.
6367 (path_oracle::query_relation): New.
6368 (path_oracle::reset_path): New.
6369 (path_oracle::dump): New.
6370 * value-relation.h (class equiv_chain): Move to here.
6371 (class path_oracle): New.
6373 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6375 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
6377 * gimple-range-fold.cc (fur_depend::register_relation): Use
6378 register_stmt/edge routines.
6379 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
6380 (equiv_oracle::equiv_oracle): Create self equivalence cache.
6381 (equiv_oracle::~equiv_oracle): Release same.
6382 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
6383 are no equivalences.
6384 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
6385 (equiv_oracle::register_relation): Rename from register_equiv.
6386 (relation_chain_head::find_relation): Relocate from dom_oracle.
6387 (relation_oracle::register_stmt): New.
6388 (relation_oracle::register_edge): New.
6389 (dom_oracle::*): Rename from relation_oracle.
6390 (dom_oracle::register_relation): Adjust to call equiv_oracle.
6391 (dom_oracle::set_one_relation): Split from register_relation.
6392 (dom_oracle::register_transitives): Consolidate 2 methods.
6393 (dom_oracle::find_relation_block): Move core to relation_chain.
6394 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
6395 * value-relation.h (class relation_oracle): New pure virtual base.
6396 (class equiv_oracle): Inherit from relation_oracle and adjust.
6397 (class dom_oracle): Rename from old relation_oracle and adjust.
6399 2021-09-17 Martin Sebor <msebor@redhat.com>
6401 PR middle-end/102200
6402 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
6403 (handle_min_max_size): Change argument. Store original SSA_NAME for
6404 operands to potentially distinct (sub)objects.
6405 (compute_objsize_r): Adjust call to the above.
6407 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6409 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
6410 (rs6000_new_builtin_vectorized_function): New function.
6411 (rs6000_new_builtin_md_vectorized_function): Likewise.
6412 (rs6000_builtin_vectorized_function): Call
6413 rs6000_new_builtin_vectorized_function.
6414 (rs6000_builtin_md_vectorized_function): Call
6415 rs6000_new_builtin_md_vectorized_function.
6417 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6419 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
6420 (ASSEMBLE_PAIR): Likewise.
6421 (BUILD_ACC): Likewise.
6422 (DISASSEMBLE_ACC): Likewise.
6423 (DISASSEMBLE_PAIR): Likewise.
6424 (PMXVBF16GER2): Likewise.
6425 (PMXVBF16GER2NN): Likewise.
6426 (PMXVBF16GER2NP): Likewise.
6427 (PMXVBF16GER2PN): Likewise.
6428 (PMXVBF16GER2PP): Likewise.
6429 (PMXVF16GER2): Likewise.
6430 (PMXVF16GER2NN): Likewise.
6431 (PMXVF16GER2NP): Likewise.
6432 (PMXVF16GER2PN): Likewise.
6433 (PMXVF16GER2PP): Likewise.
6434 (PMXVF32GER): Likewise.
6435 (PMXVF32GERNN): Likewise.
6436 (PMXVF32GERNP): Likewise.
6437 (PMXVF32GERPN): Likewise.
6438 (PMXVF32GERPP): Likewise.
6439 (PMXVF64GER): Likewise.
6440 (PMXVF64GERNN): Likewise.
6441 (PMXVF64GERNP): Likewise.
6442 (PMXVF64GERPN): Likewise.
6443 (PMXVF64GERPP): Likewise.
6444 (PMXVI16GER2): Likewise.
6445 (PMXVI16GER2PP): Likewise.
6446 (PMXVI16GER2S): Likewise.
6447 (PMXVI16GER2SPP): Likewise.
6448 (PMXVI4GER8): Likewise.
6449 (PMXVI4GER8PP): Likewise.
6450 (PMXVI8GER4): Likewise.
6451 (PMXVI8GER4PP): Likewise.
6452 (PMXVI8GER4SPP): Likewise.
6453 (XVBF16GER2): Likewise.
6454 (XVBF16GER2NN): Likewise.
6455 (XVBF16GER2NP): Likewise.
6456 (XVBF16GER2PN): Likewise.
6457 (XVBF16GER2PP): Likewise.
6458 (XVF16GER2): Likewise.
6459 (XVF16GER2NN): Likewise.
6460 (XVF16GER2NP): Likewise.
6461 (XVF16GER2PN): Likewise.
6462 (XVF16GER2PP): Likewise.
6463 (XVF32GER): Likewise.
6464 (XVF32GERNN): Likewise.
6465 (XVF32GERNP): Likewise.
6466 (XVF32GERPN): Likewise.
6467 (XVF32GERPP): Likewise.
6468 (XVF64GER): Likewise.
6469 (XVF64GERNN): Likewise.
6470 (XVF64GERNP): Likewise.
6471 (XVF64GERPN): Likewise.
6472 (XVF64GERPP): Likewise.
6473 (XVI16GER2): Likewise.
6474 (XVI16GER2PP): Likewise.
6475 (XVI16GER2S): Likewise.
6476 (XVI16GER2SPP): Likewise.
6477 (XVI4GER8): Likewise.
6478 (XVI4GER8PP): Likewise.
6479 (XVI8GER4): Likewise.
6480 (XVI8GER4PP): Likewise.
6481 (XVI8GER4SPP): Likewise.
6482 (XXMFACC): Likewise.
6483 (XXMTACC): Likewise.
6484 (XXSETACCZ): Likewise.
6485 (ASSEMBLE_PAIR_V): Likewise.
6486 (BUILD_PAIR): Likewise.
6487 (DISASSEMBLE_PAIR_V): Likewise.
6490 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
6491 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
6492 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
6493 (parse_bif_attrs): Handle ismmaint.
6494 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
6495 (write_bif_static_init): Handle ismmaint.
6497 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6499 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
6501 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
6502 (rs6000_new_builtin_valid_without_lhs): New function.
6503 (rs6000_gimple_fold_new_mma_builtin): Likewise.
6504 (rs6000_gimple_fold_new_builtin): Likewise.
6506 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
6508 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
6509 Destruct stale Value objects.
6510 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6513 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
6516 * match.pd (shift optimizations): Disable recent sign-changing
6517 optimization for shifts by zero, these will be folded later.
6519 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6521 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
6522 [power9] to [always].
6524 2021-09-17 Richard Biener <rguenther@suse.de>
6526 * tree-vect-stmts.c (vectorizable_load): Do not frob
6529 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6531 * config/i386/i386-features.c (remove_partial_avx_dependency):
6532 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
6533 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
6535 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
6537 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6538 * config/i386/i386.md (SSE FP to FP splitters): Replace
6539 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
6540 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
6541 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
6542 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
6543 * config/i386/x86-tune.def
6544 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
6545 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6547 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6550 * config/i386/i386-features.c (remove_partial_avx_dependency):
6551 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
6552 before generating vxorps.
6554 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6556 * config/i386/i386-options.c (processor_cost_table): Use
6557 tremont_cost for Tremont.
6558 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
6559 (tremont_memset): Likewise.
6560 (tremont_cost): Likewise.
6561 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
6564 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6566 * common/config/i386/i386-common.c: Use Haswell scheduling model
6568 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
6570 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
6572 (ix86_adjust_cost): Handle Tremont.
6573 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
6575 (X86_TUNE_USE_LEAVE): Likewise.
6576 (X86_TUNE_PUSH_MEMORY): Likewise.
6577 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
6578 (X86_TUNE_USE_CLTD): Likewise.
6579 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
6580 (X86_TUNE_AVOID_MFENCE): Likewise.
6581 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
6582 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
6583 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
6584 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
6585 (X86_TUNE_OPT_AGU): Likewise.
6586 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
6587 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
6588 (X86_TUNE_EXPAND_ABS): Likewise.
6589 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
6590 (X86_TUNE_SLOW_PSHUFB): Likewise.
6592 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
6594 PR rtl-optimization/102306
6595 * combine.c (try_combine): Abort the combination if we are about to
6596 duplicate volatile references.
6598 2021-09-17 liuhongt <hongtao.liu@intel.com>
6600 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
6602 (_mm256_undefined_ph): Likewise.
6603 (_mm512_undefined_ph): Likewise.
6604 (_mm_cvtsh_h): Likewise.
6605 (_mm256_cvtsh_h): Likewise.
6606 (_mm512_cvtsh_h): Likewise.
6607 (_mm512_castph_ps): Likewise.
6608 (_mm512_castph_pd): Likewise.
6609 (_mm512_castph_si512): Likewise.
6610 (_mm512_castph512_ph128): Likewise.
6611 (_mm512_castph512_ph256): Likewise.
6612 (_mm512_castph128_ph512): Likewise.
6613 (_mm512_castph256_ph512): Likewise.
6614 (_mm512_zextph128_ph512): Likewise.
6615 (_mm512_zextph256_ph512): Likewise.
6616 (_mm512_castps_ph): Likewise.
6617 (_mm512_castpd_ph): Likewise.
6618 (_mm512_castsi512_ph): Likewise.
6619 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
6621 (_mm256_castph_ps): Likewise.
6622 (_mm_castph_pd): Likewise.
6623 (_mm256_castph_pd): Likewise.
6624 (_mm_castph_si128): Likewise.
6625 (_mm256_castph_si256): Likewise.
6626 (_mm_castps_ph): Likewise.
6627 (_mm256_castps_ph): Likewise.
6628 (_mm_castpd_ph): Likewise.
6629 (_mm256_castpd_ph): Likewise.
6630 (_mm_castsi128_ph): Likewise.
6631 (_mm256_castsi256_ph): Likewise.
6632 (_mm256_castph256_ph128): Likewise.
6633 (_mm256_castph128_ph256): Likewise.
6634 (_mm256_zextph128_ph256): Likewise.
6636 2021-09-17 liuhongt <hongtao.liu@intel.com>
6638 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
6640 (_mm_mask_cvtsh_ss): Likewise.
6641 (_mm_maskz_cvtsh_ss): Likewise.
6642 (_mm_cvtsh_sd): Likewise.
6643 (_mm_mask_cvtsh_sd): Likewise.
6644 (_mm_maskz_cvtsh_sd): Likewise.
6645 (_mm_cvt_roundsh_ss): Likewise.
6646 (_mm_mask_cvt_roundsh_ss): Likewise.
6647 (_mm_maskz_cvt_roundsh_ss): Likewise.
6648 (_mm_cvt_roundsh_sd): Likewise.
6649 (_mm_mask_cvt_roundsh_sd): Likewise.
6650 (_mm_maskz_cvt_roundsh_sd): Likewise.
6651 (_mm_cvtss_sh): Likewise.
6652 (_mm_mask_cvtss_sh): Likewise.
6653 (_mm_maskz_cvtss_sh): Likewise.
6654 (_mm_cvtsd_sh): Likewise.
6655 (_mm_mask_cvtsd_sh): Likewise.
6656 (_mm_maskz_cvtsd_sh): Likewise.
6657 (_mm_cvt_roundss_sh): Likewise.
6658 (_mm_mask_cvt_roundss_sh): Likewise.
6659 (_mm_maskz_cvt_roundss_sh): Likewise.
6660 (_mm_cvt_roundsd_sh): Likewise.
6661 (_mm_mask_cvt_roundsd_sh): Likewise.
6662 (_mm_maskz_cvt_roundsd_sh): Likewise.
6663 * config/i386/i386-builtin-types.def
6664 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
6665 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
6666 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
6667 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
6668 * config/i386/i386-builtin.def: Add corrresponding new builtins.
6669 * config/i386/i386-expand.c: Handle new builtin types.
6670 * config/i386/sse.md (VF48_128): New mode iterator.
6671 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
6673 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
6676 2021-09-17 liuhongt <hongtao.liu@intel.com>
6678 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
6680 (_mm512_mask_cvtph_pd): Likewise.
6681 (_mm512_maskz_cvtph_pd): Likewise.
6682 (_mm512_cvt_roundph_pd): Likewise.
6683 (_mm512_mask_cvt_roundph_pd): Likewise.
6684 (_mm512_maskz_cvt_roundph_pd): Likewise.
6685 (_mm512_cvtxph_ps): Likewise.
6686 (_mm512_mask_cvtxph_ps): Likewise.
6687 (_mm512_maskz_cvtxph_ps): Likewise.
6688 (_mm512_cvtx_roundph_ps): Likewise.
6689 (_mm512_mask_cvtx_roundph_ps): Likewise.
6690 (_mm512_maskz_cvtx_roundph_ps): Likewise.
6691 (_mm512_cvtxps_ph): Likewise.
6692 (_mm512_mask_cvtxps_ph): Likewise.
6693 (_mm512_maskz_cvtxps_ph): Likewise.
6694 (_mm512_cvtx_roundps_ph): Likewise.
6695 (_mm512_mask_cvtx_roundps_ph): Likewise.
6696 (_mm512_maskz_cvtx_roundps_ph): Likewise.
6697 (_mm512_cvtpd_ph): Likewise.
6698 (_mm512_mask_cvtpd_ph): Likewise.
6699 (_mm512_maskz_cvtpd_ph): Likewise.
6700 (_mm512_cvt_roundpd_ph): Likewise.
6701 (_mm512_mask_cvt_roundpd_ph): Likewise.
6702 (_mm512_maskz_cvt_roundpd_ph): Likewise.
6703 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
6705 (_mm_mask_cvtph_pd): Likewise.
6706 (_mm_maskz_cvtph_pd): Likewise.
6707 (_mm256_cvtph_pd): Likewise.
6708 (_mm256_mask_cvtph_pd): Likewise.
6709 (_mm256_maskz_cvtph_pd): Likewise.
6710 (_mm_cvtxph_ps): Likewise.
6711 (_mm_mask_cvtxph_ps): Likewise.
6712 (_mm_maskz_cvtxph_ps): Likewise.
6713 (_mm256_cvtxph_ps): Likewise.
6714 (_mm256_mask_cvtxph_ps): Likewise.
6715 (_mm256_maskz_cvtxph_ps): Likewise.
6716 (_mm_cvtxps_ph): Likewise.
6717 (_mm_mask_cvtxps_ph): Likewise.
6718 (_mm_maskz_cvtxps_ph): Likewise.
6719 (_mm256_cvtxps_ph): Likewise.
6720 (_mm256_mask_cvtxps_ph): Likewise.
6721 (_mm256_maskz_cvtxps_ph): Likewise.
6722 (_mm_cvtpd_ph): Likewise.
6723 (_mm_mask_cvtpd_ph): Likewise.
6724 (_mm_maskz_cvtpd_ph): Likewise.
6725 (_mm256_cvtpd_ph): Likewise.
6726 (_mm256_mask_cvtpd_ph): Likewise.
6727 (_mm256_maskz_cvtpd_ph): Likewise.
6728 * config/i386/i386-builtin.def: Add corresponding new builtins.
6729 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6730 * config/i386/i386-expand.c: Handle new builtin types.
6731 * config/i386/sse.md
6732 (VF4_128_8_256): New.
6733 (VF48H_AVX512VL): Ditto.
6734 (ssePHmode): Add HF vector modes.
6735 (castmode): Add new convertable modes.
6738 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
6739 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
6740 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
6741 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
6742 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
6743 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
6744 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
6746 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
6747 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
6748 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
6749 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
6751 2021-09-17 liuhongt <hongtao.liu@intel.com>
6753 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
6755 (_mm_cvttsh_u32): Likewise.
6756 (_mm_cvtt_roundsh_i32): Likewise.
6757 (_mm_cvtt_roundsh_u32): Likewise.
6758 (_mm_cvttsh_i64): Likewise.
6759 (_mm_cvttsh_u64): Likewise.
6760 (_mm_cvtt_roundsh_i64): Likewise.
6761 (_mm_cvtt_roundsh_u64): Likewise.
6762 * config/i386/i386-builtin.def: Add corresponding new builtins.
6763 * config/i386/sse.md
6764 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
6767 2021-09-17 liuhongt <hongtao.liu@intel.com>
6769 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
6771 (_mm512_mask_cvttph_epi32): Likewise.
6772 (_mm512_maskz_cvttph_epi32): Likewise.
6773 (_mm512_cvtt_roundph_epi32): Likewise.
6774 (_mm512_mask_cvtt_roundph_epi32): Likewise.
6775 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
6776 (_mm512_cvttph_epu32): Likewise.
6777 (_mm512_mask_cvttph_epu32): Likewise.
6778 (_mm512_maskz_cvttph_epu32): Likewise.
6779 (_mm512_cvtt_roundph_epu32): Likewise.
6780 (_mm512_mask_cvtt_roundph_epu32): Likewise.
6781 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
6782 (_mm512_cvttph_epi64): Likewise.
6783 (_mm512_mask_cvttph_epi64): Likewise.
6784 (_mm512_maskz_cvttph_epi64): Likewise.
6785 (_mm512_cvtt_roundph_epi64): Likewise.
6786 (_mm512_mask_cvtt_roundph_epi64): Likewise.
6787 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
6788 (_mm512_cvttph_epu64): Likewise.
6789 (_mm512_mask_cvttph_epu64): Likewise.
6790 (_mm512_maskz_cvttph_epu64): Likewise.
6791 (_mm512_cvtt_roundph_epu64): Likewise.
6792 (_mm512_mask_cvtt_roundph_epu64): Likewise.
6793 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
6794 (_mm512_cvttph_epi16): Likewise.
6795 (_mm512_mask_cvttph_epi16): Likewise.
6796 (_mm512_maskz_cvttph_epi16): Likewise.
6797 (_mm512_cvtt_roundph_epi16): Likewise.
6798 (_mm512_mask_cvtt_roundph_epi16): Likewise.
6799 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
6800 (_mm512_cvttph_epu16): Likewise.
6801 (_mm512_mask_cvttph_epu16): Likewise.
6802 (_mm512_maskz_cvttph_epu16): Likewise.
6803 (_mm512_cvtt_roundph_epu16): Likewise.
6804 (_mm512_mask_cvtt_roundph_epu16): Likewise.
6805 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
6806 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
6808 (_mm_mask_cvttph_epi32): Likewise.
6809 (_mm_maskz_cvttph_epi32): Likewise.
6810 (_mm256_cvttph_epi32): Likewise.
6811 (_mm256_mask_cvttph_epi32): Likewise.
6812 (_mm256_maskz_cvttph_epi32): Likewise.
6813 (_mm_cvttph_epu32): Likewise.
6814 (_mm_mask_cvttph_epu32): Likewise.
6815 (_mm_maskz_cvttph_epu32): Likewise.
6816 (_mm256_cvttph_epu32): Likewise.
6817 (_mm256_mask_cvttph_epu32): Likewise.
6818 (_mm256_maskz_cvttph_epu32): Likewise.
6819 (_mm_cvttph_epi64): Likewise.
6820 (_mm_mask_cvttph_epi64): Likewise.
6821 (_mm_maskz_cvttph_epi64): Likewise.
6822 (_mm256_cvttph_epi64): Likewise.
6823 (_mm256_mask_cvttph_epi64): Likewise.
6824 (_mm256_maskz_cvttph_epi64): Likewise.
6825 (_mm_cvttph_epu64): Likewise.
6826 (_mm_mask_cvttph_epu64): Likewise.
6827 (_mm_maskz_cvttph_epu64): Likewise.
6828 (_mm256_cvttph_epu64): Likewise.
6829 (_mm256_mask_cvttph_epu64): Likewise.
6830 (_mm256_maskz_cvttph_epu64): Likewise.
6831 (_mm_cvttph_epi16): Likewise.
6832 (_mm_mask_cvttph_epi16): Likewise.
6833 (_mm_maskz_cvttph_epi16): Likewise.
6834 (_mm256_cvttph_epi16): Likewise.
6835 (_mm256_mask_cvttph_epi16): Likewise.
6836 (_mm256_maskz_cvttph_epi16): Likewise.
6837 (_mm_cvttph_epu16): Likewise.
6838 (_mm_mask_cvttph_epu16): Likewise.
6839 (_mm_maskz_cvttph_epu16): Likewise.
6840 (_mm256_cvttph_epu16): Likewise.
6841 (_mm256_mask_cvttph_epu16): Likewise.
6842 (_mm256_maskz_cvttph_epu16): Likewise.
6843 * config/i386/i386-builtin.def: Add new builtins.
6844 * config/i386/sse.md
6845 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
6847 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
6848 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
6849 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
6850 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
6852 2021-09-17 liuhongt <hongtao.liu@intel.com>
6854 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
6855 (_mm_cvtsh_u32): Likewise.
6856 (_mm_cvt_roundsh_i32): Likewise.
6857 (_mm_cvt_roundsh_u32): Likewise.
6858 (_mm_cvtsh_i64): Likewise.
6859 (_mm_cvtsh_u64): Likewise.
6860 (_mm_cvt_roundsh_i64): Likewise.
6861 (_mm_cvt_roundsh_u64): Likewise.
6862 (_mm_cvti32_sh): Likewise.
6863 (_mm_cvtu32_sh): Likewise.
6864 (_mm_cvt_roundi32_sh): Likewise.
6865 (_mm_cvt_roundu32_sh): Likewise.
6866 (_mm_cvti64_sh): Likewise.
6867 (_mm_cvtu64_sh): Likewise.
6868 (_mm_cvt_roundi64_sh): Likewise.
6869 (_mm_cvt_roundu64_sh): Likewise.
6870 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6871 * config/i386/i386-builtin.def: Add corresponding new builtins.
6872 * config/i386/i386-expand.c (ix86_expand_round_builtin):
6873 Handle new builtin types.
6874 * config/i386/sse.md
6875 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
6877 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
6878 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
6880 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
6882 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
6883 (altivec_resolve_new_overloaded_builtin): New forward decl.
6884 (rs6000_new_builtin_type_compatible): New function.
6885 (altivec_resolve_overloaded_builtin): Call
6886 altivec_resolve_new_overloaded_builtin.
6887 (altivec_build_new_resolved_builtin): New function.
6888 (altivec_resolve_new_overloaded_builtin): Likewise.
6889 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
6891 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
6892 name of rs6000_new_builtin_is_supported.
6894 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
6896 * config/i386/i386-protos.h (ix86_decompose_address):
6897 Change return type to bool.
6898 * config/i386/i386.c (ix86_decompose_address): Ditto.
6900 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
6903 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
6904 Added 'build/' to target, use build/%.o rule.
6905 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
6906 '$(build_exeext)' to target and 'build/' for the *.o files.
6907 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
6910 2021-09-16 Martin Jambor <mjambor@suse.cz>
6912 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
6913 the body has been removed. Remove trailing whitespace.
6915 2021-09-16 Richard Biener <rguenther@suse.de>
6917 PR middle-end/102360
6918 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
6919 of non-memory more robust.
6921 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6923 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
6924 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
6925 (leon5_adjust_cost): Increase cost of store with data dependency
6926 on ALU instruction and FPU anti-dependencies.
6927 (sparc_option_override): Add LEON5 costs
6928 (sparc_adjust_cost): Add LEON5 cost adjustments
6929 * config/sparc/sparc.h: Add LEON5
6930 * config/sparc/sparc.md: Include LEON5 scheduling information
6931 * config/sparc/sparc.opt: Add LEON5
6932 * doc/invoke.texi: Add LEON5
6933 * config/sparc/leon5.md: New file.
6935 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6937 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
6938 sensitive sequence for B2BST errata workaround.
6940 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6942 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
6943 functions with atomic instruction in the UT700 errata workaround.
6945 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6947 * config/sparc/sparc.c (next_active_non_empty_insn): New function
6948 that returns next active non empty assembly instruction.
6949 (sparc_do_work_around_errata): Use new function.
6951 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6953 * config/sparc/sparc.c (store_insn_p): Add predicate for store
6955 (load_insn_p): Add predicate for load attributes.
6956 (sparc_do_work_around_errata): Use new predicates.
6958 2021-09-16 Andreas Larsson <andreas@gaisler.com>
6960 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
6963 2021-09-16 Martin Liska <mliska@suse.cz>
6965 * config/mips/netbsd.h: Fix typo in name of a macro.
6967 2021-09-16 liuhongt <hongtao.liu@intel.com>
6969 PR middle-end/102080
6970 * match.pd: Check mask type when doing cond_op related gimple
6972 * tree.c (is_truth_type_for): New function.
6973 * tree.h (is_truth_type_for): New declaration.
6975 2021-09-16 liuhongt <hongtao.liu@intel.com>
6977 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
6979 (_mm512_mask_cvtepi32_ph): Likewise.
6980 (_mm512_maskz_cvtepi32_ph): Likewise.
6981 (_mm512_cvt_roundepi32_ph): Likewise.
6982 (_mm512_mask_cvt_roundepi32_ph): Likewise.
6983 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
6984 (_mm512_cvtepu32_ph): Likewise.
6985 (_mm512_mask_cvtepu32_ph): Likewise.
6986 (_mm512_maskz_cvtepu32_ph): Likewise.
6987 (_mm512_cvt_roundepu32_ph): Likewise.
6988 (_mm512_mask_cvt_roundepu32_ph): Likewise.
6989 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
6990 (_mm512_cvtepi64_ph): Likewise.
6991 (_mm512_mask_cvtepi64_ph): Likewise.
6992 (_mm512_maskz_cvtepi64_ph): Likewise.
6993 (_mm512_cvt_roundepi64_ph): Likewise.
6994 (_mm512_mask_cvt_roundepi64_ph): Likewise.
6995 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
6996 (_mm512_cvtepu64_ph): Likewise.
6997 (_mm512_mask_cvtepu64_ph): Likewise.
6998 (_mm512_maskz_cvtepu64_ph): Likewise.
6999 (_mm512_cvt_roundepu64_ph): Likewise.
7000 (_mm512_mask_cvt_roundepu64_ph): Likewise.
7001 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
7002 (_mm512_cvtepi16_ph): Likewise.
7003 (_mm512_mask_cvtepi16_ph): Likewise.
7004 (_mm512_maskz_cvtepi16_ph): Likewise.
7005 (_mm512_cvt_roundepi16_ph): Likewise.
7006 (_mm512_mask_cvt_roundepi16_ph): Likewise.
7007 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
7008 (_mm512_cvtepu16_ph): Likewise.
7009 (_mm512_mask_cvtepu16_ph): Likewise.
7010 (_mm512_maskz_cvtepu16_ph): Likewise.
7011 (_mm512_cvt_roundepu16_ph): Likewise.
7012 (_mm512_mask_cvt_roundepu16_ph): Likewise.
7013 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
7014 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
7016 (_mm_mask_cvtepi32_ph): Likewise.
7017 (_mm_maskz_cvtepi32_ph): Likewise.
7018 (_mm256_cvtepi32_ph): Likewise.
7019 (_mm256_mask_cvtepi32_ph): Likewise.
7020 (_mm256_maskz_cvtepi32_ph): Likewise.
7021 (_mm_cvtepu32_ph): Likewise.
7022 (_mm_mask_cvtepu32_ph): Likewise.
7023 (_mm_maskz_cvtepu32_ph): Likewise.
7024 (_mm256_cvtepu32_ph): Likewise.
7025 (_mm256_mask_cvtepu32_ph): Likewise.
7026 (_mm256_maskz_cvtepu32_ph): Likewise.
7027 (_mm_cvtepi64_ph): Likewise.
7028 (_mm_mask_cvtepi64_ph): Likewise.
7029 (_mm_maskz_cvtepi64_ph): Likewise.
7030 (_mm256_cvtepi64_ph): Likewise.
7031 (_mm256_mask_cvtepi64_ph): Likewise.
7032 (_mm256_maskz_cvtepi64_ph): Likewise.
7033 (_mm_cvtepu64_ph): Likewise.
7034 (_mm_mask_cvtepu64_ph): Likewise.
7035 (_mm_maskz_cvtepu64_ph): Likewise.
7036 (_mm256_cvtepu64_ph): Likewise.
7037 (_mm256_mask_cvtepu64_ph): Likewise.
7038 (_mm256_maskz_cvtepu64_ph): Likewise.
7039 (_mm_cvtepi16_ph): Likewise.
7040 (_mm_mask_cvtepi16_ph): Likewise.
7041 (_mm_maskz_cvtepi16_ph): Likewise.
7042 (_mm256_cvtepi16_ph): Likewise.
7043 (_mm256_mask_cvtepi16_ph): Likewise.
7044 (_mm256_maskz_cvtepi16_ph): Likewise.
7045 (_mm_cvtepu16_ph): Likewise.
7046 (_mm_mask_cvtepu16_ph): Likewise.
7047 (_mm_maskz_cvtepu16_ph): Likewise.
7048 (_mm256_cvtepu16_ph): Likewise.
7049 (_mm256_mask_cvtepu16_ph): Likewise.
7050 (_mm256_maskz_cvtepu16_ph): Likewise.
7051 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7052 * config/i386/i386-builtin.def: Add corresponding new builtins.
7053 * config/i386/i386-expand.c
7054 (ix86_expand_args_builtin): Handle new builtin types.
7055 (ix86_expand_round_builtin): Ditto.
7056 * config/i386/i386-modes.def: Declare V2HF and V6HF.
7057 * config/i386/sse.md (VI2H_AVX512VL): New.
7059 (sseintvecmode): Add HF vector modes.
7060 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
7062 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
7063 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
7064 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
7065 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
7066 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
7067 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
7068 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
7069 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
7070 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
7071 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
7072 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
7074 2021-09-16 liuhongt <hongtao.liu@intel.com>
7076 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
7078 (_mm512_mask_cvtph_epi32): Likewise.
7079 (_mm512_maskz_cvtph_epi32): Likewise.
7080 (_mm512_cvt_roundph_epi32): Likewise.
7081 (_mm512_mask_cvt_roundph_epi32): Likewise.
7082 (_mm512_maskz_cvt_roundph_epi32): Likewise.
7083 (_mm512_cvtph_epu32): Likewise.
7084 (_mm512_mask_cvtph_epu32): Likewise.
7085 (_mm512_maskz_cvtph_epu32): Likewise.
7086 (_mm512_cvt_roundph_epu32): Likewise.
7087 (_mm512_mask_cvt_roundph_epu32): Likewise.
7088 (_mm512_maskz_cvt_roundph_epu32): Likewise.
7089 (_mm512_cvtph_epi64): Likewise.
7090 (_mm512_mask_cvtph_epi64): Likewise.
7091 (_mm512_maskz_cvtph_epi64): Likewise.
7092 (_mm512_cvt_roundph_epi64): Likewise.
7093 (_mm512_mask_cvt_roundph_epi64): Likewise.
7094 (_mm512_maskz_cvt_roundph_epi64): Likewise.
7095 (_mm512_cvtph_epu64): Likewise.
7096 (_mm512_mask_cvtph_epu64): Likewise.
7097 (_mm512_maskz_cvtph_epu64): Likewise.
7098 (_mm512_cvt_roundph_epu64): Likewise.
7099 (_mm512_mask_cvt_roundph_epu64): Likewise.
7100 (_mm512_maskz_cvt_roundph_epu64): Likewise.
7101 (_mm512_cvtph_epi16): Likewise.
7102 (_mm512_mask_cvtph_epi16): Likewise.
7103 (_mm512_maskz_cvtph_epi16): Likewise.
7104 (_mm512_cvt_roundph_epi16): Likewise.
7105 (_mm512_mask_cvt_roundph_epi16): Likewise.
7106 (_mm512_maskz_cvt_roundph_epi16): Likewise.
7107 (_mm512_cvtph_epu16): Likewise.
7108 (_mm512_mask_cvtph_epu16): Likewise.
7109 (_mm512_maskz_cvtph_epu16): Likewise.
7110 (_mm512_cvt_roundph_epu16): Likewise.
7111 (_mm512_mask_cvt_roundph_epu16): Likewise.
7112 (_mm512_maskz_cvt_roundph_epu16): Likewise.
7113 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
7115 (_mm_mask_cvtph_epi32): Likewise.
7116 (_mm_maskz_cvtph_epi32): Likewise.
7117 (_mm256_cvtph_epi32): Likewise.
7118 (_mm256_mask_cvtph_epi32): Likewise.
7119 (_mm256_maskz_cvtph_epi32): Likewise.
7120 (_mm_cvtph_epu32): Likewise.
7121 (_mm_mask_cvtph_epu32): Likewise.
7122 (_mm_maskz_cvtph_epu32): Likewise.
7123 (_mm256_cvtph_epu32): Likewise.
7124 (_mm256_mask_cvtph_epu32): Likewise.
7125 (_mm256_maskz_cvtph_epu32): Likewise.
7126 (_mm_cvtph_epi64): Likewise.
7127 (_mm_mask_cvtph_epi64): Likewise.
7128 (_mm_maskz_cvtph_epi64): Likewise.
7129 (_mm256_cvtph_epi64): Likewise.
7130 (_mm256_mask_cvtph_epi64): Likewise.
7131 (_mm256_maskz_cvtph_epi64): Likewise.
7132 (_mm_cvtph_epu64): Likewise.
7133 (_mm_mask_cvtph_epu64): Likewise.
7134 (_mm_maskz_cvtph_epu64): Likewise.
7135 (_mm256_cvtph_epu64): Likewise.
7136 (_mm256_mask_cvtph_epu64): Likewise.
7137 (_mm256_maskz_cvtph_epu64): Likewise.
7138 (_mm_cvtph_epi16): Likewise.
7139 (_mm_mask_cvtph_epi16): Likewise.
7140 (_mm_maskz_cvtph_epi16): Likewise.
7141 (_mm256_cvtph_epi16): Likewise.
7142 (_mm256_mask_cvtph_epi16): Likewise.
7143 (_mm256_maskz_cvtph_epi16): Likewise.
7144 (_mm_cvtph_epu16): Likewise.
7145 (_mm_mask_cvtph_epu16): Likewise.
7146 (_mm_maskz_cvtph_epu16): Likewise.
7147 (_mm256_cvtph_epu16): Likewise.
7148 (_mm256_mask_cvtph_epu16): Likewise.
7149 (_mm256_maskz_cvtph_epu16): Likewise.
7150 * config/i386/i386-builtin-types.def: Add new builtin types.
7151 * config/i386/i386-builtin.def: Add new builtins.
7152 * config/i386/i386-expand.c
7153 (ix86_expand_args_builtin): Handle new builtin types.
7154 (ix86_expand_round_builtin): Ditto.
7155 * config/i386/sse.md (sseintconvert): New.
7157 (UNSPEC_US_FIX_NOTRUNC): Ditto.
7158 (sseintconvertsignprefix): Ditto.
7159 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
7162 2021-09-16 liuhongt <hongtao.liu@intel.com>
7164 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
7166 (_mm_cvtsi128_si16): Likewise.
7167 (_mm_mask_load_sh): Likewise.
7168 (_mm_maskz_load_sh): Likewise.
7169 (_mm_mask_store_sh): Likewise.
7170 (_mm_move_sh): Likewise.
7171 (_mm_mask_move_sh): Likewise.
7172 (_mm_maskz_move_sh): Likewise.
7173 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7174 * config/i386/i386-builtin.def: Add corresponding new builtins.
7175 * config/i386/i386-expand.c
7176 (ix86_expand_special_args_builtin): Handle new builtin types.
7177 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
7178 * config/i386/sse.md (VI2F): New mode iterator.
7179 (vec_set<mode>_0): Use new mode iterator.
7180 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
7181 (avx512f_store<mode>_mask): Ditto.
7183 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
7185 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
7187 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
7189 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
7190 Proceed if no symbol summary or the symbol alias flag is false.
7192 2021-09-15 Jakub Jelinek <jakub@redhat.com>
7196 * varasm.c (output_constructor_regular_field): Instead of assertion
7197 that array_size_for_constructor result is equal to size of
7198 TREE_TYPE (local->val) in bytes, assert that the type size is greater
7199 or equal to array_size_for_constructor result and use type size as
7202 2021-09-15 Martin Liska <mliska@suse.cz>
7205 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
7207 2021-09-15 Martin Liska <mliska@suse.cz>
7210 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
7211 Check that we have a symbol summary for a symbol.
7213 2021-09-15 Richard Biener <rguenther@suse.de>
7216 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
7217 to inherit from elfos.h
7219 2021-09-15 liuhongt <hongtao.liu@intel.com>
7222 * config/i386/i386-expand.c
7223 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
7225 (ix86_expand_vector_set): Use blendw instead of pinsrw.
7226 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
7227 AVX512FP16 which supports 16bit vector load.
7228 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
7230 (avx512bw_interleave_high<mode><mask_name>): .. this, and
7231 extend to V32HFmode.
7232 (avx2_interleave_highv16hi<mask_name>): Rename to ..
7233 (avx2_interleave_high<mode><mask_name>): .. this, and extend
7235 (vec_interleave_highv8hi<mask_name>): Rename to ..
7236 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
7237 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
7239 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
7240 this, and extend to V32HFmode.
7241 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
7242 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
7243 (vec_interleave_lowv8hi<mask_name>): Rename to ..
7244 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
7245 (sse4_1_pblendw): Rename to ..
7246 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
7247 (avx2_pblendph): New define_expand.
7248 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
7249 sseintmodesuffix instead of ssemodesuffix.
7250 (blendsuf): New mode attr.
7252 2021-09-15 Richard Biener <rguenther@suse.de>
7254 * tree-vectorizer.h (dr_misalignment): Move out of line.
7255 (dr_target_alignment): New.
7256 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
7257 (set_dr_target_alignment): New.
7258 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
7259 * tree-vect-data-refs.c (dr_misalignment): Compute and
7260 return the group members misalignment.
7261 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
7262 (vect_analyze_data_refs_alignment): Compute alignment only
7263 for the first element of a DR group.
7264 (vect_slp_analyze_node_alignment): Likewise.
7266 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
7268 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
7269 * config/i386/avx512fp16vlintrin.h: Likewise.
7270 * config/i386/i386-builtin.def: Adjust builtin name and
7271 enumeration to match AVX512F style.
7273 2021-09-15 Richard Biener <rguenther@suse.de>
7275 PR tree-optimization/102318
7276 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
7277 previous change and do the mode conversion separately from
7278 the sign conversion.
7280 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
7281 Peter Cordes <peter@cordes.ca>
7284 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
7285 (*vec_extract<mode><ssescalarmodelower>_valign): Output
7286 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
7289 2021-09-15 Richard Biener <rguenther@suse.de>
7291 * config.gcc: Remove vax-*-openbsd* configuration.
7293 2021-09-15 Richard Biener <rguenther@suse.de>
7295 * config.gcc: Remove m68k-openbsd.
7297 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
7300 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
7302 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
7304 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
7305 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
7306 (*mma_xxsetaccz): Delete.
7307 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
7308 Use UNSPECV_MMA_XXSETACCZ. Update comment.
7309 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
7311 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
7313 * Makefile.in: Remove variables related to applying no-PIE
7314 to the exes on $build.
7315 * configure: Regenerate.
7316 * configure.ac: Remove configuration related to applying
7317 no-PIE to the exes on $build.
7319 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
7321 * config/arc/arc.md (doloop_end): Add missing mode.
7322 (loop_end): Likewise.
7324 2021-09-14 Jakub Jelinek <jakub@redhat.com>
7326 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
7327 it to recursive calls, for depth above 7 just gimplify or return.
7328 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
7329 __builtin_clear_padding and TARGET_EXPR.
7330 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
7332 2021-09-14 liuhongt <hongtao.liu@intel.com>
7334 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
7336 (_mm_mask_fpclass_sh_mask): Likewise.
7337 (_mm512_mask_fpclass_ph_mask): Likewise.
7338 (_mm512_fpclass_ph_mask): Likewise.
7339 (_mm_getexp_sh): Likewise.
7340 (_mm_mask_getexp_sh): Likewise.
7341 (_mm_maskz_getexp_sh): Likewise.
7342 (_mm512_getexp_ph): Likewise.
7343 (_mm512_mask_getexp_ph): Likewise.
7344 (_mm512_maskz_getexp_ph): Likewise.
7345 (_mm_getexp_round_sh): Likewise.
7346 (_mm_mask_getexp_round_sh): Likewise.
7347 (_mm_maskz_getexp_round_sh): Likewise.
7348 (_mm512_getexp_round_ph): Likewise.
7349 (_mm512_mask_getexp_round_ph): Likewise.
7350 (_mm512_maskz_getexp_round_ph): Likewise.
7351 (_mm_getmant_sh): Likewise.
7352 (_mm_mask_getmant_sh): Likewise.
7353 (_mm_maskz_getmant_sh): Likewise.
7354 (_mm512_getmant_ph): Likewise.
7355 (_mm512_mask_getmant_ph): Likewise.
7356 (_mm512_maskz_getmant_ph): Likewise.
7357 (_mm_getmant_round_sh): Likewise.
7358 (_mm_mask_getmant_round_sh): Likewise.
7359 (_mm_maskz_getmant_round_sh): Likewise.
7360 (_mm512_getmant_round_ph): Likewise.
7361 (_mm512_mask_getmant_round_ph): Likewise.
7362 (_mm512_maskz_getmant_round_ph): Likewise.
7363 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
7365 (_mm_fpclass_ph_mask): Likewise.
7366 (_mm256_mask_fpclass_ph_mask): Likewise.
7367 (_mm256_fpclass_ph_mask): Likewise.
7368 (_mm256_getexp_ph): Likewise.
7369 (_mm256_mask_getexp_ph): Likewise.
7370 (_mm256_maskz_getexp_ph): Likewise.
7371 (_mm_getexp_ph): Likewise.
7372 (_mm_mask_getexp_ph): Likewise.
7373 (_mm_maskz_getexp_ph): Likewise.
7374 (_mm256_getmant_ph): Likewise.
7375 (_mm256_mask_getmant_ph): Likewise.
7376 (_mm256_maskz_getmant_ph): Likewise.
7377 (_mm_getmant_ph): Likewise.
7378 (_mm_mask_getmant_ph): Likewise.
7379 (_mm_maskz_getmant_ph): Likewise.
7380 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7381 * config/i386/i386-builtin.def: Add corresponding new builtins.
7382 * config/i386/i386-expand.c
7383 (ix86_expand_args_builtin): Handle new builtin types.
7384 (ix86_expand_round_builtin): Ditto.
7385 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
7386 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
7387 to support HF vector modes.
7388 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
7390 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
7391 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
7392 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
7393 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7396 2021-09-14 liuhongt <hongtao.liu@intel.com>
7398 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
7400 (_mm512_mask_reduce_ph): Likewise.
7401 (_mm512_maskz_reduce_ph): Likewise.
7402 (_mm512_reduce_round_ph): Likewise.
7403 (_mm512_mask_reduce_round_ph): Likewise.
7404 (_mm512_maskz_reduce_round_ph): Likewise.
7405 (_mm_reduce_sh): Likewise.
7406 (_mm_mask_reduce_sh): Likewise.
7407 (_mm_maskz_reduce_sh): Likewise.
7408 (_mm_reduce_round_sh): Likewise.
7409 (_mm_mask_reduce_round_sh): Likewise.
7410 (_mm_maskz_reduce_round_sh): Likewise.
7411 (_mm512_roundscale_ph): Likewise.
7412 (_mm512_mask_roundscale_ph): Likewise.
7413 (_mm512_maskz_roundscale_ph): Likewise.
7414 (_mm512_roundscale_round_ph): Likewise.
7415 (_mm512_mask_roundscale_round_ph): Likewise.
7416 (_mm512_maskz_roundscale_round_ph): Likewise.
7417 (_mm_roundscale_sh): Likewise.
7418 (_mm_mask_roundscale_sh): Likewise.
7419 (_mm_maskz_roundscale_sh): Likewise.
7420 (_mm_roundscale_round_sh): Likewise.
7421 (_mm_mask_roundscale_round_sh): Likewise.
7422 (_mm_maskz_roundscale_round_sh): Likewise.
7423 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
7425 (_mm_mask_reduce_ph): Likewise.
7426 (_mm_maskz_reduce_ph): Likewise.
7427 (_mm256_reduce_ph): Likewise.
7428 (_mm256_mask_reduce_ph): Likewise.
7429 (_mm256_maskz_reduce_ph): Likewise.
7430 (_mm_roundscale_ph): Likewise.
7431 (_mm_mask_roundscale_ph): Likewise.
7432 (_mm_maskz_roundscale_ph): Likewise.
7433 (_mm256_roundscale_ph): Likewise.
7434 (_mm256_mask_roundscale_ph): Likewise.
7435 (_mm256_maskz_roundscale_ph): Likewise.
7436 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7437 * config/i386/i386-builtin.def: Add corresponding new builtins.
7438 * config/i386/i386-expand.c
7439 (ix86_expand_args_builtin): Handle new builtin types.
7440 (ix86_expand_round_builtin): Ditto.
7441 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
7443 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
7444 ... this, and adjust for round operands.
7445 (reduces<mode><mask_scalar_name>): Likewise, with ...
7446 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
7448 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
7449 Adjust for HF vector modes.
7450 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
7452 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
7454 2021-09-14 liuhongt <hongtao.liu@intel.com>
7456 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
7458 (_mm512_mask_rcp_ph): Likewise.
7459 (_mm512_maskz_rcp_ph): Likewise.
7460 (_mm_rcp_sh): Likewise.
7461 (_mm_mask_rcp_sh): Likewise.
7462 (_mm_maskz_rcp_sh): Likewise.
7463 (_mm512_scalef_ph): Likewise.
7464 (_mm512_mask_scalef_ph): Likewise.
7465 (_mm512_maskz_scalef_ph): Likewise.
7466 (_mm512_scalef_round_ph): Likewise.
7467 (_mm512_mask_scalef_round_ph): Likewise.
7468 (_mm512_maskz_scalef_round_ph): Likewise.
7469 (_mm_scalef_sh): Likewise.
7470 (_mm_mask_scalef_sh): Likewise.
7471 (_mm_maskz_scalef_sh): Likewise.
7472 (_mm_scalef_round_sh): Likewise.
7473 (_mm_mask_scalef_round_sh): Likewise.
7474 (_mm_maskz_scalef_round_sh): Likewise.
7475 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
7477 (_mm256_rcp_ph): Likewise.
7478 (_mm_mask_rcp_ph): Likewise.
7479 (_mm256_mask_rcp_ph): Likewise.
7480 (_mm_maskz_rcp_ph): Likewise.
7481 (_mm256_maskz_rcp_ph): Likewise.
7482 (_mm_scalef_ph): Likewise.
7483 (_mm256_scalef_ph): Likewise.
7484 (_mm_mask_scalef_ph): Likewise.
7485 (_mm256_mask_scalef_ph): Likewise.
7486 (_mm_maskz_scalef_ph): Likewise.
7487 (_mm256_maskz_scalef_ph): Likewise.
7488 * config/i386/i386-builtin.def: Add new builtins.
7489 * config/i386/sse.md (VFH_AVX512VL): New.
7490 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
7491 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
7492 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
7493 Adjust to support HF vector modes.
7494 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
7496 2021-09-14 liuhongt <hongtao.liu@intel.com>
7498 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
7500 (_mm512_mask_sqrt_ph): Likewise.
7501 (_mm512_maskz_sqrt_ph): Likewise.
7502 (_mm512_sqrt_round_ph): Likewise.
7503 (_mm512_mask_sqrt_round_ph): Likewise.
7504 (_mm512_maskz_sqrt_round_ph): Likewise.
7505 (_mm512_rsqrt_ph): Likewise.
7506 (_mm512_mask_rsqrt_ph): Likewise.
7507 (_mm512_maskz_rsqrt_ph): Likewise.
7508 (_mm_rsqrt_sh): Likewise.
7509 (_mm_mask_rsqrt_sh): Likewise.
7510 (_mm_maskz_rsqrt_sh): Likewise.
7511 (_mm_sqrt_sh): Likewise.
7512 (_mm_mask_sqrt_sh): Likewise.
7513 (_mm_maskz_sqrt_sh): Likewise.
7514 (_mm_sqrt_round_sh): Likewise.
7515 (_mm_mask_sqrt_round_sh): Likewise.
7516 (_mm_maskz_sqrt_round_sh): Likewise.
7517 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
7518 (_mm256_sqrt_ph): Likewise.
7519 (_mm_mask_sqrt_ph): Likewise.
7520 (_mm256_mask_sqrt_ph): Likewise.
7521 (_mm_maskz_sqrt_ph): Likewise.
7522 (_mm256_maskz_sqrt_ph): Likewise.
7523 (_mm_rsqrt_ph): Likewise.
7524 (_mm256_rsqrt_ph): Likewise.
7525 (_mm_mask_rsqrt_ph): Likewise.
7526 (_mm256_mask_rsqrt_ph): Likewise.
7527 (_mm_maskz_rsqrt_ph): Likewise.
7528 (_mm256_maskz_rsqrt_ph): Likewise.
7529 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7530 * config/i386/i386-builtin.def: Add corresponding new builtins.
7531 * config/i386/i386-expand.c
7532 (ix86_expand_args_builtin): Handle new builtins.
7533 (ix86_expand_round_builtin): Ditto.
7534 * config/i386/sse.md (VF_AVX512FP16VL): New.
7535 (sqrt<mode>2): Adjust for HF vector modes.
7536 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
7537 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
7539 (<sse>_rsqrt<mode>2<mask_name>): New.
7540 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
7542 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7545 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
7546 'RESERVED_LOCATION_P' locations.
7547 * warning-control.cc (get_nowarn_spec, suppress_warning)
7548 (copy_warning): Likewise.
7550 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7552 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
7553 (typedef key_type_t): ... this. Remove.
7554 (nowarn_map): Document.
7555 * diagnostic-spec.c (nowarn_map): Likewise.
7556 * warning-control.cc (convert_to_key): Evolve functions into...
7557 (get_location): ... these. Adjust all users.
7559 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7561 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
7563 2021-09-13 Jason Merrill <jason@redhat.com>
7565 * params.opt: Add destructive-interference-size and
7566 constructive-interference-size.
7567 * doc/invoke.texi: Document them.
7568 * config/aarch64/aarch64.c (aarch64_override_options_internal):
7570 * config/arm/arm.c (arm_option_override): Set them.
7571 * config/i386/i386-options.c (ix86_option_override_internal):
7574 2021-09-13 Martin Liska <mliska@suse.cz>
7575 H.J. Lu <hjl.tools@gmail.com>
7578 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
7579 for x86-64 micro levels for __builtin_cpu_supports.
7580 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
7581 Add priorities for the micro-arch levels.
7582 (enum processor_features): Add new features.
7583 * common/config/i386/i386-isas.h: Add micro-arch features.
7584 * config/i386/i386-builtins.c (get_builtin_code_for_version):
7585 Support the micro-arch levels by callsing
7586 __builtin_cpu_supports.
7587 * doc/extend.texi: Document that the levels are support by
7588 __builtin_cpu_supports.
7590 2021-09-13 Andrew Pinski <apinski@marvell.com>
7593 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
7595 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
7596 (aarch64_general_gimple_fold_builtin): Likewise.
7598 2021-09-13 Andrew Pinski <apinski@marvell.com>
7600 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
7601 to the Unsupported targets list.
7602 Remove support for m32r-*-linux* and m32rle-*-linux*.
7603 * config/m32r/linux.h: Removed.
7604 * config/m32r/t-linux: Removed.
7606 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7609 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
7610 register index for SVE predicate modes.
7612 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7614 * tree-ssa-threadbackward.c
7615 (back_threader_profitability::profitable_path_p): Remove FSM
7617 (back_threader_registry::register_path): Same.
7618 * tree-ssa-threadedge.c
7619 (jump_threader::simplify_control_stmt_condition): Same.
7620 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
7621 Add backedge_threads argument.
7622 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
7623 backedge_threads argument.
7624 (back_jt_path_registry::back_jt_path_registry): Same.
7625 (dump_jump_thread_path): Adjust for FSM removal.
7626 (back_jt_path_registry::rewire_first_differing_edge): Same.
7627 (back_jt_path_registry::adjust_paths_after_duplication): Same.
7628 (back_jt_path_registry::update_cfg): Same.
7629 (jt_path_registry::register_jump_thread): Same.
7630 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
7632 (class back_jt_path_registry): Add backedge_threads to
7635 2021-09-13 Martin Liska <mliska@suse.cz>
7638 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
7640 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
7643 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
7644 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
7645 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
7646 TARGET_AVX256_STORE_BY_PIECES instead of
7647 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
7648 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7649 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
7650 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7651 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
7652 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
7654 2021-09-13 liuhongt <hongtao.liu@intel.com>
7657 * expmed.c (extract_bit_field_using_extv): Use
7658 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
7660 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7662 * Makefile.in (OBJS): Add value-pointer-equiv.o.
7663 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
7664 value-pointer-equiv.*.
7665 (ssa_equiv_stack::ssa_equiv_stack): Same.
7666 (ssa_equiv_stack::enter): Same.
7667 (ssa_equiv_stack::leave): Same.
7668 (ssa_equiv_stack::push_replacement): Same.
7669 (ssa_equiv_stack::get_replacement): Same.
7670 (is_pointer_ssa): Same.
7671 (class pointer_equiv_analyzer): Same.
7672 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7673 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
7674 (pointer_equiv_analyzer::set_global_equiv): Same.
7675 (pointer_equiv_analyzer::set_cond_equiv): Same.
7676 (pointer_equiv_analyzer::get_equiv): Same.
7677 (pointer_equiv_analyzer::enter): Same.
7678 (pointer_equiv_analyzer::leave): Same.
7679 (pointer_equiv_analyzer::get_equiv_expr): Same.
7680 (pta_valueize): Same.
7681 (pointer_equiv_analyzer::visit_stmt): Same.
7682 (pointer_equiv_analyzer::visit_edge): Same.
7683 (hybrid_folder::value_of_expr): Same.
7684 (hybrid_folder::value_on_edge): Same.
7685 * value-pointer-equiv.cc: New file.
7686 * value-pointer-equiv.h: New file.
7688 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7691 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
7692 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
7694 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7697 * config/arm/arm.md (movmisaligndi): New define_expand.
7698 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
7700 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7703 * emit-rtl.c (gen_highpart): Use adjust_address to handle
7704 MEM rather than calling simplify_gen_subreg.
7706 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
7708 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
7709 define into a block.
7711 2021-09-13 Richard Biener <rguenther@suse.de>
7713 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
7714 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
7715 change based on DARWIN_PREFER_DWARF not being defined.
7717 2021-09-13 Richard Biener <rguenther@suse.de>
7719 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
7720 to inherit from elfos.h
7722 2021-09-13 Richard Biener <rguenther@suse.de>
7724 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
7726 2021-09-13 Richard Biener <rguenther@suse.de>
7728 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
7729 override, pick up DWARF2_DEBUG define from elfos.h
7731 2021-09-13 Richard Biener <rguenther@suse.de>
7733 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
7736 2021-09-13 Richard Biener <rguenther@suse.de>
7738 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
7741 2021-09-13 Richard Biener <rguenther@suse.de>
7743 * config/i386/cygming.h: Always default to DWARF2 debugging.
7744 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
7746 * doc/install.texi: Document binutils 2.16 as minimum
7747 requirement for mingw.
7749 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
7751 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
7752 nstmts, nloads and extra_ctor_cost.
7753 (rs6000_density_test): Add load density related heuristics. Do
7754 extra costing on vector construction statements if need.
7755 (rs6000_init_cost): Init new members.
7756 (rs6000_update_target_cost_per_stmt): New function.
7757 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
7758 rs6000_update_target_cost_per_stmt and call it.
7760 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
7762 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
7763 (rs6000_init_cost): Adjust.
7765 2021-09-13 liuhongt <hongtao.liu@intel.com>
7767 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
7768 (UNSPEC_XORSIGN): Ditto.
7770 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
7772 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
7773 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
7776 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
7778 * tree-ssa-threadbackward.c (class back_threader_registry): Use
7779 back_jt_path_registry.
7780 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
7781 fwd_jt_path_registry.
7782 * tree-ssa-threadedge.h (class jump_threader): Same..
7783 * tree-ssa-threadupdate.c
7784 (jump_thread_path_registry::jump_thread_path_registry): Rename...
7785 (jt_path_registry::jt_path_registry): ...to this.
7786 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
7787 (jt_path_registry::~jt_path_registry): ...this.
7788 (fwd_jt_path_registry::fwd_jt_path_registry): New.
7789 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
7790 (jump_thread_path_registry::allocate_thread_edge): Rename...
7791 (jt_path_registry::allocate_thread_edge): ...to this.
7792 (jump_thread_path_registry::allocate_thread_path): Rename...
7793 (jt_path_registry::allocate_thread_path): ...to this.
7794 (jump_thread_path_registry::lookup_redirection_data): Rename...
7795 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
7796 (jump_thread_path_registry::thread_block_1): Rename...
7797 (fwd_jt_path_registry::thread_block_1): ...to this.
7798 (jump_thread_path_registry::thread_block): Rename...
7799 (fwd_jt_path_registry::thread_block): ...to this.
7800 (jt_path_registry::thread_through_loop_header): Rename...
7801 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
7802 (jump_thread_path_registry::mark_threaded_blocks): Rename...
7803 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
7804 (jump_thread_path_registry::debug_path): Rename...
7805 (jt_path_registry::debug_path): ...to this.
7806 (jump_thread_path_registry::dump): Rename...
7807 (jt_path_registry::debug): ...to this.
7808 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
7809 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
7810 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
7811 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
7812 (jump_thread_path_registry::duplicate_thread_path): Rename...
7813 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
7814 drop ill-formed candidates.
7815 (jump_thread_path_registry::remove_jump_threads_including): Rename...
7816 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
7817 (jt_path_registry::thread_through_all_blocks): New.
7818 (back_jt_path_registry::update_cfg): New.
7819 (fwd_jt_path_registry::update_cfg): New.
7820 (jump_thread_path_registry::register_jump_thread): Rename...
7821 (jt_path_registry::register_jump_thread): ...to this.
7822 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7824 (class jt_path_registry): ...here.
7825 (class fwd_jt_path_registry): New.
7826 (class back_jt_path_registry): New.
7828 2021-09-10 liuhongt <hongtao.liu@intel.com>
7831 2021-09-01 liuhongt <hongtao.liu@intel.com>
7833 * emit-rtl.c (validate_subreg): Get rid of all float-int
7836 2021-09-10 Jakub Jelinek <jakub@redhat.com>
7838 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
7839 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
7840 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
7841 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
7842 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
7843 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
7844 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
7845 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
7847 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
7848 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
7850 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
7852 * tree.h (OMP_ATOMIC_WEAK): Define.
7853 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
7854 fail memory order being encoded in the same enum and also print
7855 fail clause if present.
7856 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
7857 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
7858 handle pre_p == NULL case as a test mode that only returns value
7859 but doesn't change gimplify nor change anything otherwise, adjust
7860 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
7861 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
7862 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
7864 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
7865 COND_EXPR rhs. Set weak flag on gimple load/store for
7867 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
7868 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
7870 (expand_omp_atomic_cas): New function.
7871 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
7873 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
7874 using expand_omp_atomic_cas.
7876 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
7877 Michael Matz <matz@suse.de>
7879 * tree-pass.h (PROP_loop_opts_done): New.
7880 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
7881 Intersect with global range.
7882 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
7883 * tree-ssa-threadbackward.c
7884 (back_threader_profitability::profitable_path_p): Disable
7885 threading through latches until after loop optimizations have run.
7887 2021-09-10 David Faust <david.faust@oracle.com>
7889 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
7892 2021-09-10 David Faust <david.faust@oracle.com>
7894 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
7895 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
7896 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
7897 (bpf_expand_cbranch): New function.
7898 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
7900 (zero_extendsidi2): Only use mov32 instruction if it is available.
7901 (SIM mode iterator): Conditionalize support for SI mode.
7902 (JM mode iterator): New.
7903 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
7904 (*branch_on_di): Update name, use new JM iterator.
7905 * config/bpf/bpf.opt: (mjmpext): New option.
7909 (bpf_isa): New enum.
7911 2021-09-10 David Faust <david.faust@oracle.com>
7913 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
7914 for register-to-register extensions.
7915 (zero_extendqidi2): Likewise.
7917 2021-09-10 Richard Biener <rguenther@suse.de>
7919 PR middle-end/102273
7920 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
7922 2021-09-10 Richard Biener <rguenther@suse.de>
7924 PR middle-end/102269
7925 * gimplify.c (is_var_need_auto_init): Empty types do not need
7928 2021-09-10 Richard Biener <rguenther@suse.de>
7930 * configure.ac (--with-stabs): Remove.
7931 * configure: Regenerate.
7932 * doc/install.texi: Remove --with-stabs documentation.
7934 2021-09-10 liuhongt <hongtao.liu@intel.com>
7936 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
7938 (_mm512_mask_cmp_ph_mask): Likewise.
7939 (_mm512_cmp_round_ph_mask): Likewise.
7940 (_mm512_mask_cmp_round_ph_mask): Likewise.
7941 (_mm_cmp_sh_mask): Likewise.
7942 (_mm_mask_cmp_sh_mask): Likewise.
7943 (_mm_cmp_round_sh_mask): Likewise.
7944 (_mm_mask_cmp_round_sh_mask): Likewise.
7945 (_mm_comieq_sh): Likewise.
7946 (_mm_comilt_sh): Likewise.
7947 (_mm_comile_sh): Likewise.
7948 (_mm_comigt_sh): Likewise.
7949 (_mm_comige_sh): Likewise.
7950 (_mm_comineq_sh): Likewise.
7951 (_mm_ucomieq_sh): Likewise.
7952 (_mm_ucomilt_sh): Likewise.
7953 (_mm_ucomile_sh): Likewise.
7954 (_mm_ucomigt_sh): Likewise.
7955 (_mm_ucomige_sh): Likewise.
7956 (_mm_ucomineq_sh): Likewise.
7957 (_mm_comi_round_sh): Likewise.
7958 (_mm_comi_sh): Likewise.
7959 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
7960 (_mm_mask_cmp_ph_mask): Likewise.
7961 (_mm256_cmp_ph_mask): Likewise.
7962 (_mm256_mask_cmp_ph_mask): Likewise.
7963 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7964 * config/i386/i386-builtin.def: Add corresponding new builtins.
7965 * config/i386/i386-expand.c
7966 (ix86_expand_args_builtin): Handle new builtin types.
7967 (ix86_expand_round_builtin): Ditto.
7968 * config/i386/i386.md (ssevecmode): Add HF mode.
7969 (MODEFH): New mode iterator.
7970 * config/i386/sse.md
7971 (V48H_AVX512VL): New mode iterator to support HF vector modes.
7972 Ajdust corresponding description.
7973 (ssecmpintprefix): New.
7974 (VI12_AVX512VL): Adjust to support HF vector modes.
7975 (cmp_imm_predicate): Likewise.
7976 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
7978 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
7979 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
7980 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
7982 2021-09-10 liuhongt <hongtao.liu@intel.com>
7984 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
7985 (_mm512_mask_max_ph): Likewise.
7986 (_mm512_maskz_max_ph): Likewise.
7987 (_mm512_min_ph): Likewise.
7988 (_mm512_mask_min_ph): Likewise.
7989 (_mm512_maskz_min_ph): Likewise.
7990 (_mm512_max_round_ph): Likewise.
7991 (_mm512_mask_max_round_ph): Likewise.
7992 (_mm512_maskz_max_round_ph): Likewise.
7993 (_mm512_min_round_ph): Likewise.
7994 (_mm512_mask_min_round_ph): Likewise.
7995 (_mm512_maskz_min_round_ph): Likewise.
7996 (_mm_max_sh): Likewise.
7997 (_mm_mask_max_sh): Likewise.
7998 (_mm_maskz_max_sh): Likewise.
7999 (_mm_min_sh): Likewise.
8000 (_mm_mask_min_sh): Likewise.
8001 (_mm_maskz_min_sh): Likewise.
8002 (_mm_max_round_sh): Likewise.
8003 (_mm_mask_max_round_sh): Likewise.
8004 (_mm_maskz_max_round_sh): Likewise.
8005 (_mm_min_round_sh): Likewise.
8006 (_mm_mask_min_round_sh): Likewise.
8007 (_mm_maskz_min_round_sh): Likewise.
8008 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
8009 (_mm256_max_ph): Likewise.
8010 (_mm_mask_max_ph): Likewise.
8011 (_mm256_mask_max_ph): Likewise.
8012 (_mm_maskz_max_ph): Likewise.
8013 (_mm256_maskz_max_ph): Likewise.
8014 (_mm_min_ph): Likewise.
8015 (_mm256_min_ph): Likewise.
8016 (_mm_mask_min_ph): Likewise.
8017 (_mm256_mask_min_ph): Likewise.
8018 (_mm_maskz_min_ph): Likewise.
8019 (_mm256_maskz_min_ph): Likewise.
8020 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8021 * config/i386/i386-builtin.def: Add corresponding new builtins.
8022 * config/i386/i386-expand.c
8023 (ix86_expand_args_builtin): Handle new builtin types.
8024 * config/i386/sse.md
8025 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
8026 support HF vector modes.
8027 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
8028 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
8030 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
8032 * config/i386/subst.md (round_saeonly_mode512bit_condition):
8033 Adjust for HF vector modes.
8035 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
8037 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
8038 (_mm_mask_add_sh): Likewise.
8039 (_mm_maskz_add_sh): Likewise.
8040 (_mm_sub_sh): Likewise.
8041 (_mm_mask_sub_sh): Likewise.
8042 (_mm_maskz_sub_sh): Likewise.
8043 (_mm_mul_sh): Likewise.
8044 (_mm_mask_mul_sh): Likewise.
8045 (_mm_maskz_mul_sh): Likewise.
8046 (_mm_div_sh): Likewise.
8047 (_mm_mask_div_sh): Likewise.
8048 (_mm_maskz_div_sh): Likewise.
8049 (_mm_add_round_sh): Likewise.
8050 (_mm_mask_add_round_sh): Likewise.
8051 (_mm_maskz_add_round_sh): Likewise.
8052 (_mm_sub_round_sh): Likewise.
8053 (_mm_mask_sub_round_sh): Likewise.
8054 (_mm_maskz_sub_round_sh): Likewise.
8055 (_mm_mul_round_sh): Likewise.
8056 (_mm_mask_mul_round_sh): Likewise.
8057 (_mm_maskz_mul_round_sh): Likewise.
8058 (_mm_div_round_sh): Likewise.
8059 (_mm_mask_div_round_sh): Likewise.
8060 (_mm_maskz_div_round_sh): Likewise.
8061 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8062 * config/i386/i386-builtin.def: Add corresponding new builtins.
8063 * config/i386/i386-expand.c
8064 (ix86_expand_round_builtin): Handle new builtins.
8065 * config/i386/sse.md (VF_128): Change description.
8066 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
8067 Adjust to support HF vector modes.
8068 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
8071 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
8073 * config/i386/i386-expand.c
8074 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
8075 * config/i386/i386.c
8076 (ix86_preferred_simd_mode): Handle HF mode.
8077 * config/i386/sse.md (V_256H): New mode iterator.
8078 (avx_vextractf128<mode>): Use it.
8079 (VEC_INIT_MODE): Align vector HFmode condition to vector
8080 HImodes since there're no real HF instruction used.
8081 (VEC_INIT_HALF_MODE): Ditto.
8083 (VIHF_AVX512BW): Ditto.
8084 (*vec_extracthf): Ditto.
8085 (VEC_EXTRACT_MODE): Ditto.
8087 2021-09-10 Richard Biener <rguenther@suse.de>
8090 * config/dbx.h: Remove.
8091 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
8092 * config/lynx.h: Likewise.
8094 2021-09-10 liuhongt <hongtao.liu@intel.com>
8096 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
8097 right into ANDNOT + AND + IOR, using paradoxical subregs.
8098 (ix86_split_copysign_const): Remove.
8099 (ix86_split_copysign_var): Ditto.
8100 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
8101 (ix86_split_copysign_var): Ditto.
8102 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
8103 (@copysign<mode>3_var): Ditto.
8105 2021-09-09 qing zhao <qing.zhao@oracle.com>
8107 * builtins.c (expand_builtin_memset): Make external visible.
8108 * builtins.h (expand_builtin_memset): Declare extern.
8109 * common.opt (ftrivial-auto-var-init=): New option.
8110 * doc/extend.texi: Document the uninitialized attribute.
8111 * doc/invoke.texi: Document -ftrivial-auto-var-init.
8112 * flag-types.h (enum auto_init_type): New enumerated type
8114 * gimple-fold.c (clear_padding_type): Add one new parameter.
8115 (clear_padding_union): Likewise.
8116 (clear_padding_emit_loop): Likewise.
8117 (clear_type_padding_in_mask): Likewise.
8118 (gimple_fold_builtin_clear_padding): Handle this new parameter.
8119 * gimplify.c (gimple_add_init_for_auto_var): New function.
8120 (gimple_add_padding_init_for_auto_var): New function.
8121 (is_var_need_auto_init): New function.
8122 (gimplify_decl_expr): Add initialization to automatic variables per
8124 (gimplify_call_expr): Add one new parameter for call to
8125 __builtin_clear_padding.
8126 (gimplify_init_constructor): Add padding initialization in the end.
8127 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
8128 (expand_DEFERRED_INIT): New function.
8129 * internal-fn.def (DEFERRED_INIT): New internal function.
8130 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
8131 * tree-sra.c (generate_subtree_deferred_init): New function.
8132 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
8133 calls to .DEFERRED_INIT.
8134 (sra_modify_deferred_init): New function.
8135 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
8136 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
8137 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
8139 (check_defs): Likewise.
8140 (warn_uninitialized_vars): Likewise.
8141 * tree-ssa.c (ssa_undefined_value_p): Likewise.
8142 * tree.c (build_common_builtin_nodes): Build tree node for
8143 BUILT_IN_CLEAR_PADDING when needed.
8145 2021-09-09 Richard Biener <rguenther@suse.de>
8147 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
8150 2021-09-09 Richard Biener <rguenther@suse.de>
8152 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
8153 DOM walk from get_loop_body_in_dom_order using a worklist
8156 2021-09-09 liuhongt <hongtao.liu@intel.com>
8158 * config.gcc: Add avx512fp16vlintrin.h.
8159 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
8160 (_mm512_mask_add_ph): Likewise.
8161 (_mm512_maskz_add_ph): Likewise.
8162 (_mm512_sub_ph): Likewise.
8163 (_mm512_mask_sub_ph): Likewise.
8164 (_mm512_maskz_sub_ph): Likewise.
8165 (_mm512_mul_ph): Likewise.
8166 (_mm512_mask_mul_ph): Likewise.
8167 (_mm512_maskz_mul_ph): Likewise.
8168 (_mm512_div_ph): Likewise.
8169 (_mm512_mask_div_ph): Likewise.
8170 (_mm512_maskz_div_ph): Likewise.
8171 (_mm512_add_round_ph): Likewise.
8172 (_mm512_mask_add_round_ph): Likewise.
8173 (_mm512_maskz_add_round_ph): Likewise.
8174 (_mm512_sub_round_ph): Likewise.
8175 (_mm512_mask_sub_round_ph): Likewise.
8176 (_mm512_maskz_sub_round_ph): Likewise.
8177 (_mm512_mul_round_ph): Likewise.
8178 (_mm512_mask_mul_round_ph): Likewise.
8179 (_mm512_maskz_mul_round_ph): Likewise.
8180 (_mm512_div_round_ph): Likewise.
8181 (_mm512_mask_div_round_ph): Likewise.
8182 (_mm512_maskz_div_round_ph): Likewise.
8183 * config/i386/avx512fp16vlintrin.h: New header.
8184 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
8185 Add new builtin types.
8186 * config/i386/i386-builtin.def: Add corresponding builtins.
8187 * config/i386/i386-expand.c
8188 (ix86_expand_args_builtin): Handle new builtin types.
8189 (ix86_expand_round_builtin): Likewise.
8190 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
8191 * config/i386/sse.md (VFH): New mode_iterator.
8193 (avx512fmaskmode): Add HF vector modes.
8194 (avx512fmaskhalfmode): Likewise.
8195 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
8197 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
8198 (mul<mode>3<mask_name><round_name>): Likewise.
8199 (*mul<mode>3<mask_name><round_name>): Likewise.
8200 (div<mode>3): Likewise.
8201 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
8202 * config/i386/subst.md (SUBST_V): Add HF vector modes.
8203 (SUBST_A): Likewise.
8204 (round_mode512bit_condition): Adjust for V32HFmode.
8206 2021-09-09 liuhongt <hongtao.liu@intel.com>
8209 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
8210 (reduc_plus_scal_v4sf): .. this, New define_expand.
8211 (reduc_plus_scal_v2df): .. and this, New define_expand.
8213 2021-09-09 liuhongt <hongtao.liu@intel.com>
8216 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
8219 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
8222 * doc/trouble.texi (Copy Assignment): Fix description of
8223 behaviour and fix code in example.
8225 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
8228 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
8229 r11 instead of r12 for restoring CR.
8231 2021-09-08 Jakub Jelinek <jakub@redhat.com>
8232 liuhongt <hongtao.liu@intel.com>
8235 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
8236 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
8237 into AND with mask and XOR, using paradoxical subregs.
8238 (ix86_split_xorsign): Remove.
8239 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
8241 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
8243 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
8245 2021-09-08 Jakub Jelinek <jakub@redhat.com>
8248 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
8249 operands[2], emit abs<mode>2 instead.
8250 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
8251 first alternative even for avx, add another alternative with
8252 =&Yv <- 0, Yv, Yvm constraints.
8253 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
8254 to op1, emit vpandn instead.
8256 2021-09-08 liuhongt <hongtao.liu@intel.com>
8258 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
8259 (_mm256_set_ph): Likewise.
8260 (_mm512_set_ph): Likewise.
8261 (_mm_setr_ph): Likewise.
8262 (_mm256_setr_ph): Likewise.
8263 (_mm512_setr_ph): Likewise.
8264 (_mm_set1_ph): Likewise.
8265 (_mm256_set1_ph): Likewise.
8266 (_mm512_set1_ph): Likewise.
8267 (_mm_setzero_ph): Likewise.
8268 (_mm256_setzero_ph): Likewise.
8269 (_mm512_setzero_ph): Likewise.
8270 (_mm_set_sh): Likewise.
8271 (_mm_load_sh): Likewise.
8272 (_mm_store_sh): Likewise.
8273 * config/i386/i386-builtin-types.def (V8HF): New type.
8274 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
8275 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
8276 Support vector HFmodes.
8277 (ix86_expand_vector_init_one_nonzero): Likewise.
8278 (ix86_expand_vector_init_one_var): Likewise.
8279 (ix86_expand_vector_init_interleave): Likewise.
8280 (ix86_expand_vector_init_general): Likewise.
8281 (ix86_expand_vector_set): Likewise.
8282 (ix86_expand_vector_extract): Likewise.
8283 (ix86_expand_vector_init_concat): Likewise.
8284 (ix86_expand_sse_movcc): Handle vector HFmodes.
8285 (ix86_expand_vector_set_var): Ditto.
8286 * config/i386/i386-modes.def: Add HF vector modes in comment.
8287 * config/i386/i386.c (classify_argument): Add HF vector modes.
8288 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
8289 (ix86_vector_mode_supported_p): Likewise.
8290 (ix86_set_reg_reg_cost): Handle vector HFmode.
8291 (ix86_get_ssemov): Handle vector HFmode.
8292 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
8294 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
8296 (function_arg_advance_32): Ditto.
8297 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
8298 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
8299 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
8300 (VALID_SSE2_REG_VHF_MODE): New.
8301 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
8302 (SSE_REG_MODE_P): Add vector HFmode.
8303 * config/i386/i386.md (mode): Add HF vector modes.
8304 (MODE_SIZE): Likewise.
8305 (ssemodesuffix): Add ph suffix for HF vector modes.
8306 * config/i386/sse.md (VFH_128): New mode iterator.
8307 (VMOVE): Adjust for HF vector modes.
8309 (V_256_512): Likewise.
8311 (avx512fmaskmode): Likewise.
8312 (shuffletype): Likewise.
8313 (sseinsnmode): Likewise.
8314 (ssedoublevecmode): Likewise.
8315 (ssehalfvecmode): Likewise.
8316 (ssehalfvecmodelower): Likewise.
8317 (ssePScmode): Likewise.
8318 (ssescalarmode): Likewise.
8319 (ssescalarmodelower): Likewise.
8320 (sseintprefix): Likewise.
8322 (bcstscalarsuff): Likewise.
8323 (xtg_mode): Likewise.
8324 (VI12HF_AVX512VL): New mode_iterator.
8325 (VF_AVX512FP16): Likewise.
8327 (VIHF_256): Likewise.
8328 (VIHF_AVX512BW): Likewise.
8329 (V16_256): Likewise.
8330 (V32_512): Likewise.
8331 (sseintmodesuffix): New mode_attr.
8332 (sse): Add scalar and vector HFmodes.
8333 (ssescalarmode): Add vector HFmode mapping.
8334 (ssescalarmodesuffix): Add sh suffix for HFmode.
8335 (*<sse>_vm<insn><mode>3): Use VFH_128.
8336 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
8337 (*ieee_<ieee_maxmin><mode>3): Likewise.
8338 (<avx512>_blendm<mode>): New define_insn.
8339 (vec_setv8hf): New define_expand.
8340 (vec_set<mode>_0): New define_insn for HF vector set.
8341 (*avx512fp16_movsh): Likewise.
8342 (avx512fp16_movsh): Likewise.
8343 (vec_extract_lo_v32hi): Rename to ...
8344 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
8346 (vec_extract_hi_v32hi): Likewise.
8347 (vec_extract_hi_<mode>): Likewise.
8348 (vec_extract_lo_v16hi): Likewise.
8349 (vec_extract_lo_<mode>): Likewise.
8350 (vec_extract_hi_v16hi): Likewise.
8351 (vec_extract_hi_<mode>): Likewise.
8352 (vec_set_hi_v16hi): Likewise.
8353 (vec_set_hi_<mode>): Likewise.
8354 (vec_set_lo_v16hi): Likewise.
8355 (vec_set_lo_<mode>): Likewise.
8356 (*vec_extract<mode>_0): New define_insn_and_split for HF
8358 (*vec_extracthf): New define_insn.
8359 (VEC_EXTRACT_MODE): Add HF vector modes.
8360 (PINSR_MODE): Add V8HF.
8361 (sse2p4_1): Likewise.
8362 (pinsr_evex_isa): Likewise.
8363 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
8364 insert for V8HFmode.
8365 (pbroadcast_evex_isa): Add HF vector modes.
8366 (AVX2_VEC_DUP_MODE): Likewise.
8367 (VEC_INIT_MODE): Likewise.
8368 (VEC_INIT_HALF_MODE): Likewise.
8369 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
8371 (avx2_pbroadcast<mode>_1): Likewise.
8372 (<avx512>_vec_dup<mode>_1): Likewise.
8373 (<avx512>_vec_dup<mode><mask_name>): Likewise.
8374 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
8377 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
8378 H.J. Lu <hongjiu.lu@intel.com>
8379 Liu Hongtao <hongtao.liu@intel.com>
8380 Wang Hongyu <hongyu.wang@intel.com>
8381 Xu Dianhong <dianhong.xu@intel.com>
8383 * common/config/i386/cpuinfo.h (get_available_features):
8384 Detect FEATURE_AVX512FP16.
8385 * common/config/i386/i386-common.c
8386 (OPTION_MASK_ISA_AVX512FP16_SET,
8387 OPTION_MASK_ISA_AVX512FP16_UNSET,
8388 OPTION_MASK_ISA2_AVX512FP16_SET,
8389 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
8390 (OPTION_MASK_ISA2_AVX512BW_UNSET,
8391 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
8392 (ix86_handle_option): Handle -mavx512fp16.
8393 * common/config/i386/i386-cpuinfo.h (enum processor_features):
8394 Add FEATURE_AVX512FP16.
8395 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
8396 * config.gcc: Add avx512fp16intrin.h.
8397 * config/i386/avx512fp16intrin.h: New intrinsic header.
8398 * config/i386/cpuid.h: Add bit_AVX512FP16.
8399 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
8400 * config/i386/i386-builtins.c: Support _Float16 type for i386
8402 (ix86_register_float16_builtin_type): New function.
8403 (ix86_float16_type_node): New.
8404 * config/i386/i386-c.c (ix86_target_macros_internal): Define
8406 * config/i386/i386-expand.c (ix86_expand_branch): Support
8408 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
8409 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8410 (ix86_expand_fp_movcc): Ditto.
8411 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
8412 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
8413 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
8414 * config/i386/i386.c (ix86_get_ssemov): Use
8415 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
8416 (ix86_get_excess_precision): Use
8417 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
8419 (sse_store_index): Use SFmode cost for HFmode cost.
8420 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
8421 GPR cost for HFmode.
8422 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
8423 (ix86_mangle_type): Add manlging for _Float16 type.
8424 (inline_secondary_memory_needed): No memory is needed for
8425 16bit movement between gpr and sse reg under
8427 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
8428 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8429 (ix86_division_cost): Ditto.
8430 (ix86_rtx_costs): Ditto.
8431 (ix86_add_stmt_cost): Ditto.
8432 (ix86_optab_supported_p): Ditto.
8433 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
8434 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
8435 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
8436 * config/i386/i386.md (mode): Add HFmode.
8437 (MODE_SIZE): Add HFmode.
8438 (isa): Add avx512fp16.
8439 (enabled): Handle avx512fp16.
8440 (ssemodesuffix): Add sh suffix for HFmode.
8441 (comm): Add mult, div.
8442 (plusminusmultdiv): New code iterator.
8443 (insn): Add mult, div.
8444 (*movhf_internal): Adjust for avx512fp16 instruction.
8445 (*movhi_internal): Ditto.
8446 (*cmpi<unord>hf): New define_insn for HFmode.
8447 (*ieee_s<ieee_maxmin>hf3): Likewise.
8448 (extendhf<mode>2): Likewise.
8449 (trunc<mode>hf2): Likewise.
8450 (float<floatunssuffix><mode>hf2): Likewise.
8451 (*<insn>hf): Likewise.
8452 (cbranchhf4): New expander.
8453 (movhfcc): Likewise.
8454 (<insn>hf3): Likewise.
8457 * config/i386/i386.opt: Add mavx512fp16.
8458 * config/i386/immintrin.h: Include avx512fp16intrin.h.
8459 * doc/invoke.texi: Add mavx512fp16.
8460 * doc/extend.texi: Add avx512fp16 Usage Notes.
8462 2021-09-08 liuhongt <hongtao.liu@intel.com>
8464 * common.opt: Support -fexcess-precision=16.
8465 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
8466 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
8467 EXCESS_PRECISION_TYPE_FLOAT16.
8468 * config/arm/arm.c (arm_excess_precision): Ditto.
8469 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
8470 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
8471 when EXCESS_PRECISION_TYPE_FLOAT16.
8472 * config/s390/s390.c (s390_excess_precision): Ditto.
8473 * coretypes.h (enum excess_precision_type): Add
8474 EXCESS_PRECISION_TYPE_FLOAT16.
8475 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
8476 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
8477 * doc/extend.texi (Half-Precision): Document
8478 -fexcess-precision=16.
8479 * flag-types.h (enum excess_precision): Add
8480 EXCESS_PRECISION_FLOAT16.
8481 * target.def (excess_precision): Update document.
8482 * tree.c (excess_precision_type): Set excess_precision_type to
8483 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
8485 2021-09-08 liuhongt <hongtao.liu@intel.com>
8487 * doc/extend.texi: (@node Floating Types): Adjust the wording.
8488 (@node Half-Precision): Ditto.
8490 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
8493 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
8494 'CONST_INT_P (src)' to the condition of the block that tries to
8495 eliminate literal when loading integer contant.
8497 2021-09-07 David Faust <david.faust@oracle.com>
8499 * doc/extend.texi (BPF Type Attributes) New node.
8500 Document new preserve_access_index attribute.
8501 Document new preserve_access_index builtin.
8502 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
8504 2021-09-07 David Faust <david.faust@oracle.com>
8506 * config/bpf/bpf.c: Adjust includes.
8507 (bpf_handle_preserve_access_index_attribute): New function.
8508 (bpf_attribute_table): Use it here.
8509 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
8510 (bpf_option_override): Handle "-mco-re" option.
8511 (bpf_asm_init_sections): New.
8512 (TARGET_ASM_INIT_SECTIONS): Redefine.
8513 (bpf_file_end): New.
8514 (TARGET_ASM_FILE_END): Redefine.
8515 (bpf_init_builtins): Add "__builtin_preserve_access_index".
8516 (bpf_core_compute, bpf_core_get_index): New.
8517 (is_attr_preserve_access): New.
8518 (bpf_expand_builtin): Handle new builtins.
8519 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
8520 (bpf_core_walk): New.
8521 (bpf_resolve_overloaded_builtin): New.
8522 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
8524 (pass_bpf_core_attr): New RTL pass.
8525 * config/bpf/bpf-passes.def: New file.
8526 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
8527 * config/bpf/coreout.c: New file.
8528 * config/bpf/coreout.h: Likewise.
8529 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
8530 (coreout.o): New rule.
8531 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
8532 * config.gcc (bpf): Add coreout.h to extra_headers.
8533 Add coreout.o to extra_objs.
8534 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
8536 2021-09-07 David Faust <david.faust@oracle.com>
8538 * btfout.c (get_btf_id): Function is no longer static.
8539 * ctfc.h: Expose it here.
8541 2021-09-07 David Faust <david.faust@oracle.com>
8543 * ctfc.c (ctf_lookup_tree_type): New function.
8546 2021-09-07 David Faust <david.faust@oracle.com>
8548 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
8549 * ctfc.h: Analogous change.
8551 2021-09-07 David Faust <david.faust@oracle.com>
8553 * dwarf2out.c (lookup_type_die): Function is no longer static.
8554 * dwarf2out.h: Expose it here.
8556 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8558 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
8559 (ctf_debug_early_finish): New definition.
8560 (ctf_debug_finish): Likewise.
8561 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
8562 (ctf_debug_early_finish): New declaration.
8563 (ctf_debug_finish): Likewise.
8564 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
8565 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
8567 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8569 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
8570 support when compiling for CO-RE.
8571 * config/bpf/bpf.opt: Add new command line option -mco-re.
8573 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8575 * flag-types.h (enum debug_info_type): Add new enum
8576 DINFO_TYPE_BTF_WITH_CORE.
8577 (BTF_WITH_CORE_DEBUG): New bitmask.
8578 * flags.h (btf_with_core_debuginfo_p): New declaration.
8579 * opts.c (btf_with_core_debuginfo_p): New definition.
8581 2021-09-07 Jason Merrill <jason@redhat.com>
8583 * tree.h (error_operand_p): Change to inline function.
8585 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
8587 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
8588 (empty_block_with_phis_p): ...this.
8589 (potentially_threadable_block): Same.
8590 (jump_threader::thread_through_normal_block): Same.
8592 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
8595 * dwarf2out.c (mark_base_types): New overloaded function.
8596 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
8597 as the compilation unit, and call move_marked_base_types afterward.
8599 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
8602 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
8604 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
8606 2021-09-07 Richard Biener <rguenther@suse.de>
8608 PR tree-optimization/102226
8609 * tree-vect-loop.c (vect_transform_cycle_phi): Record
8610 the converted value for the epilogue PHI use.
8612 2021-09-07 Martin Liska <mliska@suse.cz>
8614 PR gcov-profile/80223
8615 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
8616 options, do not inline when no_profile_instrument_function
8617 attributes are different in early inliner. It's fine to inline
8618 it after PGO instrumentation.
8620 2021-09-07 Richard Biener <rguenther@suse.de>
8622 PR tree-optimization/101555
8623 * tree-ssa-pre.c (translate_vuse_through_block): Do not
8624 perform an alias walk to determine the validity of the
8625 mem at the start of the block which is already guaranteed
8626 by means of prune_clobbered_mems.
8627 (phi_translate_1): Pass edge to translate_vuse_through_block.
8629 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
8632 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
8633 (remainder<mode>3): Likewise.
8635 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
8637 * config/mips/mips.c (mips_file_start): add .module for
8640 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
8642 * wide-int.cc (wi::clz): Reorder tests to ensure the result
8643 is zero for all negative values.
8645 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
8647 * doc/invoke.texi (-foffload-options): Fix @opindex.
8649 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
8652 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
8653 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
8654 source alternative for AVX.
8656 2021-09-06 liuhongt <hongtao.liu@intel.com>
8658 PR middle-end/102182
8659 * optabs.c (expand_fix): Add from1 to avoid from being
8662 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
8664 * dwarf2out.c (modified_type_die): Deal with all array types earlier
8665 and use local variable consistently throughout the function.
8667 2021-09-06 Jakub Jelinek <jakub@redhat.com>
8669 PR tree-optimization/102207
8670 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8671 were promoted from signed to wider unsigned type.
8673 2021-09-06 Andrew Pinski <apinski@marvell.com>
8675 PR tree-optimization/63184
8676 * match.pd: Add simplification of pointer_diff of two pointer_plus
8677 with addr_expr in the first operand of each pointer_plus.
8678 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
8679 in the first operand of each pointer_plus.
8681 2021-09-06 Richard Biener <rguenther@suse.de>
8683 PR tree-optimization/102176
8684 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
8686 (vect_bb_slp_scalar_cost): Use the computed set of
8687 vectorized scalar stmts instead of relying on the out-of-date
8688 and not accurate PURE_SLP_STMT.
8689 (vect_bb_vectorization_profitable_p): Compute the set
8690 of vectorized scalar stmts.
8692 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8694 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
8695 GIMPLE_COND special casing.
8696 (path_range_query::range_defined_in_block): Use range_of_stmt
8697 instead of calling fold_range directly.
8699 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8701 * gimple-range-path.cc (path_range_query::range_of_expr): Set
8702 m_undefined_path when appropriate.
8703 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
8704 (path_range_query::unreachable_path_p): New.
8705 (path_range_query::precompute_ranges): Set m_undefined_path.
8706 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
8707 (path_range_query::internal_range_of_expr): New.
8708 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
8709 Use unreachable_path_p.
8711 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8713 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8714 Remove argument and call find_taken_edge.
8715 (back_threader::resolve_phi): Do not calculate taken edge before
8716 calling maybe_register_path.
8717 (back_threader::find_paths_to_names): Same.
8719 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
8721 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
8722 * config/h8300/testcompare.md (store_c): Update name, use new
8724 (store_neg_c, store_shifted_c): New patterns.
8726 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
8729 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
8730 instead of r12 for CR save, in all cases.
8732 2021-09-03 Andrew Pinski <apinski@marvell.com>
8734 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
8735 Handle error_mark_node as the type of the type_decl.
8737 2021-09-03 Andrew Pinski <apinski@marvell.com>
8739 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
8741 (aarch64_simd_types): Likewise.
8742 (aarch64_simd_intOI_type_node): Likewise.
8743 (aarch64_simd_intCI_type_node): Likewise.
8744 (aarch64_simd_intXI_type_node): Likewise.
8745 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
8746 (aarch64_fp16_ptr_type_node): Likewise.
8747 (aarch64_bf16_type_node): Likewise.
8748 (aarch64_bf16_ptr_type_node): Likewise.
8750 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8752 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
8754 (minus_op1_op2_relation_effect): ...here.
8755 (class operator_pointer_diff): New.
8756 (operator_pointer_diff::op1_op2_relation_effect): Call
8757 minus_op1_op2_relation_effect.
8758 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
8760 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8762 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
8763 Add may_peel_loop_headers.
8764 (back_threader_registry::thread_through_all_blocks): Same.
8765 (try_thread_blocks): Pass may_peel_loop_headers argument.
8766 (pass_early_thread_jumps::execute): Same.
8768 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8770 * tree-ssa-threadedge.c (has_phis_p): New.
8771 (forwarder_block_p): New.
8772 (potentially_threadable_block): Call forwarder_block_p.
8773 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
8774 (jump_threader::thread_through_normal_block): Call
8777 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8779 * tree-ssa-threadbackward.c (back_threader::dump): New.
8780 (back_threader::debug): New.
8781 (back_threader_profitability::profitable_path_p): Dump blocks
8782 even if we are bailing early.
8784 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8786 * tree-ssa-threadupdate.c (cancel_thread): New.
8787 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
8788 (jump_thread_path_registry::mark_threaded_blocks): Same.
8789 (jump_thread_path_registry::register_jump_thread): Same.
8791 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8793 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
8794 which objects are available.
8795 (jt_state::pop): Same.
8796 (jt_state::register_equiv): Same.
8797 (jt_state::register_equivs_on_edge): Same.
8799 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8801 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
8802 Move pop until after a thread is registered.
8804 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8806 * tree-ssa-threadupdate.c (debug): New.
8808 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8810 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
8811 (push_dump_file::~push_dump_file): New.
8812 (dump_ranger): Change dump_file temporarily while dumping
8814 * gimple-range-trace.h (class push_dump_file): New.
8816 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8818 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
8819 (dump_ranger): Dump function name.
8821 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8823 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8824 Adjust for non-null.
8825 (path_range_query::adjust_for_non_null_uses): New.
8826 (path_range_query::precompute_ranges): Call
8827 adjust_for_non_null_uses.
8828 * gimple-range-path.h: Add m_non_null and
8829 adjust_for_non_null_uses.
8831 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8833 * gimple-range-path.cc (path_range_query::dump): Dump path
8835 (path_range_query::precompute_ranges): Dump entire path.
8837 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8839 * value-relation.cc (relation_oracle::debug): New.
8840 * value-relation.h (relation_oracle::debug): New.
8842 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8844 * tree-ssa-loop-ch.c: Remove unnecessary include file.
8846 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8848 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
8849 Skip statements with no defining BB.
8850 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8851 Do not get confused by statements with no defining BB.
8853 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8855 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
8856 gimple_range_adjustment. Add support for constants.
8857 (adjust_realpart_expr): New.
8858 (gimple_range_adjustment): Move IMAGPART_EXPR code to
8859 adjust_imagpart_expr.
8860 * range-op.cc (integral_table::integral_table): Add entry for
8863 2021-09-03 Jakub Jelinek <jakub@redhat.com>
8865 * omp-expand.c (expand_omp_atomic_pipeline): Use
8866 IFN_ATOMIC_COMPARE_EXCHANGE instead of
8867 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
8870 2021-09-03 Jakub Jelinek <jakub@redhat.com>
8873 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
8874 that is false if DECL_BIT_FIELD.
8875 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
8876 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
8877 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
8878 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
8879 of DECL_FIELD_ABI_IGNORED.
8880 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
8881 SET_DECL_FIELD_ABI_IGNORED instead of writing to
8882 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
8883 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
8884 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
8885 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
8887 2021-09-03 liuhongt <hongtao.liu@intel.com>
8890 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
8891 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
8892 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
8894 2021-09-02 Martin Sebor <msebor@redhat.com>
8896 PR tree-optimization/17506
8898 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
8900 2021-09-02 Richard Biener <rguenther@suse.de>
8902 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
8903 fix for PR78185 and continue processing when leaving
8906 2021-09-02 Jakub Jelinek <jakub@redhat.com>
8908 PR tree-optimization/99591
8909 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8912 2021-09-02 Richard Biener <rguenther@suse.de>
8915 2021-09-02 Richard Biener <rguenther@suse.de>
8917 PR tree-optimization/102155
8918 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
8919 over a part of the RPO array and do not recurse here.
8920 Dump blocks marked as always executed.
8921 (fill_always_executed_in): Walk over the RPO array and
8922 process loops whose header we run into.
8923 (loop_invariant_motion_in_fun): Compute the first RPO
8924 using rev_post_order_and_mark_dfs_back_seme in iteration
8925 order and pass that to fill_always_executed_in.
8927 2021-09-02 liuhongt <hongtao.liu@intel.com>
8929 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
8930 * config/i386/i386.c (enum x86_64_reg_class): Add
8932 (merge_classes): Handle X86_64_SSEHF_CLASS.
8933 (examine_argument): Ditto.
8934 (construct_container): Ditto.
8935 (classify_argument): Ditto, and set HFmode/HCmode to
8937 (function_value_32): Return _FLoat16/Complex Float16 by
8939 (function_value_64): Return _Float16/Complex Float16 by SSE
8941 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
8942 (ix86_secondary_reload): Require gpr as intermediate register
8943 to store _Float16 from sse register when sse4 is not
8945 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
8947 (ix86_scalar_mode_supported_p): Ditto.
8948 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
8949 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
8950 (VALID_INT_MODE_P): Add HFmode and HCmode.
8951 * config/i386/i386.md (*pushhf_rex64): New define_insn.
8953 (*movhf_internal): Ditto.
8954 * doc/extend.texi (Half-Precision Floating Point): Documemt
8957 2021-09-02 Richard Biener <rguenther@suse.de>
8959 PR tree-optimization/102155
8960 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
8961 over a part of the RPO array and do not recurse here.
8962 Dump blocks marked as always executed.
8963 (fill_always_executed_in): Walk over the RPO array and
8964 process loops whose header we run into.
8965 (loop_invariant_motion_in_fun): Compute the first RPO
8966 using rev_post_order_and_mark_dfs_back_seme in iteration
8967 order and pass that to fill_always_executed_in.
8969 2021-09-02 YunQiang Su <syq@debian.org>
8972 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
8974 * config/mips/mips.c (mips_module_isa_name): New.
8975 mips_file_start: add .module mipsREV to all asm output
8977 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
8979 PR tree-optimization/102152
8980 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
8981 comparison to a scalar comparison before calling
8982 update_stmt_if_modified.
8984 2021-09-01 Andrew Pinski <apinski@marvell.com>
8987 * config/aarch64/aarch64.c (aarch64_expand_setmem):
8988 Check STRICT_ALIGNMENT before creating an overlapping
8991 2021-09-01 Martin Sebor <msebor@redhat.com>
8993 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
8994 (check_access): Pass additional argument.
8995 (check_memop_access): Remove template and make a member function.
8996 (maybe_check_dealloc_call): Make a pass_waccess member function.
8997 (class pass_waccess): Add, rename, and remove members.
8998 (pass_waccess::pass_waccess): Adjust to name change.
8999 (pass_waccess::~pass_waccess): Same.
9000 (check_alloca): Make a member function.
9001 (check_alloc_size_call): Same.
9002 (check_strcat): Same.
9003 (check_strncat): Same.
9004 (check_stxcpy): Same.
9005 (check_stxncpy): Same.
9006 (check_strncmp): Same.
9007 (maybe_warn_rdwr_sizes): Rename...
9008 (pass_waccess::maybe_check_access_sizes): ...to this.
9009 (pass_waccess::check_call): Adjust to name changes.
9010 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
9012 (pass_waccess::execute): Adjust to name changes.
9013 * gimple-ssa-warn-access.h (check_memop_access): Remove.
9014 * pointer-query.cc (access_ref::phi): Handle null pointer.
9015 (access_ref::inform_access): Same.
9016 (pointer_query::put_ref): Modify a cached value, not a copy of it.
9017 (pointer_query::dump): New function.
9018 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
9020 * pointer-query.h (pointer_query::dump): Declare.
9021 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
9022 (dump_strlen_info): Use function query.
9023 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
9025 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
9027 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
9029 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
9031 * doc/extend.texi: Document unavailable attribute.
9032 * print-tree.c (print_node): Handle unavailable attribute.
9033 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
9034 * tree.c (error_unavailable_use): New.
9035 * tree.h (TREE_UNAVAILABLE): New.
9036 (error_unavailable_use): New.
9038 2021-09-01 Jakub Jelinek <jakub@redhat.com>
9040 PR tree-optimization/102124
9041 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
9042 MINUS_EXPR, if itype is unsigned with smaller precision than type,
9043 add an extra cast to signed variant of itype to ensure sign-extension.
9045 2021-09-01 Martin Liska <mliska@suse.cz>
9047 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
9048 edges and rather use colors for TRUE and FALSE edges.
9050 2021-09-01 Richard Biener <rguenther@suse.de>
9052 PR tree-optimization/93491
9053 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
9054 after processing the stmt itself. Do not consider
9055 pure functions possibly not returning. Properly avoid
9056 adding possibly trapping calls to EXP_GEN when there's
9057 a preceeding possibly not returning call.
9058 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
9061 2021-09-01 Richard Biener <rguenther@suse.de>
9063 PR tree-optimization/102139
9064 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
9065 type to record a std::pair of the stmt-info and the innermost
9067 (dr_vec_info::group): New member.
9068 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
9069 (vect_compute_data_ref_alignment): Verify the recorded
9070 base alignment can be used.
9071 (data_ref_pair): Remove.
9072 (dr_group_sort_cmp): Adjust.
9073 (vect_analyze_data_ref_accesses): Store the group-ID in the
9074 dr_vec_info and operate on a vector of dr_vec_infos.
9076 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
9078 * read-md.c (md_reader::handle_enum): support value assignation.
9079 * doc/md.texi: record define_c_enum value assignation support.
9081 2021-09-01 Jakub Jelinek <jakub@redhat.com>
9083 PR tree-optimization/102141
9084 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
9085 argument. If false, emit stmts after gsi instead of before, and
9087 (bswap_replace): Adjust callers. When converting output of bswap,
9088 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
9091 2021-09-01 liuhongt <hongtao.liu@intel.com>
9093 * emit-rtl.c (validate_subreg): Get rid of all float-int
9096 2021-09-01 liuhongt <hongtao.liu@intel.com>
9099 2021-08-30 liuhongt <hongtao.liu@intel.com>
9101 * expmed.c (extract_bit_field_1): Make sure we're playing with
9102 integral modes before call extract_integral_bit_field.
9103 (extract_integral_bit_field): Add a parameter of type
9104 scalar_int_mode which corresponds to of tmode.
9105 And call extract_and_convert_fixed_bit_field instead of
9106 extract_fixed_bit_field and convert_extracted_bit_field.
9107 (extract_and_convert_fixed_bit_field): New function, it's a
9108 combination of extract_fixed_bit_field and
9109 convert_extracted_bit_field.
9111 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
9113 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
9115 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
9117 * omp-general.h (omp_is_reference): Rename to...
9118 (omp_privatize_by_reference): ... this. Adjust all users...
9119 * omp-general.c: ... here, ...
9120 * gimplify.c: ... here, ...
9121 * omp-expand.c: ... here, ...
9122 * omp-low.c: ... here.
9124 2021-08-31 Martin Sebor <msebor@redhat.com>
9126 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
9127 pointer element for equality to zero, not that of the cotaining
9130 2021-08-31 Martin Sebor <msebor@redhat.com>
9132 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
9134 2021-08-31 Martin Sebor <msebor@redhat.com>
9136 * function.h (function): Add comments.
9137 (get_range_query): Same. Add attribute returns nonnull.
9139 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9141 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
9142 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
9143 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
9145 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
9147 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
9149 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
9150 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
9151 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
9152 (m32c-*-elf*, m32c-*-rtems*): Likewise.
9153 * config/cris/cris.h (DBX_NO_XREFS): Remove.
9154 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
9155 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
9156 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
9157 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
9158 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
9160 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
9162 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
9163 allowed on target constructs and only with particular other clauses.
9164 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
9166 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
9167 structs when 'ancestor' is used.
9168 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
9170 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
9171 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
9173 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9175 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
9176 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9178 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
9179 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
9180 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
9181 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
9182 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
9183 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9184 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
9185 would be paradoxical.
9187 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9189 * combine.c (combine_simplify_rtx): Avoid converting an explicit
9190 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
9191 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9193 2021-08-31 Richard Biener <rguenther@suse.de>
9195 PR tree-optimization/102142
9196 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
9197 condition under which to unset the visited flag.
9199 2021-08-31 Richard Biener <rguenther@suse.de>
9201 PR middle-end/102129
9202 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
9203 possibly trapping expressions across calls.
9205 2021-08-31 Jakub Jelinek <jakub@redhat.com>
9207 PR tree-optimization/102134
9208 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
9209 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
9212 2021-08-31 Andrew Pinski <apinski@marvell.com>
9215 * collect-utils.c (setup_signals): New declaration.
9216 * collect-utils.h (setup_signals): New function.
9217 * collect2.c (handler): Delete.
9218 (main): Instead of manually setting up the signals,
9219 just call setup_signals.
9220 * lto-wrapper.c (main): Likewise.
9222 2021-08-31 Andrew Pinski <apinski@marvell.com>
9225 * config/i386/i386-protos.h (x86_output_aligned_bss):
9226 Change align argument to unsigned type.
9227 (x86_elf_aligned_decl_common): Likewise.
9228 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
9229 (x86_output_aligned_bss): Likewise.
9231 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9233 * config/mips/mips.c (mips_module_isa_name): New.
9234 mips_file_start: add .module mipsREV to all asm output
9236 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9238 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
9239 use enum instead of int for 'isa' member.
9240 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
9241 config/mips/netbsd.h: replace hardcoded numbers with enum.
9243 2021-08-31 liuhongt <hongtao.liu@intel.com>
9245 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
9246 define_split to define_insn_and_split.
9247 (*avx2_eq<mode>3): Removed.
9248 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
9249 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
9250 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
9252 (*avx2_gt<mode>3): Removed.
9253 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
9254 define_insn to define_expand, and adjust pattern.
9255 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
9257 2021-08-30 David Malcolm <dmalcolm@redhat.com>
9260 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
9262 2021-08-30 Jason Merrill <jason@redhat.com>
9264 * doc/invoke.texi: Document -Wmissing-requires.
9266 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9268 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
9269 TARGET_EXTRA_BUILTINS guard.
9271 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9273 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
9274 initialization of V2DI_type_node and unsigned_V2DI_type_node.
9276 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9278 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
9279 decl when new_builtins_are_live.
9280 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
9283 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
9285 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
9286 OPTION_MASK_P10_FUSION_2STORE.
9287 (POWERPC_MASKS): Likewise.
9288 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
9289 store fusion for Power10.
9290 (is_fusable_store): New.
9291 (power10_sched_reorder): Likewise.
9292 (rs6000_sched_reorder): Do Power10 specific reordering.
9293 (rs6000_sched_reorder2): Likewise.
9294 * config/rs6000/rs6000.opt: Add new option.
9296 2021-08-30 Richard Biener <rguenther@suse.de>
9298 PR tree-optimization/102128
9299 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9300 Move scanning for if-converted scalar code to the caller
9301 and instead delay clearing the visited flag for profitable
9303 (vect_slp_region): Cost all subgraphs before scheduling.
9304 For if-converted BB vectorization scan for scalar COND_EXPRs
9305 and do not vectorize if any found and the cost model is
9308 2021-08-30 Richard Biener <rguenther@suse.de>
9310 * common.opt (fexceptions): Mark
9311 EnabledBy(fnon-call-exceptions).
9312 * doc/invoke.texi (fnon-call-exceptions): Document this
9313 enables -fexceptions.
9315 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
9317 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
9318 is not already defined.
9320 2021-08-30 liuhongt <hongtao.liu@intel.com>
9322 * expmed.c (extract_bit_field_1): Make sure we're playing with
9323 integral modes before call extract_integral_bit_field.
9324 (extract_integral_bit_field): Add a parameter of type
9325 scalar_int_mode which corresponds to of tmode.
9326 And call extract_and_convert_fixed_bit_field instead of
9327 extract_fixed_bit_field and convert_extracted_bit_field.
9328 (extract_and_convert_fixed_bit_field): New function, it's a
9329 combination of extract_fixed_bit_field and
9330 convert_extracted_bit_field.
9332 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9334 * config/darwin.c (darwin_libc_has_function): Do not run
9335 the checks for x86 or modern Darwin. Make sure that there
9336 is a value set for darwin_macosx_version_min before testing.
9338 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9340 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
9342 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
9344 * ipa-modref-tree.h (modref_access_node::merge): Break out
9345 logic combining offsets and logic merging ranges to ...
9346 (modref_access_node::combined_offsets): ... here
9347 (modref_access_node::update2): ... here
9348 (modref_access_node::closer_pair_p): New member function.
9349 (modref_access_node::forced_merge): New member function.
9350 (modre_ref_node::insert): Do merging when table is full.
9352 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
9355 * config.gcc: MIPS: use N64 ABI by default if the triple end
9356 with -gnuabi64, which is used by Debian since 2013.
9358 2021-08-28 Alexandre Oliva <oliva@adacore.com>
9360 * ipa-modref.c (analyze_function): Skip debug stmts.
9361 * tree-inline.c (estimate_num_insn): Consider builtins even
9362 without a cgraph_node.
9364 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9366 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
9367 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
9368 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
9369 * config/h8300/h8300.md (eqne): New code iterator.
9370 (geultu, geultu_to_c): Similarly.
9371 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
9372 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
9373 (cmp<mode>_c): New pattern
9375 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9377 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
9379 (dom_opt_dom_walker::optimize_stmt): Use it.
9381 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9383 * config/darwin.c (finalize_ctors): Add a section-start linker-
9385 (finalize_dtors): Likewise.
9386 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
9388 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
9390 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
9391 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
9392 old initialization logic when new builtins are enabled.
9393 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
9394 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
9395 (write_init_file): Likewise.
9397 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9399 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
9400 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
9401 (amdgcn-* | gcn-*) Likewise.
9403 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
9405 * symbol-summary.h: Added missing template keyword.
9407 2021-08-27 Richard Biener <rguenther@suse.de>
9409 PR tree-optimization/45178
9410 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
9411 infinite loops without exit do not mark control dependent
9412 edges of the latch necessary.
9414 2021-08-27 konglin1 <lingling.kong@intel.com>
9417 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
9419 (<avx512>scattersi<mode>): Likewise.
9420 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
9421 (*avx512f_scatterdi<VI48F:mode>): Likewise
9423 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9425 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
9426 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
9427 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
9428 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
9429 P10_BUILTIN_PEXTD on Power10.
9431 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9433 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
9434 signedness for some Power10 bifs.
9436 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
9439 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
9440 computed alignment if the entire struct has attribute packed.
9442 2021-08-27 liuhongt <hongtao.liu@intel.com>
9446 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
9447 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
9448 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
9449 IX86_BUILTIN_SHUFPS256.
9450 (ix86_masked_all_ones): New function.
9452 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9454 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
9455 (conditional moves with memory inputs splitters): Ditto.
9456 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
9458 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9460 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
9461 search after merging.
9463 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9465 * config/rs6000/rs6000-overload.def: Add remaining overloads.
9467 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9469 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
9471 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9473 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
9474 crypto, and htm stanzas.
9476 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9478 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
9480 2021-08-26 Martin Sebor <msebor@redhat.com>
9482 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
9483 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
9484 (warn_uninitialized_vars): Same. Reduce visibility of locals.
9485 (warn_uninitialized_phi): Same.
9487 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9489 * tree-ssa-ccp.c (get_individual_bits): Helper function to
9490 extract the individual bits from a widest_int constant (mask).
9491 (gray_code_bit_flips): New read-only table for effiently
9492 enumerating permutations/combinations of bits.
9493 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
9494 by unknown counts that are guaranteed less than the target
9495 precision and four or fewer unknown bits by enumeration.
9496 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
9497 enumeration under the same conditions. Handle remaining
9498 shifts as a mask based upon the minimum possible shift value.
9500 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9501 Richard Biener <rguenther@suse.de>
9503 * match.pd (shift transformations): Remove a redundant
9504 !POINTER_TYPE_P check.
9506 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9509 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
9510 Set all_regs to true in the call to replace_rtx.
9512 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9514 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
9515 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
9516 hiting --param modref-max-refs limit.
9517 (modref_tree:insert_base): Be smarter when hitting
9518 --param modref-max-bases limit. Add new parameter REF.
9519 (modref_tree:insert): Update.
9520 (modref_tree:merge): Update.
9521 * ipa-modref.c (read_modref_records): Update.
9523 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9525 * params.opt: (modref-max-adjustments): Add full stop.
9527 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9529 * ipa-modref-tree.h (modref_ref_node::verify): New member
9531 (modref_ref_node::insert): Use it.
9532 (modref_ref_node::try_mere_with): Fix off by one error.
9534 2021-08-26 Martin Liska <mliska@suse.cz>
9535 Stefan Kneifel <stefan.kneifel@bluewin.ch>
9537 * cgraph.h (create_version_clone_with_body): Add new parameter.
9538 * cgraphclones.c: Likewise.
9539 * multiple_target.c (create_dispatcher_calls): Do not use
9541 (create_target_clone): Likewise here.
9543 2021-08-26 Jonathan Yong <10walls@gmail.com>
9545 * doc/extend.texi: Add note about reserved priorities
9546 to the constructor attribute.
9548 2021-08-25 Martin Sebor <msebor@redhat.com>
9550 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
9551 range table header alone.
9552 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
9554 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
9556 * doc/invoke.texi: Document --param modref-max-adjustments.
9557 * ipa-modref-tree.c (test_insert_search_collapse): Update.
9558 (test_merge): Update.
9559 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
9560 (modref_access_node::operator==): Fix handling of access ranges.
9561 (modref_access_node::contains): Constify parameter; handle also
9562 mismatched parm offsets.
9563 (modref_access_node::update): New function.
9564 (modref_access_node::merge): New function.
9565 (unspecified_modref_access_node): Update constructor.
9566 (modref_ref_node::insert_access): Add record_adjustments parameter;
9568 (modref_ref_node::try_merge_with): New private function.
9569 (modref_tree::insert): New record_adjustments parameter.
9570 (modref_tree::merge): New record_adjustments parameter.
9571 (modref_tree::copy_from): Update.
9572 * ipa-modref.c (dump_access): Dump adjustments field.
9573 (get_access): Update constructor.
9574 (record_access): Update call of insert.
9575 (record_access_lto): Update call of insert.
9576 (merge_call_side_effects): Add record_adjustments parameter.
9577 (get_access_for_fnspec): Update.
9578 (process_fnspec): Update.
9579 (analyze_call): Update.
9580 (analyze_function): Update.
9581 (read_modref_records): Update.
9582 (ipa_merge_modref_summary_after_inlining): Update.
9583 (propagate_unknown_call): Update.
9584 (modref_propagate_in_scc): Update.
9585 * params.opt (param-max-modref-adjustments=): New.
9587 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
9589 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
9591 (xxspltiw_v4si): Use vecperm type attribute.
9592 (xxspltiw_v4si_inst): Use vecperm type attribute.
9593 (xxspltiw_v4sf_inst): Likewise.
9594 (xxspltidp_v2df): Use vecperm type attribute. Use
9595 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
9596 (xxspltidp_v2df_inst): Likewise.
9597 (xxsplti32dx_v4si): Use vecperm type attribute.
9598 (xxsplti32dx_v4si_inst): Likewise.
9599 (xxsplti32dx_v4sf_inst): Likewise.
9600 (xxblend_<mode>): Likewise.
9601 (xxpermx): Likewise.
9602 (xxpermx_inst): Likewise.
9605 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
9608 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
9609 * diagnostic.c (diagnostic_initialize_input_context): New function.
9610 * diagnostic.h (diagnostic_initialize_input_context): Declare.
9611 * input.c (default_charset_callback): New function.
9612 (file_cache::initialize_input_context): New function.
9613 (file_cache_slot::create): Added ability to convert the input
9614 according to the input context.
9615 (file_cache::file_cache): Initialize the new input context.
9616 (class file_cache_slot): Added new m_alloc_offset member.
9617 (file_cache_slot::file_cache_slot): Initialize the new member.
9618 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
9619 (file_cache_slot::maybe_grow): Likewise.
9620 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
9621 (file_cache_slot::get_next_line): Likewise.
9622 * input.h (class file_cache): Added input context member.
9624 2021-08-25 Richard Biener <rguenther@suse.de>
9626 PR tree-optimization/102046
9627 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
9628 update ->any_pattern when swapping operands.
9630 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
9633 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
9634 (ix86_decompose_address): Remove retval check for ASHIFT,
9635 allow non-canonical zero extend if AND mask covers ASHIFT
9637 (ix86_legitimate_address_p): Adjust condition for decompose.
9638 (ix86_rtx_costs): Adjust cost for lea with non-canonical
9640 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
9642 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
9644 PR tree-optimization/101145
9645 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
9647 (number_of_iterations_lt): Invoke above function.
9648 (adjust_cond_for_loop_until_wrap):
9649 Merge to number_of_iterations_until_wrap.
9650 (number_of_iterations_cond): Update invokes for
9651 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
9653 2021-08-25 konglin1 <lingling.kong@intel.com>
9656 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
9658 (_mm512_mask_fpclass_ps_mask): Ditto.
9660 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
9662 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
9663 (vec_unpacku_hi_v8hi): Likewise.
9664 (vec_unpacku_lo_v16qi): Likewise.
9665 (vec_unpacku_lo_v8hi): Likewise.
9666 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
9667 (vec_unpacku_lo_<VP_small_lc>): Likewise.
9669 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
9671 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
9672 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9673 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9674 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
9676 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9678 PR middle-end/102031
9679 * simplify-rtx.c (simplify_truncation): When comparing precisions
9680 use "subreg_prec" variable, not "subreg_mode".
9682 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9684 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
9687 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9689 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
9690 various pointer type nodes.
9691 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
9692 values for various pointer types.
9693 (ptr_V16QI_type_node): New macro.
9694 (ptr_V1TI_type_node): New macro.
9695 (ptr_V2DI_type_node): New macro.
9696 (ptr_V2DF_type_node): New macro.
9697 (ptr_V4SI_type_node): New macro.
9698 (ptr_V4SF_type_node): New macro.
9699 (ptr_V8HI_type_node): New macro.
9700 (ptr_unsigned_V16QI_type_node): New macro.
9701 (ptr_unsigned_V1TI_type_node): New macro.
9702 (ptr_unsigned_V8HI_type_node): New macro.
9703 (ptr_unsigned_V4SI_type_node): New macro.
9704 (ptr_unsigned_V2DI_type_node): New macro.
9705 (ptr_bool_V16QI_type_node): New macro.
9706 (ptr_bool_V8HI_type_node): New macro.
9707 (ptr_bool_V4SI_type_node): New macro.
9708 (ptr_bool_V2DI_type_node): New macro.
9709 (ptr_bool_V1TI_type_node): New macro.
9710 (ptr_pixel_type_node): New macro.
9711 (ptr_intQI_type_node): New macro.
9712 (ptr_uintQI_type_node): New macro.
9713 (ptr_intHI_type_node): New macro.
9714 (ptr_uintHI_type_node): New macro.
9715 (ptr_intSI_type_node): New macro.
9716 (ptr_uintSI_type_node): New macro.
9717 (ptr_intDI_type_node): New macro.
9718 (ptr_uintDI_type_node): New macro.
9719 (ptr_intTI_type_node): New macro.
9720 (ptr_uintTI_type_node): New macro.
9721 (ptr_long_integer_type_node): New macro.
9722 (ptr_long_unsigned_type_node): New macro.
9723 (ptr_float_type_node): New macro.
9724 (ptr_double_type_node): New macro.
9725 (ptr_long_double_type_node): New macro.
9726 (ptr_dfloat64_type_node): New macro.
9727 (ptr_dfloat128_type_node): New macro.
9728 (ptr_ieee128_type_node): New macro.
9729 (ptr_ibm128_type_node): New macro.
9730 (ptr_vector_pair_type_node): New macro.
9731 (ptr_vector_quad_type_node): New macro.
9732 (ptr_long_long_integer_type_node): New macro.
9733 (ptr_long_long_unsigned_type_node): New macro.
9735 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9737 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
9738 and power9-64 stanzas.
9740 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9741 Tom de Vries <tdevries@suse.de>
9743 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
9744 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
9745 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
9746 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
9747 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
9748 * config/nvptx/nvptx-c.c: New source file.
9749 (nvptx_cpu_cpp_builtins): Move implementation here.
9751 2021-08-24 Martin Sebor <msebor@redhat.com>
9753 PR middle-end/101600
9754 PR middle-end/101977
9755 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
9756 the phrasing of a warning.
9757 (check_access): Use the remaining size after subtracting any offset
9758 rather than the whole object size.
9759 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
9760 clear for any nonnull PHI argument.
9761 (compute_objsize): Clear argument.
9763 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9765 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
9767 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9769 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
9772 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
9774 * value-relation.cc (rr_transitive_table): New.
9775 (relation_transitive): New.
9776 (value_relation::swap): Remove.
9777 (value_relation::apply_transitive): New.
9778 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
9779 (relation_oracle::register_relation): Call register_transitives.
9780 (relation_oracle::register_transitives): New.
9781 * value-relation.h (relation_oracle): Add new temporary bitmap and
9784 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
9787 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
9788 from integer to a pseudo vector register.
9790 2021-08-24 Richard Biener <rguenther@suse.de>
9792 PR tree-optimization/100089
9793 * tree-vectorizer.h (vect_slp_bb): Rename to ...
9794 (vect_slp_if_converted_bb): ... this and get the original
9795 loop as new argument.
9796 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
9797 pass original loop to vect_slp_if_converted_bb.
9798 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9799 If orig_loop was passed scan the not vectorized stmts
9800 for COND_EXPRs and force not profitable if found.
9801 (vect_slp_region): Pass down all SLP instances to costing
9802 if orig_loop was specified.
9803 (vect_slp_bbs): Pass through orig_loop.
9804 (vect_slp_bb): Rename to ...
9805 (vect_slp_if_converted_bb): ... this and get the original
9806 loop as new argument.
9807 (vect_slp_function): Adjust.
9809 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9812 * config/arm/arm.md (attribute arch): Add fix_vlldm.
9813 (arch_enabled): Use it.
9814 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
9815 use when erratum mitigation is needed.
9817 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9820 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
9821 * doc/invoke.texi (Arm Options): Document it.
9822 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
9823 (ALL_QUIRKS): Add quirk_vlldm.
9824 (cortex-m33): Add quirk_vlldm.
9825 (cortex-m35p, cortex-m55): Likewise.
9826 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
9827 targetting an affected CPU and not explicitly controlled on
9830 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9832 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
9833 (lazy_load_multiple_insn): Likewise.
9835 2021-08-24 liuhongt <hongtao.liu@intel.com>
9838 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
9839 Enable avx512 embedded broadcast.
9840 (*<avx512>_vternlog<mode>_all): Ditto.
9841 (<avx512>_vternlog<mode>_mask): Ditto.
9843 2021-08-24 liuhongt <hongtao.liu@intel.com>
9846 * config/i386/i386.c (ix86_rtx_costs): Define cost for
9848 * config/i386/i386.h (STRIP_UNARY): New macro.
9849 * config/i386/predicates.md (reg_or_notreg_operand): New
9851 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
9852 (*<avx512>_vternlog<mode>_1): New pre_reload
9853 define_insn_and_split.
9854 (*<avx512>_vternlog<mode>_2): Ditto.
9855 (*<avx512>_vternlog<mode>_3): Ditto.
9856 (any_logic1,any_logic2): New code iterator.
9857 (logic_op): New code attribute.
9858 (ternlogsuffix): Extend to VNxDF and VNxSF.
9860 2021-08-24 Richard Biener <rguenther@suse.de>
9862 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
9863 * params.opt (--param vect-inner-loop-cost-factor): Adjust
9865 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
9866 inner_loop_cost_factor to the minimum of the estimated number
9867 of iterations of the inner loop and vect-inner-loop-cost-factor.
9869 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9870 Richard Biener <rguenther@suse.de>
9872 * config/i386/i386-features.c (compute_convert_gain): Provide
9873 more accurate values for CONST_INT, when optimizing for size.
9874 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
9875 * config/i386/i386.h (COSTS_N_BYTES): to here.
9877 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9878 Jakub Jelinek <jakub@redhat.com>
9880 PR middle-end/102029
9881 * match.pd (shift transformations): Add an additional check for
9882 !POINTER_TYPE_P in the recently added left shift transformation.
9884 2021-08-24 liuhongt <hongtao.liu@intel.com>
9886 PR tree-optimization/100089
9887 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
9888 loop vectorizer when cost model is very-cheap.
9890 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9892 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
9893 asprintf, which is not available on AIX.
9895 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9897 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
9898 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
9900 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9902 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
9903 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
9905 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
9906 (rbtree.o): Likewise.
9907 (rs6000-gen-builtins): Likewise.
9908 (rs6000-builtins.c): Likewise.
9909 (rs6000-builtins.h): Likewise.
9910 (rs6000.o): Add dependency.
9911 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
9912 (rs6000-vecdefines.h): New target.
9913 (rs6000-builtins.o): Likewise.
9914 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
9915 (rs6000-c.o): Likewise.
9917 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9920 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
9921 Diagnose buffer overrun.
9922 (safe_inc_pos): Fix overrun detection.
9923 (match_identifier): Diagnose buffer overrun.
9924 (match_integer): Likewise.
9925 (match_to_right_bracket): Likewise.
9927 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9929 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
9930 Improve range compare.
9931 (modref_access_node::contains): New member function.
9932 (modref_access_node::search): Remove.
9933 (modref_access_node::insert): Be smarter about subaccesses.
9935 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
9937 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
9938 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
9939 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
9942 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
9944 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
9946 * config/h8300/jumpcall.md (call, call_value): Restrict to
9947 !SIBLING_CALL_P cases.
9948 (subcall, sibcall_value): New patterns & expanders.
9949 * config/h8300/proepi.md (epilogue): Pass new argument to
9950 h8300_expand_epilogue.
9951 (sibcall_epilogue): New expander.
9952 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
9954 (h8300_ok_for_sibcall_p): New function.
9955 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
9957 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9959 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
9960 Handle case where the operand is already the desired mode.
9962 2021-08-23 Richard Biener <rguenther@suse.de>
9965 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
9966 in addition to has_gimple_body.
9968 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9970 PR middle-end/101949
9971 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
9974 2021-08-23 Martin Liska <mliska@suse.cz>
9976 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
9979 2021-08-23 Richard Biener <rguenther@suse.de>
9981 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9982 Properly scale the inner loop cost only once.
9984 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9986 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
9987 Provide bounds for unsigned (and signed with non-negative operands)
9988 division and modulus.
9990 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9992 * simplify-rtx.c (simplify_truncation): Generalize simplification
9993 of (truncate:A (subreg:B X)).
9994 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
9995 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
9996 already has the desired machine mode.
9997 (test_scalar_int_ops): Add tests that useless extensions and
9998 truncations are optimized away.
9999 (test_scalar_int_ext_ops): New self-test function to confirm
10000 that truncations of extensions are correctly simplified.
10001 (test_scalar_int_ext_ops2): New self-test function to check
10002 truncations of truncations, extensions of extensions, and
10003 truncations of extensions.
10004 (test_scalar_ops): Call the above two functions with a
10005 representative sampling of integer machine modes.
10007 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
10009 * match.pd (shift transformations): Change the sign of an
10010 LSHIFT_EXPR if it reduces the number of explicit conversions.
10012 2021-08-23 Jakub Jelinek <jakub@redhat.com>
10014 PR tree-optimization/86723
10015 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
10016 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
10017 non-memory permutation of 64-bit src which only has bytes of
10018 0 or [5..8] and n->range is 4.
10019 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
10020 find_bswap_or_nop_finalize caller, support bswap with some bytes
10021 zeroed, as long as at least two bytes are not zeroed.
10022 (bswap_replace): Add mask argument and handle masking of bswap
10024 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
10025 caller, punt if cast64_to_32 or mask is not all ones.
10026 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
10027 caller, for now punt if cast64_to_32.
10029 2021-08-23 Richard Biener <rguenther@suse.de>
10031 PR tree-optimization/79334
10032 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
10033 a type also for COMPONENT_REFs.
10034 (vn_reference_may_trap): Check ARRAY_REF with constant index
10035 against the array domain.
10037 2021-08-23 liuhongt <hongtao.liu@intel.com>
10040 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
10041 TARGET_AVX512BW to condition.
10043 2021-08-23 Jakub Jelinek <jakub@redhat.com>
10046 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
10047 register variables already during early_dwarf if possible.
10049 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
10051 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
10053 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
10056 * config/arm/arm.opt: Fix typo.
10057 * config/arm/t-rmprofile: Fix typo.
10059 2021-08-23 Jakub Jelinek <jakub@redhat.com>
10061 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
10062 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
10063 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
10064 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
10065 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
10066 if either grainsize or num_tasks clause has the strict modifier.
10068 2021-08-23 Martin Liska <mliska@suse.cz>
10070 * dbgcnt.def (DEBUG_COUNTER): New counter.
10071 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
10073 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
10075 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
10077 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
10080 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
10082 * config/mips/mips.c (mips_msa_output_shift_immediate): New
10084 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
10085 vlshr<mode>3): Call it.
10087 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
10088 Martin Liska <mliska@suse.cz>
10090 PR middle-end/101949
10091 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
10094 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
10096 * config/mips/mips.c (mips_function_rodata_section,
10097 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
10099 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
10101 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
10103 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10105 * configure.ac (thread-local storage support): Remove tls_first_major
10106 and tls_first_minor. Use "$conftest_s" to check support.
10107 * configure: Regenerate.
10109 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10111 * configure.ac: Fixup formatting.
10113 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10115 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
10116 * configure.ac: ... update all callers.
10118 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10121 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
10122 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
10123 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
10124 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
10125 Remove remaining checks for in-tree assembler.
10126 * configure: Regenerate.
10128 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
10130 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
10131 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
10133 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
10134 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
10135 Improve logical shifts by 27 bits for H8/S.
10136 (get_shift_alg): Corresponding changes.
10137 (h8300_option_override): Revert to loops for -Os when profitable.
10139 2021-08-20 Richard Biener <rguenther@suse.de>
10141 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
10143 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
10144 index as group_id when dataref_groups were not computed.
10145 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
10146 we advace to the next BB.
10148 2021-08-20 Jakub Jelinek <jakub@redhat.com>
10150 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
10153 2021-08-20 Martin Liska <mliska@suse.cz>
10155 PR gcov-profile/89961
10156 * gcov.c (make_gcov_file_name): Rewrite using std::string.
10157 (mangle_name): Simplify, do not used the second argument.
10158 (strip_extention): New function.
10159 (get_md5sum): Likewise.
10160 (get_gcov_intermediate_filename): Handle properly -p and -x
10162 (output_gcov_file): Use string type.
10163 (generate_results): Likewise.
10164 (md5sum_to_hex): Remove.
10166 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
10168 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
10169 (UNSPEC_XXSPLTIW): Move to vsx.md.
10170 (UNSPEC_XXSPLTID): Move to vsx.md.
10171 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
10172 (UNSPEC_XXBLEND): Move to vsx.md.
10173 (UNSPEC_XXPERMX): Move to vsx.md.
10174 (VM3): Move to vsx.md.
10175 (VM3_char): Move to vsx.md.
10176 (xxspltiw_v4si): Move to vsx.md.
10177 (xxspltiw_v4sf): Move to vsx.md.
10178 (xxspltiw_v4sf_inst): Move to vsx.md.
10179 (xxspltidp_v2df): Move to vsx.md.
10180 (xxspltidp_v2df_inst): Move to vsx.md.
10181 (xxsplti32dx_v4si_inst): Move to vsx.md.
10182 (xxsplti32dx_v4sf): Move to vsx.md.
10183 (xxsplti32dx_v4sf_inst): Move to vsx.md.
10184 (xxblend_<mode>): Move to vsx.md.
10185 (xxpermx): Move to vsx.md.
10186 (xxpermx_inst): Move to vsx.md.
10187 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
10188 (UNSPEC_XXSPLTIW): Move from altivec.md.
10189 (UNSPEC_XXSPLTID): Move from altivec.md.
10190 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
10191 (UNSPEC_XXBLEND): Move from altivec.md.
10192 (UNSPEC_XXPERMX): Move from altivec.md.
10193 (VM3): Move from altivec.md.
10194 (VM3_char): Move from altivec.md.
10195 (xxspltiw_v4si): Move from altivec.md.
10196 (xxspltiw_v4sf): Move from altivec.md.
10197 (xxspltiw_v4sf_inst): Move from altivec.md.
10198 (xxspltidp_v2df): Move from altivec.md.
10199 (xxspltidp_v2df_inst): Move from altivec.md.
10200 (xxsplti32dx_v4si_inst): Move from altivec.md.
10201 (xxsplti32dx_v4sf): Move from altivec.md.
10202 (xxsplti32dx_v4sf_inst): Move from altivec.md.
10203 (xxblend_<mode>): Move from altivec.md.
10204 (xxpermx): Move from altivec.md.
10205 (xxpermx_inst): Move from altivec.md.
10207 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
10209 * tree-vect-generic.c (expand_vector_operations_1): Use either
10210 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
10211 when constructing scalar splat expressions.
10213 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
10216 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
10217 pointer to __vector_pair *.
10219 2021-08-19 Martin Sebor <msebor@redhat.com>
10221 * gimple-range.cc: Add comments.
10222 * gimple-range.h: Same.
10224 2021-08-19 Martin Sebor <msebor@redhat.com>
10226 PR middle-end/101984
10227 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
10230 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
10232 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
10233 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
10234 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
10236 2021-08-19 Jakub Jelinek <jakub@redhat.com>
10238 PR middle-end/101950
10239 * optabs.c (expand_clrsb_using_clz): New function.
10240 (expand_unop): Use it as another clrsb expansion fallback.
10242 2021-08-19 liuhongt <hongtao.liu@intel.com>
10245 2021-07-28 liuhongt <hongtao.liu@intel.com>
10248 * config/i386/i386.h (processor_costs): Add new member
10250 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
10251 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
10252 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
10253 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
10254 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
10255 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
10256 generic_cost, core_cost): Initialize integer_to_sse same value
10258 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
10259 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10260 Use integer_to_sse instead of sse_op to calculate the cost of
10263 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
10265 * config.gcc: Include rpath.opt for Darwin.
10266 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
10268 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10270 PR bootstrap/101959
10271 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
10274 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
10276 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
10277 pattern and type cast.
10278 (vld4_lane_f32): Use float RTL pattern.
10279 (vld4q_lane_f64): Use float type cast.
10281 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
10283 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
10286 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10288 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
10289 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
10290 (hash_map_tests_c_tests): Call it.
10292 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10294 * ggc.h (enum ggc_collect): New.
10295 (ggc_collect): Use it.
10296 * ggc-page.c: Adjust.
10297 * ggc-common.c: Likewise.
10298 * ggc-tests.c: Likewise.
10299 * read-rtl-function.c: Likewise.
10300 * selftest-run-tests.c: Likewise.
10301 * doc/gty.texi (Invoking the garbage collector): Likewise.
10303 2021-08-18 liuhongt <hongtao.liu@intel.com>
10306 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
10308 * config/i386/sse.md (*sse3_haddv2df3_low): Add
10309 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
10310 (*sse3_hsubv2df3_low): Ditto.
10311 * config/i386/x86-tune.def
10312 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
10314 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10316 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
10317 (gori_compute::compute_operand_range): Add tracing.
10318 (gori_compute::logical_combine): Ditto.
10319 (gori_compute::compute_logical_operands): Ditto.
10320 (gori_compute::compute_operand1_range): Ditto.
10321 (gori_compute::compute_operand2_range): Ditto.
10322 (gori_compute::outgoing_edge_range_p): Ditto.
10323 * gimple-range-gori.h (class gori_compute): Add range_tracer.
10325 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10327 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
10328 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
10329 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
10330 * params.opt (--param=evrp-mode): Adjust options.
10332 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10334 * Makefile.in (OBJS): Add gimple-range-trace.o.
10335 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
10336 * gimple-range-fold.cc: Adjust headers.
10337 * gimple-range-trace.cc: New.
10338 * gimple-range-trace.h: New.
10339 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
10340 (gimple_ranger::range_of_expr): Add tracing.
10341 (gimple_ranger::range_on_entry): Ditto.
10342 (gimple_ranger::range_on_exit): Ditto.
10343 (gimple_ranger::range_on_edge): Ditto.
10344 (gimple_ranger::fold_range_internal): Ditto.
10345 (gimple_ranger::dump_bb): Do not calculate edge range twice.
10346 (trace_ranger::*): Remove.
10347 (enable_ranger): Never create a trace_ranger.
10348 (debug_seed_ranger): Move to gimple-range-trace.cc.
10349 (dump_ranger): Ditto.
10350 (debug_ranger): Ditto.
10351 * gimple-range.h: Include gimple-range-trace.h.
10352 (range_on_entry, range_on_exit): No longer virtual.
10353 (class trace_ranger): Remove.
10354 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
10356 2021-08-17 Martin Sebor <msebor@redhat.com>
10358 PR middle-end/101854
10359 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
10360 in gimple-ssa-warn-access.cc.
10361 * calls.c (alloc_max_size): Move code to check_alloca.
10362 (get_size_range): Move to pointer-query.cc.
10363 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
10364 (get_attr_nonstring_decl): Move to tree.c.
10365 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
10366 (append_attrname): Same.
10367 (maybe_warn_rdwr_sizes): Same.
10368 (initialize_argument_information): Move code to
10369 gimple-ssa-warn-access.cc.
10370 * calls.h (maybe_warn_alloc_args_overflow): Move to
10371 gimple-ssa-warn-access.h.
10372 (get_attr_nonstring_decl): Move to tree.h.
10373 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
10374 (enum size_range_flags): Move to pointer-query.h.
10375 (get_size_range): Same.
10376 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
10377 to avoid Clang -Wunused-function.
10378 (get_size_range): Declare static.
10379 (maybe_emit_free_warning): Rename...
10380 (maybe_check_dealloc_call): ...to this for consistency.
10381 (class pass_waccess): Add members.
10382 (pass_waccess::~pass_waccess): Defined.
10383 (alloc_max_size): Move here from calls.c.
10384 (maybe_warn_alloc_args_overflow): Same.
10385 (check_alloca): New function.
10386 (check_alloc_size_call): New function.
10387 (check_strncat): Handle another warning flag.
10388 (pass_waccess::check_builtin): Handle alloca.
10389 (fntype_argno_type): Move here from calls.c.
10390 (append_attrname): Same.
10391 (maybe_warn_rdwr_sizes): Same.
10392 (pass_waccess::check_call): Define.
10393 (check_nonstring_args): New function.
10394 (pass_waccess::check): Call new member functions.
10395 (pass_waccess::execute): Enable ranger.
10396 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
10397 (maybe_warn_nonstring_arg): Same.
10398 * gimple-ssa-warn-restrict.c: Remove #include.
10399 * pointer-query.cc (get_size_range): Move here from calls.c.
10400 * pointer-query.h (enum size_range_flags): Same.
10401 (get_size_range): Same.
10402 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
10403 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
10405 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
10407 * ggc.h (ggc_collect): Add 'force_collect' parameter.
10408 * ggc-page.c (ggc_collect): Use that one instead of global
10409 'ggc_force_collect'. Adjust all users.
10410 * doc/gty.texi (Invoking the garbage collector): Update.
10411 * ggc-internal.h (ggc_force_collect): Remove.
10412 * ggc-common.c (ggc_force_collect): Likewise.
10413 * selftest.h (forcibly_ggc_collect): Remove.
10414 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
10415 * read-rtl-function.c (test_loading_labels): Adjust.
10416 * selftest-run-tests.c (run_tests): Likewise.
10418 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10420 * config/darwin.c (darwin_file_end): Reset and reclaim the
10421 section names table at the end of compile.
10423 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10426 * config.in: Regenerate.
10427 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
10428 (ASM_SPEC): Pass options to disable branch shortening where
10430 * configure: Regenerate.
10431 * configure.ac: Detect versions of 'as' that support the
10432 optimisation which has the bug.
10434 2021-08-17 Richard Biener <rguenther@suse.de>
10436 * optabs-query.c (supports_vec_gather_load_p): Also check
10438 (supports_vec_scatter_store_p): Likewise.
10439 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
10440 back to masked variants if non-masked are not supported.
10441 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
10442 When we need to use masked gather/scatter but do not have
10443 a mask set up a constant true one.
10444 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
10445 non-SSA_NAME masks.
10447 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10449 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
10450 algorithm as PLUS_EXPR to improve subtraction bit bounds.
10451 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
10453 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10455 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
10456 calculate the mask-value pair result of a multiplication by an
10458 (bit_value_binop) [MULT_EXPR]: Call it from here for
10459 multiplications by (sparse) non-negative constants.
10461 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
10464 * config.gcc (gcc_cv_initfini_array): Leave undefined for
10465 uclinuxfdpiceabi targets.
10467 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10469 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
10470 reverse debug stmts.
10472 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10474 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
10476 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10478 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
10479 (__LD4Q_LANE_FUNC): Likewise.
10480 (vld4_lane_u8): Define without macro.
10481 (vld4_lane_u16): Likewise.
10482 (vld4_lane_u32): Likewise.
10483 (vld4_lane_u64): Likewise.
10484 (vld4_lane_s8): Likewise.
10485 (vld4_lane_s16): Likewise.
10486 (vld4_lane_s32): Likewise.
10487 (vld4_lane_s64): Likewise.
10488 (vld4_lane_f16): Likewise.
10489 (vld4_lane_f32): Likewise.
10490 (vld4_lane_f64): Likewise.
10491 (vld4_lane_p8): Likewise.
10492 (vld4_lane_p16): Likewise.
10493 (vld4_lane_p64): Likewise.
10494 (vld4q_lane_u8): Likewise.
10495 (vld4q_lane_u16): Likewise.
10496 (vld4q_lane_u32): Likewise.
10497 (vld4q_lane_u64): Likewise.
10498 (vld4q_lane_s8): Likewise.
10499 (vld4q_lane_s16): Likewise.
10500 (vld4q_lane_s32): Likewise.
10501 (vld4q_lane_s64): Likewise.
10502 (vld4q_lane_f16): Likewise.
10503 (vld4q_lane_f32): Likewise.
10504 (vld4q_lane_f64): Likewise.
10505 (vld4q_lane_p8): Likewise.
10506 (vld4q_lane_p16): Likewise.
10507 (vld4q_lane_p64): Likewise.
10508 (vld4_lane_bf16): Likewise.
10509 (vld4q_lane_bf16): Likewise.
10511 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10513 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
10514 (__LD3Q_LANE_FUNC): Delete.
10515 (vld3_lane_u8): Define without macro.
10516 (vld3_lane_u16): Likewise.
10517 (vld3_lane_u32): Likewise.
10518 (vld3_lane_u64): Likewise.
10519 (vld3_lane_s8): Likewise.
10520 (vld3_lane_s16): Likewise.
10521 (vld3_lane_s32): Likewise.
10522 (vld3_lane_s64): Likewise.
10523 (vld3_lane_f16): Likewise.
10524 (vld3_lane_f32): Likewise.
10525 (vld3_lane_f64): Likewise.
10526 (vld3_lane_p8): Likewise.
10527 (vld3_lane_p16): Likewise.
10528 (vld3_lane_p64): Likewise.
10529 (vld3q_lane_u8): Likewise.
10530 (vld3q_lane_u16): Likewise.
10531 (vld3q_lane_u32): Likewise.
10532 (vld3q_lane_u64): Likewise.
10533 (vld3q_lane_s8): Likewise.
10534 (vld3q_lane_s16): Likewise.
10535 (vld3q_lane_s32): Likewise.
10536 (vld3q_lane_s64): Likewise.
10537 (vld3q_lane_f16): Likewise.
10538 (vld3q_lane_f32): Likewise.
10539 (vld3q_lane_f64): Likewise.
10540 (vld3q_lane_p8): Likewise.
10541 (vld3q_lane_p16): Likewise.
10542 (vld3q_lane_p64): Likewise.
10543 (vld3_lane_bf16): Likewise.
10544 (vld3q_lane_bf16): Likewise.
10546 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10548 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
10549 (__LD2Q_LANE_FUNC): Likewise.
10550 (vld2_lane_u8): Define without macro.
10551 (vld2_lane_u16): Likewise.
10552 (vld2_lane_u32): Likewise.
10553 (vld2_lane_u64): Likewise.
10554 (vld2_lane_s8): Likewise.
10555 (vld2_lane_s16): Likewise.
10556 (vld2_lane_s32): Likewise.
10557 (vld2_lane_s64): Likewise.
10558 (vld2_lane_f16): Likewise.
10559 (vld2_lane_f32): Likewise.
10560 (vld2_lane_f64): Likewise.
10561 (vld2_lane_p8): Likewise.
10562 (vld2_lane_p16): Likewise.
10563 (vld2_lane_p64): Likewise.
10564 (vld2q_lane_u8): Likewise.
10565 (vld2q_lane_u16): Likewise.
10566 (vld2q_lane_u32): Likewise.
10567 (vld2q_lane_u64): Likewise.
10568 (vld2q_lane_s8): Likewise.
10569 (vld2q_lane_s16): Likewise.
10570 (vld2q_lane_s32): Likewise.
10571 (vld2q_lane_s64): Likewise.
10572 (vld2q_lane_f16): Likewise.
10573 (vld2q_lane_f32): Likewise.
10574 (vld2q_lane_f64): Likewise.
10575 (vld2q_lane_p8): Likewise.
10576 (vld2q_lane_p16): Likewise.
10577 (vld2q_lane_p64): Likewise.
10578 (vld2_lane_bf16): Likewise.
10579 (vld2q_lane_bf16): Likewise.
10581 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10583 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
10586 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10588 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
10589 (rank_for_schedule): Use it.
10591 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10593 PR rtl-optimization/91598
10594 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
10595 insns after memory reads and before memory writes.
10597 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
10599 * rtl.h (CONST_VECTOR_P): New macro.
10600 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
10601 code testing macros.
10602 (aarch64_ptrue_all_mode): Likewise.
10603 (aarch64_expand_mov_immediate): Likewise.
10604 (aarch64_const_vec_all_in_range_p): Likewise.
10605 (aarch64_rtx_costs): Likewise.
10606 (aarch64_legitimate_constant_p): Likewise.
10607 (aarch64_simd_valid_immediate): Likewise.
10608 (aarch64_simd_make_constant): Likewise.
10609 (aarch64_convert_mult_to_shift): Likewise.
10610 (aarch64_expand_sve_vec_perm): Likewise.
10611 (aarch64_vec_fpconst_pow_of_2): Likewise.
10613 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10615 PR tree-optimization/101938
10616 * range-op.cc (operator_abs::op1_range): Special case
10617 -TYPE_MIN_VALUE for flag_wrapv.
10619 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
10621 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
10624 2021-08-17 Jakub Jelinek <jakub@redhat.com>
10626 * tree.def (OMP_SCOPE): New tree code.
10627 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
10628 * tree-nested.c (convert_nonlocal_reference_stmt,
10629 convert_local_reference_stmt, convert_gimple_call): Handle
10631 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
10632 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
10633 * gimple.c (gimple_build_omp_scope): New function.
10634 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
10635 * gimple.h (gimple_build_omp_scope): Declare.
10636 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
10637 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
10638 gimple_omp_scope_set_clauses): New inline functions.
10639 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
10640 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
10641 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
10642 * gimple-walk.c (walk_gimple_stmt): Likewise.
10643 * gimple-low.c (lower_stmt): Likewise.
10644 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10645 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
10646 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
10647 allowing task reductions. Reject inscan reductions on scope.
10648 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
10649 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
10650 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
10651 (estimate_num_insns): Likewise.
10652 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
10653 contexts if var isn't privatized there.
10654 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
10655 (scan_omp_1_stmt): Likewise.
10656 (maybe_add_implicit_barrier_cancel): Look through outer
10658 (lower_omp_scope): New function.
10659 (lower_omp_task_reductions): Handle OMP_SCOPE.
10660 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
10661 (diagnose_sb_1, diagnose_sb_2): Likewise.
10662 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
10663 (expand_omp): Handle GIMPLE_OMP_SCOPE.
10664 (omp_make_gimple_edges): Likewise.
10665 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
10667 2021-08-17 Richard Biener <rguenther@suse.de>
10669 PR tree-optimization/101925
10670 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
10671 reverse on COMPONENT_REF and ARRAY_REF according to
10672 what reverse_storage_order_for_component_p does.
10673 (vn_reference_eq): Compare reversed on reference ops.
10674 (reverse_storage_order_for_component_p): New overload.
10675 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
10676 on the reference looked up.
10678 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
10680 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
10681 shifts on the H8/S.
10682 (h8300_option_override): Use loops on H8/S more often when optimizing
10684 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
10685 accordingly. Handle various arithmetic right shifts with special
10686 sequences that we couldn't handle before.
10688 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
10690 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
10692 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
10694 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
10695 * config/sparc/sparc.c (sparc_gcov_type_size): New.
10696 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
10697 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
10698 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
10699 * doc/tm.texi.in: Regenerate.
10700 * target.def (gcov_type_size): New target hook.
10701 * targhooks.c (default_gcov_type_size): New.
10702 * targhooks.h (default_gcov_type_size): Declare.
10703 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
10705 (gimple_gen_time_profiler): Likewise.
10707 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
10709 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
10711 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
10713 PR middle-end/101931
10714 * omp-low.c (omp_runtime_api_call): Update for routines
10715 added in the meanwhile.
10717 2021-08-16 Martin Liska <mliska@suse.cz>
10719 PR tree-optimization/100393
10720 * tree-switch-conversion.c (group_cluster::dump): Use
10721 get_comparison_count.
10722 (jump_table_cluster::find_jump_tables): Pre-compute number of
10723 comparisons and then decrement it. Cache also max_ratio.
10724 (jump_table_cluster::can_be_handled): Change signature.
10725 * tree-switch-conversion.h (get_comparison_count): New.
10727 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
10729 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
10730 encodings only when -fgnat-encodings=all is specified.
10731 (add_bound_info): Likewise.
10732 (add_byte_size_attribute): Likewise.
10733 (gen_member_die): Likewise.
10735 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10737 * omp-oacc-neuter-broadcast.cc
10738 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
10740 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10742 * omp-oacc-neuter-broadcast.cc
10743 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
10746 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10748 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
10749 (execute_omp_oacc_neuter_broadcast): ... here.
10750 (install_var_field, build_receiver_ref, build_sender_ref): Take
10751 'field_map_t *' parameter. Adjust all users.
10752 (worker_single_copy, neuter_worker_single): Take a
10753 'record_field_map_t *' parameter. Adjust all users.
10755 2021-08-16 liuhongt <hongtao.liu@intel.com>
10758 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
10761 2021-08-16 Martin Liska <mliska@suse.cz>
10764 * multiple_target.c (create_dispatcher_calls): Make default
10765 function local only if it is a definition.
10767 2021-08-16 Martin Liska <mliska@suse.cz>
10770 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
10771 consider equal SSA_NAMEs when one is a param.
10773 2021-08-16 liuhongt <hongtao.liu@intel.com>
10776 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
10777 Support vpermi2b for V32QI/V16QImode.
10778 (ix86_extract_perm_from_pool_constant): New function.
10779 (ix86_expand_vec_one_operand_perm_avx512): Support
10780 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
10781 (expand_vec_perm_1): Adjust comments for upper.
10782 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
10784 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
10785 (pshufb_truncv4siv4hi_operand): Ditto.
10786 (pshufb_truncv8hiv8qi_operand): Ditto.
10787 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
10788 New pre_reload define_insn_and_split.
10789 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
10790 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
10791 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
10792 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
10793 (*avx512f_permvar_truncv4div4si_1): Ditto.
10794 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
10795 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
10796 (*avx512f_pshufd_truncv2div2si_1): Ditto.
10798 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
10800 * config/riscv/multilib-generator: Support code model option for
10802 * doc/install.texi: Add document of new option for
10803 --with-multilib-generator.
10805 2021-08-15 Clément Chigot <clement.chigot@atos.net>
10807 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
10808 (rs6000_legitimize_tls_address_aix): Use it.
10809 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
10810 xcoff_tls_exec_model_detected is true.
10812 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
10814 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
10815 to allow a bit more code growth, saving many dozens of cycles.
10816 (h8300_option_override): Adjus shift_alg_si if optimizing for
10818 (get_shift_alg): Use special + inline shifts for residuals
10821 2021-08-14 Stafford Horne <shorne@gmail.com>
10824 * config/or1k/or1k-opts.h: New file.
10825 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
10826 Support generating gotha relocations if -mcmodel=large is
10828 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
10830 * config/or1k/or1k.opt (mcmodel=): New option.
10831 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
10833 2021-08-14 Martin Sebor <msebor@redhat.com>
10835 PR middle-end/101791
10836 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
10837 to valid_new_delete_pair_p.
10838 * tree.c (valid_new_delete_pair_p): Add argument.
10839 * tree.h (valid_new_delete_pair_p): Same.
10841 2021-08-14 Jakub Jelinek <jakub@redhat.com>
10844 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
10845 <case E_V64QImode>: For this mode assert
10846 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
10848 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
10851 * config/rs6000/altivec.md (xxeval): Use register_predicate
10852 instead of altivec_register_predicate.
10854 2021-08-13 Martin Sebor <msebor@redhat.com>
10856 PR middle-end/101734
10857 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
10858 (maybe_warn_operand): Call it.
10860 2021-08-13 Martin Liska <mliska@suse.cz>
10863 * attribs.c (decl_attributes): Make naked functions "noipa"
10866 2021-08-13 Martin Liska <mliska@suse.cz>
10869 * symtab.c (symtab_node::noninterposable_alias): Do not create
10870 local aliases for target_clone functions as the clonning pass
10873 2021-08-13 Martin Liska <mliska@suse.cz>
10875 * opts.c (LIVE_PATCHING_OPTION): Define.
10876 (control_options_for_live_patching): Use it in error messages.
10878 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
10880 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
10881 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
10882 ignore_stores_eaf_flags): New constants.
10883 (remove_useless_eaf_flags): New function.
10884 (eaf_flags_useful_p): Use it.
10885 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
10887 (modref_lattice::init): Add EAF_NOREAD.
10888 (modref_lattice::add_escape_point): Do not reacord escape point if
10890 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
10891 use remove_useless_eaf_flags.
10892 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
10893 (modref_lattice::merge_direct_load): Add EAF_NOREAD
10894 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
10895 (analyze_parms): Use remove_useless_eaf_flags.
10896 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
10897 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
10898 use remove_useless_eaf_flags.
10899 (modref_propagate_flags_in_scc): Update.
10900 * ipa-modref.h: Turn eaf_flags_t back to char.
10901 * tree-core.h (EAF_NOT_RETURNED): Fix.
10902 (EAF_NOREAD): New constant
10903 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
10905 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
10906 (handle_pure_call): Likewise.
10908 2021-08-12 Jakub Jelinek <jakub@redhat.com>
10910 * tree.def (OMP_MASKED): New tree code.
10911 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
10912 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
10913 OMP_CLAUSE_FILTER_EXPR): Define.
10914 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
10915 (omp_clause_code_name): Likewise.
10916 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
10917 * tree-nested.c (convert_nonlocal_omp_clauses,
10918 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
10919 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
10920 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
10921 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
10922 (dump_generic_node): Handle OMP_MASTER.
10923 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
10924 * gimple.c (gimple_build_omp_masked): New function.
10925 (gimple_copy): Handle GIMPLE_OMP_MASKED.
10926 * gimple.h (gimple_build_omp_masked): Declare.
10927 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
10928 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
10929 gimple_omp_masked_set_clauses): New inline functions.
10930 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
10931 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
10932 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
10933 * gimple-walk.c (walk_gimple_stmt): Likewise.
10934 * gimple-low.c (lower_stmt): Likewise.
10935 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10936 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
10937 that take one expression rather than decl or constant, force
10938 gimplification of that into a SSA_NAME or temporary unless min
10940 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
10941 (gimplify_expr): Handle OMP_MASKED.
10942 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
10943 (estimate_num_insns): Likewise.
10944 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
10945 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
10946 diagnostics for existence of masked construct.
10947 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
10948 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
10949 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
10952 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
10955 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
10956 (ldexp<mode>3): Use avx512f_scalef<mode>2.
10957 (UNSPEC_SCALEF): Move from ...
10958 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
10960 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
10962 * ipa-split.c (consider_split): Fix condition testing void functions.
10964 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
10966 * doc/invoke.texi: Remove docs for threader-mode param.
10967 * flag-types.h (enum threader_mode): Remove.
10968 * params.opt: Remove threader-mode param.
10969 * tree-ssa-threadbackward.c (class back_threader): Remove
10970 path_is_unreachable_p.
10971 Make find_paths private.
10972 Add maybe_thread and thread_through_all_blocks.
10973 Remove reference marker for m_registry.
10974 Remove reference marker for m_profit.
10975 (back_threader::back_threader): Adjust for registry and profit not
10977 (dump_path): Move down.
10978 (debug): Move down.
10979 (class thread_jumps): Remove.
10980 (class back_threader_registry): Remove m_all_paths.
10982 (thread_jumps::thread_through_all_blocks): Move to back_threader
10984 (fsm_find_thread_path): Remove
10985 (back_threader::maybe_thread): New.
10986 (back_threader::thread_through_all_blocks): Move from
10988 (back_threader_registry::back_threader_registry): Remove
10990 (back_threader_registry::~back_threader_registry): Remove.
10991 (thread_jumps::find_taken_edge): Remove.
10992 (thread_jumps::check_subpath_and_update_thread_path): Remove.
10993 (thread_jumps::maybe_register_path): Remove.
10994 (thread_jumps::handle_phi): Remove.
10995 (handle_assignment_p): Remove.
10996 (thread_jumps::handle_assignment): Remove.
10997 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
10998 (thread_jumps::find_jump_threads_backwards): Remove.
10999 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
11000 (try_thread_blocks): Rename find_jump_threads_backwards to
11002 (pass_early_thread_jumps::execute): Same.
11004 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
11006 * tree-core.h (omp_clause_proc_bind_kind): Add
11007 OMP_CLAUSE_PROC_BIND_PRIMARY.
11008 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
11009 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
11011 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
11013 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
11014 fno-common reference.
11015 * config/arc/arc.c (arc_override_options): Remove overriding of
11018 2021-08-12 Jakub Jelinek <jakub@redhat.com>
11021 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
11022 If d->testing_p, return true after performing checks instead of
11023 actually expanding the insn.
11024 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
11025 !TARGET_AVX512BW and return false.
11027 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
11029 * configure.ac (PE linker --disable-dynamicbase support): New check.
11030 * configure: Regenerate.
11031 * config.in: Likewise.
11032 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
11033 (LINK_SPEC): Use it.
11034 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
11035 (LINK_SPEC): Likewise.
11037 2021-08-12 liuhongt <hongtao.liu@intel.com>
11040 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
11041 post_reload define_insn_and_split.
11042 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
11043 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
11044 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
11045 (*avx2_zero_extendv8hiv8si2_2): Ditto.
11046 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
11047 (*avx512f_zero_extendv8siv8di2_2): Ditto.
11048 (*avx2_zero_extendv4siv4di2_2): Ditto.
11049 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
11050 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
11053 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
11055 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
11058 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
11060 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
11062 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
11064 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
11065 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
11066 initialization of pcvoid_type_node here...
11067 (altivec_init_builtins): ...from here.
11068 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
11069 RS6000_BTI_const_ptr_void.
11070 (pcvoid_type_node): New macro.
11072 2021-08-11 Richard Biener <rguenther@suse.de>
11075 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
11076 hard-register accesses.
11078 2021-08-11 Richard Biener <rguenther@suse.de>
11080 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
11081 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
11082 to determine has_volatile_ops.
11084 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
11086 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
11088 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
11089 Alexandre Oliva <oliva@adacore.com>
11091 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
11093 (modref_lattice::merge_deref): Fix handling of indirect scape points.
11094 (update_escape_summary_1): Likewise.
11095 (update_escape_summary): Likewise.
11096 (ipa_merge_modref_summary_after_inlining): Likewise.
11098 2021-08-11 Richard Biener <rguenther@suse.de>
11100 PR middle-end/101858
11101 * fold-const.c (fold_binary_loc): Guard simplification
11102 of X < (cast) (1 << Y) to integer types.
11104 2021-08-11 Richard Biener <rguenther@suse.de>
11106 PR tree-optimization/101861
11107 * tree-vect-stmts.c (vectorizable_load): Fix error in
11108 previous change with regard to gather vectorization.
11110 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11113 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
11115 (vdup_n_s16): Likewise.
11116 (vdup_n_s32): Likewise.
11117 (vdup_n_s64): Likewise.
11118 (vdup_n_u8): Likewise.
11119 (vdup_n_u16): Likewise.
11120 (vdup_n_u32): Likewise.
11121 (vdup_n_u64): Likewise.
11122 (vdup_n_p8): Likewise.
11123 (vdup_n_p16): Likewise.
11124 (vdup_n_p64): Likewise.
11125 (vdup_n_f16): Likewise.
11126 (vdup_n_f32): Likewise.
11127 (vdupq_n_s8): Likewise.
11128 (vdupq_n_s16): Likewise.
11129 (vdupq_n_s32): Likewise.
11130 (vdupq_n_s64): Likewise.
11131 (vdupq_n_u8): Likewise.
11132 (vdupq_n_u16): Likewise.
11133 (vdupq_n_u32): Likewise.
11134 (vdupq_n_u64): Likewise.
11135 (vdupq_n_p8): Likewise.
11136 (vdupq_n_p16): Likewise.
11137 (vdupq_n_p64): Likewise.
11138 (vdupq_n_f16): Likewise.
11139 (vdupq_n_f32): Likewise.
11140 (vmov_n_s8): Replace call to builtin with call to corresponding
11142 (vmov_n_s16): Likewise.
11143 (vmov_n_s32): Likewise.
11144 (vmov_n_s64): Likewise.
11145 (vmov_n_u8): Likewise.
11146 (vmov_n_u16): Likewise.
11147 (vmov_n_u32): Likewise.
11148 (vmov_n_u64): Likewise.
11149 (vmov_n_p8): Likewise.
11150 (vmov_n_p16): Likewise.
11151 (vmov_n_f16): Likewise.
11152 (vmov_n_f32): Likewise.
11153 (vmovq_n_s8): Likewise.
11154 (vmovq_n_s16): Likewise.
11155 (vmovq_n_s32): Likewise.
11156 (vmovq_n_s64): Likewise.
11157 (vmovq_n_u8): Likewise.
11158 (vmovq_n_u16): Likewise.
11159 (vmovq_n_u32): Likewise.
11160 (vmovq_n_u64): Likewise.
11161 (vmovq_n_p8): Likewise.
11162 (vmovq_n_p16): Likewise.
11163 (vmovq_n_f16): Likewise.
11164 (vmovq_n_f32): Likewise.
11165 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
11167 2021-08-11 liuhongt <hongtao.liu@intel.com>
11170 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
11171 when TARGET_AVX512F and TARGET_SSE_MATH.
11173 2021-08-10 Jakub Jelinek <jakub@redhat.com>
11176 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
11177 for V32HImode if !TARGET_AVX512BW.
11178 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
11179 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
11180 early, but actually check the permutation.
11182 2021-08-10 Richard Biener <rguenther@suse.de>
11184 PR tree-optimization/101809
11185 * tree-vect-stmts.c (get_load_store_type): Allow emulated
11186 gathers with offset vector nunits being a constant multiple
11187 of the data vector nunits.
11188 (vect_get_gather_scatter_ops): Use the appropriate nunits
11189 for the offset vector defs.
11190 (vectorizable_store): Adjust call to
11191 vect_get_gather_scatter_ops.
11192 (vectorizable_load): Likewise. Handle the case of less
11193 offset vectors than data vectors.
11195 2021-08-10 Jakub Jelinek <jakub@redhat.com>
11198 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
11199 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
11202 2021-08-10 Richard Biener <rguenther@suse.de>
11204 PR tree-optimization/101801
11205 PR tree-optimization/101819
11206 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
11207 * tree-vect-loop.c (vect_emulated_vector_p): New function.
11208 (vectorizable_reduction): Re-instantiate a check for emulated
11210 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11211 (vectorizable_operation): Likewise. Cost emulated vector
11212 operations according to the scalar sequence synthesized by
11215 2021-08-10 Richard Biener <rguenther@suse.de>
11217 PR middle-end/101824
11218 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
11219 volatile in case the variable was.
11221 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
11224 * config/i386/constraints.md (BC): Document for integer SSE
11225 constant all bits set operand.
11226 (BF): New constraint for const floating-point all bits set
11228 * config/i386/i386.c (standard_sse_constant_p): Likewise.
11229 (standard_sse_constant_opcode): Likewise.
11230 * config/i386/sse.md (sseconstm1): New mode attribute.
11231 (mov<mode>_internal): Replace BC with <sseconstm1>.
11233 2021-08-10 liuhongt <hongtao.liu@intel.com>
11235 * config/i386/sse.md (cond_<insn><mode>): New expander.
11236 (VI248_AVX512VLBW): New mode iterator.
11237 * config/i386/predicates.md
11238 (nonimmediate_or_const_vec_dup_operand): New predicate.
11240 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
11242 PR tree-optimization/101741
11243 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
11244 type of parameter for toupper/tolower.
11246 2021-08-09 Martin Jambor <mjambor@suse.cz>
11248 PR testsuite/101654
11249 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
11251 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
11253 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
11255 (is_store_insn1): Verify source is a register.
11257 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
11260 * config/i386/mmx.md (<any_logic:code>v2sf3):
11261 Rename from *mmx_<any_logic:code>v2sf3
11263 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11265 * config/nvptx/nvptx.c: Cross-reference parts adapted in
11266 'gcc/omp-oacc-neuter-broadcast.cc'.
11267 * omp-low.c: Likewise.
11268 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
11271 2021-08-09 Julian Brown <julian@codesourcery.com>
11272 Kwok Cheung Yeung <kcy@codesourcery.com>
11273 Thomas Schwinge <thomas@codesourcery.com>
11275 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
11276 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
11277 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
11278 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
11279 (gcn_fork_join): Update comment.
11280 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
11281 (macc_experimental_workers): Remove unused option.
11283 2021-08-09 Julian Brown <julian@codesourcery.com>
11284 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
11285 Kwok Cheung Yeung <kcy@codesourcery.com>
11286 Thomas Schwinge <thomas@codesourcery.com>
11288 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
11289 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
11290 Add documentation hook.
11291 * doc/tm.texi: Regenerate.
11292 * omp-oacc-neuter-broadcast.cc: New file.
11293 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
11294 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
11295 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
11296 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
11297 * target.def (goacc.create_worker_broadcast_record): Add target
11299 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
11301 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
11302 Rename prototype to...
11303 (gcn_goacc_create_worker_broadcast_record): ... this.
11304 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
11306 (gcn_goacc_create_worker_broadcast_record): ... this.
11307 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
11309 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
11311 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
11314 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
11315 the right iterator.
11317 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11319 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
11321 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11323 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
11324 consider '-foffload-abi'.
11325 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
11326 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
11329 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11331 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
11334 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11336 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
11338 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11340 * doc/gty.texi (Files): Update.
11342 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11344 * doc/gty.texi (Files): Fix GTY header file example.
11346 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
11348 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
11349 determine the upper and lower bounds from a mask-value pair.
11350 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
11351 absolute value and unsigned absolute value expressions.
11352 (bit_value_binop): Initialize *VAL's precision.
11353 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
11354 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
11355 support when the operands are unknown but potentially equal.
11356 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
11358 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
11360 * config/aarch64/aarch64.md
11361 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
11363 2021-08-08 Sergei Trofimovich <siarheit@google.com>
11365 * lra-constraints.c: Fix s/otput/output/ typo.
11367 2021-08-06 Martin Sebor <msebor@redhat.com>
11369 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
11370 (expand_builtin_strcat): Same.
11371 (expand_builtin_stpncpy): Same.
11372 (expand_builtin_strncat): Same.
11373 (check_read_access): Same.
11374 (check_memop_access): Same.
11375 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
11376 (expand_builtin_strnlen): Same.
11377 (expand_builtin_memcpy): Same.
11378 (expand_builtin_memmove): Same.
11379 (expand_builtin_mempcpy): Same.
11380 (expand_builtin_strcpy): Same.
11381 (expand_builtin_strcpy_args): Same.
11382 (expand_builtin_stpcpy_1): Same.
11383 (expand_builtin_strncpy): Same.
11384 (expand_builtin_memset): Same.
11385 (expand_builtin_bzero): Same.
11386 (expand_builtin_strcmp): Same.
11387 (expand_builtin_strncmp): Same.
11388 (expand_builtin): Remove handlers.
11389 (fold_builtin_strlen): Add a comment.
11390 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
11391 * calls.c (maybe_warn_nonstring_arg): Same.
11392 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
11393 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
11394 (gimple_fold_builtin_stpcpy): Same.
11395 * gimple-ssa-warn-access.cc (has_location): New function.
11396 (get_location): Same.
11397 (get_callee_fndecl): Same.
11398 (call_nargs): Same.
11400 (warn_string_no_nul): Define.
11401 (unterminated_array): Same.
11402 (check_nul_terminated_array): Same.
11403 (maybe_warn_nonstring_arg): Same.
11404 (maybe_warn_for_bound): Same.
11405 (warn_for_access): Same.
11406 (check_access): Same.
11407 (check_memop_access): Same.
11408 (check_read_access): Same.
11409 (warn_dealloc_offset): Use helper functions.
11410 (maybe_emit_free_warning): Same.
11411 (class pass_waccess): Add members.
11412 (check_strcat): New function.
11413 (check_strncat): New function.
11414 (check_stxcpy): New function.
11415 (check_stxncpy): New function.
11416 (check_strncmp): New function.
11417 (pass_waccess::check_builtin): New function.
11418 (pass_waccess::check): Call it.
11419 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
11421 (maybe_warn_for_bound): Same.
11422 (check_access): Same.
11423 (check_memop_access): Same.
11424 (check_read_access): Same.
11425 * pointer-query.h (struct access_data): Define a ctor overload.
11427 2021-08-06 Richard Biener <rguenther@suse.de>
11429 PR tree-optimization/101801
11430 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
11431 (vect_can_vectorize_without_simd_p): ... to this.
11432 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
11433 (vect_can_vectorize_without_simd_p): ... to this and fold
11434 in vect_min_worthwhile_factor.
11435 (vect_min_worthwhile_factor): Remove.
11436 (vectorizable_reduction): Adjust and remove the cost part.
11437 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11438 (vectorizable_operation): Likewise.
11440 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
11443 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
11444 Add general_gr_operand predicate to operand 3.
11446 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
11448 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
11449 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
11451 2021-08-06 Tamar Christina <tamar.christina@arm.com>
11453 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
11454 register_svprfop): Pass vec<> by pointer.
11455 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
11456 * langhooks.c (lhd_simulate_enum_decl): Likewise.
11457 * langhooks.h (struct lang_hooks_for_types): Likewise.
11459 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11461 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
11462 __builtin_memcpy instead of constructing an additional
11463 __builtin_aarch64_simd_oi one vector at a time.
11464 (vst1q_bf16_x2): Likewise.
11465 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
11466 an additional __builtin_aarch64_simd_ci one vector at a time.
11467 (vst1q_bf16_x3): Likewise.
11468 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
11469 (vst1q_bf16_x4): Likewise.
11470 (vst2_bf16): Use __builtin_memcpy instead of constructing an
11471 additional __builtin_aarch64_simd_oi one vector at a time.
11472 (vst2q_bf16): Likewise.
11473 (vst3_bf16): Use __builtin_memcpy instead of constructing an
11474 additional __builtin_aarch64_simd_ci mode one vector at a
11476 (vst3q_bf16): Likewise.
11477 (vst4_bf16): Use __builtin_memcpy instead of constructing an
11478 additional __builtin_aarch64_simd_xi one vector at a time.
11479 (vst4q_bf16): Likewise.
11481 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11483 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
11484 (__ST2Q_LANE_FUNC): Delete.
11485 (vst2_lane_f16): Use __builtin_memcpy to copy vector
11486 structure instead of constructing __builtin_aarch64_simd_oi
11487 one vector at a time.
11488 (vst2_lane_f32): Likewise.
11489 (vst2_lane_f64): Likewise.
11490 (vst2_lane_p8): Likewise.
11491 (vst2_lane_p16): Likewise.
11492 (vst2_lane_p64): Likewise.
11493 (vst2_lane_s8): Likewise.
11494 (vst2_lane_s16): Likewise.
11495 (vst2_lane_s32): Likewise.
11496 (vst2_lane_s64): Likewise.
11497 (vst2_lane_u8): Likewise.
11498 (vst2_lane_u16): Likewise.
11499 (vst2_lane_u32): Likewise.
11500 (vst2_lane_u64): Likewise.
11501 (vst2_lane_bf16): Likewise.
11502 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
11503 structure instead of using a union.
11504 (vst2q_lane_f32): Likewise.
11505 (vst2q_lane_f64): Likewise.
11506 (vst2q_lane_p8): Likewise.
11507 (vst2q_lane_p16): Likewise.
11508 (vst2q_lane_p64): Likewise.
11509 (vst2q_lane_s8): Likewise.
11510 (vst2q_lane_s16): Likewise.
11511 (vst2q_lane_s32): Likewise.
11512 (vst2q_lane_s64): Likewise.
11513 (vst2q_lane_u8): Likewise.
11514 (vst2q_lane_u16): Likewise.
11515 (vst2q_lane_u32): Likewise.
11516 (vst2q_lane_u64): Likewise.
11517 (vst2q_lane_bf16): Likewise.
11519 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11521 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
11522 (__ST3Q_LANE_FUNC): Delete.
11523 (vst3_lane_f16): Use __builtin_memcpy to copy vector
11524 structure instead of constructing __builtin_aarch64_simd_ci
11525 one vector at a time.
11526 (vst3_lane_f32): Likewise.
11527 (vst3_lane_f64): Likewise.
11528 (vst3_lane_p8): Likewise.
11529 (vst3_lane_p16): Likewise.
11530 (vst3_lane_p64): Likewise.
11531 (vst3_lane_s8): Likewise.
11532 (vst3_lane_s16): Likewise.
11533 (vst3_lane_s32): Likewise.
11534 (vst3_lane_s64): Likewise.
11535 (vst3_lane_u8): Likewise.
11536 (vst3_lane_u16): Likewise.
11537 (vst3_lane_u32): Likewise.
11538 (vst3_lane_u64): Likewise.
11539 (vst3_lane_bf16): Likewise.
11540 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
11541 structure instead of using a union.
11542 (vst3q_lane_f32): Likewise.
11543 (vst3q_lane_f64): Likewise.
11544 (vst3q_lane_p8): Likewise.
11545 (vst3q_lane_p16): Likewise.
11546 (vst3q_lane_p64): Likewise.
11547 (vst3q_lane_s8): Likewise.
11548 (vst3q_lane_s16): Likewise.
11549 (vst3q_lane_s32): Likewise.
11550 (vst3q_lane_s64): Likewise.
11551 (vst3q_lane_u8): Likewise.
11552 (vst3q_lane_u16): Likewise.
11553 (vst3q_lane_u32): Likewise.
11554 (vst3q_lane_u64): Likewise.
11555 (vst3q_lane_bf16): Likewise.
11557 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11559 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
11560 (__ST4Q_LANE_FUNC): Delete.
11561 (vst4_lane_f16): Use __builtin_memcpy to copy vector
11562 structure instead of constructing __builtin_aarch64_simd_xi
11563 one vector at a time.
11564 (vst4_lane_f32): Likewise.
11565 (vst4_lane_f64): Likewise.
11566 (vst4_lane_p8): Likewise.
11567 (vst4_lane_p16): Likewise.
11568 (vst4_lane_p64): Likewise.
11569 (vst4_lane_s8): Likewise.
11570 (vst4_lane_s16): Likewise.
11571 (vst4_lane_s32): Likewise.
11572 (vst4_lane_s64): Likewise.
11573 (vst4_lane_u8): Likewise.
11574 (vst4_lane_u16): Likewise.
11575 (vst4_lane_u32): Likewise.
11576 (vst4_lane_u64): Likewise.
11577 (vst4_lane_bf16): Likewise.
11578 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
11579 structure instead of using a union.
11580 (vst4q_lane_f32): Likewise.
11581 (vst4q_lane_f64): Likewise.
11582 (vst4q_lane_p8): Likewise.
11583 (vst4q_lane_p16): Likewise.
11584 (vst4q_lane_p64): Likewise.
11585 (vst4q_lane_s8): Likewise.
11586 (vst4q_lane_s16): Likewise.
11587 (vst4q_lane_s32): Likewise.
11588 (vst4q_lane_s64): Likewise.
11589 (vst4q_lane_u8): Likewise.
11590 (vst4q_lane_u16): Likewise.
11591 (vst4q_lane_u32): Likewise.
11592 (vst4q_lane_u64): Likewise.
11593 (vst4q_lane_bf16): Likewise.
11595 2021-08-06 Martin Liska <mliska@suse.cz>
11597 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
11598 a target option is restored, it can have
11599 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
11600 and error should not be emitted.
11602 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
11604 * gcov-io.h (gcov_write): Declare.
11605 * gcov-io.c (gcov_write): New.
11606 (gcov_write_counter): Remove.
11607 (gcov_write_tag_length): Likewise.
11608 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
11609 gcov_write_unsigned().
11610 * doc/invoke.texi (fprofile-info-section): Mention
11611 __gcov_info_to_gdca().
11613 2021-08-06 Martin Sebor <msebor@redhat.com>
11615 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
11616 arguments to by-reference.
11617 (iterate_fix_dominators): Same.
11618 * dominance.h (iterate_fix_dominators): Same.
11619 * ipa-prop.h: Call auto_vec::to_vec_legacy.
11620 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
11621 arguments to by-reference.
11622 (debug_data_dependence_relation): Same.
11623 (dump_data_dependence_relations): Same.
11624 * tree-data-ref.h (debug_data_dependence_relation): Same.
11625 (dump_data_dependence_relations): Same.
11626 * tree-predcom.c (dump_chains): Same.
11627 (initialize_root_vars_lm): Same.
11628 (determine_unroll_factor): Same.
11629 (replace_phis_by_defined_names): Same.
11630 (insert_init_seqs): Same.
11631 (pcom_worker::tree_predictive_commoning_loop): Call
11632 auto_vec::to_vec_legacy.
11633 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
11634 arguments to by-reference.
11635 * tree-ssa-threadbackward.c (populate_worklist): Same.
11636 (back_threader::resolve_def): Same.
11637 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
11638 (vect_enhance_data_refs_alignment): Same.
11639 (vect_check_lower_bound): Same.
11640 (vect_prune_runtime_alias_test_list): Same.
11641 (vect_permute_store_chain): Same.
11642 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
11643 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
11644 * tree-vectorizer.h (vect_permute_store_chain): Same.
11645 * vec.c (test_init): New function.
11646 (vec_c_tests): Call new function.
11647 * vec.h (vec): Declare ctors, dtor, and assignment.
11648 (auto_vec::vec_to_legacy): New function.
11649 (vec::copy): Adjust initialization.
11651 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
11654 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
11655 callee only uses GPRs.
11656 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
11657 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
11658 * config/i386/x86gprintrin.h: Add
11659 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
11660 to disable non-GPR ISAs.
11662 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11664 PR middle-end/101787
11665 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
11667 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11669 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
11670 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
11671 (vect_is_extending_load, vect_is_integer_truncation): New functions,
11672 moved from aarch64.c but given different names.
11673 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
11674 (aarch64_is_reduction, aarch64_reduc_type)
11675 (aarch64_embedded_comparison_type, aarch64_comparison_type)
11676 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
11677 in favor of the above. Update callers accordingly.
11679 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11682 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
11683 writing .cpu directive in asm output.
11684 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
11685 (arm_last_printed_arch_string): Delete.
11686 (arm_last-printed_fpu_string): Delete.
11687 (arm_configure_build_target): If use of floating-point/SIMD is
11688 disabled, remove all fp/simd related features from the target ISA.
11689 (last_arm_targ_options): New variable.
11690 (arm_print_asm_arch_directives): Add new parameters. Change order
11691 of emitted directives and handle all cases here.
11692 (arm_file_start): Always call arm_print_asm_arch_directives, move
11693 all generation of .arch/.arch_extension here.
11694 (arm_file_end): Call arm_print_asm_arch.
11695 (arm_declare_function_name): Call arm_print_asm_arch_directives
11696 instead of printing .arch/.fpu directives directly.
11698 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11700 * config/arm/arm.c (arm_configure_build_target): Don't call
11701 arm_option_reconfigure_globals.
11702 (arm_option_restore): Call arm_option_reconfigure_globals after
11703 reconfiguring the target.
11704 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
11706 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11708 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
11709 arch_name is always set.
11711 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
11713 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
11714 of vec_select high-half from being added into Neon subtract
11717 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
11719 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
11720 of vec_select high-half from being added into Neon add cost.
11722 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
11724 * cfgloop.h (loops_list::loops_list): Add one optional argument
11725 root and adjust accordingly, update loop tree walking and factor
11727 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
11729 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
11731 PR tree-optimization/101626
11732 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
11733 reverse scalar storage order on a pointer or vector component.
11735 2021-08-05 liuhongt <hongtao.liu@intel.com>
11737 * config/i386/sse.md (cond_<code><mode>): New expander.
11739 2021-08-05 liuhongt <hongtao.liu@intel.com>
11741 * config/i386/sse.md (cond_<code><mode>): New expander.
11743 2021-08-05 liuhongt <hongtao.liu@intel.com>
11745 * config/i386/sse.md (cond_<code><mode>): New expander.
11747 2021-08-04 David Malcolm <dmalcolm@redhat.com>
11750 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
11752 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11755 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
11756 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
11758 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11761 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
11762 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
11763 data with SSE register from one memory location to another.
11765 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11767 * config/s390/s390.c (expand_perm_with_vpdi): New function.
11768 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
11769 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
11770 parameterized expander.
11771 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
11773 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11775 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
11776 (struct expand_vec_perm_d): Define struct.
11777 (expand_perm_with_merge): New function.
11778 (vectorize_vec_perm_const_1): New function.
11779 (s390_vectorize_vec_perm_const): New function.
11780 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
11782 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11784 * config/s390/vector.md (V_HW_64): Remove mode iterator.
11785 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
11786 * config/s390/vx-builtins.md
11787 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
11790 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11792 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
11794 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
11796 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
11797 instead of an unspec.
11799 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11801 * config/s390/s390-modes.def: Add more vector modes to support
11802 concatenation of two vectors.
11803 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
11805 (s390_expand_merge): Likewise.
11806 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
11807 (s390_expand_merge): New function.
11808 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
11809 Remove constant definitions.
11810 * config/s390/vector.md (V_HW_2): Add mode iterators.
11811 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
11812 (vec_2x_nelts, vec_2x_wide): New mode attributes.
11813 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
11814 New pattern definitions.
11815 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
11816 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
11817 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
11818 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
11820 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
11822 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
11823 emit vec merge pattern.
11825 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
11827 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
11829 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
11830 vec_select high-half from being added into Neon multiply
11832 * rtlanal.c (vec_series_highpart_p): Define.
11833 * rtlanal.h (vec_series_highpart_p): Declare.
11835 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
11837 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
11839 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
11840 vec_select cost from being added into Neon multiply cost.
11842 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11844 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
11845 which old_loop_vinfo is an epilogue loop that handles a constant
11846 number of iterations.
11848 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11850 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
11851 when a reanalyzed loop fails to be cheaper than the current
11854 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11856 * config/aarch64/aarch64.c: Fix a typo.
11858 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
11860 PR gcov-profile/101773
11861 * gcov-io.c (gcov_close): Check return code of a fclose.
11863 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
11866 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
11867 .file statement when needed.
11869 2021-08-04 Richard Biener <rguenther@suse.de>
11871 * tree-vect-data-refs.c (vect_check_gather_scatter):
11872 Include widening conversions only when the result is
11873 still handed by native gather or the current offset
11874 size not already matches the data size.
11875 Also succeed analysis in case there's no native support,
11876 noted by a IFN_LAST ifn and a NULL decl.
11877 (vect_analyze_data_refs): Always consider gathers.
11878 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
11879 Test for no IFN gather rather than decl gather.
11880 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
11881 gather-scatter info and cost emulated gathers accordingly.
11882 (vect_truncate_gather_scatter_offset): Properly test for
11884 (vect_use_strided_gather_scatters_p): Likewise.
11885 (get_load_store_type): Handle emulated gathers and its
11887 (vectorizable_load): Likewise. Emulate them by extracting
11888 scalar offsets, doing scalar loads and a vector construct.
11890 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11893 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
11894 argument to set m_max_size.
11895 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
11896 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
11897 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
11899 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
11900 Marc Glisse <marc.glisse@inria.fr>
11902 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
11903 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
11904 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
11906 2021-08-04 Richard Biener <rguenther@suse.de>
11908 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
11909 out code to decompose vector loads ...
11910 (optimize_vector_load): ... here. Generalize it to
11911 handle intermediate widening and TARGET_MEM_REF loads
11912 and apply it to loads with a supported vector mode as well.
11914 2021-08-04 Richard Biener <rguenther@suse.de>
11916 PR tree-optimization/101756
11917 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
11918 the result of the reduction epilogue is compatible to the original
11921 2021-08-04 liuhongt <hongtao.liu@intel.com>
11924 * config/i386/i386.md (peephole2): Refine predicate from
11925 register_operand to general_reg_operand.
11927 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
11929 * gimple-range-path.h (path_range_query::dump): Mark override.
11931 2021-08-04 Richard Biener <rguenther@suse.de>
11933 PR tree-optimization/101769
11934 * tree-tailcall.c (eliminate_tail_call): Add the created loop
11935 for the first recursion and return it via the new output parameter.
11936 (optimize_tail_call): Pass through new output param.
11937 (tree_optimize_tail_calls_1): After creating all latches,
11938 add the created loop to the loop tree. Do not mark loops for fixup.
11940 2021-08-04 Martin Liska <mliska@suse.cz>
11942 * doc/invoke.texi: Document threader-mode param.
11944 2021-08-04 liuhongt <hongtao.liu@intel.com>
11946 * config/i386/sse.md (cond_fma<mode>): New expander.
11947 (cond_fms<mode>): Ditto.
11948 (cond_fnma<mode>): Ditto.
11949 (cond_fnms<mode>): Ditto.
11951 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
11953 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
11955 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
11957 * config/rs6000/constraints.md: Remove "e" from the list of available
11958 constraint characters.
11960 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
11962 PR gcov-profile/71672
11963 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
11965 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
11967 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
11969 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
11971 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
11972 try XMM31 to avoid vzeroupper.
11974 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11976 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
11977 -mcpu=neoverse-512tvb.
11978 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
11979 * config/aarch64/aarch64-tune.md: Regenerate.
11980 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
11981 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
11982 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
11983 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
11984 (aarch64_adjust_body_cost): Likewise.
11986 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11988 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
11989 record issue information for operations that occur in the
11992 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11994 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
11995 parameter. Detect cases in which an Advanced SIMD MLA would almost
11996 certainly require a MOV.
11997 (aarch64_count_ops): Update accordingly.
11999 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12001 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
12002 function, split out from...
12003 (aarch64_detect_vector_stmt_subtype): ...here.
12004 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
12006 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12008 * config/aarch64/aarch64-protos.h (sve_vec_cost):
12009 Add gather_load_x32_cost and gather_load_x64_cost.
12010 * config/aarch64/aarch64.c (generic_sve_vector_cost)
12011 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
12012 accordingly, using the values given by the scalar_load * number
12013 of elements calculation that we used previously.
12014 (aarch64_detect_vector_stmt_subtype): Use the new fields.
12016 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12018 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
12019 function, split out from...
12020 (aarch64_adjust_body_cost): ...here.
12022 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12024 * config/aarch64/fractional-cost.h: New file.
12025 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
12026 and cost_fraction.h.
12027 (vec_cost_fraction): New typedef.
12028 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
12029 (aarch64_detect_vector_stmt_subtype): Likewise.
12030 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
12031 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
12033 (aarch64_adjust_body_cost): Likewise.
12034 (aarch64_test_cost_fraction): New function.
12035 (aarch64_run_selftests): Call it.
12037 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12039 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
12041 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
12043 (aarch64_estimated_poly_value): Likewise. Use the least significant
12044 set bit for the minimum and likely values. Use the most significant
12045 set bit for the maximum value.
12047 2021-08-03 liuhongt <hongtao.liu@intel.com>
12049 * config/i386/sse.md (cond_<insn><mode>): New expander.
12050 (cond_mul<mode>): Ditto.
12052 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
12054 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
12056 2021-08-03 liuhongt <hongtao.liu@intel.com>
12058 * config/i386/sse.md (cond_<insn><mode>):New expander.
12059 (cond_mul<mode>): Ditto.
12060 (cond_div<mode>): Ditto.
12062 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
12064 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
12065 check stack_realign_needed for stack realignment.
12066 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
12067 than the largest integer supported by vector register.
12068 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
12069 (MOVE_MAX): Set to bytes of the largest integer supported by
12071 (STORE_MAX_PIECES): New.
12073 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
12075 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
12076 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
12077 data from one memory location to another.
12079 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
12081 PR middle-end/90773
12082 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
12084 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
12086 PR tree-optimization/101724
12087 * params.opt: Remove --param=threader-iterative.
12088 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
12091 2021-08-02 Tom de Vries <tdevries@suse.de>
12093 PR middle-end/101665
12094 * doc/extend.texi (nonnull attribute): Improve documentation.
12096 2021-08-02 Andrew Pinski <apinski@marvell.com>
12098 PR rtl-optimization/101683
12099 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
12101 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
12103 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
12104 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
12105 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
12106 (tree_ssa_phiop_worker): Update call to function above.
12108 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
12111 * config/i386/i386.md (bsr_rex64_1_zext): New.
12112 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
12113 with gen_bsr_rex64_1_zext.
12115 2021-07-31 Jakub Jelinek <jakub@redhat.com>
12118 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
12119 define_insn patterns.
12120 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
12121 Add combine splitters for constant - clz.
12122 (clz<mode>2): Use a temporary pseudo for bsr result.
12124 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12126 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
12127 _mm_floor_sd, _mm_floor_ss): New.
12129 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12131 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
12132 _mm_ceil_sd, _mm_ceil_ss): New.
12134 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12136 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
12137 _mm_blend_ps, _mm_blendv_ps): New.
12139 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
12140 Uroš Bizjak <ubizjak@gmail.com>
12142 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
12143 to generate a conditional move using the carry flag after sub $1.
12144 (peephole2): Eliminate a register-to-register move by inverting
12145 the condition of a conditional move.
12147 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
12149 * config/mmix/mmix.md ("call", "call_value", "*call_real")
12150 ("*call_value_real"): Don't generate rtx mentioning the generic
12151 operands 1 and 2 to "call", and similarly for "call_value".
12152 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
12153 (mmix_print_operand): Use '!' instead of 'p'.
12155 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
12157 * doc/md.texi (call): Correct information about operand 2.
12158 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
12160 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12162 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
12164 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12166 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
12168 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12170 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
12171 const_basic_block..
12172 (*::get_bb_range): Ditto.
12173 (*::bb_range_p): Ditto.
12174 * gimple-range-cache.h: Change prototypes.
12176 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12178 PR middle-end/90773
12179 * builtins.c (builtin_memcpy_read_str): Change the mode argument
12180 from scalar_int_mode to fixed_size_mode.
12181 (builtin_strncpy_read_str): Likewise.
12182 (gen_memset_value_from_prev): New function.
12183 (builtin_memset_read_str): Change the mode argument from
12184 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
12185 and support CONST_VECTOR.
12186 (builtin_memset_gen_str): Likewise.
12187 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
12189 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
12190 with fixed_size_mode.
12191 (builtin_memset_read_str): Likewise.
12192 * expr.c (widest_int_mode_for_size): Renamed to ...
12193 (widest_fixed_size_mode_for_size): Add a bool argument to
12194 indicate if QI vector mode can be used.
12195 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
12196 instead of widest_int_mode_for_size.
12197 (pieces_addr::adjust): Change the mode argument from
12198 scalar_int_mode to fixed_size_mode.
12199 (op_by_pieces_d): Make m_len read-only. Add a bool member,
12200 m_qi_vector_mode, to indicate that QI vector mode can be used.
12201 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
12202 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
12203 instead of widest_int_mode_for_size.
12204 (op_by_pieces_d::get_usable_mode): Change the mode argument from
12205 scalar_int_mode to fixed_size_mode. Call
12206 widest_fixed_size_mode_for_size instead of
12207 widest_int_mode_for_size.
12208 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
12209 function to return the smallest integer or QI vector mode.
12210 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
12211 instead of widest_int_mode_for_size. Call
12212 smallest_fixed_size_mode_for_size instead of
12213 smallest_int_mode_for_size.
12214 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
12215 indicate that QI vector mode can be used and pass it to
12216 op_by_pieces_d::op_by_pieces_d.
12217 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
12218 instead of widest_int_mode_for_size. Pass memsetp to
12219 widest_fixed_size_mode_for_size to support QI vector mode.
12220 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
12221 (store_by_pieces): Pass memsetp to
12222 store_by_pieces_d::store_by_pieces_d.
12223 (clear_by_pieces_1): Removed.
12224 (clear_by_pieces): Replace clear_by_pieces_1 with
12225 builtin_memset_read_str and pass true to store_by_pieces_d to
12226 support vector mode broadcast.
12227 (string_cst_read_str): Change the mode argument from
12228 scalar_int_mode to fixed_size_mode.
12229 * expr.h (by_pieces_constfn): Change scalar_int_mode to
12231 (by_pieces_prev): Likewise.
12232 * rtl.h (lowpart_subreg_regno): New.
12233 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
12234 simplify_subreg_regno.
12235 * target.def (gen_memset_scratch_rtx): New hook.
12236 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
12237 * doc/tm.texi: Regenerated.
12239 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12242 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
12243 TARGET_EXPR instead of MODIFY_EXPR.
12245 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12248 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
12249 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
12250 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
12252 (vec_cmpu<IMSA:mode><mode_i>): New expander.
12254 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12257 * config/i386/i386-options.c (ix86_option_override_internal):
12258 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
12260 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12263 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
12264 explicitly dereferencing __a.
12265 (vld1_s64): Likewise.
12266 (vld1_u64): Likewise.
12267 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
12268 and change to VAR13.
12270 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
12272 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
12273 use of m_range_analyzer.
12274 (loop_versioning::lv_dom_walker::before_dom_children): Same.
12275 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
12276 (loop_versioning::prune_loop_conditions): Replace vr_values use
12277 with range_query interface.
12278 (pass_loop_versioning::execute): Use ranger.
12280 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12283 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
12284 enum values, and emit a warning if they mismatch.
12286 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
12288 * cfgloop.h (as_const): New function.
12289 (class loop_iterator): Rename to ...
12290 (class loops_list): ... this.
12291 (loop_iterator::next): Rename to ...
12292 (loops_list::Iter::fill_curr_loop): ... this and adjust.
12293 (loop_iterator::loop_iterator): Rename to ...
12294 (loops_list::loops_list): ... this and adjust.
12295 (loops_list::Iter): New class.
12296 (loops_list::iterator): New type.
12297 (loops_list::const_iterator): New type.
12298 (loops_list::begin): New function.
12299 (loops_list::end): Likewise.
12300 (loops_list::begin const): Likewise.
12301 (loops_list::end const): Likewise.
12302 (FOR_EACH_LOOP): Remove.
12303 (FOR_EACH_LOOP_FN): Remove.
12304 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
12305 for loop with loops_list instance.
12306 (sort_sibling_loops): Likewise.
12307 (disambiguate_loops_with_multiple_latches): Likewise.
12308 (verify_loop_structure): Likewise.
12309 * cfgloopmanip.c (create_preheaders): Likewise.
12310 (force_single_succ_latches): Likewise.
12311 * config/aarch64/falkor-tag-collision-avoidance.c
12312 (execute_tag_collision_avoidance): Likewise.
12313 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
12314 * config/s390/s390.c (s390_adjust_loops): Likewise.
12315 * doc/loop.texi: Likewise.
12316 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
12317 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
12318 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
12319 (loop_versioning::make_versioning_decisions): Likewise.
12320 * gimple-ssa-split-paths.c (split_paths): Likewise.
12321 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
12322 * graphite.c (canonicalize_loop_form): Likewise.
12323 (graphite_transform_loops): Likewise.
12324 * ipa-fnsummary.c (analyze_function_body): Likewise.
12325 * ipa-pure-const.c (analyze_function): Likewise.
12326 * loop-doloop.c (doloop_optimize_loops): Likewise.
12327 * loop-init.c (loop_optimizer_finalize): Likewise.
12328 (fix_loop_structure): Likewise.
12329 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
12330 (move_loop_invariants): Likewise.
12331 * loop-unroll.c (decide_unrolling): Likewise.
12332 (unroll_loops): Likewise.
12333 * modulo-sched.c (sms_schedule): Likewise.
12334 * predict.c (predict_loops): Likewise.
12335 (pass_profile::execute): Likewise.
12336 * profile.c (branch_prob): Likewise.
12337 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
12338 (sel_find_rgns): Likewise.
12339 * tree-cfg.c (replace_loop_annotate): Likewise.
12340 (replace_uses_by): Likewise.
12341 (move_sese_region_to_fn): Likewise.
12342 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
12343 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
12344 * tree-parloops.c (parallelize_loops): Likewise.
12345 * tree-predcom.c (tree_predictive_commoning): Likewise.
12346 * tree-scalar-evolution.c (scev_initialize): Likewise.
12347 (scev_reset): Likewise.
12348 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
12349 * tree-ssa-live.c (remove_unused_locals): Likewise.
12350 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
12351 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
12352 (tree_ssa_lim_initialize): Likewise.
12353 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
12354 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
12355 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
12356 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
12357 (free_numbers_of_iterations_estimates): Likewise.
12358 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
12359 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
12360 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
12361 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
12362 (pass_scev_cprop::execute): Likewise.
12363 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12364 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
12365 * tree-ssa-threadupdate.c
12366 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
12367 * tree-vectorizer.c (vectorize_loops): Likewise.
12368 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
12370 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
12372 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
12373 generating a VOIDmode register for e.g the
12374 function_arg_info::end_marker.
12376 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
12378 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
12379 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
12380 (cc mode_attr): Similarly.
12381 (ccz subst_attr): Similarly.
12382 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
12383 * config/h8300/testcompare.md: Remove various cc0 based patterns
12384 that had been commented out. Add pattern to set CCZ from a bit
12387 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
12388 Julian Brown <julian@codesourcery.com>
12389 Kwok Cheung Yeung <kcy@codesourcery.com>
12391 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
12392 'update_stmt' after modification.
12393 (pass_oacc_loop_designation): New function, extracted out of...
12394 (pass_oacc_device_lower): ... this.
12395 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
12396 (make_pass_oacc_loop_designation): New
12397 * passes.def: Add it.
12398 * tree-parloops.c (create_parallel_loop): Adjust.
12399 * tree-pass.h (make_pass_oacc_loop_designation): New.
12401 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
12403 * flag-types.h (enum threader_mode): New.
12404 * params.opt: Add entry for --param=threader-mode.
12405 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
12406 (class back_threader): New.
12407 (back_threader::back_threader): New.
12408 (back_threader::~back_threader): New.
12409 (back_threader::maybe_register_path): New.
12410 (back_threader::find_taken_edge): New.
12411 (back_threader::find_taken_edge_switch): New.
12412 (back_threader::find_taken_edge_cond): New.
12413 (back_threader::resolve_def): New.
12414 (back_threader::resolve_phi): New.
12415 (back_threader::find_paths_to_names): New.
12416 (back_threader::find_paths): New.
12419 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
12420 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
12421 (pass_thread_jumps::execute): Abstract out code...
12422 (try_thread_blocks): ...here.
12423 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
12424 Abstract out threading candidate code to...
12425 (single_succ_to_potentially_threadable_block): ...here.
12426 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
12428 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
12429 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
12430 Return bool from register_jump_thread.
12432 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
12434 * target.def: in0 and in1 do not need to be registers.
12435 * doc/tm.texi: Regenerate.
12437 2021-07-29 liuhongt <hongtao.liu@intel.com>
12440 * config/i386/i386.c (ix86_widen_mult_cost): New function.
12441 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
12444 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
12447 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
12448 (rs6000_preferred_doloop_mode): New hook.
12449 * doc/tm.texi: Regenerate.
12450 * doc/tm.texi.in: Add hook preferred_doloop_mode.
12451 * target.def (preferred_doloop_mode): New hook.
12452 * targhooks.c (default_preferred_doloop_mode): New hook.
12453 * targhooks.h (default_preferred_doloop_mode): New hook.
12454 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
12455 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
12456 and compute_doloop_base_on_mode.
12458 2021-07-28 Martin Sebor <msebor@redhat.com>
12460 PR middle-end/101494
12461 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
12462 and size computation.
12464 2021-07-28 Martin Sebor <msebor@redhat.com>
12466 PR middle-end/101601
12467 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
12469 Handle pointers to functions.
12471 2021-07-28 Martin Sebor <msebor@redhat.com>
12473 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
12474 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
12475 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
12476 (access_ref::access_ref): Same.
12477 (access_ref::phi): Same.
12478 (access_ref::get_ref): Same.
12479 (access_ref::size_remaining): Same.
12480 (access_ref::offset_in_range): Same.
12481 (access_ref::add_offset): Same.
12482 (access_ref::inform_access): Same.
12483 (ssa_name_limit_t::visit_phi): Same.
12484 (ssa_name_limit_t::leave_phi): Same.
12485 (ssa_name_limit_t::next): Same.
12486 (ssa_name_limit_t::next_phi): Same.
12487 (ssa_name_limit_t::~ssa_name_limit_t): Same.
12488 (pointer_query::pointer_query): Same.
12489 (pointer_query::get_ref): Same.
12490 (pointer_query::put_ref): Same.
12491 (pointer_query::flush_cache): Same.
12492 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
12493 (check_nul_terminated_array): Same.
12494 (unterminated_array): Same.
12495 (maybe_warn_for_bound): Same.
12496 (check_read_access): Same.
12497 (warn_for_access): Same.
12498 (get_size_range): Same.
12499 (check_access): Same.
12500 (gimple_call_alloc_size): Move to tree.c.
12501 (gimple_parm_array_size): Move to pointer-query.cc.
12502 (get_offset_range): Same.
12503 (gimple_call_return_array): Same.
12504 (handle_min_max_size): Same.
12505 (handle_array_ref): Same.
12506 (handle_mem_ref): Same.
12507 (compute_objsize): Same.
12508 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
12509 (call_dealloc_argno): Same.
12510 (fndecl_dealloc_argno): Same.
12511 (new_delete_mismatch_p): Same.
12512 (matching_alloc_calls_p): Same.
12513 (warn_dealloc_offset): Same.
12514 (maybe_emit_free_warning): Same.
12515 * builtins.h (check_nul_terminated_array): Move to
12516 gimple-ssa-warn-access.h.
12517 (check_nul_terminated_array): Same.
12518 (warn_string_no_nul): Same.
12519 (unterminated_array): Same.
12520 (class ssa_name_limit_t): Same.
12521 (class pointer_query): Same.
12522 (struct access_ref): Same.
12523 (class range_query): Same.
12524 (struct access_data): Same.
12525 (gimple_call_alloc_size): Same.
12526 (gimple_parm_array_size): Same.
12527 (compute_objsize): Same.
12528 (class access_data): Same.
12529 (maybe_emit_free_warning): Same.
12530 * calls.c (initialize_argument_information): Remove call to
12531 maybe_emit_free_warning.
12532 * gimple-array-bounds.cc: Include new header..
12533 * gimple-fold.c: Same.
12534 * gimple-ssa-sprintf.c: Same.
12535 * gimple-ssa-warn-restrict.c: Same.
12536 * passes.def: Add pass_warn_access.
12537 * tree-pass.h (make_pass_warn_access): Declare.
12538 * tree-ssa-strlen.c: Include new headers.
12539 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
12540 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
12541 * gimple-ssa-warn-access.cc: New file.
12542 * gimple-ssa-warn-access.h: New file.
12543 * pointer-query.cc: New file.
12544 * pointer-query.h: New file.
12546 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12548 PR middle-end/101624
12549 * ubsan.c (maybe_instrument_pointer_overflow,
12550 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
12551 PARM_DECLs or RESULT_DECLs.
12552 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
12554 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12556 PR middle-end/101642
12557 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
12558 to type of bswap16 for comparison.
12559 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
12561 2021-07-28 Richard Biener <rguenther@suse.de>
12563 PR tree-optimization/101615
12564 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
12565 at CTOR SLP graph entries.
12567 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12569 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
12570 Add "r,w" alternative.
12572 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
12575 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
12576 AVX_U128_DIRTY when all bits are zero.
12578 2021-07-28 Richard Biener <rguenther@suse.de>
12580 PR tree-optimization/101615
12581 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
12582 external nodes cannot be permuted so make them perm_out 0.
12584 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
12587 * config.in: Regenerate.
12588 * config/gcn/gcn-hsa.h (A_FIJI): New define.
12589 (A_900): New define.
12590 (A_906): New define.
12591 (A_908): New define.
12592 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
12593 * config/gcn/gcn.c (output_file_start): Adjust attributes according
12594 to the assembler capabilities.
12595 * config/gcn/mkoffload.c (main): Likewise.
12596 * configure: Regenerate.
12597 * configure.ac: Add tests for LLVM assembler attribute features.
12599 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
12601 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
12602 cond_false and cond_true on branches.
12604 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12606 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
12609 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12611 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
12613 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12616 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
12617 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
12618 and has special !TARGET_AVX512VL expansion.
12619 (vashrv2di3<mask_name>): Rename to ...
12620 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
12621 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
12623 2021-07-28 Martin Uecker <muecker@gwdg.de>
12625 * calls.c (maybe_warn_rdwr_sizes): Correct argument
12626 numbers in warning that were switched.
12628 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
12630 PR tree-optimization/101596
12631 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
12632 by using new_type's precision instead.
12634 2021-07-28 liuhongt <hongtao.liu@intel.com>
12637 * config/i386/i386.h (processor_costs): Add new member
12639 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
12640 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
12641 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
12642 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
12643 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
12644 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
12645 generic_cost, core_cost): Initialize integer_to_sse same value
12647 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
12648 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12649 Use integer_to_sse instead of sse_op to calculate the cost of
12652 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12654 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
12656 (write_init_file): Call write_ovld_static_init.
12658 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12660 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
12662 (write_init_file): Call write_bif_static_init.
12664 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12666 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
12667 (TYPE_MAP_SIZE): New macro.
12668 (type_map): New initialized variable.
12669 (typemap_cmp): New function.
12670 (write_type_node): Likewise.
12671 (write_fntype_init): Implement.
12673 2021-07-27 Martin Sebor <msebor@redhat.com>
12675 PR tree-optimization/101584
12676 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
12677 (check_defs): Call it.
12679 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12681 * tree-ssa-dom.c (dom_jump_threader_simplifier):
12682 Put avail_exprs_stack in the class, instead of passing it to
12683 jump_threader_simplifier.
12684 (dom_jump_threader_simplifier::simplify): Add state argument.
12685 (dom_opt_dom_walker): Add state.
12686 (pass_dominator::execute): Pass state to threader.
12687 (dom_opt_dom_walker::before_dom_children): Use state.
12688 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
12689 arguments by state.
12690 (jump_threader::record_temporary_equivalences_from_phis):
12691 Register equivalences through the state variable.
12692 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
12693 Record ranges in a statement through the state variable.
12694 (jump_threader::simplify_control_stmt_condition): Pass state to
12696 (jump_threader::simplify_control_stmt_condition_1): Same.
12697 (jump_threader::thread_around_empty_blocks): Remove obsolete
12699 (jump_threader::thread_through_normal_block): Record equivalences
12700 on edge through the state variable.
12701 (jump_threader::thread_across_edge): Abstract state pushing.
12702 (jt_state::jt_state): New.
12703 (jt_state::push): New.
12704 (jt_state::pop): New.
12705 (jt_state::register_equiv): New.
12706 (jt_state::record_ranges_from_stmt): New.
12707 (jt_state::register_equivs_on_edge): New.
12708 (jump_threader_simplifier::jump_threader_simplifier): Move from
12710 (jump_threader_simplifier::simplify): Add state argument.
12711 * tree-ssa-threadedge.h (class jt_state): New.
12712 (class jump_threader): Add state to constructor.
12713 (class jump_threader_simplifier): Add state to simplify. Remove
12714 avail_exprs_stack from class.
12715 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
12717 (vrp_jump_threader::vrp_jump_threader): Add state.
12718 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
12720 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12722 * Makefile.in (OBJS): Add gimple-range-path.o.
12723 * gimple-range-path.cc: New file.
12724 * gimple-range-path.h: New file.
12726 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
12728 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
12729 inside vec_duplicate for all patterns.
12730 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
12731 Push sign/zero-extension inside vec_duplicate.
12733 2021-07-27 Richard Biener <rguenther@suse.de>
12735 PR tree-optimization/101573
12736 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
12737 looking at uninitialized PHI arg defs in some constrained cases.
12738 (warn_uninitialized_vars): Call it.
12739 (execute_early_warn_uninitialized): Calculate dominators.
12741 2021-07-27 Richard Biener <rguenther@suse.de>
12743 PR tree-optimization/39821
12744 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
12745 vector_stmt for widening arithmetic.
12746 (vectorizable_conversion): Adjust.
12748 2021-07-27 Martin Jambor <mjambor@suse.cz>
12750 * cgraph.h (ipa_replace_map): New field force_load_ref.
12751 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
12752 aded new flag load_dereferenced, adjusted comments.
12753 (ipa_get_param_dereferenced): New function.
12754 (ipa_set_param_dereferenced): Likewise.
12755 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
12756 * ipa-cp.c: Include gimple.h.
12757 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
12758 (get_replacement_map): New parameter force_load_ref, set the
12759 appropriate flag in ipa_replace_map if set.
12760 (struct symbol_and_index_together): New type.
12761 (adjust_refs_in_act_callers): New function.
12762 (adjust_references_in_caller): Likewise.
12763 (create_specialized_node): When appropriate, call
12764 adjust_references_in_caller and force only load references.
12765 * ipa-prop.c (load_from_dereferenced_name): New function.
12766 (ipa_analyze_controlled_uses): Also detect loads from a
12767 dereference, harden testing of call statements.
12768 (ipa_write_node_info): Stream the dereferenced flag.
12769 (ipa_read_node_info): Likewise.
12770 (ipa_set_jf_constant): Also create refdesc when jump function
12771 references a variable.
12772 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
12773 also on references of variables and return a symtab_node. Adjust
12775 (propagate_controlled_uses): Also remove references to VAR_DECLs.
12777 2021-07-27 Jakub Jelinek <jakub@redhat.com>
12779 PR middle-end/101586
12780 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
12781 positions above or equal to sz except for diagnostics of flexible
12784 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
12786 PR tree-optimization/78888
12787 * gimple-range-fold.cc (get_letter_range): New.
12788 (fold_using_range::range_of_builtin_call): Call get_letter_range.
12790 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
12792 PR tree-optimization/78888
12793 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
12794 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
12796 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
12797 Marc Glisse <marc.glisse@inria.fr>
12799 * match.pd (rotate): Simplify equality/inequality of rotations.
12800 (bswap): Simplify equality/inequality tests of byte swapping.
12802 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12804 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
12807 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12809 * range-op.cc (operator_lshift::fold_range): Pass rel to
12810 base class fold_range.
12811 (operator_rshift::fold_range): Same.
12813 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
12816 * toplev.h (min_align_loops_log): Remove declaration.
12817 (min_align_jumps_log, min_align_labels_log): Likewise.
12818 (min_align_functions_log): Likewise.
12820 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12822 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
12824 (execute_vrp): Abstract out simplification of conditionals...
12825 (simplify_casted_conds): ...here.
12827 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12829 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
12830 Add gimple argument.
12831 (array_bounds_checker::check_array_ref): Same.
12832 (array_bounds_checker::check_addr_expr): Same.
12833 (array_bounds_checker::check_array_bounds): Pass statement to
12834 check_array_bounds and check_addr_expr.
12835 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
12836 (check_addr_expr): Same.
12837 (get_value_range): Same.
12839 2021-07-26 Tamar Christina <tamar.christina@arm.com>
12841 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
12842 (sdot_prod, udot_prod): ... This.
12843 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
12845 (<sur>dot_prod<vsi2qi>): ... this.
12846 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
12847 Change operands order.
12848 (<sur>sadv16qi): Use new operands order.
12849 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
12850 vdotq_s32): Use new RTL ordering.
12852 2021-07-26 Tamar Christina <tamar.christina@arm.com>
12854 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
12855 aarch64_types_ternop_suss_qualifiers): New.
12856 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
12857 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
12858 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
12860 2021-07-23 Jakub Jelinek <jakub@redhat.com>
12862 PR rtl-optimization/101562
12863 * expmed.c (store_integral_bit_field): Only use movstrict_optab
12864 if the operand isn't paradoxical.
12866 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
12868 * gimple-array-bounds.h (class array_bounds_checker): Change
12869 ranges type to range_query.
12871 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12873 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
12874 __builtin_memcpy instead of constructing
12875 __builtin_aarch64_simd_oi one vector at a time.
12876 (vst1_u64_x2): Likewise.
12877 (vst1_f64_x2): Likewise.
12878 (vst1_s8_x2): Likewise.
12879 (vst1_p8_x2): Likewise.
12880 (vst1_s16_x2): Likewise.
12881 (vst1_p16_x2): Likewise.
12882 (vst1_s32_x2): Likewise.
12883 (vst1_u8_x2): Likewise.
12884 (vst1_u16_x2): Likewise.
12885 (vst1_u32_x2): Likewise.
12886 (vst1_f16_x2): Likewise.
12887 (vst1_f32_x2): Likewise.
12888 (vst1_p64_x2): Likewise.
12889 (vst1q_s8_x2): Likewise.
12890 (vst1q_p8_x2): Likewise.
12891 (vst1q_s16_x2): Likewise.
12892 (vst1q_p16_x2): Likewise.
12893 (vst1q_s32_x2): Likewise.
12894 (vst1q_s64_x2): Likewise.
12895 (vst1q_u8_x2): Likewise.
12896 (vst1q_u16_x2): Likewise.
12897 (vst1q_u32_x2): Likewise.
12898 (vst1q_u64_x2): Likewise.
12899 (vst1q_f16_x2): Likewise.
12900 (vst1q_f32_x2): Likewise.
12901 (vst1q_f64_x2): Likewise.
12902 (vst1q_p64_x2): Likewise.
12904 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12906 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
12907 __builtin_memcpy instead of constructing
12908 __builtin_aarch64_simd_ci one vector at a time.
12909 (vst1_u64_x3): Likewise.
12910 (vst1_f64_x3): Likewise.
12911 (vst1_s8_x3): Likewise.
12912 (vst1_p8_x3): Likewise.
12913 (vst1_s16_x3): Likewise.
12914 (vst1_p16_x3): Likewise.
12915 (vst1_s32_x3): Likewise.
12916 (vst1_u8_x3): Likewise.
12917 (vst1_u16_x3): Likewise.
12918 (vst1_u32_x3): Likewise.
12919 (vst1_f16_x3): Likewise.
12920 (vst1_f32_x3): Likewise.
12921 (vst1_p64_x3): Likewise.
12922 (vst1q_s8_x3): Likewise.
12923 (vst1q_p8_x3): Likewise.
12924 (vst1q_s16_x3): Likewise.
12925 (vst1q_p16_x3): Likewise.
12926 (vst1q_s32_x3): Likewise.
12927 (vst1q_s64_x3): Likewise.
12928 (vst1q_u8_x3): Likewise.
12929 (vst1q_u16_x3): Likewise.
12930 (vst1q_u32_x3): Likewise.
12931 (vst1q_u64_x3): Likewise.
12932 (vst1q_f16_x3): Likewise.
12933 (vst1q_f32_x3): Likewise.
12934 (vst1q_f64_x3): Likewise.
12935 (vst1q_p64_x3): Likewise.
12937 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
12940 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
12941 hard register when LRA is in progress.
12943 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12945 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
12946 __builtin_memcpy instead of using a union.
12947 (vst1q_s8_x4): Likewise.
12948 (vst1_s16_x4): Likewise.
12949 (vst1q_s16_x4): Likewise.
12950 (vst1_s32_x4): Likewise.
12951 (vst1q_s32_x4): Likewise.
12952 (vst1_u8_x4): Likewise.
12953 (vst1q_u8_x4): Likewise.
12954 (vst1_u16_x4): Likewise.
12955 (vst1q_u16_x4): Likewise.
12956 (vst1_u32_x4): Likewise.
12957 (vst1q_u32_x4): Likewise.
12958 (vst1_f16_x4): Likewise.
12959 (vst1q_f16_x4): Likewise.
12960 (vst1_f32_x4): Likewise.
12961 (vst1q_f32_x4): Likewise.
12962 (vst1_p8_x4): Likewise.
12963 (vst1q_p8_x4): Likewise.
12964 (vst1_p16_x4): Likewise.
12965 (vst1q_p16_x4): Likewise.
12966 (vst1_s64_x4): Likewise.
12967 (vst1_u64_x4): Likewise.
12968 (vst1_p64_x4): Likewise.
12969 (vst1q_s64_x4): Likewise.
12970 (vst1q_u64_x4): Likewise.
12971 (vst1q_p64_x4): Likewise.
12972 (vst1_f64_x4): Likewise.
12973 (vst1q_f64_x4): Likewise.
12975 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
12977 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
12978 instead of constructing __builtin_aarch64_simd_oi one vector
12980 (vst2_u64): Likewise.
12981 (vst2_f64): Likewise.
12982 (vst2_s8): Likewise.
12983 (vst2_p8): Likewise.
12984 (vst2_s16): Likewise.
12985 (vst2_p16): Likewise.
12986 (vst2_s32): Likewise.
12987 (vst2_u8): Likewise.
12988 (vst2_u16): Likewise.
12989 (vst2_u32): Likewise.
12990 (vst2_f16): Likewise.
12991 (vst2_f32): Likewise.
12992 (vst2_p64): Likewise.
12993 (vst2q_s8): Likewise.
12994 (vst2q_p8): Likewise.
12995 (vst2q_s16): Likewise.
12996 (vst2q_p16): Likewise.
12997 (vst2q_s32): Likewise.
12998 (vst2q_s64): Likewise.
12999 (vst2q_u8): Likewise.
13000 (vst2q_u16): Likewise.
13001 (vst2q_u32): Likewise.
13002 (vst2q_u64): Likewise.
13003 (vst2q_f16): Likewise.
13004 (vst2q_f32): Likewise.
13005 (vst2q_f64): Likewise.
13006 (vst2q_p64): Likewise.
13008 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13010 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
13011 instead of constructing __builtin_aarch64_simd_ci one vector
13013 (vst3_u64): Likewise.
13014 (vst3_f64): Likewise.
13015 (vst3_s8): Likewise.
13016 (vst3_p8): Likewise.
13017 (vst3_s16): Likewise.
13018 (vst3_p16): Likewise.
13019 (vst3_s32): Likewise.
13020 (vst3_u8): Likewise.
13021 (vst3_u16): Likewise.
13022 (vst3_u32): Likewise.
13023 (vst3_f16): Likewise.
13024 (vst3_f32): Likewise.
13025 (vst3_p64): Likewise.
13026 (vst3q_s8): Likewise.
13027 (vst3q_p8): Likewise.
13028 (vst3q_s16): Likewise.
13029 (vst3q_p16): Likewise.
13030 (vst3q_s32): Likewise.
13031 (vst3q_s64): Likewise.
13032 (vst3q_u8): Likewise.
13033 (vst3q_u16): Likewise.
13034 (vst3q_u32): Likewise.
13035 (vst3q_u64): Likewise.
13036 (vst3q_f16): Likewise.
13037 (vst3q_f32): Likewise.
13038 (vst3q_f64): Likewise.
13039 (vst3q_p64): Likewise.
13041 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13043 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
13044 instead of constructing __builtin_aarch64_simd_xi one vector
13046 (vst4_u64): Likewise.
13047 (vst4_f64): Likewise.
13048 (vst4_s8): Likewise.
13049 (vst4_p8): Likewise.
13050 (vst4_s16): Likewise.
13051 (vst4_p16): Likewise.
13052 (vst4_s32): Likewise.
13053 (vst4_u8): Likewise.
13054 (vst4_u16): Likewise.
13055 (vst4_u32): Likewise.
13056 (vst4_f16): Likewise.
13057 (vst4_f32): Likewise.
13058 (vst4_p64): Likewise.
13059 (vst4q_s8): Likewise.
13060 (vst4q_p8): Likewise.
13061 (vst4q_s16): Likewise.
13062 (vst4q_p16): Likewise.
13063 (vst4q_s32): Likewise.
13064 (vst4q_s64): Likewise.
13065 (vst4q_u8): Likewise.
13066 (vst4q_u16): Likewise.
13067 (vst4q_u32): Likewise.
13068 (vst4q_u64): Likewise.
13069 (vst4q_f16): Likewise.
13070 (vst4q_f32): Likewise.
13071 (vst4q_f64): Likewise.
13072 (vst4q_p64): Likewise.
13074 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13076 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
13077 instead of constructing __builtin_aarch64_simd_oi one vector
13079 (vtbx4_u8): Likewise.
13080 (vtbx4_p8): Likewise.
13082 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13084 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
13085 instead of constructing __builtin_aarch64_simd_oi one vector
13087 (vtbl3_u8): Likewise.
13088 (vtbl3_p8): Likewise.
13089 (vtbl4_s8): Likewise.
13090 (vtbl4_u8): Likewise.
13091 (vtbl4_p8): Likewise.
13093 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13095 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
13096 instead of constructing __builtin_aarch64_simd_oi one vector
13098 (vqtbx2_u8): Likewise.
13099 (vqtbx2_p8): Likewise.
13100 (vqtbx2q_s8): Likewise.
13101 (vqtbx2q_u8): Likewise.
13102 (vqtbx2q_p8): Likewise.
13103 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
13104 __builtin_aarch64_simd_ci one vector at a time.
13105 (vqtbx3_u8): Likewise.
13106 (vqtbx3_p8): Likewise.
13107 (vqtbx3q_s8): Likewise.
13108 (vqtbx3q_u8): Likewise.
13109 (vqtbx3q_p8): Likewise.
13110 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
13111 __builtin_aarch64_simd_xi one vector at a time.
13112 (vqtbx4_u8): Likewise.
13113 (vqtbx4_p8): Likewise.
13114 (vqtbx4q_s8): Likewise.
13115 (vqtbx4q_u8): Likewise.
13116 (vqtbx4q_p8): Likewise.
13118 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13120 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
13121 instead of constructing __builtin_aarch64_simd_oi one vector
13123 (vqtbl2_u8): Likewise.
13124 (vqtbl2_p8): Likewise.
13125 (vqtbl2q_s8): Likewise.
13126 (vqtbl2q_u8): Likewise.
13127 (vqtbl2q_p8): Likewise.
13128 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
13129 __builtin_aarch64_simd_ci one vector at a time.
13130 (vqtbl3_u8): Likewise.
13131 (vqtbl3_p8): Likewise.
13132 (vqtbl3q_s8): Likewise.
13133 (vqtbl3q_u8): Likewise.
13134 (vqtbl3q_p8): Likewise.
13135 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
13136 __builtin_aarch64_simd_xi one vector at a time.
13137 (vqtbl4_u8): Likewise.
13138 (vqtbl4_p8): Likewise.
13139 (vqtbl4q_s8): Likewise.
13140 (vqtbl4q_u8): Likewise.
13141 (vqtbl4q_p8): Likewise.
13143 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
13146 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
13148 2021-07-22 Andrew Pinski <apinski@marvell.com>
13150 PR tree-optimization/10153
13151 * tree-tailcall.c (create_tailcall_accumulator):
13152 Don't call fold_convert as the type should be correct already.
13153 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
13154 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
13156 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
13158 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
13159 varying_p check for null/non-null check.
13161 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13163 PR tree-optimization/101511
13164 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
13165 is in ssa2's equiv set, and don't trap if so.
13167 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13169 PR tree-optimization/101497
13170 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
13173 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13175 PR tree-optimization/101496
13176 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
13177 first, then vrp_visit_cond_Stmt.
13179 2021-07-22 liuhongt <hongtao.liu@intel.com>
13181 * config/i386/i386-expand.c
13182 (ix86_broadcast_from_integer_constant): Rename to ..
13183 (ix86_broadcast_from_constant): .. this, and extend it to
13185 (ix86_expand_vector_move): Extend to float mode.
13186 * config/i386/i386-features.c
13187 (replace_constant_pool_with_broadcast): Remove.
13188 (remove_partial_avx_dependency_gate): Ditto.
13189 (constant_pool_broadcast): Ditto.
13190 (class pass_constant_pool_broadcast): Ditto.
13191 (make_pass_constant_pool_broadcast): Ditto.
13192 (remove_partial_avx_dependency): Adjust gate.
13193 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
13194 * config/i386/i386-protos.h
13195 (make_pass_constant_pool_broadcast): Remove.
13197 2021-07-22 liuhongt <hongtao.liu@intel.com>
13199 * config/i386/constraints.md (Wb): New constraint.
13201 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
13203 (*ashlqi3_1): Ditto.
13204 (*<insn><mode>3_1): Split to ..
13205 (*ashr<mode>3_1): this, ...
13206 (*lshr<mode>3_1): and this, also extend this pattern to avx512
13208 (*<insn><mode>3_1): Split to ..
13209 (*ashr<mode>3_1): this, ...
13210 (*lshrqi3_1): and this, also extend this pattern to avx512
13212 (*lshrhi3_1): And this, also extend this pattern to avx512
13214 * config/i386/sse.md (k<code><mode>): New define_split after
13215 it to convert generic shift pattern to mask shift ones.
13217 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
13218 Joseph Myers <joseph@codesourcery.com>
13219 Cesar Philippidis <cesar@codesourcery.com>
13221 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
13222 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
13224 * tree-pretty-print.c (dump_omp_clause): Likewise.
13225 * omp-general.c (oacc_verify_routine_clauses): Likewise.
13226 * gimplify.c (gimplify_scan_omp_clauses)
13227 (gimplify_adjust_omp_clauses): Likewise.
13228 * tree-nested.c (convert_nonlocal_omp_clauses)
13229 (convert_local_omp_clauses): Likewise.
13230 * omp-low.c (scan_sharing_clauses): Likewise.
13231 * omp-offload.c (execute_oacc_device_lower): Update.
13233 2021-07-21 Martin Sebor <msebor@redhat.com>
13235 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
13237 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13239 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
13242 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13244 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
13246 (write_fntype_init): New stub function.
13247 (write_init_bif_table): Likewise.
13248 (write_init_ovld_table): New function.
13249 (write_init_file): Implement.
13251 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13253 * config/rs6000/rs6000-gen-builtins.c
13254 (write_autogenerated_header): New function.
13255 (write_decls): Likewise.
13256 (write_extern_fntype): New callback function.
13257 (write_header_file): Implement.
13259 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13261 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
13264 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13266 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
13268 (complete_base_type): Likewise.
13269 (construct_fntype_id): Likewise.
13270 (parse_bif_entry): Call contruct_fntype_id.
13271 (parse_ovld_entry): Likewise.
13273 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13275 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
13276 (MAXOVLDSTANZAS): New macro.
13277 (ovld_stanzas): New variable.
13278 (curr_ovld_stanza): Likewise.
13279 (MAXOVLDS): New macro.
13280 (ovlddata): New struct.
13281 (ovlds): New variable.
13282 (curr_ovld): Likewise.
13283 (max_ovld_args): Likewise.
13284 (parse_ovld_entry): New function.
13285 (parse_ovld_stanza): Likewise.
13286 (parse_ovld): Implement.
13288 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13290 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
13293 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13295 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
13296 (parse_prototype): Implement.
13298 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13300 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
13301 (curr_bif_stanza): New variable.
13302 (stanza_entry): New struct.
13303 (stanza_map): New initialized variable.
13304 (enable_string): Likewise.
13305 (fnkinds): New enum.
13306 (typelist): New struct.
13307 (attrinfo): Likewise.
13308 (MAXRESTROPNDS): New macro.
13309 (prototype): New struct.
13310 (MAXBIFS): New macro.
13311 (bifdata): New struct.
13312 (bifs): New variable.
13313 (curr_bif): Likewise.
13314 (bif_order): Likewise.
13315 (bif_index): Likewise.
13316 (fatal): New function.
13317 (stanza_name_to_stanza): Likewise.
13318 (parse_bif_attrs): New stub function.
13319 (parse_prototype): Likewise.
13320 (parse_bif_entry): New function.
13321 (parse_bif_stanza): Likewise.
13322 (parse_bif): Implement.
13323 (set_bif_order): New function.
13324 (create_bif_order): Implement.
13326 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13328 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
13329 (num_bifs): New variable.
13330 (num_ovld_stanzas): Likewise.
13331 (num_ovlds): Likewise.
13332 (parse_codes): New enum.
13333 (bif_rbt): New variable.
13334 (ovld_rbt): Likewise.
13335 (fntype_rbt): Likewise.
13336 (bifo_rbt): Likewise.
13337 (parse_bif): New stub function.
13338 (create_bif_order): Likewise.
13339 (parse_ovld): Likewise.
13340 (write_header_file): Likewise.
13341 (write_init_file): Likewise.
13342 (write_defines_file): Likewise.
13343 (delete_output_files): New function.
13346 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
13349 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
13350 from CRC32 _builtin functions.
13352 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
13354 * coverage.c (build_gcov_info_var_registration): Mark the object placed
13355 in the linker set as referenced so that it does not get optimized away.
13357 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
13360 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13362 * config.gcc (riscv*-*-*): Detect which python is available.
13364 2021-07-21 Jakub Jelinek <jakub@redhat.com>
13366 PR middle-end/101535
13367 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
13368 contexts in which decl isn't privatized and for ORT_TARGET return
13369 false if decl is mapped.
13371 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13373 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
13374 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
13375 has been successfully unrolled.
13377 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13379 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
13380 exit condition if no tail loop is needed, and if the original exit
13381 condition should therefore be kept as-is.
13382 (tree_transform_and_unroll_loop): Handle that case here too.
13384 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
13386 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
13388 (free_data_refs): Likewise.
13389 * tree-data-ref.h (free_dependence_relations): Likewise.
13390 (free_data_refs): Likewise.
13391 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
13393 (struct component): Likewise.
13394 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
13395 (pcom_worker::~pcom_worker): Likewise.
13396 (pcom_worker::release_chain): Adjust as auto_vec changes.
13397 (pcom_worker::loop): Rename to ...
13398 (pcom_worker::m_loop): ... this.
13399 (pcom_worker::datarefs): Rename to ...
13400 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
13401 (pcom_worker::dependences): Rename to ...
13402 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
13403 (pcom_worker::chains): Rename to ...
13404 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
13405 (pcom_worker::looparound_phis): Rename to ...
13406 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
13408 (pcom_worker::cache): Rename to ...
13409 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
13410 (pcom_worker::release_chain): Adjust for auto_vec changes.
13411 (pcom_worker::release_chains): Adjust for auto_vec and renaming
13413 (release_component): Remove.
13414 (release_components): Adjust for release_component removal.
13415 (component_of): Adjust to use vec.
13416 (merge_comps): Likewise.
13417 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
13418 (pcom_worker::determine_offset): Likewise.
13419 (class comp_ptrs): Remove.
13420 (pcom_worker::split_data_refs_to_components): Adjust for renaming
13421 changes, for comp_ptrs removal with auto_vec.
13422 (pcom_worker::suitable_component_p): Adjust for renaming changes.
13423 (pcom_worker::filter_suitable_components): Adjust for release_component
13425 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
13426 (pcom_worker::find_looparound_phi): Likewise.
13427 (pcom_worker::add_looparound_copies): Likewise.
13428 (pcom_worker::determine_roots_comp): Likewise.
13429 (pcom_worker::single_nonlooparound_use): Likewise.
13430 (pcom_worker::execute_pred_commoning_chain): Likewise.
13431 (pcom_worker::execute_pred_commoning): Likewise.
13432 (pcom_worker::try_combine_chains): Likewise.
13433 (pcom_worker::prepare_initializers_chain): Likewise.
13434 (pcom_worker::prepare_initializers): Likewise.
13435 (pcom_worker::prepare_finalizers_chain): Likewise.
13436 (pcom_worker::prepare_finalizers): Likewise.
13437 (pcom_worker::tree_predictive_commoning_loop): Likewise.
13439 2021-07-20 Martin Sebor <msebor@redhat.com>
13441 PR middle-end/101397
13442 * builtins.c (gimple_call_return_array): Add argument. Correct
13443 offsets for memchr, mempcpy, stpcpy, and stpncpy.
13444 (compute_objsize_r): Adjust offset computation for argument returning
13447 2021-07-20 Martin Sebor <msebor@redhat.com>
13449 PR middle-end/101300
13450 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
13452 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
13454 * function.c (assign_parm_setup_block): Use adjust_address instead
13455 of change_address to preserve MEM_EXPR and friends.
13457 2021-07-20 Martin Sebor <msebor@redhat.com>
13459 * cfgloop.h (single_likely_exit): Adjust by-value argument to
13460 by-const-reference.
13461 * cfgloopanal.c (single_likely_exit): Same.
13462 * cgraph.h (struct cgraph_node): Same.
13463 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
13464 * genautomata.c (merge_states): Same.
13465 * genextract.c (VEC_char_to_string): Same.
13466 * genmatch.c (dt_node::gen_kids_1): Same.
13467 (walk_captures): Adjust by-value argument to by-reference.
13468 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
13469 to by-const-reference.
13470 * gimple.c (gimple_build_call_vec): Same.
13471 (gimple_build_call_internal_vec): Same.
13472 (gimple_build_switch): Same.
13473 (sort_case_labels): Same.
13474 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13476 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
13477 by-const-reference.
13478 (gimple_build_call_internal_vec): Same.
13479 (gimple_build_switch): Same.
13480 (sort_case_labels): Same.
13481 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13483 * haifa-sched.c (calc_priorities): Adjust by-value argument to
13484 by-const-reference.
13485 (sched_init_luids): Same.
13486 (haifa_init_h_i_d): Same.
13487 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
13488 (adjust_callers_for_value_intersection): Adjust by-value argument to
13490 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
13491 by-const-reference.
13492 (find_more_contexts_for_caller_subset): Same.
13493 (find_aggregate_values_for_callers_subset): Same.
13494 (copy_useful_known_contexts): Same.
13495 * ipa-fnsummary.c (remap_edge_summaries): Same.
13496 (remap_freqcounting_predicate): Same.
13497 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
13499 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
13500 to by-const-reference.
13501 * ipa-predicate.h (predicate::remap_after_inlining): Same.
13502 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
13503 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
13504 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
13505 * read-rtl.c (add_overload_instance): Same.
13506 * rtl.h (native_decode_rtx): Same.
13507 (native_decode_vector_rtx): Same.
13508 * sched-int.h (sched_init_luids): Same.
13509 (haifa_init_h_i_d): Same.
13510 * simplify-rtx.c (native_decode_vector_rtx): Same.
13511 (native_decode_rtx): Same.
13512 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
13513 (shrink_wrap_one_built_in_call_with_conds): Same.
13514 (shrink_wrap_conditional_dead_built_in_calls): Same.
13515 * tree-data-ref.c (create_runtime_alias_checks): Same.
13516 (compute_all_dependences): Same.
13517 * tree-data-ref.h (compute_all_dependences): Same.
13518 (create_runtime_alias_checks): Same.
13519 (index_in_loop_nest): Same.
13520 * tree-if-conv.c (mask_exists): Same.
13521 * tree-loop-distribution.c (class loop_distribution): Same.
13522 (loop_distribution::create_rdg_vertices): Same.
13523 (dump_rdg_partitions): Same.
13524 (debug_rdg_partitions): Same.
13525 (partition_contains_all_rw): Same.
13526 (loop_distribution::distribute_loop): Same.
13527 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
13528 (oacc_entry_exit_single_gang): Same.
13529 * tree-ssa-loop-im.c (hoist_memory_references): Same.
13530 (loop_suitable_for_sm): Same.
13531 * tree-ssa-loop-niter.c (bound_index): Same.
13532 * tree-ssa-reassoc.c (update_ops): Same.
13533 (swap_ops_for_binary_stmt): Same.
13534 (rewrite_expr_tree): Same.
13535 (rewrite_expr_tree_parallel): Same.
13536 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
13537 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
13538 * tree-ssa-structalias.c (process_all_all_constraints): Same.
13539 (make_constraints_to): Same.
13540 (handle_lhs_call): Same.
13541 (find_func_aliases_for_builtin_call): Same.
13542 (sort_fieldstack): Same.
13543 (check_for_overlaps): Same.
13544 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
13545 (vect_create_cond_for_unequal_addrs): Same.
13546 (vect_create_cond_for_lower_bounds): Same.
13547 (vect_create_cond_for_alias_checks): Same.
13548 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
13549 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
13550 (vect_make_slp_decision): Same.
13551 (vect_slp_bbs): Same.
13552 (duplicate_and_interleave): Same.
13553 (vect_transform_slp_perm_load): Same.
13554 (vect_schedule_slp): Same.
13555 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
13556 (vect_schedule_slp): Same.
13557 (duplicate_and_interleave): Same.
13558 * tree.c (build_vector_from_ctor): Same.
13559 (build_vector): Same.
13560 (check_vector_cst): Same.
13561 (check_vector_cst_duplicate): Same.
13562 (check_vector_cst_fill): Same.
13563 (check_vector_cst_stepped): Same.
13564 * tree.h (build_vector_from_ctor): Same.
13566 2021-07-20 Jakub Jelinek <jakub@redhat.com>
13569 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
13570 type from bool to int.
13571 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
13572 EASY_VECTOR_MSB case if either step or copies is not 1.
13573 (vspltis_shifted): Fix comment typo.
13574 (easy_altivec_constant): Change return type from bool to int, instead
13575 of returning true return byte size of the element mode that should be
13576 used to synthetize the constant.
13577 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
13578 that vspltis_shifted is 0, handle the case where easy_altivec_constant
13579 assumes using different vector mode from CONST_VECTOR's mode.
13580 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
13581 easy_altivec_constant to determine mode in which -1 >> -1 should be
13582 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
13584 2021-07-20 Richard Biener <rguenther@suse.de>
13587 * dwarf2out.h (dwarf_file_data): Add key member.
13588 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
13589 (dwarf_file_hasher::hash): Hash key.
13590 (lookup_filename): Remap the filename and store it in the
13591 filename member of dwarf_file_data when creating a new
13593 (file_name_acquire): Do not remap the filename again.
13594 (maybe_emit_file): Likewise.
13596 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
13598 * config/aarch64/aarch64-simd-builtins.def: Use two variant
13599 generators for all TBL/TBX intrinsics and rename to
13600 consistent forms: qtbl[1234] or qtbx[1234].
13601 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
13603 (aarch64_qtbl1<mode>): This.
13604 (aarch64_tbx1<mode>): Rename to...
13605 (aarch64_qtbx1<mode>): This.
13606 (aarch64_tbl2v16qi): Delete.
13607 (aarch64_tbl3<mode>): Rename to...
13608 (aarch64_qtbl2<mode>): This.
13609 (aarch64_tbx4<mode>): Rename to...
13610 (aarch64_qtbx2<mode>): This.
13611 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
13612 renamed qtbl1 and qtbl2 RTL patterns.
13613 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
13615 (vqtbl1_s8): Likewise.
13616 (vqtbl1_u8): Likewise.
13617 (vqtbl1q_p8): Likewise.
13618 (vqtbl1q_s8): Likewise.
13619 (vqtbl1q_u8): Likewise.
13620 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
13621 (vqtbx1_u8): Likewise.
13622 (vqtbx1_p8): Likewise.
13623 (vqtbx1q_s8): Likewise.
13624 (vqtbx1q_u8): Likewise.
13625 (vqtbx1q_p8): Likewise.
13626 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
13627 (vtbl1_u8): Likewise.
13628 (vtbl1_p8): Likewise.
13629 (vtbl2_s8): Likewise
13630 (vtbl2_u8): Likewise.
13631 (vtbl2_p8): Likewise.
13632 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
13633 (vtbl3_u8): Likewise.
13634 (vtbl3_p8): Likewise.
13635 (vtbl4_s8): Likewise.
13636 (vtbl4_u8): Likewise.
13637 (vtbl4_p8): Likewise.
13638 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
13639 (vtbx2_u8): Likewise.
13640 (vtbx2_p8): Likewise.
13641 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
13642 (vqtbl2_u8): Likewise.
13643 (vqtbl2_p8): Likewise.
13644 (vqtbl2q_s8): Likewise.
13645 (vqtbl2q_u8): Likewise.
13646 (vqtbl2q_p8): Likewise.
13647 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
13648 (vqtbx2_u8): Likewise.
13649 (vqtbx2_p8): Likewise.
13650 (vqtbx2q_s8): Likewise.
13651 (vqtbx2q_u8): Likewise.
13652 (vqtbx2q_p8): Likewise.
13653 (vtbx4_s8): Likewise.
13654 (vtbx4_u8): Likewise.
13655 (vtbx4_p8): Likewise.
13657 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
13660 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
13662 (define_peephole2 atomic_loaddi_fpu): Ditto.
13664 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13666 * config.gcc (riscv*-*-*): Detect which python is available.
13668 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13670 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
13671 (smul<mode>3_highpart): ... this.
13672 (mulhu_<mode>): Rename to...
13673 (umul<mode>3_highpart): ... this.
13674 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
13675 MULHU_V2DI, MULHU_V4SI): Adjust.
13677 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13679 PR tree-optimization/100696
13680 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
13681 * internal-fn.def (IFN_MULH): New internal function.
13682 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
13683 recog normal multiply highpart as IFN_MULH.
13684 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
13687 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13689 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
13690 (BTF_DEBUGGING_INFO): Likewise.
13691 * doc/tm.texi.in: Document the new macros.
13692 * doc/tm.texi: Regenerated.
13693 * toplev.c: Guard initialization of debug hooks.
13695 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13697 * flags.h (ctf_debuginfo_p): New function declaration.
13698 * opts.c (ctf_debuginfo_p): New function definition.
13700 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
13703 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
13704 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
13705 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
13706 * config/gcn/gcn-valu.md: Add a warning comment.
13707 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
13708 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
13709 * config/gcn/gcn.md: Add a warning comment.
13710 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
13711 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
13713 (EF_AMDGPU_XNACK): New.
13714 (EF_AMDGPU_SRAM_ECC): New.
13716 (copy_early_debug_info): Use elf_flags.
13717 (main): Handle -mxnack and -msram-ecc options.
13718 * doc/invoke.texi: Document -mxnack and -msram-ecc.
13720 2021-07-19 Andrew Pinski <apinski@marvell.com>
13723 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
13724 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
13726 2021-07-19 Richard Biener <rguenther@suse.de>
13728 PR tree-optimization/101505
13729 * tree-vect-patterns.c (vect_determine_precisions): Walk
13730 PHIs also for loop vectorization.
13732 2021-07-19 Richard Biener <rguenther@suse.de>
13734 * gimple.h (gimple_expr_type): Remove.
13735 * doc/gimple.texi: Remove gimple_expr_type documentation.
13737 2021-07-19 Richard Biener <rguenther@suse.de>
13739 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
13740 (ao_ref_init_from_vn_reference): Likewise.
13741 (fully_constant_reference): Likewise.
13742 (vn_reference_lookup_call): Do not set vr->type to random
13744 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
13746 * tree-vect-generic.c (expand_vector_piecewise): Pass in
13747 whether we expanded parallel.
13748 (expand_vector_parallel): Adjust.
13749 (expand_vector_addition): Likewise.
13750 (expand_vector_comparison): Likewise.
13751 (expand_vector_operation): Likewise.
13752 (expand_vector_scalar_condition): Likewise.
13753 (expand_vector_conversion): Likewise.
13755 2021-07-19 Richard Biener <rguenther@suse.de>
13757 * tree-vrp.c (register_edge_assert_for_2): Use the
13759 (vrp_folder::fold_predicate_in): Likewise.
13760 * vr-values.c (gimple_assign_nonzero_p): Likewise.
13761 (vr_values::extract_range_from_comparison): Likewise.
13762 (vr_values::extract_range_from_ubsan_builtin): Use the
13763 type of the first operand.
13764 (vr_values::extract_range_basic): Push down type
13765 computation, use the appropriate LHS.
13766 (vr_values::extract_range_from_assignment): Use the
13769 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
13772 * common/config/i386/i386-common.c (ix86_handle_option): For
13773 -mgeneral-regs-only, enable the GPR only instructions which are
13774 enabled implicitly by SSE ISAs unless they have been disabled
13777 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
13780 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
13781 ix86_avx_u128_mode_needed.
13782 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
13783 returns AVX register.
13785 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
13787 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
13788 (handle_const_call): Liekise
13789 (handle_pure_call): Liekise
13791 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
13793 PR tree-optimization/96542
13794 * range-op.cc (range_operator::wi_fold_in_parts): New.
13795 (range_operator::fold_range): Call wi_fold_in_parts.
13796 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
13797 * range-op.h (wi_fold_in_parts): Add prototype.
13799 2021-07-16 David Malcolm <dmalcolm@redhat.com>
13801 * doc/analyzer.texi: Add __analyzer_dump_state.
13803 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13805 * config/rs6000/rbtree.c: New file.
13806 * config/rs6000/rbtree.h: New file.
13808 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13810 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
13811 (typeinfo): Add restr field.
13812 (match_bracketed_pair): New function.
13813 (match_const_restriction): Implement.
13815 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13817 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
13819 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13821 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
13822 (basetype): Likewise.
13823 (typeinfo): Likewise.
13824 (handle_pointer): New function.
13825 (match_basetype): New stub function.
13826 (match_const_restriction): Likewise.
13827 (match_type): New function.
13829 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13831 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
13833 (advance_line): Likewise.
13834 (safe_inc_pos): Likewise.
13835 (match_identifier): Likewise.
13836 (match_integer): Likewise.
13837 (match_to_right_bracket): Likewise.
13839 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13841 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
13842 (ovld_file): Likewise.
13843 (header_file): Likewise.
13844 (init_file): Likewise.
13845 (defines_file): Likewise.
13846 (pgm_path): Likewise.
13847 (bif_path): Likewise.
13848 (ovld_path): Likewise.
13849 (header_path): Likewise.
13850 (init_path): Likewise.
13851 (defines_path): Likewise.
13852 (LINELEN): New macro.
13853 (linebuf): New variable.
13857 (bif_diag): New function.
13858 (ovld_diag): Likewise.
13860 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13862 * config/rs6000/rs6000-builtin-new.def: New.
13863 * config/rs6000/rs6000-overload.def: New.
13865 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13867 * config/rs6000/rs6000-gen-builtins.c: New.
13869 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13871 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
13872 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
13873 * gengtype-state.c (state_writer::write_state_file_list): Add a
13874 parameter to the fileslist expression for the number of build
13876 (read_state_files_list): Detect build headers and strip the
13877 initial "./" or ".\" from their names.
13878 * gengtype.c (build_headers): New global variable.
13879 (num_build_headers): Likewise.
13880 (open_base_files): Emit #include for each build header.
13881 (main): Detect and count build headers.
13882 * gengtype.h (build_headers): New extern variable.
13883 (num_build_headers): Likewise.
13885 2021-07-16 Richard Biener <rguenther@suse.de>
13887 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
13888 the type of the LHS.
13889 (find_bswap_or_nop_1): Likewise.
13890 (find_bswap_or_nop): Likewise.
13891 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
13893 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13894 Remove unused parameters, pass in the scalar type. Fix
13895 internal store function handling.
13896 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
13897 (vect_get_vector_types_for_stmt): Move down check for
13898 existing vector stmt after we've determined a scalar type.
13899 Pass down the used scalar type to vect_get_smallest_scalar_type.
13900 * tree-vect-generic.c (expand_vector_condition): Use
13901 the type of the LHS.
13902 (expand_vector_scalar_condition): Likewise.
13903 (expand_vector_operations_1): Likewise.
13904 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
13905 (vect_recog_dot_prod_pattern): Likewise.
13906 (vect_recog_sad_pattern): Likewise.
13907 (vect_recog_widen_op_pattern): Likewise.
13908 (vect_recog_widen_sum_pattern): Likewise.
13909 (vect_recog_mixed_size_cond_pattern): Likewise.
13911 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
13913 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
13914 (dump_eaf_flags): Dump EAF_NOT_RETURNED
13915 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
13916 and EAF_NOT_RETURNED.
13917 (modref_summary::useful_p): Likewise.
13918 (modref_summary_lto::useful_p): Likewise.
13919 (struct) modref_summary_lto: Use eaf_fleags_t.
13920 (deref_flags): Handle EAF_NOT_RETURNED.
13921 (struct escape_point): Use min_flags.
13922 (modref_lattice::init): Add EAF_NOT_RETURNED.
13923 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
13924 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
13926 (analyze_parms): Also analyze const functions; update conition on
13928 (modref_write): Update streaming.
13929 (read_section): Update streaming.
13930 (remap_arg_flags): Use eaf_flags_t.
13931 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
13932 * ipa-modref.h: (eaf_flags_t): New typedef.
13933 (struct modref_summary): Use eaf_flags_t.
13934 * tree-core.h (EAF_NOT_RETURNED): New constant.
13936 2021-07-16 Richard Biener <rguenther@suse.de>
13938 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
13939 the type of the LHS.
13940 (gimple_assign_nonnegative_warnv_p): Likewise.
13941 (gimple_call_nonnegative_warnv_p): Likewise. Return false
13942 if the call has no LHS.
13943 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
13944 * tree-eh.c (stmt_could_throw_1_p): Likewise.
13945 * tree-inline.c (insert_init_stmt): Likewise.
13946 * tree-ssa-loop-niter.c (get_val_for): Likewise.
13947 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
13949 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
13950 gassign *. Use the type of the lhs.
13951 (vn_nary_op_lookup_stmt): Adjust.
13952 (vn_nary_op_insert_stmt): Likewise.
13954 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
13956 * config/s390/predicates.md (bras_sym_operand): Accept all
13957 functions in 64-bit mode, use UNSPEC_PLT31.
13958 (larl_operand): Use UNSPEC_PLT31.
13959 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
13960 (legitimize_pic_address): Likewise.
13961 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
13963 (s390_delegitimize_address): Use UNSPEC_PLT31.
13964 (s390_output_addr_const_extra): Likewise.
13965 (print_operand): Add @PLT to TLS calls, handle %K.
13966 (s390_function_profiler): Mark __fentry__/_mcount as function,
13967 use %K, use UNSPEC_PLT31.
13968 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
13969 (s390_emit_call): Use UNSPEC_PLT31.
13970 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
13971 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
13972 (*movdi_64): Use %K.
13973 (reload_base_64): Likewise.
13974 (*sibcall_brc): Likewise.
13975 (*sibcall_brcl): Likewise.
13976 (*sibcall_value_brc): Likewise.
13977 (*sibcall_value_brcl): Likewise.
13979 (*brasl): Likewise.
13980 (*bras_r): Likewise.
13981 (*brasl_r): Likewise.
13982 (*bras_tls): Likewise.
13983 (*brasl_tls): Likewise.
13984 (main_base_64): Likewise.
13985 (reload_base_64): Likewise.
13986 (@split_stack_call<mode>): Likewise.
13988 2021-07-16 Richard Biener <rguenther@suse.de>
13990 PR tree-optimization/101467
13991 * tree-vect-stmts.c (vect_gen_while): Properly guard
13992 make_temp_ssa_name usage.
13994 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
13996 * config.gcc: Don't use forked print-sysroot-suffix.sh and
13997 t-sysroot-suffix for C-SKY.
13998 * config/csky/print-sysroot-suffix.sh: Delete.
13999 * config/csky/t-csky-linux: Delete.
14000 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
14001 instead of CSKY_MULTILIB_DIRNAMES.
14003 2021-07-16 Richard Biener <rguenther@suse.de>
14005 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
14006 conversion issues with the partial reduction of the reused
14007 vector accumulator.
14009 2021-07-16 Richard Biener <rguenther@suse.de>
14011 * config/i386/i386-options.c (ix86_option_override_internal): Set
14012 param_vect_partial_vector_usage to zero if not set.
14014 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
14017 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
14018 (VALID_INT_MODE_P): Add SDmode and DDmode.
14019 Add TDmode for TARGET_64BIT.
14020 (VALID_DFP_MODE_P): Remove.
14021 * config/i386/i386.c (ix86_hard_regno_mode_ok):
14022 Do not use VALID_DFP_MODE_P.
14024 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
14026 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
14028 (fold_using_range::fold_stmt): Ditto.
14029 (fold_using_range::range_of_range_op): Ditto.
14030 (fold_using_range::range_of_phi): Ditto.
14031 (fold_using_range::range_of_call): Ditto.
14032 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
14033 (fold_using_range::range_of_builtin_call): Ditto.
14034 (fold_using_range::range_of_cond_expr): Ditto.
14035 * gimple-range-fold.h (gimple_range_type): New.
14037 2021-07-15 Martin Sebor <msebor@redhat.com>
14039 PR middle-end/97027
14040 * tree-ssa-strlen.c (handle_assign): New function.
14041 (maybe_warn_overflow): Add argument.
14042 (nonzero_bytes_for_type): New function.
14043 (count_nonzero_bytes): Handle more tree types. Call
14044 nonzero_bytes_for_tye.
14045 (count_nonzero_bytes): Handle types.
14046 (handle_store): Handle stores from function calls.
14047 (strlen_check_and_optimize_call): Move code to handle_assign. Call
14048 it for assignments from function calls.
14050 2021-07-15 David Malcolm <dmalcolm@redhat.com>
14055 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
14057 2021-07-15 David Malcolm <dmalcolm@redhat.com>
14059 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
14061 2021-07-15 Martin Sebor <msebor@redhat.com>
14065 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
14066 (operand_compare::verify_hash_value): Same.
14067 * tree-core.h (OEP_DECL_NAME): New.
14069 2021-07-15 Martin Jambor <mjambor@suse.cz>
14071 * profile-count.h (profile_count::value): Change the return type to
14073 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
14075 * tree-cfg.c (dump_function_to_file): Likewise.
14077 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
14080 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
14081 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
14084 2021-07-15 Richard Biener <rguenther@suse.de>
14086 * tree-vectorizer.h (vect_gen_while): Match up with
14087 vect_gen_while_not.
14088 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
14089 of vect_gen_while_not.
14090 (vect_gen_while_not): Adjust.
14091 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
14093 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
14095 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
14096 (ranger_cache::range_of_def): Call adjust_range.
14097 (ranger_cache::entry_range): Same.
14098 * gimple-range-cache.h (non_null_ref::adjust_range): New.
14099 * gimple-range.cc (gimple_ranger::range_of_expr): Call
14101 (gimple_ranger::range_on_entry): Same.
14103 2021-07-15 Tamar Christina <tamar.christina@arm.com>
14106 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14108 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
14110 2021-07-15 Tamar Christina <tamar.christina@arm.com>
14113 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14115 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
14116 (sdot_prod, udot_prod): ...These.
14117 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
14118 (aarch64_<sur>dot<vsi2qi>): Rename to...
14119 (<sur>dot_prod<vsi2qi>): ...This.
14120 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
14123 2021-07-15 Jakub Jelinek <jakub@redhat.com>
14125 PR middle-end/101437
14126 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
14127 types even if they have non-BLKmode TYPE_MODE.
14129 2021-07-15 Richard Biener <rguenther@suse.de>
14132 * gcc.c (process_command): Process -gtoggle like process_options
14133 would after parsing options.
14135 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14137 * cfgexpand.c (expand_asm_loc): Adjust.
14138 (expand_asm_stmt): Likewise.
14139 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
14140 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
14141 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
14142 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
14143 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
14144 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
14145 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
14146 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
14147 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
14148 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
14149 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
14150 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
14151 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
14152 * doc/tm.texi: Regenerate.
14153 * target.def: Add location argument to md_asm_adjust.
14155 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14157 * tree-diagnostic.c (diagnostic_report_current_function): Use the
14158 diagnostic's location, not input_location.
14160 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14162 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
14164 (expand_asm_stmt): Likewise.
14166 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
14168 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
14169 addressed memory rtx, if any.
14170 (rs6000_split_multireg_move): Fix code formatting.
14171 Handle MMA build built-ins with operands in adjacent memory locations.
14173 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
14175 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
14178 2021-07-14 Jason Merrill <jason@redhat.com>
14180 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
14182 2021-07-14 Jason Merrill <jason@redhat.com>
14184 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
14187 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14189 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
14190 Remove erroneous line.
14192 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
14194 * params.opt (param_evrp_mode): Change default.
14196 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14198 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
14199 (sdot_prod, udot_prod): ...These.
14200 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
14201 (aarch64_<sur>dot<vsi2qi>): Rename to...
14202 (<sur>dot_prod<vsi2qi>): ...This.
14203 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
14206 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14208 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
14210 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14212 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
14214 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14216 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
14218 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14220 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
14221 (usdot_prod<vsi2qi>): ... This.
14222 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
14223 (usdot_prod): ...This.
14224 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
14225 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
14227 (@<sur>dot_prod<vsi2qi>): ...This.
14228 * config/aarch64/aarch64-sve-builtins-base.cc
14229 (svusdot_impl::expand): Use it.
14231 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14233 * optabs.def (usdot_prod_optab): New.
14234 * doc/md.texi: Document it and clarify other dot prod optabs.
14235 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
14236 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
14237 * optabs.c (expand_widen_pattern_expr): Likewise.
14238 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14239 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
14240 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
14242 (vect_widened_op_tree): Optionally ignore
14244 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
14246 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
14249 * config/i386/driver-i386.c (host_detect_local_cpu): Check
14250 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
14251 Enable UINTR only for 64-bit codegen.
14252 * config/i386/i386-options.c
14253 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
14255 * config/i386/i386.h (ARCH_ARG): New.
14256 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
14257 "[arch|tune] 64" for 64-bit codegen.
14259 2021-07-14 Richard Biener <rguenther@suse.de>
14261 PR tree-optimization/101445
14262 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
14263 of the IV in the correct direction for negative stride
14266 2021-07-14 Jakub Jelinek <jakub@redhat.com>
14269 * godump.c (godump_str_hash): New type.
14270 (godump_container::pot_dummy_types): Use string_hash instead of
14271 ptr_hash in the hash_set.
14273 2021-07-14 Richard Biener <rguenther@suse.de>
14275 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
14276 vector types where the old vector type has a multiple of
14277 the new vector type elements.
14278 (vect_create_partial_epilog): New function, split out from...
14279 (vect_create_epilog_for_reduction): ... here.
14280 (vect_transform_cycle_phi): Reduce the re-used accumulator
14281 to the new vector type.
14283 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14285 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
14286 non-'t'-sized arg check.
14288 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14290 * tree-cfg.c (cleanup_dead_labels_eh): Update
14291 post_landing_pad label upon change of landing pad block's
14293 (cleanup_dead_labels): Check that a removed label is not that
14296 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
14298 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
14300 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
14301 Add Neon to general purpose register case for zero-extend
14303 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
14304 case to prevent some cases opting to go through memory.
14305 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
14306 * rtl.c (rtvec_series_p): Define predicate to determine
14307 whether a vector contains a linear series of integers.
14308 * rtl.h (rtvec_series_p): Define.
14309 * rtlanal.c (vec_series_lowpart_p): Define predicate to
14310 determine if a vector selection is equivalent to the low part
14312 * rtlanal.h (vec_series_lowpart_p): Define.
14313 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
14314 Add vec_select -> subreg simplification.
14316 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
14318 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
14319 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
14320 _mm_test_mix_ones_zeros): New.
14322 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
14323 Richard Biener <rguenther@suse.de>
14325 * gimple.c (gimple_could_trap_p_1): Make S argument a
14326 "const gimple*". Preserve constness in call to
14327 gimple_asm_volatile_p.
14328 (gimple_could_trap_p): Make S argument a "const gimple*".
14329 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
14330 Update function prototypes.
14332 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14334 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
14335 (_loop_vec_info::main_loop_edge): New field.
14336 (_loop_vec_info::skip_main_loop_edge): Likewise.
14337 (_loop_vec_info::skip_this_loop_edge): Likewise.
14338 (_loop_vec_info::reusable_accumulators): Likewise.
14339 (_stmt_vec_info::reduc_scalar_results): Likewise.
14340 (_stmt_vec_info::reused_accumulator): Likewise.
14341 (vect_get_main_loop_result): Declare.
14342 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
14343 reduc_scalar_inputs.
14344 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
14345 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
14346 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
14347 skip_main_loop_edge and skip_this_loop_edge fields.
14348 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
14349 (vect_emit_reduction_init_stmts): New function.
14350 (get_initial_def_for_reduction): Use it.
14351 (get_initial_defs_for_reduction): Likewise. Change the vinfo
14352 parameter to a loop_vec_info.
14353 (vect_create_epilog_for_reduction): Store the scalar results
14354 in the reduc_info. If an epilogue loop is reusing an accumulator
14355 from the main loop, and if the epilogue loop can also be skipped,
14356 try to place the reduction code in the join block. Record
14357 accumulators that could potentially be reused by epilogue loops.
14358 (vect_transform_cycle_phi): When vectorizing epilogue loops,
14359 try to reuse accumulators from the main loop. Record the initial
14360 value in reduc_info for non-SLP reductions too.
14362 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14364 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
14365 adjustment handling. Take the neutral value as an argument,
14366 in place of the code argument.
14367 (vect_transform_cycle_phi): Update accordingly. Handle the
14368 initial values of cond reductions separately from code reductions.
14369 Choose the adjustment here rather than in
14370 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
14372 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14374 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
14375 (neutral_op_for_reduction): ...this, providing a more general
14377 (vect_create_epilog_for_reduction): Update accordingly.
14378 (vectorizable_reduction): Likewise.
14379 (vect_transform_cycle_phi): Likewise.
14381 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14383 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
14384 reduc_info instead of the original stmt_vec_info.
14385 (vect_transform_cycle_phi): Update accordingly.
14387 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14389 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
14390 reduc_info as an additional parameter.
14391 (vect_transform_cycle_phi): Update accordingly.
14393 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14395 * tree-vectorizer.h: Include tree-ssa-operands.h.
14396 (vect_phi_initial_value): New function.
14397 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
14398 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
14399 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14400 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
14402 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14404 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
14405 the phi results to vectype after creating them. Remove later
14406 conversion code that thus becomes redundant.
14408 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14410 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
14411 the new_phis vector with a reduc_inputs vector. Combine handling
14412 of reduction chains and ncopies > 1.
14414 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14416 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
14417 scalar_results to group_size elements after reducing down from
14418 N*group_size elements. Construct an array_slice of the live-out
14419 stmts and assert that there is one stmt per scalar result.
14421 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14423 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
14424 nested_in_vect_loop and use double_reduc everywhere. Remove dead
14425 assignment to "loop".
14427 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14429 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
14430 vector types first. For scalar types, consider both the preferred
14431 vector mode and the alternative vector modes.
14432 * optabs-query.c (can_vec_mask_load_store_p): Use the same
14433 structure as above, in particular using related_vector_mode
14434 for modes provided by autovectorize_vector_modes.
14436 2021-07-13 Jakub Jelinek <jakub@redhat.com>
14437 Richard Biener <rguenther@suse.de>
14439 PR tree-optimization/101419
14440 * tree-pass.h (PROP_objsz): Define.
14441 (make_pass_early_object_sizes): Declare.
14442 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
14443 there to pass_early_object_sizes, drop parameter.
14444 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
14445 drop parameter, move pass_post_ipa_warn right after that.
14446 * tree-object-size.c (pass_object_sizes::execute): Rename to...
14447 (object_sizes_execute): ... this. Add insert_min_max_p argument.
14448 (pass_data_object_sizes): Move after object_sizes_execute.
14449 (pass_object_sizes): Likewise. In execute method call
14450 object_sizes_execute, drop set_pass_param method and insert_min_max_p
14451 non-static data member and its initializer in the ctor.
14452 (pass_data_early_object_sizes, pass_early_object_sizes,
14453 make_pass_early_object_sizes): New.
14454 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
14455 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
14457 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
14460 * config/riscv/constraints.md ("S"): Update description and remove
14462 * doc/md.texi (Machine Constraints): Document the 'S' constraints
14465 2021-07-13 Richard Biener <rguenther@suse.de>
14468 2021-07-12 Richard Biener <rguenther@suse.de>
14470 * tree-vect-slp.c (vect_slp_region): Show the number of
14471 SLP graph entries in the optimization message.
14473 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
14475 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
14477 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
14478 return type to long.
14479 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
14482 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14484 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
14485 Query relation between the 2 operands and use it.
14487 2021-07-12 Sergei Trofimovich <siarheit@google.com>
14489 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
14491 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14494 * config/i386/predicates.md (vec_setm_sse41_operand):
14495 Rename from vec_setm_operand.
14496 (vec_setm_avx2_operand): New predicate.
14497 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
14498 Use vec_setm_sse41_operand as operand 2 predicate.
14499 (vec_set<V_256_512:mode): New expander.
14500 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
14501 as operand 2 predicate.
14503 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14505 PR tree-optimization/101335
14506 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
14508 2021-07-12 Andrew Pinski <apinski@marvell.com>
14510 * tree-ssa-phiopt.c (match_simplify_replacement): Move
14511 insert of the sequence before the movement of the
14512 statement. Check if to see if the statement is used
14513 outside of the original phi to see if we should move it.
14515 2021-07-12 Richard Biener <rguenther@suse.de>
14517 * dump-context.h (debug_dump_context::debug_dump_context):
14518 Add FILE * parameter defaulted to stderr.
14519 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
14520 * tree-vect-slp.c (dot_slp_tree): New functions.
14522 2021-07-12 Richard Biener <rguenther@suse.de>
14524 PR tree-optimization/101373
14525 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
14526 references when the BB may not return.
14527 (compute_avail): Pass in the function we're working on and
14528 replace cfun references with it. Externally throwing
14529 const calls also possibly terminate the function.
14530 (pass_pre::execute): Pass down the function we're working on.
14531 * gcse.c (compute_hash_table_work): Externally throwing
14532 const/pure calls also need record_last_mem_set_info.
14533 * postreload-gcse.c (record_opr_changes): Looping or externally
14534 throwing const/pure calls also need record_last_mem_set_info.
14536 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14538 * recog.c (memory_address_addr_space_p): Change the type to bool.
14539 Return true/false instead of 1/0.
14540 (offsettable_memref_p): Ditto.
14541 (offsettable_nonstrict_memref_p): Ditto.
14542 (offsettable_address_addr_space_p): Ditto.
14543 Change the type of addressp indirect function to bool.
14544 * recog.h (memory_address_addr_space_p): Change the type to bool.
14545 (strict_memory_address_addr_space_p): Ditto.
14546 (offsettable_memref_p): Ditto.
14547 (offsettable_nonstrict_memref_p): Ditto.
14548 (offsettable_address_addr_space_p): Ditto.
14549 * reload.c (maybe_memory_address_addr_space_p): Ditto.
14550 (strict_memory_address_addr_space_p): Change the type to bool.
14551 Return true/false instead of 1/0.
14552 (maybe_memory_address_addr_space_p): Change the type to bool.
14554 2021-07-12 Richard Biener <rguenther@suse.de>
14556 * tree-vect-slp.c (vect_slp_region): Show the number of
14557 SLP graph entries in the optimization message.
14559 2021-07-12 Richard Biener <rguenther@suse.de>
14561 PR tree-optimization/101394
14562 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
14563 copies from abnormals for a full redundancy.
14565 2021-07-12 Richard Biener <rguenther@suse.de>
14567 PR middle-end/101423
14568 * gimple.c (gimple_could_trap_p_1): Internal function calls
14570 * tree-eh.c (tree_could_trap_p): Likewise.
14572 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14575 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
14577 (vmulq_n_u32): Likewise.
14578 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
14579 (vmulq_n_f32): Likewise.
14580 (vmul_n_f16): Likewise.
14581 (vmulq_n_f16): Likewise.
14583 2021-07-12 Martin Liska <mliska@suse.cz>
14585 PR sanitizer/101425
14586 * gcc.c (check_offload_target_name): Call
14587 candidates_list_and_hint only if we have a candidate.
14589 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14592 * config/arm/neon.md (vec_init): Move to ...
14593 * config/arm/vec-common.md (vec_init): ... here.
14594 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
14596 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
14598 PR tree-optimization/101403
14599 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
14600 signedness of the shift is not the same as the signedness of
14601 the type extension.
14603 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
14604 Uroš Bizjak <ubizjak@gmail.com>
14606 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
14607 divmod of a constant numerator with new define_insn_and_split.
14609 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
14612 * config/i386/i386-expand.c (ix86_expand_call): If a call is
14613 to a non-local-binding, or local but to a public symbol, then
14614 assume that it might be indirected via the lazy symbol binder.
14615 Mark R10 and R10 as clobbered in that case.
14617 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
14620 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
14621 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
14622 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
14624 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
14626 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
14627 log2 range of operands[3] to [1,31].
14628 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
14630 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14632 * doc/md.texi: Don't split @smallexample in multiple @groups.
14634 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14636 * doc/md.texi: Add missing 'see' word.
14638 2021-07-09 Andrew Pinski <apinski@marvell.com>
14640 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
14641 to take sequence and gimple_match_op. Accept the case where
14642 op is a SSA_NAME and one statement in the sequence.
14643 Also allow constants.
14644 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
14645 Update call to phiopt_early_allow. Discard the sequence if not
14648 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
14653 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
14654 instead of ISA_HAS_MSA.
14655 (mips_expand_vec_unpack): Likewise.
14656 (mips_expand_vector_init): Likewise.
14658 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
14660 * config/rs6000/vsx.md (mods_<mode>): Rename to...
14661 (mod<mode>3): ... this.
14662 (modu_<mode>): Rename to...
14663 (umod<mode>3): ... this.
14664 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
14665 MODU_V4SI): Adjust.
14667 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
14669 * config/h8300/shiftrotate.md (variable shifts): Expose condition
14670 code handling for the test before the loop.
14672 2021-07-08 Martin Jambor <mjambor@suse.cz>
14675 * ipa-sra.c (class isra_call_summary): New member
14676 m_before_any_store, initialize it in the constructor.
14677 (isra_call_summary::dump): Dump the new field.
14678 (ipa_sra_call_summaries::duplicate): Copy it.
14679 (process_scan_results): Set it.
14680 (isra_write_edge_summary): Stream it.
14681 (isra_read_edge_summary): Likewise.
14682 (param_splitting_across_edge): Only override
14683 safe_to_import_accesses if m_before_any_store is set.
14685 2021-07-08 Martin Sebor <msebor@redhat.com>
14687 PR bootstrap/101374
14688 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
14689 Use Object Size Type 0 instead of 1.
14691 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
14693 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
14696 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
14698 * match.pd: Simplify an extend-operate-truncate sequence involving
14701 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
14702 Richard Biener <rguenther@suse.de>
14704 PR tree-optimization/40210
14705 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
14706 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
14707 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
14709 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
14712 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
14714 * config/i386/mmx.md (V_32): New mode iterator.
14715 (mov<V_32:mode>): Use V_32 mode iterator.
14716 (*mov<V_32:mode>_internal): Ditto.
14717 (*push<V_32:mode>2_rex64): Ditto.
14718 (*push<V_32:mode>2): Ditto.
14719 (movmisalign<V_32:mode>): Ditto.
14720 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
14721 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
14722 (vec_unpacks_lo_v4qi): New expander.
14723 (vec_unpacks_hi_v4qi): Ditto.
14724 (vec_unpacku_lo_v4qi): Ditto.
14725 (vec_unpacku_hi_v4qi): Ditto.
14726 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
14727 (VALID_INT_MODE_P): Ditto.
14729 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
14732 * config/rs6000/rs6000.md (udivti3): New insn.
14733 (divti3): New insn.
14734 (umodti3): New insn.
14735 (modti3): New insn.
14737 2021-07-07 Martin Sebor <msebor@redhat.com>
14739 PR tree-optimization/100137
14740 PR tree-optimization/99121
14741 PR tree-optimization/97027
14742 * builtins.c (access_ref::access_ref): Also set offmax.
14743 (access_ref::offset_in_range): Define new function.
14744 (access_ref::add_offset): Set offmax.
14745 (access_ref::inform_access): Handle access_none.
14746 (handle_mem_ref): Clear ostype.
14747 (compute_objsize_r): Handle ASSERT_EXPR.
14748 * builtins.h (struct access_ref): Add offmax member.
14749 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
14750 compute_objsize() and simplify.
14752 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
14754 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
14755 and VSX_BUILTIN_STXVP.
14757 2021-07-07 Martin Sebor <msebor@redhat.com>
14760 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
14761 a stray %K from error_at() missed in r12-2088.
14763 2021-07-07 Richard Biener <rguenther@suse.de>
14765 PR tree-optimization/99728
14766 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
14768 (mem_refs_may_alias_p): Add assert we handled aggregate
14770 (sm_seq_valid_bb): Give up when running into aggregate copies.
14771 (ref_indep_loop_p): Handle aggregate copies as never
14772 being invariant themselves but allow other refs to be
14773 disambiguated against them.
14774 (can_sm_ref_p): Do not try to apply store-motion to aggregate
14777 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
14780 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
14783 2021-07-06 Martin Sebor <msebor@redhat.com>
14785 * gimple-pretty-print.c (percent_G_format): Remove.
14786 * tree-diagnostic.c (default_tree_printer): Remove calls.
14787 * tree-pretty-print.c (percent_K_format): Remove.
14788 * tree-pretty-print.h (percent_K_format): Remove.
14790 2021-07-06 Martin Sebor <msebor@redhat.com>
14792 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
14793 Remove %K and use error_at.
14794 (aarch64_expand_fcmla_builtin): Same.
14795 (aarch64_expand_builtin_tme): Same.
14796 (aarch64_expand_builtin_memtag): Same.
14797 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
14798 (arm_expand_builtin): Same.
14799 * config/arm/arm.c (bounds_check): Same.
14801 2021-07-06 Martin Sebor <msebor@redhat.com>
14803 * builtins.c (warn_string_no_nul): Remove %G.
14804 (maybe_warn_for_bound): Same.
14805 (warn_for_access): Same.
14806 (check_access): Same.
14807 (check_strncat_sizes): Same.
14808 (expand_builtin_strncat): Same.
14809 (expand_builtin_strncmp): Same.
14810 (expand_builtin): Same.
14811 (expand_builtin_object_size): Same.
14812 (warn_dealloc_offset): Same.
14813 (maybe_emit_free_warning): Same.
14814 * calls.c (maybe_warn_alloc_args_overflow): Same.
14815 (maybe_warn_nonstring_arg): Same.
14816 (maybe_warn_rdwr_sizes): Same.
14817 * expr.c (expand_expr_real_1): Remove %K.
14818 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
14819 (gimple_fold_builtin_strncat): Same.
14820 * gimple-ssa-sprintf.c (format_directive): Same.
14821 (handle_printf_call): Same.
14822 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
14823 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
14824 (maybe_diag_access_bounds): Same. Call gimple_location.
14825 (check_bounds_or_overlap): Same.
14826 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
14827 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
14828 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
14829 (maybe_diag_stxncpy_trunc): Same.
14830 (handle_builtin_stxncpy_strncat): Same.
14831 (maybe_warn_pointless_strcmp): Same.
14832 * tree-ssa-uninit.c (maybe_warn_operand): Same.
14834 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
14837 * config/i386/predicates.md (vec_setm_operand): Enable
14838 register_operand for TARGET_SSE4_1.
14839 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
14840 as operand 2 predicate. Call ix86_expand_vector_set_var
14841 for non-constant index operand.
14842 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
14843 Call ix86_expand_vector_set_var for non-constant index operand.
14845 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
14847 * config/h8300/jumpcall.md (*branch): When possible, generate
14848 the comparison in CCZN mode.
14849 * config/h8300/predicates.md (simple_memory_operand): Reject all
14850 auto-increment addressing modes.
14852 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
14854 PR bootstrap/100246
14855 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
14858 2021-07-06 Richard Biener <rguenther@suse.de>
14860 * doc/md.texi (vec_fmaddsub<mode>4): Document.
14861 (vec_fmsubadd<mode>4): Likewise.
14862 * optabs.def (vec_fmaddsub$a4): Add.
14863 (vec_fmsubadd$a4): Likewise.
14864 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
14865 (IFN_VEC_FMSUBADD): Likewise.
14866 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
14867 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
14868 (addsub_pattern::build): Likewise.
14869 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
14870 and CFN_VEC_FMSUBADD are not transparent for permutes.
14871 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
14872 (vec_fmsubadd<mode>4): Likewise.
14874 2021-07-06 Richard Biener <rguenther@suse.de>
14876 * doc/invoke.texi (fmove-loop-stores): Document.
14877 * common.opt (fmove-loop-stores): New option.
14878 * opts.c (default_options_table): Enable -fmove-loop-stores
14879 at -O1 but not -Og.
14880 * tree-ssa-loop-im.c (pass_lim::execute): Pass
14881 flag_move_loop_stores instead of true to
14882 loop_invariant_motion_in_fun.
14884 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
14886 * doc/install.texi: Document --with-dsymutil.
14888 2021-07-06 Andrew Pinski <apinski@marvell.com>
14890 PR tree-optimization/101256
14891 * dbgcnt.def (phiopt_edge_range): New counter.
14892 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
14893 Check to make sure the new name is defined in the same
14894 bb as the conditional before duplicating range info.
14895 Also add debug counter.
14897 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
14899 PR rtl-optimization/100328
14900 * config/i386/i386-options.c (ix86_option_override_internal):
14901 Set param_ira_consider_dup_in_all_alts to 0.
14903 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
14905 PR rtl-optimization/100328
14906 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
14908 * ira.c (ira_get_dup_out_num): Adjust as parameter
14909 param_ira_consider_dup_in_all_alts.
14910 * params.opt (ira-consider-dup-in-all-alts): New.
14911 * ira-conflicts.c (process_regs_for_copy): Add one parameter
14912 single_input_op_has_cstr_p.
14913 (get_freq_for_shuffle_copy): New function.
14914 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
14915 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
14917 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
14919 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
14920 generate condition code aware RTL directly.
14922 2021-07-05 Andrew Pinski <apinski@marvell.com>
14924 PR tree-optimization/101039
14925 * match.pd (A CMP 0 ? A : -A): New patterns.
14926 * tree-ssa-phiopt.c (abs_replacement): Delete function.
14927 (tree_ssa_phiopt_worker): Don't call abs_replacement.
14928 Update comment about abs_replacement.
14930 2021-07-05 Andrew Pinski <apinski@marvell.com>
14932 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
14933 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
14935 2021-07-05 Andrew Pinski <apinski@marvell.com>
14937 * tree-ssa-phiopt.c (match_simplify_replacement):
14938 Add early_p argument. Call gimple_simplify_phiopt
14939 instead of gimple_simplify.
14940 (tree_ssa_phiopt_worker): Update call to
14941 match_simplify_replacement and allow unconditionally.
14942 (phiopt_early_allow): New function.
14943 (gimple_simplify_phiopt): New function.
14945 2021-07-05 Andrew Pinski <apinski@marvell.com>
14947 PR middle-end/101237
14948 * fold-const.c (negate_expr_p): Remove call to element_mode
14949 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
14950 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
14951 (fold_negate_expr_1): Likewise.
14952 (const_unop): Likewise.
14953 (fold_cond_expr_with_comparison): Likewise.
14954 (fold_binary_loc): Likewise.
14955 (fold_ternary_loc): Likewise.
14956 (tree_call_nonnegative_warnv_p): Likewise.
14957 * match.pd (-(A + B) -> (-B) - A): Likewise.
14959 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
14961 * configure.ac: Handle --with-dsymutil in the same way as we
14962 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
14963 Extract the type and version for the dsymutil configured or
14964 found by the default searches.
14965 * config.in: Regenerated.
14966 * configure: Regenerated.
14967 * collect2.c (do_dsymutil): Handle locating dsymutil in the
14968 same way as for the assembler and linker.
14969 * config/darwin.h (DSYMUTIL): Delete.
14970 * gcc.c: Report a configured dsymutil correctly.
14971 * exec-tool.in: Allow for dsymutil.
14973 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
14975 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
14976 Handle V4QI and V2HI modes.
14977 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
14978 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
14979 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
14980 Handle 4-byte dual operands with TARGET_XOP and single operands
14981 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
14982 mmx_pshufbv4qi3 for TARGET_SSSE3.
14983 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
14984 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
14985 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
14986 (expand_vec_perm_even_odd_1): Handle V4QI mode.
14987 (expand_vec_perm_broadcast_1): Handle V4QI mode.
14988 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
14989 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
14990 (mmx_pshufbv4qi3): Ditto.
14991 (*mmx_pblendw32): Ditto.
14992 (*mmx_pblendw64): Rename from *mmx_pblendw.
14993 (mmx_punpckhbw_low): New insn_and_split pattern.
14994 (mmx_punpcklbw_low): Ditto.
14996 2021-07-05 Richard Biener <rguenther@suse.de>
14998 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
14999 set LOOP_C_INFINITE on the vectorized loop.
15001 2021-07-05 Richard Biener <rguenther@suse.de>
15003 PR middle-end/101291
15004 * cfgloopmanip.c (loop_version): Set the loop copy of the
15005 versioned loop to the new loop.
15007 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
15010 * config.gcc: Ensure that Darwin biarch definitions are
15011 added before i386.h.
15012 * config/i386/darwin.h (TARGET_64BIT): Remove.
15013 (PR80556_WORKAROUND): New.
15014 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
15015 (DARWIN_SUBARCH_SPEC): New.
15016 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
15017 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
15018 (REAL_LIBGCC_SPEC): Remove.
15019 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
15020 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
15021 (REAL_LIBGCC_SPEC): Remove.
15023 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
15025 PR middle-end/101294
15026 * expr.c (store_constructor): Don't use vec_duplicate on vector.
15028 2021-07-02 Martin Sebor <msebor@redhat.com>
15030 PR middle-end/98871
15031 PR middle-end/98512
15032 * diagnostic.c (get_any_inlining_info): New.
15033 (update_effective_level_from_pragmas): Handle inlining context.
15034 (diagnostic_enabled): Same.
15035 (diagnostic_report_diagnostic): Same.
15036 * diagnostic.h (struct diagnostic_info): Add ctor.
15037 (struct diagnostic_context): Add new member.
15038 * tree-diagnostic.c (set_inlining_locations): New.
15039 (tree_diagnostics_defaults): Set new callback pointer.
15041 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
15043 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
15045 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
15046 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
15047 lxvp and stxvp built-ins.
15048 (mma_init_builtins): Handle lxvp and stxvp built-ins.
15049 (builtin_function_type): Likewise.
15050 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
15052 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15054 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
15055 additional argument for the code.
15056 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
15057 argument for the code. Just return if the ZN bits are useful or
15058 not rather than the old style CC_* enums.
15059 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
15060 more generic shiftqi patterns.
15061 (shifthi_noscratch, shiftsi_noscratch): Similarly.
15062 (shiftqi_noscratch_set_flags): New pattern.
15063 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
15065 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
15067 PR tree-optimization/101223
15068 * range-op.cc (build_lt): Add -1 for signed values.
15069 (built_gt): Subtract -1 for signed values.
15071 2021-07-02 David Faust <david.faust@oracle.com>
15073 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
15074 (btf_asm_type): Likewise.
15076 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15078 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
15079 an array of rtx rather than a pointer to rtx. Add code argument.
15080 (compute_a_shift_length): Similarly.
15081 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
15082 shift itself isn't an operand. Create dummy operand[0] to carry
15083 a mode and pass a suitable rtx code to compute_a_shift_length.
15084 (get_shift_alg): Adjust operand number of clobber in output templates.
15085 (output_a_shift): Make first argument an array of rtx rather than
15086 a pointer to rtx. Add code argument for the type of shift.
15087 Adjust now that the shift itself is no longer an operand.
15088 (compute_a_shift_length): Similarly.
15089 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
15090 iterator rather than nshift_operator.
15091 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
15092 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
15093 and compute_a_shift_length.
15094 (shiftqi_noscratch_clobber_flags): Likewise.
15095 (shifthi_noscratch_clobber_flags): Likewise.
15096 (shiftsi_noscratch_clobber_flags): Likewise.
15098 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
15101 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
15102 dsymutil for BTF/CTF.
15104 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
15107 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
15108 segment to include BTF.
15109 (BTF_INFO_SECTION_NAME): New.
15111 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15113 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
15114 (small_data_operand, memreg_operand, small_insn_p): Likewise.
15115 * config/m32r/m32r.c (call_operand): Adjust return type.
15116 (small_data_operand, memreg_operand): Likewise.
15118 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15120 * config/frv/frv-protos.h (integer_register_operand): Adjust return
15122 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
15123 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
15124 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
15125 (move_source_operand, move_destination_operand): Likewise.
15126 (condexec_source_operand, condexec_dest_operand): Likewise.
15127 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
15128 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
15129 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
15130 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
15131 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
15132 (even_fpr_operand, odd_fpr_operand): Likewise.
15133 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
15134 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
15135 (uint4_operand, uint1_operand, int_2word_operand): Likewise
15136 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
15137 (relational_operator, float_relational_operator): Likewise.
15138 (ccr_eqne_operator, minmax_operator): Likewise.
15139 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
15140 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
15141 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
15142 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
15143 (quad_acc_operand, accg_operand): Likewise.
15145 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15147 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
15148 return type to a bool.
15149 (nonimmediate_nonstack_operand): Likewise.
15150 (xstormy16_splittable_below100_operand): Likewise.
15151 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
15153 (xstormy16_splittable_below100_operand): Likewise.
15155 2021-07-02 Richard Biener <rguenther@suse.de>
15157 PR tree-optimization/101293
15158 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
15159 with combined offsets.
15160 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
15161 combined with the rest of the offset.
15163 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
15165 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
15166 PIC encodings for PE-COFF targets.
15168 2021-07-02 Jakub Jelinek <jakub@redhat.com>
15171 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
15172 Return nullptr for TImode inner mode.
15174 2021-07-02 Richard Biener <rguenther@suse.de>
15176 PR tree-optimization/101280
15177 PR tree-optimization/101173
15178 * gimple-loop-interchange.cc
15179 (tree_loop_interchange::valid_data_dependences): Properly
15180 guard all dependence checks with DDR_REVERSED_P or its
15183 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
15185 * config/i386/i386-expand.c (ix86_expand_builtin):
15186 Add branch to clear odata when ZF is set for asedecenc_expand
15187 and wideaesdecenc_expand.
15189 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
15191 * config/i386/gcc-auto-profile: regenerate
15193 2021-07-02 liuhongt <hongtao.liu@intel.com>
15195 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
15196 (trunc<mode><pmov_dst_4_lower>2): this.
15198 2021-07-01 David Malcolm <dmalcolm@redhat.com>
15200 * diagnostic.h (diagnostic_context::m_file_cache): New field.
15201 * input.c (class fcache): Rename to...
15202 (class file_cache_slot): ...this, making most members private and
15203 prefixing fields with "m_".
15204 (file_cache_slot::get_file_path): New accessor.
15205 (file_cache_slot::get_use_count): New accessor.
15206 (file_cache_slot::missing_trailing_newline_p): New accessor.
15207 (file_cache_slot::inc_use_count): New.
15208 (fcache_buffer_size): Move to...
15209 (file_cache_slot::buffer_size): ...here.
15210 (fcache_line_record_size): Move to...
15211 (file_cache_slot::line_record_size): ...here.
15212 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
15213 (fcache_tab_size): Move to file_cache::num_file_slots.
15214 (diagnostic_file_cache_init): Update for move of fcache_tab
15215 to global_dc->m_file_cache.
15216 (diagnostic_file_cache_fini): Likewise.
15217 (lookup_file_in_cache_tab): Convert to...
15218 (file_cache::lookup_file): ...this.
15219 (diagnostics_file_cache_forcibly_evict_file): Update for move of
15220 fcache_tab to global_dc->m_file_cache, moving most of
15221 implementation to...
15222 (file_cache::forcibly_evict_file): ...this new function and...
15223 (file_cache_slot::evict): ...this new function.
15224 (evicted_cache_tab_entry): Convert to...
15225 (file_cache::evicted_cache_tab_entry): ...this.
15226 (add_file_to_cache_tab): Convert to...
15227 (file_cache::add_file): ...this, moving bulk of implementation
15229 (file_cache_slot::create): ..this new function.
15230 (file_cache::file_cache): New.
15231 (file_cache::~file_cache): New.
15232 (lookup_or_add_file_to_cache_tab): Convert to...
15233 (file_cache::lookup_or_add_file): ..this new function.
15234 (fcache::fcache): Rename to...
15235 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
15237 (fcache::~fcache): Rename to...
15238 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
15240 (needs_read): Convert to...
15241 (file_cache_slot::needs_read_p): ...this.
15242 (needs_grow): Convert to...
15243 (file_cache_slot::needs_grow_p): ...this.
15244 (maybe_grow): Convert to...
15245 (file_cache_slot::maybe_grow): ...this.
15246 (read_data): Convert to...
15247 (file_cache_slot::read_data): ...this.
15248 (maybe_read_data): Convert to...
15249 (file_cache_slot::maybe_read_data): ...this.
15250 (get_next_line): Convert to...
15251 (file_cache_slot::get_next_line): ...this.
15252 (goto_next_line): Convert to...
15253 (file_cache_slot::goto_next_line): ...this.
15254 (read_line_num): Convert to...
15255 (file_cache_slot::read_line_num): ...this.
15256 (location_get_source_line): Update for moving of globals to
15257 global_dc->m_file_cache.
15258 (location_missing_trailing_newline): Likewise.
15259 * input.h (class file_cache_slot): New forward decl.
15260 (class file_cache): New.
15262 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
15264 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
15265 128-bit floating point conditional move support.
15266 (have_compare_and_set_mask): Add IEEE 128-bit floating point
15268 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
15269 (mov<mode>cc_p10, IEEE128 iterator): New insn.
15270 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
15271 (fpmask<mode>, IEEE128 iterator): New insn.
15272 (xxsel<mode>, IEEE128 iterator): New insn.
15274 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
15277 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
15279 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15281 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15283 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
15285 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
15286 (vec_duplicate<mode>): New expander.
15288 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15291 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15293 (ix86_byte_broadcast): New function.
15294 (ix86_convert_const_wide_int_to_broadcast): Likewise.
15295 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
15296 size is 16 bytes or bigger.
15297 (ix86_broadcast_from_integer_constant): New function.
15298 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
15299 to broadcast if mode size is 16 bytes or bigger.
15300 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
15302 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
15304 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15306 * config/i386/predicates.md (ix86_endbr_immediate_operand):
15307 Return true/false instead of 1/0.
15308 (movq_parallel): Ditto.
15310 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15312 * recog.c (general_operand): Return true/false instead of 1/0.
15313 (register_operand): Ditto.
15314 (immediate_operand): Ditto.
15315 (const_int_operand): Ditto.
15316 (const_scalar_int_operand): Ditto.
15317 (const_double_operand): Ditto.
15318 (push_operand): Ditto.
15319 (pop_operand): Ditto.
15320 (memory_operand): Ditto.
15321 (indirect_operand): Ditto.
15323 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15325 * genpreds.c (write_predicate_subfunction):
15326 Change the type of written subfunction to bool.
15327 (write_one_predicate_function):
15328 Change the type of written function to bool.
15329 (write_tm_preds_h): Ditto.
15330 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
15331 * recog.c (general_operand): Change the type to bool.
15332 (address_operand): Ditto.
15333 (register_operand): Ditto.
15334 (pmode_register_operand): Ditto.
15335 (scratch_operand): Ditto.
15336 (immediate_operand): Ditto.
15337 (const_int_operand): Ditto.
15338 (const_scalar_int_operand): Ditto.
15339 (const_double_operand): Ditto.
15340 (nonimmediate_operand): Ditto.
15341 (nonmemory_operand): Ditto.
15342 (push_operand): Ditto.
15343 (pop_operand): Ditto.
15344 (memory_operand): Ditto.
15345 (indirect_operand): Ditto.
15346 (ordered_comparison_operator): Ditto.
15347 (comparison_operator): Ditto.
15348 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
15349 Change the type of indirect predicate function to bool.
15350 * config/rs6000/rs6000.c (easy_vector_constant):
15351 Change the type to bool.
15352 * config/mips/mips-protos.h (m16_based_address_p):
15353 Change the type of operand 3 to bool.
15355 2021-07-01 Richard Biener <rguenther@suse.de>
15357 PR tree-optimization/101280
15358 PR tree-optimization/101173
15359 * gimple-loop-interchange.cc
15360 (tree_loop_interchange::valid_data_dependences): Revert
15361 previous change and instead correctly handle DDR_REVERSED_P
15364 2021-07-01 Richard Biener <rguenther@suse.de>
15366 PR tree-optimization/101278
15367 * tree-ssa-dse.c (dse_classify_store): First check for
15368 uses, then ignore stmt for chaining purposes.
15370 2021-07-01 Richard Biener <rguenther@suse.de>
15372 PR tree-optimization/100778
15373 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
15374 vectorized ops ahead of their scalar BB.
15376 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15379 * config/i386/i386.md (*nabs<dwi>2_doubleword):
15380 New insn_and_split pattern.
15381 (*nabs<dwi>2_1): Ditto.
15382 * config/i386/i386-features.c
15383 (general_scalar_chain::compute_convert_gain):
15384 Handle (NEG (ABS (...))) RTX. Rewrite src code
15385 scanner as switch statement.
15386 (general_scalar_chain::convert_insn):
15387 Handle (NEG (ABS (...))) RTX.
15388 (general_scalar_to_vector_candidate_p):
15389 Detect (NEG (ABS (...))) RTX. Reorder case statements
15390 for (AND (NOT (...) ...)) fallthrough.
15392 2021-07-01 Richard Biener <rguenther@suse.de>
15394 PR tree-optimization/101178
15395 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
15396 (slpg::perm_in): Add.
15397 (slpg::get_perm_in): Remove.
15398 (slpg::get_perm_materialized): Add.
15399 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
15400 during permute propagation and materialization.
15402 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15405 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
15407 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15409 PR middle-end/94366
15410 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
15411 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
15412 use boolean_type_node instead of integer_type_node as NE_EXPR type.
15413 (lower_reduction_clauses): Likewise.
15415 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15417 * config/gcn/gcn.c: Include dwarf2.h.
15418 (gcn_addr_space_debug): New function.
15419 (TARGET_ADDR_SPACE_DEBUG): New hook.
15421 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15423 * common/config/gcn/gcn-common.c
15424 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
15425 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
15427 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
15428 (gcn_frame_pointer_rqd): New function.
15429 (TARGET_FRAME_POINTER_REQUIRED): New hook.
15431 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15433 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
15434 prologue register saves.
15435 (gcn_debug_unwind_info): Use UI_DWARF2.
15436 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
15437 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
15438 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
15439 (DWARF_LINK_REGISTER): New define.
15440 (FIRST_PSEUDO_REGISTER): Increment.
15441 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
15442 (CALL_USED_REGISTERS): Likewise.
15443 (REGISTER_NAMES): Likewise.
15445 2021-06-30 Richard Biener <rguenther@suse.de>
15447 PR tree-optimization/101267
15448 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
15449 API and use SLP compatible interface of vect_is_simple_use.
15450 Reject not vectorized SLP defs for callers that do not support
15452 (vect_check_store_rhs): Handle masked stores and pass down
15453 the appropriate operator index.
15454 (vectorizable_call): Adjust.
15455 (vectorizable_store): Likewise.
15456 (vectorizable_load): Likewise. Handle SLP pecularity of
15458 (vect_is_simple_use): Remove special-casing of masked stores.
15460 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15462 * common.opt (foffload): Remove help as Driver only.
15463 * gcc.c (display_help): Add -foffload.
15465 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15467 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
15468 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
15469 a candidate list; better inform no offload target is configured
15470 and fix hint extraction when passed target is not '\0' at [len].
15471 * common.opt (foffload): Add tailing '.'.
15472 (foffload-options): Likewise; fix flag name in the help string.
15474 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
15477 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
15478 (vcage_f32): Gate comparison on __FAST_MATH__.
15479 (vcageq_f32): Likewise.
15480 (vcale_f32): Likewise.
15481 (vcaleq_f32): Likewise.
15482 (vcagt_f32): Likewise.
15483 (vcagtq_f32): Likewise.
15484 (vcalt_f32): Likewise.
15485 (vcaltq_f32): Likewise.
15486 (vcage_f16): Likewise.
15487 (vcageq_f16): Likewise.
15488 (vcale_f16): Likewise.
15489 (vcaleq_f16): Likewise.
15490 (vcagt_f16): Likewise.
15491 (vcagtq_f16): Likewise.
15492 (vcalt_f16): Likewise.
15493 (vcaltq_f16): Likewise.
15495 2021-06-30 Richard Biener <rguenther@suse.de>
15497 PR tree-optimization/101264
15498 * tree-vect-slp.c (vect_optimize_slp): Propagate the
15499 computed perm_in to all "any" permute successors
15500 we cannot de-duplicate immediately.
15502 2021-06-30 liuhongt <hongtao.liu@intel.com>
15505 * config/i386/sse.md
15506 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
15508 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
15510 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
15511 * config/i386/subst.md (maskz_scalar): New define_subst.
15512 (maskz_scalar_name): New subst_attr.
15513 (maskz_scalar_op5): Ditto.
15514 (round_saeonly_maskz_scalar_op5): Ditto.
15515 (round_saeonly_maskz_scalar_operand5): Ditto.
15517 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
15519 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
15520 Increase code CSECT alignment to at least 32 bytes.
15521 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
15522 alignment designation.
15524 2021-06-29 Sergei Trofimovich <siarheit@google.com>
15526 * doc/generic.texi: Fix s/net yet/not yet/ typo.
15528 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15530 PR tree-optimization/101254
15531 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
15532 wrapping/non-wrapping when setting the result range.
15534 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15536 * value-query.cc (gimple_range_global): Allow phis.
15538 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15540 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
15541 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
15542 statement for location context.
15543 (check_for_binary_op_overflow): Ditto.
15544 (simplify_using_ranges::get_vr_for_comparison): Ditto.
15545 (simplify_using_ranges::compare_name_with_value): Ditto.
15546 (simplify_using_ranges::compare_names): Ditto.
15547 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
15548 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
15549 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
15550 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
15551 (simplify_using_ranges::two_valued_val_range_p): Ditto.
15552 (simplify_using_ranges::simplify): Ditto.
15553 * vr-values.h: Adjust prototypes.
15555 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
15558 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
15560 2021-06-29 Julian Brown <julian@codesourcery.com>
15562 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
15563 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
15564 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
15566 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
15567 (MAX_FIXED_MODE_SIZE): Change to 128.
15569 2021-06-29 Julian Brown <julian@codesourcery.com>
15571 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
15572 (s_mnemonic): Add clrsb.
15573 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
15574 (clrsb<mode>2): Add expander for SImode/DImode.
15576 2021-06-29 Julian Brown <julian@codesourcery.com>
15578 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
15579 muldi3): Add patterns.
15581 2021-06-29 Julian Brown <julian@codesourcery.com>
15583 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
15584 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
15586 2021-06-29 Julian Brown <julian@codesourcery.com>
15588 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
15590 2021-06-29 Joseph Myers <joseph@codesourcery.com>
15592 * btfout.c, ctfout.c: Include "memmodel.h".
15594 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15596 * gcc.c (check_offload_target_name): Cast len argument to
15597 %q.*s to 'int'; avoid -Wstringop-truncation warning.
15599 2021-06-29 Richard Biener <rguenther@suse.de>
15601 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
15602 to "any" permute nodes and relax "any" permute proapgation
15603 during iterative backward propagation.
15605 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15608 * common.opt (-foffload=): Update description.
15609 (-foffload-options=): New.
15610 * doc/invoke.texi (C Language Options): Document
15611 -foffload and -foffload-options.
15612 * gcc.c (check_offload_target_name): New, split off from
15613 handle_foffload_option.
15614 (check_foffload_target_names): New.
15615 (handle_foffload_option): Handle -foffload=default.
15616 (driver_handle_option): Update for -foffload-options.
15617 * lto-opts.c (lto_write_options): Use -foffload-options
15618 instead of -foffload.
15619 * lto-wrapper.c (merge_and_complain, append_offload_options):
15621 * opts.c (common_handle_option): Likewise.
15623 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15625 * doc/invoke.texi (C Language Options): Sort options
15626 alphabetically in optlist and also the description itself.
15627 Remove leftover -fallow-single-precision from and add missing
15628 -fgnu-tm to the optlist.
15630 2021-06-29 Richard Biener <rguenther@suse.de>
15632 * tree-vect-slp.c (slpg_vertex::visited): Remove.
15633 (vect_slp_perms_eq): Handle -1 permutes.
15634 (vect_optimize_slp): Rewrite permute propagation.
15636 2021-06-29 Jakub Jelinek <jakub@redhat.com>
15639 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
15640 perform the optimization in GENERIC when sanitizing and x has a
15643 2021-06-29 Richard Biener <rguenther@suse.de>
15645 PR tree-optimization/101242
15646 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
15647 PHIs with not represented initial values as leafs.
15649 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
15651 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
15652 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
15653 "register" keyword.
15654 (pdp11_initial_elimination_offset) Remove unused variable.
15655 (pdp11_cmp_length) Ditto.
15656 (pdp11_insn_cost): Ditto, and fix signedness warning.
15658 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
15660 * btfout.c: Include tm_p.h.
15663 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15665 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
15667 (bpf_expand_epilogue): Likewise.
15668 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
15669 Do not define DBX_DEBUGGING_INFO.
15671 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15673 * doc/invoke.texi: Document the CTF and BTF debug info options.
15675 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15676 David Faust <david.faust@oracle.com>
15677 Jose E. Marchesi <jose.marchesi@oracle.com>
15678 Weimin Pan <weimin.pan@oracle.com>
15680 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
15681 GTFILES. Add new object files.
15682 * common.opt: Add CTF and BTF debug info options.
15683 * btfout.c: New file.
15684 * ctfc.c: Likewise.
15685 * ctfc.h: Likewise.
15686 * ctfout.c: Likewise.
15687 * dwarf2ctf.c: Likewise.
15688 * dwarf2ctf.h: Likewise.
15689 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
15691 * dwarf2out.c (dwarf2out_source_line): Likewise.
15692 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
15694 (debug_format_do_cu): New function.
15695 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
15697 Include dwarf2ctf.c.
15698 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
15700 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
15701 (CTF_DEBUG): New bitmask.
15702 (BTF_DEBUG): Likewise.
15703 (enum ctf_debug_info_levels): New enum.
15704 * gengtype.c (open_base_files): Handle ctfc.h.
15705 (main): Handle uint32_t type.
15706 * flags.h (btf_debuginfo_p): New definition.
15707 (dwarf_based_debuginfo_p): Likewise.
15708 * opts.c (debug_type_names): Add entries for CTF and BTF.
15709 (btf_debuginfo_p): New function.
15710 (dwarf_based_debuginfo_p): Likewise.
15711 (common_handle_option): Handle -gctfN and -gbtf options.
15712 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
15713 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
15716 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
15718 * dwarf2out.c (AT_class): Function is no longer static.
15719 (AT_int): Likewise.
15720 (AT_unsigned): Likewise.
15721 (AT_loc): Likewise.
15722 (get_AT): Likewise.
15723 (get_AT_string): Likewise.
15724 (get_AT_flag): Likewise.
15725 (get_AT_unsigned): Likewise.
15726 (get_AT_ref): Likewise.
15727 (new_die_raw): Likewise.
15728 (lookup_decl_die): Likewise.
15729 (base_type_die): Likewise.
15730 (add_name_attribute): Likewise.
15731 (add_AT_int): Likewise.
15732 (add_AT_unsigned): Likewise.
15733 (add_AT_loc): Likewise.
15734 (dw_get_die_tag): New function.
15735 (dw_get_die_child): Likewise.
15736 (dw_get_die_sib): Likewise.
15737 (struct dwarf_file_data): Move from here to dwarf2out.h
15738 (struct dw_attr_struct): Likewise.
15739 * dwarf2out.h: Analogous changes.
15741 2021-06-28 Martin Jambor <mjambor@suse.cz>
15744 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
15745 members m_dead_stmts and m_dead_ssas.
15746 * ipa-param-manipulation.c
15747 (ipa_param_body_adjustments::mark_dead_statements): New function.
15748 (ipa_param_body_adjustments::common_initialization): Call it on
15749 all removed but not split parameters.
15750 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
15752 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
15754 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
15755 dead debug statements.
15756 (copy_phis_for_bb): Do not copy dead PHI nodes.
15758 2021-06-28 Martin Jambor <mjambor@suse.cz>
15761 * symtab-clones.h (clone_info): Removed member param_adjustments.
15762 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
15763 deal with pass-through splits now.
15764 (ipa_param_performed_split): Removed.
15765 (ipa_param_adjustments::modify_call): Adjusted parameters.
15766 (class ipa_param_body_adjustments): Adjusted parameters of
15767 register_replacement, modify_gimple_stmt and modify_call_stmt.
15768 (ipa_verify_edge_has_no_modifications): Declare.
15769 (ipa_edge_modifications_finalize): Declare.
15770 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
15771 performed_splits processing, pas only edge to padjs->modify_call,
15772 check that call arguments were not modified if they should not have
15774 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
15776 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
15777 (ipa_edge_modification_info): Likewise.
15778 (ipa_edge_modification_sum): Likewise.
15779 (ipa_edge_modifications): New edge summary.
15780 (ipa_verify_edge_has_no_modifications): New function.
15781 (transitive_split_p): Removed.
15782 (transitive_split_map): Likewise.
15783 (init_transitive_splits): Likewise.
15784 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
15785 summary instead of performed_splits.
15786 (ipa_param_body_adjustments::register_replacement): Drop dummy
15787 parameter, set base_index of the created ipa_param_body_replacement.
15788 (phi_arg_will_live_p): New function.
15789 (ipa_param_body_adjustments::common_initialization): Do not create
15790 IPA_SRA dummy decls.
15791 (simple_tree_swap_info): Removed.
15792 (remap_split_decl_to_dummy): Likewise.
15793 (record_argument_state_1): New function.
15794 (record_argument_state): Likewise.
15795 (ipa_param_body_adjustments::modify_call_stmt): New parameter
15796 orig_stmt. Do not work with dummy decls, save necessary info about
15797 changes to ipa_edge_modifications.
15798 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
15799 orig_stmt, pass it to modify_call_stmt.
15800 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
15801 modify_gimple_stmt.
15802 (ipa_edge_modifications_finalize): New function.
15803 * tree-inline.c (remap_gimple_stmt): Pass original statement to
15804 modify_gimple_stmt.
15805 (copy_phis_for_bb): Do not copy dead PHI nodes.
15806 (expand_call_inline): Do not remap performed_splits.
15807 (update_clone_info): Likewise.
15808 * toplev.c: Include ipa-param-manipulation.h.
15809 (toplev::finalize): Call ipa_edge_modifications_finalize.
15811 2021-06-28 Andrew Pinski <apinski@marvell.com>
15813 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
15814 info if we're the only things setting the target PHI.
15815 (value_replacement): Don't duplicate range here.
15816 (minmax_replacement): Likewise.
15818 2021-06-28 Richard Biener <rguenther@suse.de>
15820 PR tree-optimization/101229
15821 * gimple-walk.c (gimple_walk_op): Handle PHIs.
15823 2021-06-28 Martin Liska <mliska@suse.cz>
15825 * config/v850/v850.c (construct_dispose_instruction): Allocate
15827 (construct_prepare_instruction): Likewise.
15829 2021-06-28 Martin Liska <mliska@suse.cz>
15831 * config/v850/v850.c (v850_option_override): Build default
15833 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
15834 ignored for inlining.
15835 (TARGET_CAN_INLINE_P): New.
15837 2021-06-28 Richard Biener <rguenther@suse.de>
15839 PR tree-optimization/101207
15840 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
15841 permute eliding for load permutations properly.
15843 2021-06-28 Richard Biener <rguenther@suse.de>
15845 PR tree-optimization/101173
15846 * gimple-loop-interchange.cc
15847 (tree_loop_interchange::valid_data_dependences): Disallow outer
15848 loop dependence distance of zero.
15850 2021-06-28 liuhongt <hongtao.liu@intel.com>
15853 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
15854 define_insn_and_split.
15855 (*avx_cmp<mode>3_ltint): Ditto.
15856 (*avx2_pcmp<mode>3_3): Ditto.
15857 (*avx2_pcmp<mode>3_4): Ditto.
15858 (*avx2_pcmp<mode>3_5): Ditto.
15860 2021-06-28 liuhongt <hongtao.liu@intel.com>
15862 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
15863 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
15864 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
15865 IX86_BUILTIN_PBLENDVB128): Replace icode with
15867 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
15869 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
15870 New pre_reload splitter.
15872 2021-06-27 Andrew Pinski <apinski@marvell.com>
15874 PR middle-end/101230
15875 * fold-const.c (fold_ternary_loc): Check
15876 the return value of invert_tree_comparison.
15878 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
15880 * config.gcc: Add SPDX License Identifier.
15881 (powerpc-ibm-aix789): Default to aix73.h.
15882 (powerpc-ibm-aix7.2.*.*): New stanza.
15883 * config/rs6000/aix72.h: Add SPDX License Identifier.
15884 * config/rs6000/aix73.h: New file.
15886 2021-06-26 Jason Merrill <jason@redhat.com>
15888 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
15890 2021-06-26 Andrew Pinski <apinski@marvell.com>
15892 * genmatch.c (lower_cond): Copy for_subst_vec
15893 for the simplify also.
15894 (lower): Swap the order for lower_for and lower_cond.
15896 2021-06-26 Andrew Pinski <apinski@marvell.com>
15898 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
15899 flow senatitive info on the moved ssa set.
15901 2021-06-26 Andrew Pinski <apinski@marvell.com>
15903 * fold-const.c (fold_cond_expr_with_comparison):
15904 Exand arg0 into comp_code, arg00, and arg01.
15905 (fold_ternary_loc): Use invert_tree_comparison
15906 instead of fold_invert_truthvalue for the case
15907 where we have A CMP B ? C : A.
15909 2021-06-25 Martin Sebor <msebor@redhat.com>
15911 PR middle-end/101216
15912 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
15914 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
15916 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
15918 2021-06-25 Richard Biener <rguenther@suse.de>
15920 PR tree-optimization/101202
15921 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
15924 2021-06-25 Richard Biener <rguenther@suse.de>
15926 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
15927 STMT_VINFO_REDUC_DEF from the original representative.
15929 2021-06-25 Martin Sebor <msebor@redhat.com>
15931 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
15932 gimple_no_warning_p and gimple_set_no_warning with
15933 warning_suppressed_p, and suppress_warning.
15935 (maybe_warn_for_bound): Same.
15936 (warn_for_access): Same.
15937 (check_access): Same.
15938 (expand_builtin_strncmp): Same.
15939 (fold_builtin_varargs): Same.
15940 * calls.c (maybe_warn_nonstring_arg): Same.
15941 (maybe_warn_rdwr_sizes): Same.
15942 * cfgexpand.c (expand_call_stmt): Same.
15943 * cgraphunit.c (check_global_declaration): Same.
15944 * fold-const.c (fold_undefer_overflow_warnings): Same.
15945 (fold_truth_not_expr): Same.
15946 (fold_unary_loc): Same.
15947 (fold_checksum_tree): Same.
15948 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
15949 (array_bounds_checker::check_mem_ref): Same.
15950 (array_bounds_checker::check_addr_expr): Same.
15951 (array_bounds_checker::check_array_bounds): Same.
15952 * gimple-expr.c (copy_var_decl): Same.
15953 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
15954 (gimple_fold_builtin_strncat): Same.
15955 (gimple_fold_builtin_stxcpy_chk): Same.
15956 (gimple_fold_builtin_stpcpy): Same.
15957 (gimple_fold_builtin_sprintf): Same.
15958 (fold_stmt_1): Same.
15959 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
15960 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
15961 * gimple-ssa-sprintf.c (handle_printf_call): Same.
15962 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
15963 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
15964 * gimple-ssa-warn-restrict.h: Adjust declarations.
15965 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
15966 gimple_no_warning_p and gimple_set_no_warning with
15967 warning_suppressed_p, and suppress_warning.
15968 (check_call): Same.
15969 (check_bounds_or_overlap): Same.
15970 * gimple.c (gimple_build_call_from_tree): Same.
15971 * gimplify.c (gimplify_return_expr): Same.
15972 (gimplify_cond_expr): Same.
15973 (gimplify_modify_expr_complex_part): Same.
15974 (gimplify_modify_expr): Same.
15975 (gimple_push_cleanup): Same.
15976 (gimplify_expr): Same.
15977 * omp-expand.c (expand_omp_for_generic): Same.
15978 (expand_omp_taskloop_for_outer): Same.
15979 * omp-low.c (lower_rec_input_clauses): Same.
15980 (lower_lastprivate_clauses): Same.
15981 (lower_send_clauses): Same.
15982 (lower_omp_target): Same.
15983 * tree-cfg.c (pass_warn_function_return::execute): Same.
15984 * tree-complex.c (create_one_component_var): Same.
15985 * tree-inline.c (remap_gimple_op_r): Same.
15986 (copy_tree_body_r): Same.
15987 (declare_return_variable): Same.
15988 (expand_call_inline): Same.
15989 * tree-nested.c (lookup_field_for_decl): Same.
15990 * tree-sra.c (create_access_replacement): Same.
15991 (generate_subtree_copies): Same.
15992 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
15993 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
15994 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
15995 * tree-ssa-loop-im.c (execute_sm): Same.
15996 * tree-ssa-phiopt.c (cond_store_replacement): Same.
15997 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
15998 (handle_builtin_strcpy): Same.
15999 (maybe_diag_stxncpy_trunc): Same.
16000 (handle_builtin_stxncpy_strncat): Same.
16001 (handle_builtin_strcat): Same.
16002 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
16003 (set_no_uninit_warning): Same.
16004 (uninit_undefined_value_p): Same.
16005 (warn_uninit): Same.
16006 (maybe_warn_operand): Same.
16007 * tree-vrp.c (compare_values_warnv): Same.
16008 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
16009 (test_for_singularity): Same.
16010 * gimple.h (warning_suppressed_p): New function.
16011 (suppress_warning): Same.
16012 (copy_no_warning): Same.
16013 (gimple_set_block): Call gimple_set_location.
16014 (gimple_set_location): Call copy_warning.
16016 2021-06-25 Martin Sebor <msebor@redhat.com>
16018 * tree.h (warning_suppressed_at, copy_warning,
16019 warning_suppressed_p, suppress_warning): New functions.
16021 2021-06-25 Martin Sebor <msebor@redhat.com>
16023 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
16024 * gengtype.c (open_base_files): Add diagnostic-spec.h.
16025 * diagnostic-spec.c: New file.
16026 * diagnostic-spec.h: New file.
16027 * tree.h (no_warning, all_warnings, suppress_warning_at): New
16029 * warning-control.cc: New file.
16031 2021-06-25 liuhongt <hongtao.liu@intel.com>
16034 * config/i386/i386.c (x86_order_regs_for_local_alloc):
16037 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16039 PR tree-optimization/101189
16040 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
16041 LHS range of condition to postfold routine.
16042 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
16043 FALSE edge if the LHS range supports it being taken.
16044 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
16046 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16048 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
16049 (relation_oracle::find_relation_block): Check correct bitmap.
16050 (relation_oracle::dump): Do not dump NULL blocks.
16052 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16054 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
16055 range_on_edge instead of manually calculating.
16057 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16059 * range-op.cc: Fix comment.
16061 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
16064 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
16065 Handle V8QI and V4HI modes.
16066 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
16068 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
16069 (mmxpackmode): New mode attribute.
16070 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
16071 (mmxunpackmode): New mode attribute.
16072 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
16073 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
16074 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
16075 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
16076 * config/i386/i386.md (extsuffix): Move from ...
16077 * config/i386/sse.md: ... here.
16079 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
16081 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
16082 (dwarf2out_finish): ...instead of here.
16084 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
16086 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
16087 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
16088 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
16089 * configure: Regenerate.
16091 2021-06-24 Richard Biener <rguenther@suse.de>
16093 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
16094 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
16095 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
16097 2021-06-24 Richard Biener <rguenther@suse.de>
16099 * config/i386/sse.md (avx_addsubv4df3): Rename to
16101 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
16102 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
16103 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
16104 * config/i386/i386-builtin.def: Adjust.
16105 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
16106 * optabs.def (vec_addsub_optab): New optab.
16107 * tree-vect-slp-patterns.c (class addsub_pattern): New.
16108 (slp_patterns): Add addsub_pattern.
16109 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
16110 across CFN_VEC_ADDSUB.
16111 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
16113 * doc/md.texi (vec_addsub<mode>3): Document.
16115 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16117 PR middle-end/101170
16118 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
16119 where regno + subreg_regno_offset wraps around use 0 as starting
16122 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16124 PR middle-end/101172
16125 * stor-layout.c (finish_bitfield_representative): If nextf has
16126 error_mark_node type, set repr type to error_mark_node too.
16128 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
16130 * config/s390/s390.c (s390_function_profiler): Ignore labelno
16132 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
16134 2021-06-24 Richard Biener <rguenther@suse.de>
16136 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
16137 across operations that have different semantics on different
16140 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16142 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
16143 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
16144 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
16145 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
16146 OMP_TARGET user outer_ctx instead of ctx for placeholders and
16147 initializer/combiner gimplification.
16148 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
16149 on target constructs.
16150 (lower_rec_input_clauses): Likewise.
16151 (lower_omp_target): Likewise.
16152 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
16153 on target if in_reduction is present.
16155 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
16157 * tree-predcom.c (class pcom_worker): New class.
16158 (release_chain): Renamed to...
16159 (pcom_worker::release_chain): ...this.
16160 (release_chains): Renamed to...
16161 (pcom_worker::release_chains): ...this.
16162 (aff_combination_dr_offset): Renamed to...
16163 (pcom_worker::aff_combination_dr_offset): ...this.
16164 (determine_offset): Renamed to...
16165 (pcom_worker::determine_offset): ...this.
16166 (class comp_ptrs): New class.
16167 (split_data_refs_to_components): Renamed to...
16168 (pcom_worker::split_data_refs_to_components): ...this,
16169 and update with class comp_ptrs.
16170 (suitable_component_p): Renamed to...
16171 (pcom_worker::suitable_component_p): ...this.
16172 (filter_suitable_components): Renamed to...
16173 (pcom_worker::filter_suitable_components): ...this.
16174 (valid_initializer_p): Renamed to...
16175 (pcom_worker::valid_initializer_p): ...this.
16176 (find_looparound_phi): Renamed to...
16177 (pcom_worker::find_looparound_phi): ...this.
16178 (add_looparound_copies): Renamed to...
16179 (pcom_worker::add_looparound_copies): ...this.
16180 (determine_roots_comp): Renamed to...
16181 (pcom_worker::determine_roots_comp): ...this.
16182 (determine_roots): Renamed to...
16183 (pcom_worker::determine_roots): ...this.
16184 (single_nonlooparound_use): Renamed to...
16185 (pcom_worker::single_nonlooparound_use): ...this.
16186 (remove_stmt): Renamed to...
16187 (pcom_worker::remove_stmt): ...this.
16188 (execute_pred_commoning_chain): Renamed to...
16189 (pcom_worker::execute_pred_commoning_chain): ...this.
16190 (execute_pred_commoning): Renamed to...
16191 (pcom_worker::execute_pred_commoning): ...this.
16192 (struct epcc_data): New member worker.
16193 (execute_pred_commoning_cbck): Call execute_pred_commoning
16194 with pcom_worker pointer.
16195 (find_use_stmt): Renamed to...
16196 (pcom_worker::find_use_stmt): ...this.
16197 (find_associative_operation_root): Renamed to...
16198 (pcom_worker::find_associative_operation_root): ...this.
16199 (find_common_use_stmt): Renamed to...
16200 (pcom_worker::find_common_use_stmt): ...this.
16201 (combinable_refs_p): Renamed to...
16202 (pcom_worker::combinable_refs_p): ...this.
16203 (reassociate_to_the_same_stmt): Renamed to...
16204 (pcom_worker::reassociate_to_the_same_stmt): ...this.
16205 (stmt_combining_refs): Renamed to...
16206 (pcom_worker::stmt_combining_refs): ...this.
16207 (combine_chains): Renamed to...
16208 (pcom_worker::combine_chains): ...this.
16209 (try_combine_chains): Renamed to...
16210 (pcom_worker::try_combine_chains): ...this.
16211 (prepare_initializers_chain): Renamed to...
16212 (pcom_worker::prepare_initializers_chain): ...this.
16213 (prepare_initializers): Renamed to...
16214 (pcom_worker::prepare_initializers): ...this.
16215 (prepare_finalizers_chain): Renamed to...
16216 (pcom_worker::prepare_finalizers_chain): ...this.
16217 (prepare_finalizers): Renamed to...
16218 (pcom_worker::prepare_finalizers): ...this.
16219 (tree_predictive_commoning_loop): Renamed to...
16220 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
16221 some calls and remove some cleanup code.
16222 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
16223 (static variable looparound_phis): Remove.
16224 (static variable name_expansions): Remove.
16226 2021-06-24 Richard Biener <rguenther@suse.de>
16228 * tree-vect-slp.c (slpg_vertex): New struct.
16229 (vect_slp_build_vertices): Adjust.
16230 (vect_optimize_slp): Likewise. Maintain an outgoing permute
16231 and a materialized one.
16233 2021-06-24 Richard Biener <rguenther@suse.de>
16235 PR tree-optimization/101105
16236 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
16237 Only ignore steps when they are equal or scalar order is preserved.
16239 2021-06-24 liuhongt <hongtao.liu@intel.com>
16242 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
16243 Adjust comments for ix86_expand_vecop_qihi2.
16244 (ix86_expand_vecmul_qihi): Renamed to ..
16245 (ix86_expand_vecop_qihi2): Adjust function prototype to
16246 support shift operation, add static to definition.
16247 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
16248 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
16249 ix86_expand_vec_shift_qihi_constant.
16250 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
16251 (ix86_expand_vec_shift_qihi_constant): Deleted.
16252 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
16254 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
16255 condition TARGET_64BIT.
16256 (mul<mode>3): Ditto.
16257 (<insn><mode>3): Ditto.
16258 (vlshr<mode>3): Extend to support avx512 vlshr.
16259 (v<insn><mode>3): New expander for
16261 (v<insn>v8qi3): Ditto.
16262 (vashrv8hi3<mask_name>): Renamed to ..
16263 (vashr<mode>3): And extend to support V16QImode for avx512.
16264 (vashrv16qi3): Deleted.
16265 (vashrv2di3<mask_name>): Extend expander to support avx512
16268 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
16270 * doc/lto.texi (Design Overview): Update that slim objects are
16273 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
16275 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
16276 of OTHER_POWER10_MASKS so it will not be enabled by default.
16278 2021-06-23 Richard Biener <rguenther@suse.de>
16279 Martin Jambor <mjambor@suse.cz>
16281 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
16282 param replacement unconditionally. Adjust comment.
16284 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16286 * Makefile.in (OBJS): Add gimple-range-fold.o
16287 * gimple-range-fold.cc: New.
16288 * gimple-range-fold.h: New.
16289 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
16290 (gimple_range_calc_op2): Ditto.
16291 * gimple-range-gori.h: Move prototypes to here.
16292 * gimple-range.cc: Adjust include files.
16293 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
16294 (fur_source::get_operand): Ditto.
16295 (fur_source::get_phi_operand): Ditto.
16296 (fur_source::query_relation): Ditto.
16297 (fur_source::register_relation): Ditto.
16298 (class fur_edge): Ditto.
16299 (fur_edge::fur_edge): Ditto.
16300 (fur_edge::get_operand): Ditto.
16301 (fur_edge::get_phi_operand): Ditto.
16302 (fur_stmt::fur_stmt): Ditto.
16303 (fur_stmt::get_operand): Ditto.
16304 (fur_stmt::get_phi_operand): Ditto.
16305 (fur_stmt::query_relation): Ditto.
16306 (class fur_depend): Relocate to gimple-range-fold.h.
16307 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
16308 (fur_depend::register_relation): Ditto.
16309 (fur_depend::register_relation): Ditto.
16310 (class fur_list): Ditto.
16311 (fur_list::fur_list): Ditto.
16312 (fur_list::get_operand): Ditto.
16313 (fur_list::get_phi_operand): Ditto.
16314 (fold_range): Ditto.
16315 (adjust_pointer_diff_expr): Ditto.
16316 (gimple_range_adjustment): Ditto.
16317 (gimple_range_base_of_assignment): Ditto.
16318 (gimple_range_operand1): Ditto.
16319 (gimple_range_operand2): Ditto.
16320 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
16321 (gimple_range_calc_op2): Ditto.
16322 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
16323 (fold_using_range::range_of_range_op): Ditto.
16324 (fold_using_range::range_of_address): Ditto.
16325 (fold_using_range::range_of_phi): Ditto.
16326 (fold_using_range::range_of_call): Ditto.
16327 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
16328 (fold_using_range::range_of_builtin_call): Ditto.
16329 (fold_using_range::range_of_cond_expr): Ditto.
16330 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
16331 (fold_using_range::relation_fold_and_or): Ditto.
16332 (fold_using_range::postfold_gcond_edges): Ditto.
16333 * gimple-range.h: Add gimple-range-fold.h to include files. Change
16334 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
16335 (gimple_range_handler): Relocate to gimple-range-fold.h.
16336 (gimple_range_ssa_p): Ditto.
16337 (range_compatible_p): Ditto.
16338 (class fur_source): Ditto.
16339 (class fur_stmt): Ditto.
16340 (class fold_using_range): Ditto.
16341 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
16342 (gimple_range_calc_op2): Ditto.
16344 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16346 PR tree-optimization/101148
16347 PR tree-optimization/101014
16348 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
16349 (ranger_cache::~ranger_cache): Adjust.
16350 (ranger_cache::block_range): Check if propagation disallowed.
16351 (ranger_cache::propagate_cache): Disallow propagation if new value
16352 can't be stored properly.
16353 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
16355 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16357 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
16358 (sbr_vector::set_bb_range): Return true.
16359 (class sbr_sparse_bitmap): Adjust.
16360 (sbr_sparse_bitmap::set_bb_range): Return value.
16361 (block_range_cache::set_bb_range): Return value.
16362 (ranger_cache::propagate_cache): Use return value to print msg.
16363 * gimple-range-cache.h (class block_range_cache): Adjust.
16365 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16367 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
16369 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
16371 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
16372 so this pattern can be used for test/compare removal. Pass
16373 current insn to compute_logical_op_length and output_logical_op.
16374 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
16375 (h8300_and_costs): Add argument to compute_logical_op_length.
16376 (output_logical_op): Add new argument. Use it to determine if the
16377 condition codes are used and adjust the output accordingly.
16378 (compute_logical_op_length): Add new argument and update length
16379 computations when condition codes are used.
16380 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
16382 (output_logical_op): Likewise.
16384 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16387 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
16388 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
16389 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
16390 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
16391 * config/i386/sse.md (unspec): ... here.
16393 2021-06-23 Martin Liska <mliska@suse.cz>
16396 * optc-save-gen.awk: Put back arm_fp16_format to
16399 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16402 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
16405 (clz<mode>2): Update RTX pattern for additions.
16407 2021-06-23 Jakub Jelinek <jakub@redhat.com>
16409 PR middle-end/101167
16410 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
16411 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
16413 2021-06-22 Sergei Trofimovich <siarheit@google.com>
16415 * doc/rtl.texi: drop unbalanced parenthesis.
16417 2021-06-22 Richard Biener <rguenther@suse.de>
16419 PR middle-end/101156
16420 * gimplify.c (gimplify_expr): Remove premature incorrect
16423 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16425 PR tree-optimization/101159
16426 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
16429 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16431 PR middle-end/101160
16432 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
16433 clear crtl->return_rtx instead of keeping it referencing a pseudo.
16435 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16436 Andrew Pinski <apinski@marvell.com>
16438 PR tree-optimization/101162
16439 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
16442 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16444 * range-op.cc (range_relational_tests): New.
16445 (range_op_tests): Call range_relational_tests.
16447 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16449 * range-op.cc (operator_cast::lhs_op1_relation): New.
16450 (operator_identity::lhs_op1_relation): Mew.
16452 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16454 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
16456 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16458 * range-op.cc (operator_plus::lhs_op1_relation): New.
16459 (operator_plus::lhs_op2_relation): New.
16461 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16463 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
16464 relation_oracle if dominators exist.
16465 (ranger_cache::~ranger_cache): Dispose of oracle.
16466 (ranger_cache::dump_bb): Dump oracle.
16467 * gimple-range.cc (fur_source::fur_source): New.
16468 (fur_source::get_operand): Use mmeber query.
16469 (fur_source::get_phi_operand): Use member_query.
16470 (fur_source::query_relation): New.
16471 (fur_source::register_dependency): Delete.
16472 (fur_source::register_relation): New.
16473 (fur_edge::fur_edge): Adjust.
16474 (fur_edge::get_phi_operand): Fix comment.
16475 (fur_edge::query): Delete.
16476 (fur_stmt::fur_stmt): Adjust.
16477 (fur_stmt::query): Delete.
16478 (fur_depend::fur_depend): Adjust.
16479 (fur_depend::register_relation): New.
16480 (fur_depend::register_relation): New.
16481 (fur_list::fur_list): Adjust.
16482 (fur_list::get_operand): Use member query.
16483 (fold_using_range::range_of_range_op): Process and query relations.
16484 (fold_using_range::range_of_address): Adjust dependency call.
16485 (fold_using_range::range_of_phi): Ditto.
16486 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
16487 (fold_using_range::relation_fold_and_or): New.
16488 (fold_using_range::postfold_gcond_edges): New.
16489 * gimple-range.h (class gimple_ranger): Adjust.
16490 (class fur_source): Adjust members.
16491 (class fur_stmt): Ditto.
16492 (class fold_using_range): Ditto.
16494 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16496 * range-op.cc (range_operator::wi_fold): Apply relation effect.
16497 (range_operator::fold_range): Adjust and apply relation effect.
16498 (*::fold_range): Add relation parameters.
16499 (*::op1_range): Ditto.
16500 (*::op2_range): Ditto.
16501 (range_operator::lhs_op1_relation): New.
16502 (range_operator::lhs_op2_relation): New.
16503 (range_operator::op1_op2_relation): New.
16504 (range_operator::op1_op2_relation_effect): New.
16505 (relop_early_resolve): New.
16506 (operator_equal::op1_op2_relation): New.
16507 (operator_equal::fold_range): Call relop_early_resolve.
16508 (operator_not_equal::op1_op2_relation): New.
16509 (operator_not_equal::fold_range): Call relop_early_resolve.
16510 (operator_lt::op1_op2_relation): New.
16511 (operator_lt::fold_range): Call relop_early_resolve.
16512 (operator_le::op1_op2_relation): New.
16513 (operator_le::fold_range): Call relop_early_resolve.
16514 (operator_gt::op1_op2_relation): New.
16515 (operator_gt::fold_range): Call relop_early_resolve.
16516 (operator_ge::op1_op2_relation): New.
16517 (operator_ge::fold_range): Call relop_early_resolve.
16518 * range-op.h (class range_operator): Adjust parameters and methods.
16520 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16522 * Makefile.in (OBJS): Add value-relation.o.
16523 * gimple-range.h: Adjust include files.
16524 * tree-data-ref.c: Adjust include file order.
16525 * value-query.cc (range_query::get_value_range): Default to no oracle.
16526 (range_query::query_relation): New.
16527 (range_query::query_relation): New.
16528 * value-query.h (class range_query): Adjust.
16529 * value-relation.cc: New.
16530 * value-relation.h: New.
16532 2021-06-22 Richard Biener <rguenther@suse.de>
16534 PR tree-optimization/101151
16535 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
16538 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
16540 * config/riscv/riscv.c (thead_c906_tune_info): New.
16541 (riscv_tune_info_table): Use new tune.
16543 2021-06-22 Richard Biener <rguenther@suse.de>
16545 PR tree-optimization/101158
16546 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
16547 checking after checking for matching operation.
16549 2021-06-22 Richard Biener <rguenther@suse.de>
16551 PR tree-optimization/101159
16552 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
16553 missing NULL vectype check.
16555 2021-06-22 Richard Biener <rguenther@suse.de>
16557 PR tree-optimization/101154
16558 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
16560 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16563 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
16564 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
16565 * config/i386/i386.c (ix86_expand_prologue): Clear it.
16566 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
16567 Remove "" from match_operand. Emit new insns using emit_move_insn and
16568 set ix86_last_zero_store_uid to INSN_UID of the last store.
16569 Add peephole2s for 1/2/4 stores of const0_rtx following previous
16572 2021-06-22 Martin Liska <mliska@suse.cz>
16574 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
16577 2021-06-22 Martin Liska <mliska@suse.cz>
16579 * gcov-io.h: Remove padding entries.
16581 2021-06-22 liuhongt <hongtao.liu@intel.com>
16583 PR tree-optimization/97770
16584 * tree-vect-patterns.c (vect_recog_popcount_pattern):
16586 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
16588 2021-06-22 liuhongt <hongtao.liu@intel.com>
16591 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
16592 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
16593 (expand<mode>_mask): this ..
16594 (*expand<mode>_mask): New pre_reload splitter to transform
16595 v{,p}expand* to vmov* when mask is zero, all ones, or has all
16596 ones in it's lower part, otherwise still generate
16599 2021-06-22 liuhongt <hongtao.liu@intel.com>
16602 * config/i386/i386-expand.c
16603 (ix86_expand_special_args_builtin): Keep constm1_operand only
16604 if it satisfies insn's operand predicate.
16606 2021-06-21 Jason Merrill <jason@redhat.com>
16609 * df-scan.c (df_ref_record): Check that regno < endregno.
16610 * function.c (assign_parms, expand_function_end): Do nothing with a
16611 TYPE_EMPTY_P result.
16613 2021-06-21 Richard Biener <rguenther@suse.de>
16615 PR tree-optimization/101120
16616 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
16618 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
16619 DR chain DCE capability.
16620 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
16621 * tree-vect-stmts.c (vectorizable_load): Remove unused
16622 loads in the DR chain for SLP.
16624 2021-06-21 Jakub Jelinek <jakub@redhat.com>
16626 PR inline-asm/100785
16627 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
16628 output or input operands were already error_mark_node.
16629 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
16630 remove all inputs, outputs and clobbers from the asm and
16631 set template to "".
16633 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16635 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
16636 (vceq_s16): Likewise.
16637 (vceq_s32): Likewise.
16638 (vceq_u8): Likewise.
16639 (vceq_u16): Likewise.
16640 (vceq_u32): Likewise.
16641 (vceq_p8): Likewise.
16642 (vceqq_s8): Likewise.
16643 (vceqq_s16): Likewise.
16644 (vceqq_s32): Likewise.
16645 (vceqq_u8): Likewise.
16646 (vceqq_u16): Likewise.
16647 (vceqq_u32): Likewise.
16648 (vceqq_p8): Likewise.
16649 (vceq_f32): Gate __a == __b on __FAST_MATH__.
16650 (vceqq_f32): Likewise.
16651 (vceq_f16): Likewise.
16652 (vceqq_f16): Likewise.
16654 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16657 * config/arm/iterators.md (NEON_VACMP): Remove.
16658 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
16660 (neon_vca<cmp_op><mode>_insn): Likewise.
16661 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
16664 2021-06-21 Richard Biener <rguenther@suse.de>
16666 PR tree-optimization/101121
16667 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
16668 when we just lack a stmt with the desired op when doing permutation.
16669 (vect_build_slp_tree): When caching a failed SLP build attempt
16670 assert that at least one lane is marked as not matching.
16672 2021-06-21 liuhongt <hongtao.liu@intel.com>
16675 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
16676 register alternative.
16677 (*and<mode>_1): Ditto.
16679 (*andn<mode>_1): Ditto.
16680 (*<code><mode>_1): Ditto.
16681 (*<code>qi_1): Ditto.
16682 (*one_cmpl<mode>2_1): Ditto.
16683 (*one_cmplsi2_1_zext): Ditto.
16684 (*one_cmplqi2_1): Ditto.
16685 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
16686 the order of mask registers to be before general registers.
16688 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
16691 * config/i386/i386.md: New define_peephole2s to shrink writing
16692 1, 2 or 4 consecutive zeros to memory when optimizing for size.
16694 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
16696 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
16697 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
16698 more efficient code when the source can be trivially simplified.
16700 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
16702 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
16703 a range if global is not available.
16704 (ranger_cache::entry_range): Fallback to range_of_def.
16705 * gimple-range-cache.h (range_of_def): Adjust prototype.
16707 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
16709 PR tree-optimization/101014
16710 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
16712 (ranger_cache::~ranger_cache): Ditto.
16713 (ranger_cache::enable_new_values): Delete.
16714 (ranger_cache::push_poor_value): Delete.
16715 (ranger_cache::range_of_def): Remove poor value processing.
16716 (ranger_cache::entry_range): Ditto.
16717 (ranger_cache::fill_block_cache): Ditto.
16718 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
16719 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
16720 * gimple-range.h (class gimple_ranger): Adjust.
16722 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16725 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
16726 derived from arm_canon_arch.
16727 (arm_canon_arch_option): Call it.
16728 (arm_canon_arch_multilib_option): New function.
16729 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
16730 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
16731 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
16732 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
16733 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
16734 * config/arm/arm.opt (mlibarch): New option.
16735 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
16736 of march on RHS with mlibarch.
16738 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
16740 * config.in: Regenerate.
16741 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
16743 * configure: Regenerate.
16744 * configure.ac: Fix for global_load assembler functions.
16746 2021-06-18 Richard Biener <rguenther@suse.de>
16748 PR tree-optimization/101112
16749 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
16750 to lookup a pattern stmt def.
16752 2021-06-18 Jakub Jelinek <jakub@redhat.com>
16754 PR middle-end/101062
16755 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
16756 representatives in QUAL_UNION_TYPE.
16758 2021-06-18 Andrew Pinski <apinski@marvell.com>
16760 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
16761 Add counting of how many times it is done.
16762 (factor_out_conditional_conversion): Likewise.
16763 (match_simplify_replacement): Likewise.
16764 (value_replacement): Likewise.
16765 (spaceship_replacement): Likewise.
16766 (cond_store_replacement): Likewise.
16767 (cond_if_else_store_replacement_1): Likewise.
16768 (hoist_adjacent_loads): Likewise.
16770 2021-06-18 Andrew Pinski <apinski@marvell.com>
16772 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
16773 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
16774 (verify_gimple_assign_binary): Reject point and offset types on
16775 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
16776 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
16777 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
16779 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
16781 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
16782 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
16783 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
16786 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
16788 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
16789 earlyclobber to alts 0/1.
16790 (gen_addadd): Add earlyclobber to alts 0/1.
16791 * config/rs6000/fusion.md: Regenerate file.
16793 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16795 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
16797 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16799 * gimple-range-cache.cc: Comment cleanups.
16800 * gimple-range-gori.cc: Comment cleanups.
16801 * gimple-range.cc: Comment/spacing cleanups
16802 * value-range.h: Comment cleanups.
16804 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
16807 * calls.c (expand_call): Replace PUSH_ARGS with
16808 targetm.calls.push_argument (0).
16809 (emit_library_call_value_1): Likewise.
16810 * defaults.h (PUSH_ARGS): Removed.
16811 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
16812 targetm.calls.push_argument (0).
16813 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
16814 (emit_push_insn): Pass the number bytes to push to
16815 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
16816 * hooks.c (hook_bool_uint_true): New.
16817 * hooks.h (hook_bool_uint_true): Likewise.
16818 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
16819 targetm.calls.push_argument (0).
16820 * target.def (push_argument): Add a targetm.calls hook.
16821 * targhooks.c (default_push_argument): New.
16822 * targhooks.h (default_push_argument): Likewise.
16823 * config/bpf/bpf.h (PUSH_ARGS): Removed.
16824 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
16825 * config/cr16/cr16.h (PUSH_ARGS): Removed.
16826 * config/i386/i386.c (ix86_push_argument): New.
16827 (TARGET_PUSH_ARGUMENT): Likewise.
16828 * config/i386/i386.h (PUSH_ARGS): Removed.
16829 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
16830 * config/m32c/m32c.h (PUSH_ARGS): Removed.
16831 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
16832 * config/pru/pru.h (PUSH_ARGS): Likewise.
16833 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
16834 TARGET_PUSH_ARGUMENT hook.
16835 * doc/tm.texi: Regenerated.
16837 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
16840 * config/i386/i386-expand.c (expand_vector_set_var):
16841 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
16842 ix86_expand_vector_init_duplicate.
16843 (ix86_expand_vector_init_duplicate): Emit insv_1 for
16844 QImode for !TARGET_PARTIAL_REG_STALL.
16845 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
16846 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
16847 as operand 2 predicate. Call ix86_expand_vector_set_var
16848 for non-constant index operand.
16849 (vec_setv2si): Ditto.
16850 (vec_setv4hi): Ditto.
16851 (vec_setv8qi): ditto.
16853 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
16855 PR tree-optimization/100790
16856 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
16859 2021-06-17 Martin Liska <mliska@suse.cz>
16861 * doc/invoke.texi: Use consistently -O1 instead of -O.
16863 2021-06-17 Martin Liska <mliska@suse.cz>
16865 * gcov-io.h: Update documentation entry about string format.
16867 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
16870 * config/s390/vecintrin.h (vec_doublee): Fix to use
16871 __builtin_s390_vflls.
16872 (vec_floate): Fix to use __builtin_s390_vflrd.
16874 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16876 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
16877 * dominance.h (get_dominated_to_depth): Likewise.
16878 (get_all_dominated_blocks): Likewise.
16879 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
16880 * gcse.c (hoist_code): Likewise.
16881 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
16882 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
16883 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
16884 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
16886 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16888 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
16889 * dominance.h (get_dominated_by_region): Likewise.
16890 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
16891 (gimple_duplicate_sese_tail): Likewise.
16892 (move_sese_region_to_fn): Likewise.
16894 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16896 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
16897 * dominance.h (get_dominated_by): Likewise.
16898 * auto-profile.c (afdo_find_equiv_class): Adjust.
16899 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
16900 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
16901 * tree-cfg.c (test_linear_chain): Likewise.
16902 (test_diamond): Likewise.
16904 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16906 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
16907 * cfgloopanal.c (get_loop_hot_path): Likewise.
16908 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
16910 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16912 * cgraph.c (cgraph_node::collect_callers): Return
16913 auto_vec<cgraph_edge *>.
16914 * cgraph.h (cgraph_node::collect_callers): Likewise.
16915 * ipa-cp.c (create_specialized_node): Adjust.
16916 (decide_about_value): Likewise.
16917 (decide_whether_version_node): Likewise.
16918 * ipa-sra.c (process_isra_node_results): Likewise.
16920 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16922 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
16923 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
16925 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
16928 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
16930 * gimple-range.cc (debug_seed_ranger): New.
16931 (dump_ranger): New.
16932 (debug_ranger): New.
16934 2021-06-17 Richard Biener <rguenther@suse.de>
16936 PR tree-optimization/54400
16937 * tree-vectorizer.h (enum slp_instance_kind): Add
16938 slp_inst_kind_bb_reduc.
16939 (reduction_fn_for_scalar_code): Declare.
16940 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
16941 Check SLP_INSTANCE_KIND instead of looking at the
16943 (vect_slp_analyze_instance_alignment): Likewise.
16944 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
16945 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
16946 chain linearization from vect_build_slp_tree_2 and generalize
16947 for the use of BB reduction vectorization.
16948 (vect_build_slp_tree_2): Adjust accordingly.
16949 (vect_optimize_slp): Elide permutes at the root of BB reduction
16951 (vectorizable_bb_reduc_epilogue): New function.
16952 (vect_slp_prune_covered_roots): Likewise.
16953 (vect_slp_analyze_operations): Use them.
16954 (vect_slp_check_for_constructors): Recognize associatable
16955 chains for BB reduction vectorization.
16956 (vectorize_slp_instance_root_stmt): Generate code for the
16957 BB reduction epilogue.
16959 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16961 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
16963 (gori_compute::may_recompute_p): New.
16964 (gori_compute::outgoing_edge_range_p): Perform recomputations.
16965 * gimple-range-gori.h (class gori_compute): Add prototype.
16967 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16969 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
16970 true when a range can be calculated.
16971 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
16973 2021-06-16 Martin Sebor <msebor@redhat.com>
16975 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
16976 Correct documented defaults.
16978 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
16980 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
16981 m_new_value_p directly.
16983 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
16986 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
16987 Handle 64bit modes for TARGET_SSE4_1.
16988 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
16989 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
16990 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
16991 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
16992 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
16993 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
16995 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16997 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
16998 Change to an expander that emits the correct instruction
16999 depending on endianness.
17000 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
17001 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
17003 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17005 * config/aarch64/aarch64-simd-builtins.def: Split generator
17006 for aarch64_<su>qmovn builtins into scalar and vector
17008 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
17010 (aarch64_<su>qmovn<mode>_insn_be): Define.
17011 (aarch64_<su>qmovn<mode>): Split into scalar and vector
17012 variants. Change vector variant to an expander that emits the
17013 correct instruction depending on endianness.
17015 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17017 * config/aarch64/aarch64-simd-builtins.def: Split generator
17018 for aarch64_sqmovun builtins into scalar and vector variants.
17019 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
17020 Split into scalar and vector variants. Change vector variant
17021 to an expander that emits the correct instruction depending
17023 (aarch64_sqmovun<mode>_insn_le): Define.
17024 (aarch64_sqmovun<mode>_insn_be): Define.
17026 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17028 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
17029 Define - modeling zero-high-half semantics.
17030 (aarch64_xtn<mode>): Change to an expander that emits the
17031 appropriate instruction depending on endianness.
17032 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
17034 (aarch64_xtn2<mode>_le): Rename to...
17035 (aarch64_xtn2<mode>_insn_le): This.
17036 (aarch64_xtn2<mode>_be): Rename to...
17037 (aarch64_xtn2<mode>_insn_be): This.
17038 (vec_pack_trunc_<mode>): Emit truncation instruction instead
17040 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
17041 attribute iterator.
17043 2021-06-16 Martin Jambor <mjambor@suse.cz>
17045 PR tree-optimization/100453
17046 * tree-sra.c (create_access): Disqualify any const candidates
17047 which are written to.
17048 (sra_modify_expr): Do not store sub-replacements back to a const base.
17049 (handle_unscalarized_data_in_subtree): Likewise.
17050 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
17051 instead of constant_decl_p.
17053 2021-06-16 Jakub Jelinek <jakub@redhat.com>
17055 PR middle-end/101062
17056 * stor-layout.c (finish_bitfield_representative): For fields in unions
17057 assume nextf is always NULL.
17058 (finish_bitfield_layout): Compute bit field representatives also in
17059 unions, but handle it as if each bitfield was the only field in the
17062 2021-06-16 Richard Biener <rguenther@suse.de>
17064 PR tree-optimization/101088
17065 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
17066 supported refs on edges. Do not assert same ref but
17067 different kind stores are unsuported but mark them so.
17068 (hoist_memory_references): Only look for supported refs
17071 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
17073 PR rtl-optimization/46235
17074 * config/i386/i386.md: New define_split for bt followed by cmov.
17075 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
17076 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
17077 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
17078 by setnc with zero extension.
17080 2021-06-16 Richard Biener <rguenther@suse.de>
17082 PR tree-optimization/101083
17083 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
17084 vectype as argument.
17085 (vect_build_slp_tree_2): Adjust.
17087 2021-06-15 Martin Sebor <msebor@redhat.com>
17089 PR middle-end/100876
17090 * builtins.c: (gimple_call_return_array): Account for size_t
17091 mangling as either unsigned int or unsigned long
17093 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
17095 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
17096 up before eliminating comparisons.
17098 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
17100 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
17101 nonzero|X is nonzero.
17102 (range_op_bitwise_and_tests): Add tests for above.
17104 2021-06-15 Carl Love <cel@us.ibm.com>
17107 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
17109 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
17112 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
17115 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
17116 (struct gimplify_omp_ctx): Extend defaultmap array by one.
17117 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
17118 (omp_notice_variable): Update type classification for Fortran.
17119 (gimplify_scan_omp_clauses): Update calls for new argument; handle
17120 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
17121 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
17122 * langhooks.c (lhd_omp_scalar_p): Likewise.
17123 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
17124 (LANG_HOOKS_DECLS): Add them.
17125 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
17126 omp_scalar_p pointer type to include the new bool argument.
17128 2021-06-15 David Malcolm <dmalcolm@redhat.com>
17130 * doc/analyzer.texi
17131 (Special Functions for Debugging the Analyzer): Add
17132 __analyzer_dump_capacity.
17134 2021-06-15 Jakub Jelinek <jakub@redhat.com>
17137 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
17138 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
17140 2021-06-15 Richard Biener <rguenther@suse.de>
17142 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
17143 check to identify loop latches.
17144 * cfgloop.c (verify_loop_structure): Likewise.
17145 * loop-init.c (apply_loop_flags): Allow marked irreducible
17146 regions even with multiple latches.
17147 * predict.c (rebuild_frequencies): Simplify.
17149 2021-06-15 Richard Biener <rguenther@suse.de>
17151 * tree-ssa-threadupdate.c
17152 (jump_thread_path_registry::mark_threaded_blocks): Assert we
17153 have marked irreducible regions.
17155 2021-06-14 Martin Sebor <msebor@redhat.com>
17158 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
17159 Handle calls to placement new.
17160 (ndecl_dealloc_argno): Avoid placement delete.
17162 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
17165 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
17166 create_tmp_reg_or_ssa_name().
17168 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
17170 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
17171 (ranger_cache::enable_new_values): Set to specified value and
17172 return the old value.
17173 (ranger_cache::disable_new_values): Delete.
17174 (ranger_cache::fill_block_cache): Disable non 1st order derived
17176 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
17177 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
17179 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
17182 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
17183 Return true early when testing with V2HImode.
17184 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
17186 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
17188 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
17189 (mve_vec_unpack<US>_hi_<mode>): New pattern.
17190 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
17191 (mve_vmovntq_<supf><mode>): Prefix with '@'.
17192 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
17194 (vec_unpack<US>_lo_<mode>): Likewise.
17195 (vec_pack_trunc_<mode>): Rename to
17196 neon_quad_vec_pack_trunc_<mode>.
17197 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
17199 (vec_unpack<US>_lo_<mode>): New.
17200 (vec_pack_trunc_<mode>): New.
17202 2021-06-14 Richard Biener <rguenther@suse.de>
17204 PR tree-optimization/100934
17205 * tree-ssa-dom.c (pass_dominator::execute): Properly
17206 mark irreducible regions.
17208 2021-06-14 Martin Liska <mliska@suse.cz>
17210 * doc/invoke.texi: Put r{...} on the same line as @item.
17212 2021-06-14 Martin Liska <mliska@suse.cz>
17214 * doc/invoke.texi: Add missing newline.
17216 2021-06-14 Martin Liska <mliska@suse.cz>
17218 * doc/invoke.texi: Remove '+' charasters.
17220 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
17222 * config.gcc (arc): Add support for with_cpu option.
17223 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
17225 2021-06-14 Richard Biener <rguenther@suse.de>
17227 PR tree-optimization/101031
17228 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
17229 instead of size when accounting for a possibly string
17232 2021-06-14 Martin Liska <mliska@suse.cz>
17234 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
17236 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
17238 * value-query.cc (gimple_range_global): Call get_range_global
17239 if called after inlining.
17241 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
17244 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
17245 Emit constant permutation insn directly from here.
17247 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
17249 * attribs.c (find_attribute_namespace): Iterate over vec<> with
17251 * auto-profile.c (afdo_find_equiv_class): Likewise.
17252 * gcc.c (do_specs_vec): Likewise.
17253 (do_spec_1): Likewise.
17254 (driver::set_up_specs): Likewise.
17255 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
17256 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
17257 (imm_store_chain_info::try_coalesce_bswap): Likewise.
17258 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
17259 (get_location_for_stmts): Likewise.
17260 * graphite-poly.c (print_iteration_domains): Likewise.
17261 (free_poly_bb): Likewise.
17262 (remove_gbbs_in_scop): Likewise.
17263 (free_scop): Likewise.
17264 (dump_gbb_cases): Likewise.
17265 (dump_gbb_conditions): Likewise.
17266 (print_pdrs): Likewise.
17267 (print_scop): Likewise.
17268 * ifcvt.c (cond_move_process_if_block): Likewise.
17269 * lower-subreg.c (decompose_multiword_subregs): Likewise.
17270 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
17271 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
17272 * sel-sched-dump.c (dump_insn_vector): Likewise.
17273 * store-motion.c (store_ops_ok): Likewise.
17274 (store_killed_in_insn): Likewise.
17275 * timevar.c (timer::named_items::print): Likewise.
17276 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
17277 (cleanup_tree_cfg_noloop): Likewise.
17278 * tree-data-ref.c (dump_data_references): Likewise.
17279 (print_dir_vectors): Likewise.
17280 (print_dist_vectors): Likewise.
17281 (dump_data_dependence_relations): Likewise.
17282 (dump_dist_dir_vectors): Likewise.
17283 (dump_ddrs): Likewise.
17284 (create_runtime_alias_checks): Likewise.
17285 (free_subscripts): Likewise.
17286 (save_dist_v): Likewise.
17287 (save_dir_v): Likewise.
17288 (invariant_access_functions): Likewise.
17289 (same_access_functions): Likewise.
17290 (access_functions_are_affine_or_constant_p): Likewise.
17291 (find_data_references_in_stmt): Likewise.
17292 (graphite_find_data_references_in_stmt): Likewise.
17293 (free_dependence_relations): Likewise.
17294 (free_data_refs): Likewise.
17295 * tree-inline.c (copy_debug_stmts): Likewise.
17296 * tree-into-ssa.c (dump_currdefs): Likewise.
17297 (rewrite_update_phi_arguments): Likewise.
17298 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
17299 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
17301 (vect_slp_analyze_node_dependences): Likewise.
17302 (vect_slp_analyze_instance_dependence): Likewise.
17303 (vect_record_base_alignments): Likewise.
17304 (vect_get_peeling_costs_all_drs): Likewise.
17305 (vect_peeling_supportable): Likewise.
17306 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
17307 (vec_info::free_stmt_vec_infos): Likewise.
17309 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
17311 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
17312 (andqi3_1<cczn>): Removed.
17313 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
17314 (H8/SX bit logicals): Split out from other patterns.
17315 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
17316 mulqihi3_const_clobber_flags.
17317 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
17319 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
17322 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
17323 to true if red zone is used.
17324 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
17325 ix86_red_zone_used.
17326 * config/i386/i386.h (machine_function): Add red_zone_used.
17327 (ix86_red_zone_size): Removed.
17328 (ix86_red_zone_used): New.
17329 * config/i386/i386.md (peephole2 patterns): Replace
17330 ix86_red_zone_size with ix86_red_zone_used.
17332 2021-06-12 Jason Merrill <jason@redhat.com>
17334 * doc/extend.texi (unused variable attribute): Applies to
17335 structure fields as well.
17337 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
17339 * auto-profile.c (read_profile): fix a typo in an error string
17341 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
17343 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
17345 * tree-pretty-print.c (dump_omp_clauses): Update.
17346 (dump_generic_node) <OMP_CLAUSE>: Use it.
17348 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17351 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
17352 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
17353 the polymorphic variants matching code.
17354 (__arm_vld1q_z): Likewise.
17355 (__arm_vld2q): Likewise.
17356 (__arm_vld4q): Likewise.
17357 (__arm_vldrbq_gather_offset): Likewise.
17358 (__arm_vldrbq_gather_offset_z): Likewise.
17360 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17362 PR tree-optimization/96392
17363 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
17365 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17367 PR tree-optimization/96392
17368 * fold-const.c (fold_real_zero_addition_p): Take both arguments
17369 of the addition or subtraction, not just the zero. Use this
17370 other argument in tests for signaling NaNs and signed zeros.
17371 (tree_expr_maybe_real_minus_zero_p): New predicate.
17372 * fold-const.h (fold_real_zero_addition_p): Update prototype.
17373 (tree_expr_maybe_real_minus_zero_p): New function prototype.
17374 * match.pd: Update calls to fold_real_zero_addition_p.
17375 Replace HONOR_NANS with tree_expr_maybe_nan_p.
17376 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
17377 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
17378 * tree-ssa-reassoc.c (eliminate_using_constants): Update
17379 call to fold_real_zero_addition_p.
17381 2021-06-11 Richard Biener <rguenther@suse.de>
17383 PR tree-optimization/101025
17384 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
17385 all refs that require dependence checking.
17387 2021-06-11 Richard Biener <rguenther@suse.de>
17389 PR tree-optimization/101028
17390 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
17391 reassoc discovery fails fatally, mark appropriate lanes
17394 2021-06-11 Richard Biener <rguenther@suse.de>
17396 PR tree-optimization/101026
17397 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
17398 have a representative for the associated chain nodes.
17400 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17402 PR rtl-optimization/101008
17403 * simplify-rtx.c (relational_result): New function.
17404 (simplify_logical_relational_operation,
17405 simplify_relational_operation): Use it.
17407 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17410 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
17412 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
17415 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
17416 false if the permutation can be implemented with constant
17417 permutation instruction in wider mode.
17418 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
17419 Handle V8QImode and V4HImode.
17421 2021-06-11 Martin Liska <mliska@suse.cz>
17423 PR gcov-profile/100788
17424 * common.opt: Add new option.
17425 * coverage.c (coverage_begin_function): Emit warning instead on
17426 the internal compiler error.
17427 * doc/invoke.texi: Document the option.
17428 * toplev.c (process_options): Enable it by default.
17430 2021-06-11 Richard Biener <rguenther@suse.de>
17432 PR middle-end/101009
17433 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
17434 to set *init_b to true when we encounter a constant equal
17436 (compute_affine_dependence): Also dump the actual DR_REF.
17438 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17440 PR tree-optimization/100984
17441 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
17442 replacements table.
17443 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
17445 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
17447 * config/rs6000/rs6000.md
17448 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
17449 define_insn_and_split.
17451 2021-06-11 Richard Biener <rguenther@suse.de>
17453 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
17454 to sort operands of the associative chain.
17456 2021-06-11 Richard Biener <rguenther@suse.de>
17458 * system.h (gcc_stablesort_r): Declare.
17459 * sort.cc (gcc_sort_r): Support stable sort.
17460 (gcc_stablesort_r): Define.
17461 * vec.h (vec<>::stablesort): Add.
17463 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
17466 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
17467 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
17468 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
17469 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
17470 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
17471 (expand_vec_perm_interleave2): Handle 64bit modes.
17472 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
17473 (expand_vec_perm_even_odd_1): Ditto.
17474 (ix86_vectorize_vec_perm_const): Ditto.
17475 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
17476 * config/i386/sse.md: ... here.
17477 * config/i386/mmx.md (*vec_interleave_lowv2sf):
17478 New insn_and_split pattern.
17479 (*vec_interleave_highv2sf): Ditto.
17480 (mmx_pshufbv8qi3): New insn pattern.
17481 (*mmx_pblendw): Ditto.
17483 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
17485 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
17486 (build_acc): Likewise.
17487 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
17488 source operands in little-endian mode.
17489 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
17490 (mma_init_builtins): Likewise.
17491 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
17492 ordering for the MMA assemble and build source operands.
17493 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
17495 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
17498 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
17500 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
17502 * config/h8300/extensions.md: Replace _clobber_flags patterns
17505 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
17507 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
17508 (vcond_mask_<mode><tointvec>): this.
17510 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
17511 Thomas Schwinge <thomas@codesourcery.com>
17513 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
17514 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
17515 * gimple.h (enum gf_mask): Split
17516 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
17517 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
17518 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
17519 (is_gimple_omp_oacc): Update.
17520 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
17521 * gimplify.c (gimplify_omp_target_update): Likewise.
17522 * omp-expand.c (expand_omp_target, build_omp_regions_1)
17523 (omp_make_gimple_edges): Likewise.
17524 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
17527 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17529 * value-query.cc (value_query::value_on_edge): Rename name to
17531 (range_query::range_on_edge): Same.
17532 (range_query::value_of_expr): Same.
17533 (range_query::value_on_edge): Same.
17534 * value-query.h (class value_query): Same.
17535 (class range_query): Same.
17537 2021-06-10 Richard Biener <rguenther@suse.de>
17539 PR tree-optimization/101003
17540 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
17541 use the pattern stmt defs when linearizing a chain.
17543 2021-06-10 Jakub Jelinek <jakub@redhat.com>
17546 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
17547 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
17549 2021-06-10 Clement Chigot <clement.chigot@atos.net>
17551 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
17552 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
17554 2021-06-09 Andrew Pinski <apinski@marvell.com>
17556 PR tree-optimization/100925
17557 * match.pd (a ? CST1 : CST2): Limit transformations
17558 that would produce a negative to integeral types only.
17559 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
17561 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17564 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17566 * doc/tm.texi: Correctly update.
17568 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17570 * doc/tm.texi: Correctly update.
17572 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17575 * doc/tm.texi.in (Trampolines): Add a missing blank line.
17577 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
17580 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
17581 and -ftrampolines work only with Ada.
17582 * doc/tm.texi.in (Trampolines): Likewise.
17583 * doc/tm.texi: Regenerated.
17585 2021-06-09 Carl Love <cel@us.ibm.com>
17587 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
17588 Add define for new builtins.
17589 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
17590 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
17591 overloaded builtin definitions.
17592 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
17593 VSIGNEXTSD2Q): Add builtin expansions.
17594 (SIGNEXT): Add P10 overload definition.
17595 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
17596 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
17597 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
17598 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
17599 vsignextend_si_v2di)[VIlong]: Add define_expand.
17600 Make define_insn vsx_sign_extend_si_v2di visible.
17601 * doc/extend.texi: Add documentation for the vec_signexti,
17602 vec_signextll builtins and vec_signextq.
17604 2021-06-09 Carl Love <cel@us.ibm.com>
17606 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
17607 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
17608 __floattikf_sw, __floatuntikf_sw respectively.
17609 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
17610 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
17611 define_insn for mode IEEE 128.
17613 2021-06-09 Carl Love <cel@us.ibm.com>
17615 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
17616 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
17617 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
17618 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
17619 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
17620 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
17621 uses of VSX_TI to VEC_TI.
17623 2021-06-09 Carl Love <cel@us.ibm.com>
17625 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
17627 2021-06-09 Carl Love <cel@us.ibm.com>
17629 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
17631 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
17632 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
17633 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
17634 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
17635 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
17636 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
17637 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
17639 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
17640 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
17641 altivec_vrlqnm): New define_expands.
17642 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
17643 VCMPGTUT_P): Add macro expansions.
17644 (BU_P10V_AV_P): Add builtin predicate definition.
17645 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
17646 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
17647 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
17648 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
17649 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
17650 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
17651 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
17652 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
17653 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
17654 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
17655 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
17656 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
17657 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
17658 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
17659 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
17660 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
17661 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
17662 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
17663 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
17664 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
17665 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
17666 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
17667 P10V_BUILTIN_MODU_V1TI):
17668 New overloaded definitions.
17669 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
17670 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
17671 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
17672 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
17673 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
17674 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
17676 (altivec_init_builtins): New E_V1TImode case statement.
17677 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
17678 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
17679 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
17680 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
17681 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
17682 E_V1TImode]: New case statements.
17683 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
17684 value RS6000_BTI_bool_V1TI.
17685 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
17686 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
17687 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
17688 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
17689 vlshrv1ti3, vashrv1ti3): New define_expands.
17690 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
17691 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
17692 UNSPEC_VSX_MODUQ): New unspecs.
17693 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
17694 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
17696 (vcmpnet): New define_expand.
17697 * doc/extend.texi: Add documentation for the new builtins vec_rl,
17698 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
17699 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
17700 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
17701 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
17702 vec_any_ge, vec_any_le.
17704 2021-06-09 Carl Love <cel@us.ibm.com>
17706 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
17707 bug in argument generation.
17709 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
17711 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
17713 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
17714 remove <supf> iterator.
17715 (mve_vclzq_u<mode>): New.
17716 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
17717 (neon_vclz<mode): Move to ...
17718 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
17719 * config/arm/vec-common.md: ... here. Add support for MVE.
17721 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
17723 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
17724 (@mve_vrhaddq_<supf><mode): Likewise.
17725 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
17726 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
17727 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
17729 2021-06-09 imba-tjd <109224573@qq.com>
17731 * doc/invoke.texi: Fix typo.
17733 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
17735 PR middle-end/53267
17736 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
17737 Support evaluation of fmod/fmodf/fmodl at compile-time.
17739 2021-06-09 Richard Biener <rguenther@suse.de>
17741 PR tree-optimization/100981
17742 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
17743 gimple_get_lhs to also handle calls.
17744 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
17747 2021-06-09 Richard Biener <rguenther@suse.de>
17749 PR tree-optimization/97832
17750 * tree-vectorizer.h (_slp_tree::failed): New.
17751 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
17753 (_slp_tree::~_slp_tree): Free failed.
17754 (vect_build_slp_tree): Retain failed nodes and record
17755 matches in them, copying that back out when running
17756 into a cached fail. Dump start and end of discovery.
17757 (dt_sort_cmp): New.
17758 (vect_build_slp_tree_2): Handle associatable chains
17759 together doing more aggressive operand swapping.
17761 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17764 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
17766 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
17767 Linux and GNU targets.
17769 2021-06-09 Richard Biener <rguenther@suse.de>
17771 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
17774 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17776 * config/arc/arc.md (loop_end): Change it to
17777 define_insn_and_split.
17779 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17781 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
17782 (machi): New pattern.
17783 (umaddhisi4): Use VMAC2HU instruction.
17784 (umachi): New pattern.
17786 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17788 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
17789 * config/arc/arc.c (arc_split_move_p): New function.
17790 (arc_split_move): Clean up.
17791 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
17792 (movdf_insn): Likewise.
17793 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
17795 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
17798 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
17799 argument to "raw". Do not emit segment overrides when "raw" is true.
17801 2021-06-09 Martin Liska <mliska@suse.cz>
17803 * doc/gcov.texi: Create a proper JSON files.
17804 * doc/invoke.texi: Remove dots in order to make it a valid
17807 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
17809 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
17810 (insn_is_load_p): Use pattern_is_rotate64.
17811 (insn_is_swap_p): Likewise.
17812 (quad_aligned_load_p): Likewise.
17813 (const_load_sequence_p): Likewise.
17814 (replace_swapped_aligned_load): Likewise.
17815 (recombine_lvx_pattern): Likewise.
17816 (recombine_stvx_pattern): Likewise.
17818 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
17820 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
17821 fur_stmt source record.
17822 * gimple-range.cc (fur_source::get_operand): Generic range query.
17823 (fur_source::get_phi_operand): New.
17824 (fur_source::register_dependency): New.
17825 (fur_source::query): New.
17826 (class fur_edge): New. Edge source for operands.
17827 (fur_edge::fur_edge): New.
17828 (fur_edge::get_operand): New.
17829 (fur_edge::get_phi_operand): New.
17830 (fur_edge::query): New.
17831 (fur_stmt::fur_stmt): New.
17832 (fur_stmt::get_operand): New.
17833 (fur_stmt::get_phi_operand): New.
17834 (fur_stmt::query): New.
17835 (class fur_depend): New. Statement source and process dependencies.
17836 (fur_depend::fur_depend): New.
17837 (fur_depend::register_dependency): New.
17838 (class fur_list): New. List source for operands.
17839 (fur_list::fur_list): New.
17840 (fur_list::get_operand): New.
17841 (fur_list::get_phi_operand): New.
17842 (fold_range): New. Instantiate appropriate fur_source class and fold.
17843 (fold_using_range::range_of_range_op): Use new API.
17844 (fold_using_range::range_of_address): Ditto.
17845 (fold_using_range::range_of_phi): Ditto.
17846 (imple_ranger::fold_range_internal): Use fur_depend class.
17847 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
17848 * gimple-range.h (class fur_source): Now a base class.
17849 (class fur_stmt): New.
17850 (fold_range): New prototypes.
17851 (fur_source::fur_source): Delete.
17853 2021-06-08 Andrew Pinski <apinski@marvell.com>
17855 PR tree-optimization/25290
17856 * tree-ssa-phiopt.c (xor_replacement): Delete.
17857 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
17858 (match_simplify_replacement): Allow one cheap preparation
17859 statement that can be moved to before the if.
17861 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
17863 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
17864 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
17866 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
17868 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
17869 Create length attribute on define_insn_and_split. Only split for cases which we
17871 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
17872 fix length computation.
17873 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
17875 2021-06-08 Richard Biener <rguenther@suse.de>
17877 PR tree-optimization/100923
17878 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
17879 the operand vector to be valueized.
17880 (valueize_refs): Likewise.
17881 (valueize_shared_reference_ops_from_ref): Adjust.
17882 (valueize_shared_reference_ops_from_call): Likewise.
17883 (vn_reference_lookup_3): Likewise.
17884 (vn_reference_lookup_pieces): Likewise. Re-valueize
17885 with honoring availability when we are about to create
17886 the ao_ref and valueized before.
17887 (vn_reference_lookup): Likewise.
17888 (vn_reference_insert_pieces): Adjust.
17890 2021-06-08 Richard Biener <rguenther@suse.de>
17892 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
17893 (_slp_instance::root_stmts): ... a vector.
17894 (SLP_INSTANCE_ROOT_STMT): Rename to ...
17895 (SLP_INSTANCE_ROOT_STMTS): ... this.
17896 (slp_root::root): Change to...
17897 (slp_root::roots): ... a vector.
17898 (slp_root::slp_root): Adjust.
17899 * tree-vect-slp.c (_slp_instance::location): Adjust.
17900 (vect_free_slp_instance): Release the root stmt vector.
17901 (vect_build_slp_instance): Adjust.
17902 (vect_analyze_slp): Likewise.
17903 (_bb_vec_info::~_bb_vec_info): Likewise.
17904 (vect_slp_analyze_operations): Likewise.
17905 (vect_bb_vectorization_profitable_p): Likewise. Adjust
17906 costs for the root stmt.
17907 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
17909 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
17911 (vectorize_slp_instance_root_stmt): Adjust.
17912 (vect_schedule_slp): Likewise.
17914 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
17916 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
17917 (ssa_equiv_stack::ssa_equiv_stack): New.
17918 (ssa_equiv_stack::~ssa_equiv_stack): New.
17919 (ssa_equiv_stack::enter): New.
17920 (ssa_equiv_stack::leave): New.
17921 (ssa_equiv_stack::push_replacement): New.
17922 (ssa_equiv_stack::get_replacement): New.
17923 (is_pointer_ssa): New.
17924 (class pointer_equiv_analyzer): New.
17925 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
17926 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
17927 (pointer_equiv_analyzer::set_global_equiv): New.
17928 (pointer_equiv_analyzer::set_cond_equiv): New.
17929 (pointer_equiv_analyzer::get_equiv): New.
17930 (pointer_equiv_analyzer::enter): New.
17931 (pointer_equiv_analyzer::leave): New.
17932 (pointer_equiv_analyzer::get_equiv_expr): New.
17933 (pta_valueize): New.
17934 (pointer_equiv_analyzer::visit_stmt): New.
17935 (pointer_equiv_analyzer::visit_edge): New.
17936 (hybrid_folder::value_of_expr): Call PTA.
17937 (hybrid_folder::value_on_edge): Same.
17938 (hybrid_folder::pre_fold_bb): New.
17939 (hybrid_folder::post_fold_bb): New.
17940 (hybrid_folder::pre_fold_stmt): New.
17941 (rvrp_folder::pre_fold_bb): New.
17942 (rvrp_folder::post_fold_bb): New.
17943 (rvrp_folder::pre_fold_stmt): New.
17944 (rvrp_folder::value_of_expr): Call PTA.
17945 (rvrp_folder::value_on_edge): Same.
17947 2021-06-08 Jakub Jelinek <jakub@redhat.com>
17950 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
17951 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
17953 2021-06-08 Richard Biener <rguenther@suse.de>
17955 PR middle-end/100951
17956 * tree-vect-generic.c (expand_vector_piecewise): Build a
17957 VECTOR_CST if all elements are constant.
17958 (expand_vector_condition): Likewise.
17959 (lower_vec_perm): Likewise.
17960 (expand_vector_conversion): Likewise.
17962 2021-06-08 Martin Liska <mliska@suse.cz>
17964 * doc/invoke.texi: Document new param evrp-sparse-threshold.
17966 2021-06-08 Martin Liska <mliska@suse.cz>
17968 * genautomata.c (create_automata): Fix typo.
17970 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
17972 PR tree-optimization/100794
17973 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
17974 allow_unroll_p and only allow unrolling when it's true.
17975 (tree_predictive_commoning): Add parameter allow_unroll_p and
17977 (run_tree_predictive_commoning): Likewise.
17978 (pass_predcom::gate): Check flag_tree_loop_vectorize and
17979 global_options_set.x_flag_predictive_commoning.
17980 (pass_predcom::execute): Adjust for allow_unroll_p.
17982 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
17984 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
17985 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
17986 lambda function cleanup, remove scev_reset call, and adjust return
17988 (tree_predictive_commoning): Adjust for different changed values,
17989 only set flag TODO_update_ssa_only_virtuals if changed.
17990 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
17991 from todo_flags_finish.
17993 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
17995 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
17996 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
17997 (sbr_sparse_bitmap::bitmap_set_quad): New.
17998 (sbr_sparse_bitmap::bitmap_get_quad): New.
17999 (sbr_sparse_bitmap::set_bb_range): New.
18000 (sbr_sparse_bitmap::get_bb_range): New.
18001 (sbr_sparse_bitmap::bb_range_p): New.
18002 (block_range_cache::block_range_cache): initialize bitmap obstack.
18003 (block_range_cache::~block_range_cache): Destruct obstack.
18004 (block_range_cache::set_bb_range): Decide when to utilze the
18005 sparse on entry cache.
18006 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
18007 * params.opt (-param=evrp-sparse-threshold): New.
18009 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
18011 * bitmap.c (bitmap_set_aligned_chunk): New.
18012 (bitmap_get_aligned_chunk): New.
18013 (test_aligned_chunk): New.
18014 (bitmap_c_tests): Call test_aligned_chunk.
18015 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
18017 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
18020 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18022 (ix86_expand_vector_init_one_nonzero): Ditto.
18023 (ix86_expand_vector_init_one_var): Ditto.
18024 (ix86_expand_vector_init_general): Ditto.
18025 * config/i386/mmx.md (vec_initv4qiqi): New expander.
18027 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
18029 * config/h8300/movepush.md: Change most _clobber_flags
18030 patterns to instead use <cczn> subst.
18031 (movsi_cczn): New pattern with usable CC cases split out.
18032 (movsi_h8sx_cczn): Likewise.
18034 2021-06-07 Martin Liska <mliska@suse.cz>
18036 * common/common-target.def: Split long lines and replace them
18038 * target.def: Likewise.
18039 * doc/tm.texi: Re-generated.
18041 2021-06-07 Jakub Jelinek <jakub@redhat.com>
18044 * fold-const.c (fold_read_from_vector): Return NULL if trying to
18045 read from a CONSTRUCTOR with vector type elements.
18047 2021-06-07 Jakub Jelinek <jakub@redhat.com>
18049 PR middle-end/100898
18050 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
18051 should copy any arguments. Don't call gimple_call_num_args
18052 on id->call_stmt or call_stmt more than once.
18054 2021-06-07 liuhongt <hongtao.liu@intel.com>
18057 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
18059 (<insn>v4siv4di2): Delete constraints for define_expand.
18061 2021-06-07 liuhongt <hongtao.liu@intel.com>
18064 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
18065 assignment of cfun->machine->has_explicit_vzeroupper.
18066 * config/i386/i386-features.c
18067 (ix86_add_reg_usage_to_vzerouppers): Delete.
18068 (ix86_add_reg_usage_to_vzeroupper): Ditto.
18069 (rest_of_handle_insert_vzeroupper): Remove
18070 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
18072 (gate): Remove cfun->machine->has_explicit_vzeroupper.
18073 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
18075 * config/i386/i386.c (ix86_insn_callee_abi): New function.
18076 (ix86_initialize_callee_abi): Ditto.
18077 (ix86_expand_avx_vzeroupper): Ditto.
18078 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
18080 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
18081 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
18083 * config/i386/i386.h (struct GTY(()) machine_function): Delete
18084 has_explicit_vzeroupper.
18085 * config/i386/i386.md (enum unspec): New member
18087 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
18088 define_constants for insn callee abi index.
18089 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
18090 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
18091 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
18092 (*avx_vzeroupper): Rename to ..
18093 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
18094 call_insn which has a special vzeroupper ABI.
18095 (*avx_vzeroupper_1): Deleted.
18097 2021-06-07 liuhongt <hongtao.liu@intel.com>
18100 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
18101 it won't use stack pointer reg.
18102 * final.c (leaf_function_p): When call_insn is a fake call, it
18103 won't affect caller as a leaf function.
18104 * reg-stack.c (callee_clobbers_any_stack_reg): New.
18105 (subst_stack_regs): When call_insn doesn't clobber any stack
18106 reg, don't clear the arguments.
18107 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
18109 * shrink-wrap.c (requires_stack_frame_p): No need for stack
18110 frame for a fake call.
18111 * rtl.h (FAKE_CALL_P): New macro.
18113 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
18115 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
18117 (sparc_order_regs_for_local_alloc): ...this.
18118 (sparc_leaf_reg_remap): Declare.
18119 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
18120 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
18121 * config/sparc/sparc.c (leaf_reg_remap): Delete.
18122 (order_regs_for_local_alloc): Rename to...
18123 (sparc_order_regs_for_local_alloc): ...this.
18124 (sparc_leaf_reg_remap): New function.
18125 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
18127 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
18129 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
18130 Use assemble_name to output BSS section name.
18132 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
18134 * config/i386/constraints.md (Bs):
18135 Remove boolean operators from match_test RTX.
18138 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
18141 2021-06-06 Martin Liska <mliska@suse.cz>
18143 * doc/extend.texi: Add missing @headitem.
18144 * doc/invoke.texi: Likewise.
18145 * doc/objc.texi: Likewise.
18147 2021-06-06 Martin Liska <mliska@suse.cz>
18149 * genhooks.c (emit_findices): Remove unused function.
18150 (emit_documentation): Do not call emit_findices
18151 and do not search for @Fcode directives.
18153 2021-06-06 Martin Liska <mliska@suse.cz>
18155 * doc/invoke.texi: Remove extra character.
18157 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18159 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
18161 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18163 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
18164 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
18165 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
18166 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
18168 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18170 * config/or1k/or1k.md (*movdi): Fix empty split condition.
18172 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18174 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
18177 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18179 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
18180 *zero_extendsidi2): Fix empty split condition.
18182 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
18184 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
18186 * config/h8300/bitfield.md: Likewise.
18187 * config/h8300/combiner.md: Likewise.
18188 * config/h8300/divmod.md: Likewise.
18189 * config/h8300/extensions.md: Likewise.
18190 * config/h8300/jumpcall.md: Likewise.
18191 * config/h8300/movepush.md: Likewise.
18192 * config/h8300/multiply.md: Likewise.
18193 * config/h8300/other.md: Likewise.
18194 * config/h8300/shiftrotate.md: Likewise.
18195 * config/h8300/logical.md: Likewise. Fix split pattern to use
18196 code iterator that somehow slipped through.
18198 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
18200 PR middle-end/100905
18201 * tree-nested.c (convert_nonlocal_omp_clauses,
18202 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
18204 2021-06-04 Martin Sebor <msebor@redhat.com>
18206 PR middle-end/100732
18207 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
18208 with either source or destination argument of invalid type.
18209 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
18210 calls with arguments of invalid type.
18212 2021-06-04 Martin Sebor <msebor@redhat.com>
18214 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
18216 (attr_access::vla_bounds): Also handle VLA bounds.
18218 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
18220 * config/i386/predicates.md (GOT_memory_operand):
18221 Implement using match_code RTXes.
18222 (GOT32_symbol_operand): Ditto.
18224 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
18227 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18229 (ix86_expand_vector_init_general): Ditto.
18230 Use SImode instead of word_mode for logic operations
18231 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
18232 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
18233 implemented by expand_vec_perm_1.
18234 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
18235 should be implemented using standard shuffle patterns.
18236 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
18237 V2HI modes to modes, implementable with shuffle for one operand.
18238 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
18239 (*pshufw_1): New insn pattern.
18240 (*vec_dupv2hi): Ditto.
18241 (vec_initv2hihi): New expander.
18243 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
18245 * config/arm/vfp.md (no_literal_pool_df_immediate,
18246 no_literal_pool_sf_immediate): Fix empty split condition.
18248 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
18250 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
18251 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
18252 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
18253 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
18254 *<sse4_1_avx2>_pblendvb_lt): Likewise.
18256 2021-06-04 Jakub Jelinek <jakub@redhat.com>
18259 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
18260 concatenation from half-sized modes with TImode elements.
18262 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
18264 * config/arc/arc.c (arc_override_options): Disable millicode
18265 thunks when RF16 is on.
18267 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18269 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
18271 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18273 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
18274 Replace PROMOTE_MODE marco with its content.
18276 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18278 * config/cris/cris.md (*addi_reload): Fix empty split condition.
18280 2021-06-03 Jim Wilson <jimw@sifive.com>
18282 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
18283 turn it on for all riscv targets.
18285 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
18288 * config/i386/i386-expand.c (ix86_expand_vector_set):
18289 Handle V2HI and V4QI modes.
18290 (ix86_expand_vector_extract): Ditto.
18291 * config/i386/mmx.md (*pinsrw): New insn pattern.
18294 (*pextrw_zext): Ditto.
18296 (*pextrb_zext): Ditto.
18297 (vec_setv2hi): New expander.
18298 (vec_extractv2hihi): Ditto.
18299 (vec_setv4qi): Ditto.
18300 (vec_extractv4qiqi): Ditto.
18301 (vec_setv8qi): Enable only for TARGET_SSE4_1.
18302 (vec_extractv8qiqi): Ditto.
18304 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
18306 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
18307 order to subf instruction.
18308 * config/rs6000/fusion.md: Regenerate.
18310 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18312 * calls.c (get_size_range): Use range_of_expr instead of
18313 determine_value_range.
18314 * tree-affine.c (expr_to_aff_combination): Same.
18315 * tree-data-ref.c (split_constant_offset): Same.
18316 * tree-vrp.c (determine_value_range_1): Remove.
18317 (determine_value_range): Remove.
18318 * tree-vrp.h (determine_value_range): Remove.
18320 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18322 * function-tests.c (test_ranges): Call gimple_range_tests.
18323 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
18325 * gimple-range.cc (fur_source::get_operand): Do not call
18326 get_tree_range or gimple_range_global.
18328 (get_tree_range): Move to value-query.cc.
18329 Call get_arith_expr_range.
18330 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
18331 Include gimple-range-tests.cc.
18332 * gimple-range.h (fold_range): Add argument.
18333 (get_tree_range): Remove.
18334 * selftest.h (gimple_range_tests): New.
18335 * value-query.cc (global_range_query::range_of_expr): Add
18337 (range_query::get_tree_range): Move from gimple-range.cc.
18338 * value-query.h (class range_query): Add get_tree_range and
18339 get_arith_expr_range. Make fur_source a friend.
18340 * vr-values.c (vr_values::range_of_expr): Pass stmt to
18342 * gimple-range-tests.cc: New file.
18344 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18346 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
18347 update_global_range.
18348 * value-query.cc (update_global_range): New.
18349 * value-query.h (update_global_range): New.
18351 2021-06-03 David Malcolm <dmalcolm@redhat.com>
18353 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
18354 printing the same location twice if there are fix-it hints,
18355 multiple locations, or a label.
18357 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
18359 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
18360 thresholds to narrow the upper bound on epilogue iterations.
18362 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
18364 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
18365 (mve_vabsq_s<mode>): Likewise.
18366 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
18367 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
18368 * config/arm/vec-common.md (neg<mode>2): Rename to
18369 <absneg_str><mode>2.
18371 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
18373 * common/config/arc/arc-common.c (arc_option_optimization_table):
18374 Remove malign-call.
18375 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
18376 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
18377 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
18378 * config/arc/arc.md (abssi2_mixed): Remove pattern.
18379 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
18380 (malign-call): Likewise.
18381 (mmixed-code): Likewise.
18382 * doc/invoke.texi (ARC): Update doc.
18384 2021-06-03 Martin Liska <mliska@suse.cz>
18386 * common.opt: Use proper Enum values.
18387 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
18388 (parse_sanitizer_options): Handle only sanitizer_opts.
18389 (common_handle_option): Just assign value.
18391 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18394 * tree-inline.c (inline_forbidden_p): Remove test on return type.
18396 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18398 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
18399 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
18400 (type_byte_size): Inline into...
18401 (add_byte_size_attribute): ...this and call add_scalar_info.
18403 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18405 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
18406 (typed_binop_from_tree): New function.
18407 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
18408 turn a divide by a power of 2 into a shift.
18409 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
18410 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
18411 typed divide by calling typed_binop_from_tree.
18413 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18415 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
18416 (is_handled_procedure_type): Likewise.
18417 (struct loc_descr_context): Add strict_signedness field.
18418 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
18419 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
18420 (resolve_args_picking): Minor tweak.
18421 (function_to_dwarf_procedure): Initialize strict_signedness field.
18422 (type_byte_size): Likewise.
18423 (field_byte_offset): Likewise.
18424 (gen_descr_array_type_die): Likewise.
18425 (gen_variant_part): Likewise.
18426 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
18427 to true when a context is present before evaluating the arguments.
18428 <COND_EXPR>: Do not generate a useless comparison with zero.
18429 When dereferencing an address, if strict_signedness is true and the
18430 type is small and signed, use DW_OP_deref_type to do the dereference
18431 and then DW_OP_convert to convert back to the generic type.
18433 2021-06-03 Jakub Jelinek <jakub@redhat.com>
18436 * tree-inline.c (copy_tree_body_r): Handle iterators on
18437 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
18439 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18441 * config/arc/arc.md (*bbit_di): Remove.
18443 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
18445 PR rtl-optimization/100264
18446 * ree.c (get_sub_rtx): Ignore SET expressions without register
18447 destinations and remove assertion, as it is not valid anymore
18448 with this new behaviour.
18449 (merge_def_and_ext): Eliminate destination check for register
18450 as such SET expressions can't occur anymore.
18451 (combine_reaching_defs): Likewise.
18453 2021-06-02 Jakub Jelinek <jakub@redhat.com>
18456 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
18457 -Wtype-limits warnings.
18458 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
18459 in operands to avoid -Wsign-compare warnings.
18461 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
18463 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
18466 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
18468 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
18470 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
18472 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
18474 * config/s390/subst.md(cconly_subst): Use a single constraint
18475 in (match_scratch).
18477 2021-06-02 Martin Liska <mliska@suse.cz>
18479 * ipa-icf.h: Use auto_vec for memory_access_types.
18481 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
18483 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
18484 argument from prototype.
18485 (output_logical_op): Add rtx_code argument.
18486 (compute_logical_op_length): Likewise.
18487 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
18488 to compute_a_shift_length.
18489 (output_logical_op); New argument with the rtx code rather than
18490 extracting it from an operand. Handle QImode too.
18491 (compute_logical_op_length): Similary.
18492 (compute_a_shift_length): Drop unused argument.
18493 * config/h8300/h8300.md (logicals): New code iterator.
18494 * config/h8300/logical.md (<code><mode>3 expander): Combine
18495 the "and" expander with the "ior"/"xor" expander.
18496 (bclr<mode>msx): Combine the QI/HI mode patterns.
18497 (<logical><mode>3 insns): Use code iterator rather than match_operator.
18498 Handle QImode as well. Update call to output_logical_op and
18499 compute_logical_op_length to pass in rtx_code
18500 Fix split condition on all define_insn_and_split patterns.
18501 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
18502 the flags and setting ZN via existing define_subst.
18503 * config/h8300/shiftrotate.md: Drop unused argument from
18504 calls to compute_a_shift_length.
18505 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
18507 2021-06-01 Andrew Pinski <apinski@marvell.com>
18509 PR tree-optimization/25290
18510 * tree-ssa-phiopt.c (match_simplify_replacement):
18512 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
18513 (two_value_replacement): Change the comment about
18514 conditional_replacement.
18515 (conditional_replacement): Delete.
18517 2021-06-01 Andrew Pinski <apinski@marvell.com>
18519 PR tree-optimization/95481
18520 * tree-tailcall.c (find_tail_calls): Handle empty typed
18523 2021-06-01 Andrew Pinski <apinski@marvell.com>
18525 * gimplify.c (zero_sized_field_decl): Delete
18526 (zero_sized_type): Delete
18527 (gimplify_init_ctor_eval): Use is_empty_type instead
18528 of zero_sized_field_decl.
18529 (gimplify_modify_expr): Use is_empty_type instead of
18532 2021-06-01 Jason Merrill <jason@redhat.com>
18535 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
18537 2021-06-01 Jason Merrill <jason@redhat.com>
18540 * diagnostic.h (warning_enabled_at): Declare.
18541 * diagnostic.c (diagnostic_enabled): Factor out from...
18542 (diagnostic_report_diagnostic): ...here.
18543 (warning_enabled_at): New.
18545 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
18547 * gimple-ssa-evrp.c: Enable exporting of global ranges.
18549 2021-06-01 Martin Liska <mliska@suse.cz>
18552 * doc/invoke.texi: Mention that -fgcse-after-reload
18553 is enabled with -O3.
18555 2021-06-01 liuhongt <hongtao.liu@intel.com>
18557 PR tree-optimization/98365
18558 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
18559 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
18560 (convert_scalar_cond_reduction): Ditto.
18561 (predicate_scalar_phi): Ditto.
18563 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18565 PR tree-optimization/100781
18566 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
18567 value calculation by default.
18568 (ranger_cache::enable_new_values): New.
18569 (ranger_cache::disable_new_values): New.
18570 (ranger_cache::push_poor_value): Check if new values are allowed.
18571 * gimple-range-cache.h (class ranger_cache): New member/methods.
18572 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
18573 statement, and disable/renable new value calculation.
18575 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18577 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
18578 (ranger_cache::range_of_def): New.
18579 (ranger_cache::entry_range): New.
18580 (ranger_cache::exit_range): New.
18581 (ranger_cache::range_of_expr): Adjust.
18582 (ranger_cache::range_on_edge): Adjust.
18583 (ranger_cache::propagate_cache): Call exit_range directly.
18584 * gimple-range-cache.h (class ranger_cache): Adjust.
18586 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18588 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
18589 gori_compute being a member rather than base class.
18590 dervied call to member call.
18591 (ranger_cache::dump): No longer dump gori_map.
18592 (ranger_cache::dump_bb): New.
18593 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
18594 being a member rather than base class.
18595 (ranger_cache::set_global_range): Ditto.
18596 (ranger_cache::ssa_range_in_bb): Ditto.
18597 (ranger_cache::range_of_expr): New.
18598 (ranger_cache::range_on_edge): New.
18599 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
18600 (ranger_cache::propagate_cache): Adjust debugging output.
18601 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
18603 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
18604 member, and inherit from range_query instead.
18605 (ranger_cache::dump_bb): New. split from dump.
18606 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
18607 (gori_compute::expr_range_at_stmt): Delete.
18608 (gori_compute::compute_name_range_op): Delete.
18609 (gori_compute::compute_operand_range_switch): Add fur_source.
18610 (gori_compute::compute_operand_range): Add fur_source param, inline
18611 old compute_name_range_op and optimize_logical_operands.
18612 (struct tf_range): Delete.
18613 (gori_compute::logical_combine): Adjust
18614 (gori_compute::optimize_logical_operands): Delete.
18615 (gori_compute::compute_logical_operands_in_chain): Delete.
18616 (gori_compute::compute_logical_operands): Adjust.
18617 (gori_compute::compute_operand1_range): Adjust to fur_source.
18618 (gori_compute::compute_operand2_range): Ditto.
18619 (gori_compute::compute_operand1_and_operand2_range): Ditto.
18620 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
18621 and adjust to fur_source.
18622 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
18623 range_query and fur_source.
18624 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
18625 from the ranger_cache..
18626 (gimple_ranger::fold_range_internal): Adjust to base class change of
18628 (gimple_ranger::dump_bb): Adjust dump.
18629 * gimple-range.h (gimple_ranger):export gori computes object.
18631 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18633 PR tree-optimization/100774
18634 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
18635 Constant values are also not stale.
18636 (ranger_cache::set_global_range): Range invariant values should also
18637 have the correct timestamp.
18639 2021-05-31 Martin Liska <mliska@suse.cz>
18641 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
18642 Unpack FUNCTION_DECL_DECL_TYPE.
18643 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
18644 Stream FUNCTION_DECL_DECL_TYPE instead of
18645 DECL_IS_OPERATOR_NEW_P.
18646 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
18648 (DECL_IS_OPERATOR_NEW_P): Likewise.
18649 (DECL_IS_OPERATOR_DELETE_P): Likewise.
18650 (DECL_LAMBDA_FUNCTION_P): Likewise.
18652 2021-05-31 Richard Biener <rguenther@suse.de>
18655 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
18656 * internal-fn.def (SHUFFLEVECTOR): New.
18657 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
18658 * doc/extend.texi: Document __builtin_shufflevector.
18660 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
18663 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
18664 indexed form addresses.
18666 2021-05-29 Jeff Law <jlaw@tachyum.com>
18668 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
18669 parameter. Call callers fixed.
18671 (output_plussi): Add FALLTHRU markers.
18672 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
18674 2021-05-29 Jakub Jelinek <jakub@redhat.com>
18676 PR middle-end/99928
18677 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
18678 combined with parallel, make sure to add shared clause to
18679 parallel for explicit linear clause.
18681 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
18683 PR tree-optimization/100787
18684 * gimple-ssa-evrp.c: Disable exporting of global ranges.
18686 2021-05-28 Jason Merrill <jason@redhat.com>
18688 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
18689 operator--, operator*, operator==, and operator!=.
18690 (class tsi_range): New.
18692 2021-05-28 Richard Biener <rguenther@suse.de>
18694 PR tree-optimization/100778
18695 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
18696 trapping ops in different BBs.
18698 2021-05-28 Richard Biener <rguenther@suse.de>
18701 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
18702 copy fntype from original call.
18704 2021-05-28 Martin Liska <mliska@suse.cz>
18706 PR gcov-profile/100751
18707 * doc/gcov.texi: Revert partially a hunk that was wrong.
18709 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
18711 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
18713 (HAVE_sync_compare_and_swaphi): Likewise.
18714 (HAVE_sync_compare_and_swapsi): Likewise.
18716 2021-05-28 Jakub Jelinek <jakub@redhat.com>
18718 PR middle-end/99928
18719 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
18721 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
18723 * gimplify.c (gimplify_omp_affinity): New.
18724 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
18725 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
18726 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
18727 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
18728 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
18730 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
18731 Richard Biener <rguenther@suse.de>
18733 * match.pd <popcount & / + pattern matching>:
18734 When generating popcount directly fails, try doing it in two halves.
18736 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
18738 * Makefile.in (generated_files): Add gimple-match.c and
18741 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
18743 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
18745 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
18747 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
18749 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
18751 PR tree-optimization/99398
18752 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
18753 where the fed operands are CTOR/CST and propagated through
18754 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
18755 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
18757 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
18760 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18762 * config/i386/mmx.md (addv2sf3): Do not call
18763 ix86_fixup_binary_operands_no_copy.
18766 (<smaxmin:code>v2sf3): Ditto.
18767 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
18768 (<plusminus:insn><VI_32:mode>3): Remove expander.
18769 (<plusminus:insn><VI_32:mode>3): Rename from
18770 "*<plusminus:insn><VI_32:mode>3".
18771 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
18772 (mulv2hi3): Remove expander.
18773 (mulv2hi3): Rename from *mulv2hi3.
18774 (<s>mulv2hi3_highpart): Remove expander.
18775 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
18776 (<smaxmin:code><MMXMODE14:mode>3): Rename from
18777 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
18778 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
18779 (SMAXMIN_MMXMODEI): Remove mode iterator.
18780 (<smaxmin:code>v4hi3): New expander.
18781 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
18782 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
18783 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
18784 (SMAXMIN_VI_32): Remove mode iterator.
18785 (<umaxmin:code><MMXMODE24:mode>3): Rename from
18786 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
18787 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
18788 (UMAXMIN_MMXMODEI): Remove mode iterator.
18789 (<umaxmin:code>v8qi3): New expander.
18790 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
18791 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
18792 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
18793 (UMAXMIN_VI_32): Remove mode iterator.
18794 (<any_shift:insn>v2hi3): Remove expander.
18795 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
18796 (<any_logic:code><MMXMODEI:mode>3): Do not call
18797 ix86_fixup_binary_operands_no_copy.
18798 (<any_logic:code><VI_32:mode>3): Remove expander.
18799 (<any_logic:code><VI_32:mode>3): Rename from
18800 "*<any_logic:code><VI_32:mode>3".
18801 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
18802 * config/i386/sse.md (div<VF2:mode>3): Do not call
18803 ix86_fixup_binary_operands_no_copy.
18804 (div<VF1:mode>3): Ditto.
18805 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
18806 (smulhrsv4hi3): Ditto.
18807 (smulhrsv2hi3): Ditto.
18809 2021-05-27 Martin Sebor <msebor@redhat.com>
18811 * ggc.h (gt_ggc_mx): Add overloads for all integers.
18813 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
18815 (hash_map::operator==): New function.
18817 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18820 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
18821 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
18822 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
18823 (*xop_maskcmp<VI_32:mode>3): Ditto.
18824 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
18825 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
18827 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
18830 * config/arm/arm.c (arm_configure_build_target): Remove parameter
18831 opts_set, directly check opts parameters for being non-null.
18832 (arm_option_restore): Update call to arm_configure_build_target.
18833 (arm_option_override): Likewise.
18834 (arm_can_inline_p): Likewise.
18835 (arm_valid_target_attribute_tree): Likewise.
18836 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
18837 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
18840 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
18842 * vr-values.c (simplify_conversion_using_ranges): Use
18843 get_range_query instead of get_global_range_query.
18845 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
18847 * gimple-range.cc (get_range_global): Move to value-query.cc.
18848 (gimple_range_global): Same.
18849 (get_global_range_query): Same.
18850 (global_range_query::range_of_expr): Same.
18851 * gimple-range.h (class global_range_query): Move to
18853 (gimple_range_global): Same.
18854 * tree-ssanames.c (get_range_info): Move to value-query.cc.
18855 (get_ptr_nonnull): Same.
18856 * tree-ssanames.h (get_range_info): Remove.
18857 (get_ptr_nonnull): Remove.
18858 * value-query.cc (get_ssa_name_range_info): Move from
18860 (get_ssa_name_ptr_info_nonnull): Same.
18861 (get_range_global): Move from gimple-range.cc.
18862 (gimple_range_global): Same.
18863 (get_global_range_query): Same.
18864 (global_range_query::range_of_expr): Same.
18865 * value-query.h (class global_range_query): Move from
18867 (gimple_range_global): Same.
18869 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18872 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
18873 (uavgv2hi3_ceil): Ditto.
18875 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
18878 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
18880 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18882 * tree-ssanames.c (get_range_info): Merge both copies of
18883 get_range_info into one that works with irange.
18884 * tree-ssanames.h (get_range_info): Remove version that works on
18887 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18889 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
18890 (expand_builtin_strnlen): Same.
18891 (determine_block_size): Same.
18892 * fold-const.c (expr_not_equal_to): Same.
18893 * gimple-fold.c (size_must_be_zero_p): Same.
18894 * gimple-match-head.c: Include gimple-range.h.
18895 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
18896 * gimple-ssa-warn-restrict.c
18897 (builtin_memref::extend_offset_range): Same.
18898 * graphite-sese-to-poly.c (add_param_constraints): Same.
18899 * internal-fn.c (get_min_precision): Same.
18900 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
18901 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
18903 * tree-data-ref.c (split_constant_offset): Same.
18904 (dr_step_indicator): Same.
18905 * tree-dfa.c (get_ref_base_and_extent): Same.
18906 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
18907 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
18908 (determine_value_range): Same.
18909 (record_nonwrapping_iv): Same.
18910 (infer_loop_bounds_from_signedness): Same.
18911 (scev_var_range_cant_overflow): Same.
18912 * tree-ssa-phiopt.c (two_value_replacement): Same.
18913 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
18914 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
18915 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
18917 (dump_strlen_info): Same.
18918 (set_strlen_range): Same.
18919 (maybe_diag_stxncpy_trunc): Same.
18920 (get_len_or_size): Same.
18921 (handle_integral_assign): Same.
18922 * tree-ssa-structalias.c (find_what_p_points_to): Same.
18923 * tree-ssa-uninit.c (find_var_cmp_const): Same.
18924 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
18925 * tree-vect-patterns.c (vect_get_range_info): Same.
18926 (vect_recog_divmod_pattern): Same.
18927 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
18928 (register_edge_assert_for_2): Same.
18929 (determine_value_range_1): Same.
18930 * tree.c (get_range_pos_neg): Same.
18931 * vr-values.c (vr_values::get_lattice_entry): Same.
18932 (vr_values::update_value_range): Same.
18933 (simplify_conversion_using_ranges): Same.
18935 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18937 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
18938 get_range_query instead of query argument.
18939 (pass_walloca::execute): Enable and disable global ranger.
18941 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18943 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
18945 (rvrp_folder::~rvrp_folder): Call disable_ranger.
18946 (hybrid_folder::hybrid_folder): Call enable_ranger.
18947 (hybrid_folder::~hybrid_folder): Call disable_ranger.
18949 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18951 * function.c (allocate_struct_function): Set cfun->x_range_query.
18952 * function.h (struct function): Declare x_range_query.
18953 (get_range_query): New.
18954 (get_global_range_query): New.
18955 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
18956 Remove call to safe_grow_cleared.
18957 * gimple-range.cc (get_range_global): New.
18958 (gimple_range_global): Move from gimple-range.h.
18959 (get_global_range_query): New.
18960 (global_range_query::range_of_expr): New.
18961 (enable_ranger): New.
18962 (disable_ranger): New.
18963 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
18964 (class global_range_query): New.
18965 (enable_ranger): New.
18966 (disable_ranger): New.
18967 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
18968 dump_all_value_ranges to dump.
18969 * tree-vrp.c (vrp_prop::finalize): Same.
18970 * value-query.cc (range_query::dump): New.
18971 * value-query.h (range_query::dump): New.
18972 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
18973 (vr_values::dump): ...this.
18974 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
18975 dump and make virtual.
18977 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
18979 * config/i386/i386.c (ix86_autovectorize_vector_modes):
18980 Add V4QImode and V16QImode for TARGET_SSE2.
18981 * doc/sourcebuild.texi (Vector-specific attributes):
18982 Add vect64 and vect32 description.
18984 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
18986 * gimple-range-gori.cc (range_def_chain::register_dependency):
18987 Resize m_def_chain when needed.
18989 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
18991 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
18992 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
18993 * config/arm/vec-common.md: .. here. Add support for MVE.
18995 2021-05-26 Jakub Jelinek <jakub@redhat.com>
18997 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
18999 * config/microblaze/microblaze.c (microblaze_legitimize_address,
19001 microblaze_option_override, print_operand): Likewise.
19002 * config/microblaze/microblaze.md (call_internal_plt,
19003 call_value_intern_plt, call_value_intern): Likewise.
19004 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
19005 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
19006 call_value_multiple_internal1): Likewise.
19007 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
19009 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
19011 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
19012 arc_ccfsm_advance, symbolic_reference_mentioned_p,
19013 arc_raw_symbolic_reference_mentioned_p): Remove register
19016 2021-05-26 Jakub Jelinek <jakub@redhat.com>
19019 * omp-low.c: Include omp-offload.h.
19020 (create_omp_child_function): If current_function_decl has
19021 "omp declare target" attribute and is_gimple_omp_offloaded,
19022 remove that attribute from the copy of attribute list and
19023 add "omp target entrypoint" attribute instead.
19024 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
19025 variables for offloading if in omp_maybe_offloaded_ctx.
19026 * omp-offload.c (pass_omp_target_link::execute): Nullify second
19027 argument to GOMP_target_data_ext in offloaded code.
19029 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19031 * config/csky/csky.c (csky_can_change_mode_class): Delete.
19032 For csky, HF/SF mode use the low bits of VREGS.
19034 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
19036 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
19037 DECL which is a reference for OMP.
19039 2021-05-26 Martin Liska <mliska@suse.cz>
19041 PR gcov-profile/100751
19042 * doc/gcov.texi: Document that __gcov_dump can be called just
19043 once and that __gcov_reset resets run-time counters.
19045 2021-05-26 Martin Liska <mliska@suse.cz>
19047 * doc/install.texi: Port relevant part from install-old.texi
19048 and re-generate list of CPUs and systems.
19050 2021-05-26 Martin Liska <mliska@suse.cz>
19052 * Makefile.in: Remove it.
19053 * doc/include/fdl.texi: Update next/previous chapters.
19054 * doc/install.texi: Likewise.
19055 * doc/install-old.texi: Removed.
19057 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19059 * config/csky/csky.c (ck810_legitimate_index_p): Support
19060 "base + index" with DF mode.
19061 * config/csky/constraints.md ("Y"): New constraint for memory operands
19062 without index register.
19063 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
19064 when mov between memory and general registers, and lower their priority.
19065 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
19067 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19069 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
19071 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19073 * config/csky/csky.md (untyped_call): Emit clobber for return
19074 registers to mark them used.
19076 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19078 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
19080 2021-05-26 Andrew Pinski <apinski@marvell.com>
19082 * match.pd (x < 0 ? ~y : y): New patterns.
19084 2021-05-26 Andrew Pinski <apinski@marvell.com>
19086 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
19087 A?POW2:0 and A?0:POW2.
19089 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19091 * gimple-range-gori.cc (class logical_stmt_cache): Delete
19092 (logical_stmt_cache::logical_stmt_cache ): Delete.
19093 (logical_stmt_cache::~logical_stmt_cache): Delete.
19094 (logical_stmt_cache::cache_entry::dump): Delete.
19095 (logical_stmt_cache::get_range): Delete.
19096 (logical_stmt_cache::cached_name ): Delete.
19097 (logical_stmt_cache::same_cached_name): Delete.
19098 (logical_stmt_cache::cacheable_p): Delete.
19099 (logical_stmt_cache::slot_diagnostics ): Delete.
19100 (logical_stmt_cache::dump): Delete.
19101 (gori_compute_cache::gori_compute_cache): Delete.
19102 (gori_compute_cache::~gori_compute_cache): Delete.
19103 (gori_compute_cache::compute_operand_range): Delete.
19104 (gori_compute_cache::cache_stmt): Delete.
19105 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
19107 (class gori_compute_cache): Delete.
19109 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19111 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
19113 (fold_using_range::range_of_address): Adjust.
19114 (fold_using_range::range_of_phi): Adjust.
19115 * gimple-range.h (class fur_source): Adjust.
19116 (fur_source::fur_source): Adjust.
19118 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19120 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
19121 from expr_range_in_bb and adjust.
19122 (gori_compute::compute_name_range_op): Adjust.
19123 (gori_compute::optimize_logical_operands): Adjust.
19124 (gori_compute::compute_logical_operands_in_chain): Adjust.
19125 (gori_compute::compute_operand1_range): Adjust.
19126 (gori_compute::compute_operand2_range): Adjust.
19127 (ori_compute_cache::cache_stmt): Adjust.
19128 * gimple-range-gori.h (gori_compute): Rename prototype.
19130 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19132 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
19133 checked only after range_of_stmt, not range_on_entry.
19134 (gimple_ranger::range_on_entry): Check for non-null in any
19135 predecessor block, if it is not already non-null.
19136 (gimple_ranger::range_on_exit): DOnt check for non-null after
19137 range on entry call.
19138 (gimple_ranger::dump_bb): New. Split from dump.
19139 (gimple_ranger::dump): Adjust.
19140 * gimple-range.h (class gimple_ranger): Adjust.
19142 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19144 * gimple-range-cache.cc (struct range_timestamp): Delete.
19145 (class temporal_cache): Adjust.
19146 (temporal_cache::get_timestamp): Delete.
19147 (temporal_cache::set_dependency): Delete.
19148 (temporal_cache::temporal_value): Adjust.
19149 (temporal_cache::current_p): Take dependencies as params.
19150 (temporal_cache::set_timestamp): Adjust.
19151 (temporal_cache::set_always_current): Adjust.
19152 (ranger_cache::get_non_stale_global_range): Adjust.
19153 (ranger_cache::register_dependency): Delete.
19154 * gimple-range-cache.h (class range_cache): Adjust.
19156 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19158 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
19160 (range_def_chain::~range_def_chain): Dispose of obstack rather than
19161 each individual bitmap.
19162 (range_def_chain::set_import): New.
19163 (range_def_chain::get_imports): New.
19164 (range_def_chain::chain_import_p): New.
19165 (range_def_chain::register_dependency): Rename from build_def_chain
19167 (range_def_chain::def_chain_in_bitmap_p): New.
19168 (range_def_chain::add_def_chain_to_bitmap): New.
19169 (range_def_chain::has_def_chain): Just check first depenedence.
19170 (range_def_chain::get_def_chain): Process imports, use generic
19171 register_dependency routine.
19172 (range_def_chain::dump): New.
19173 (gori_map::gori_map): Allocate import list.
19174 (gori_map::~gori_map): Release imports.
19175 (gori_map::exports): Check for past allocated block size.
19176 (gori_map::imports): New.
19177 (gori_map::def_chain_in_export_p): Delete.
19178 (gori_map::is_import_p): New.
19179 (gori_map::maybe_add_gori): Handle imports.
19180 (gori_map::dump): Adjust output, add imports.
19181 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
19182 (gori_export_iterator::gori_export_iterator): New.
19183 (gori_export_iterator::next): New.
19184 (gori_export_iterator::get_name): New.
19185 * gimple-range-gori.h (range_def_chain): Add imports and direct
19186 dependecies via struct rdc.
19187 (range_def_chain::depend1): New.
19188 (range_def_chain::depend2): New.
19189 (class gori_map): Adjust.
19190 (FOR_EACH_GORI_IMPORT_NAME): New.
19191 (FOR_EACH_GORI_EXPORT_NAME): New.
19192 (class gori_export_iterator): New.
19194 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19196 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
19197 export cache filling to here.
19198 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
19200 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19202 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
19203 (gori_map): Move to gimple-range-gori.h.
19204 (gori_compute::gori_compute): Adjust.
19205 (gori_compute::~gori_compute): Delete.
19206 (gori_compute::compute_operand_range_switch): Adjust.
19207 (gori_compute::compute_operand_range): Adjust.
19208 (gori_compute::compute_logical_operands): Adjust.
19209 (gori_compute::has_edge_range_p ): Adjust.
19210 (gori_compute::set_range_invariant): Delete.
19211 (gori_compute::dump): Adjust.
19212 (gori_compute::outgoing_edge_range_p): Adjust.
19213 * gimple-range-gori.h (class range_def_chain): Relocate here.
19214 (class gori_map): Relocate here.
19215 (class gori_compute): Inherit from gori_map, and adjust.
19217 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
19219 * value-range.cc (range_tests_legacy): Use
19220 build_nonstandard_integer_type instead of int and short.
19222 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
19224 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
19225 when really creating an initialization statement for it.
19227 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
19229 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
19231 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
19233 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
19235 2021-05-25 Martin Liska <mliska@suse.cz>
19237 PR tree-optimization/92860
19239 * optc-save-gen.awk: Remove exceptions.
19241 2021-05-25 Martin Liska <mliska@suse.cz>
19243 * asan.h (sanitize_coverage_p): New function.
19244 * doc/extend.texi: Document it.
19245 * fold-const.c (fold_range_test): Use sanitize_flags_p
19246 instead of flag_sanitize_coverage.
19247 (fold_truth_andor): Likewise.
19248 * sancov.c: Likewise.
19249 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
19250 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
19251 -fsanitize-coverage when inlining.
19253 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19255 * config/csky/csky-modes.def : Fix copyright.
19257 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19259 * config/csky/csky-modes.def : Amend copyright.
19260 * config/csky/csky_insn_fpuv2.md : Likewise.
19261 * config/csky/csky_insn_fpuv3.md : Likewise.
19263 2021-05-25 Richard Biener <rguenther@suse.de>
19265 PR middle-end/100727
19266 * calls.c (initialize_argument_information): Explicitely test
19267 for WITH_SIZE_EXPR.
19268 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
19270 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19272 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
19273 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
19274 the signle definition. The signle definition may not work well
19275 at simplify_subreg_regno().
19276 (HARD_FRAME_POINTER_REGNUM): New.
19277 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
19278 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
19279 csky_initial_elimination_offset, csky_expand_prologue,
19280 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
19282 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19284 * config/csky/csky.c (csky_option_override):
19285 Init csky_arch_isa_features[] in advance, so TARGET_DSP
19286 and TARGET_DIV can be set well.
19288 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19290 * config/csky/constraints.md ("l", "h"): Delete.
19291 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
19292 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
19293 * config/csky/csky.c (regno_reg_classm,
19294 csky_secondary_reload, csky_register_move_cost):
19295 Use HILO_REGS instead of LO_REGS and HI_REGS.
19297 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19299 * config/csky/constraints.md ("W"): New constriant for mem operand
19300 with base reg, index register.
19301 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
19302 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
19304 ("Dv"): New constraint for const double value that can be used at
19306 * config/csky/csky-modes.def (HFmode): New mode.
19307 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
19308 to "csky_valid_mem_constraint_operand" and support new constraint
19310 (csky_get_movedouble_length): New.
19311 (fpuv3_output_move): New.
19312 (fpuv3_const_double): New.
19313 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
19314 (decompose_csky_address): Refine.
19315 (csky_print_operand): New "CONST_DOUBLE" operand.
19316 (csky_output_move): Support fpv3 instructions.
19317 (csky_get_movedouble_length): New.
19318 (fpuv3_output_move): New.
19319 (fpuv3_const_double): New.
19320 (csky_emit_compare): Cover float comparsion.
19321 (csky_emit_compare_float): Refine.
19322 (csky_vaild_fpuv2_mem_operand): Rename to
19323 "csky_valid_mem_constraint_operand" and support new constraint "W".
19324 (ck860_rtx_costs): New.
19325 (csky_rtx_costs): Add the cost calculation of CK860.
19326 (regno_reg_class): New vregs for fpuv3.
19327 (csky_dbx_regno): Likewise.
19328 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
19329 (csky_conditional_register_usage): Suporrot fpuv3.
19330 (csky_dwarf_register_span): Suporrot fpuv3.
19331 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
19332 (ck810_legitimate_index_p): Support fp16.
19333 * config/csky/csky.h (TARGET_TLS): ADD CK860.
19334 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
19335 (TARGET_SINGLE_FPU): Support fpuv3.
19336 (TARGET_SUPPORT_FPV3): New.
19337 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
19338 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
19339 REG_CLASS_CONTENTS): Support fpuv3.
19340 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
19341 (csky_movsf_fpv2): Likewise.
19342 (ck801_movsf): Likewise.
19343 (csky_movsf): Likewise.
19345 (csky_movdf_fpv2): Likewise.
19346 (ck801_movdf): Likewise.
19347 (csky_movdf): Likewise.
19348 (movsicc): Refine. Use "comparison_operatior" instead of
19349 "ordered_comparison_operatior".
19350 (addsicc): Likewise.
19351 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
19352 (call_value_internal_vh): New.
19353 * config/csky/csky_cores.def (CK860): New arch and cpu.
19358 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
19359 into emit-patterns and match-patterns, remain the emit-patterns here,
19360 and move the match-patterns to csky_insn_fpuv2.md or
19361 csky_insn_fpuv3.md.
19362 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
19363 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
19365 * config/csky/csky_isa.def (fcr): New.
19370 (CK860): New definition for ck860.
19371 * config/csky/csky_tables.opt (ck860): New processors ck860,
19372 ck860f. And new arch ck860.
19377 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
19378 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
19379 * config/csky/t-csky-elf: Support 860.
19380 * config/csky/t-csky-linux: Likewise.
19381 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
19383 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
19385 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
19386 add generation of logical-add and add-logical fusion pairs.
19387 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
19389 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
19390 logical-add and add-logical fusion by default.
19391 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
19392 -mpower10-fusion-add-logical options.
19393 * config/rs6000/fusion.md: Regenerate file.
19395 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
19397 * value-range.cc (irange::legacy_equal_p): Check type when
19398 comparing VR_VARYING types.
19399 (range_tests_legacy): Test comparing VARYING ranges of different
19402 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19404 * config/aarch64/aarch64.c (neoversen1_tunings):
19405 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
19407 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19409 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
19410 extern weak symbols. Limit symbol offsets for non-GOT symbols with
19413 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19415 * config/arm/neon.md (vec_load_lanesxi<mode>)
19416 (vec_store_lanexoi<mode>): Move ...
19417 * config/arm/vec-common.md: here.
19419 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19421 * config/arm/neon.md (vec_load_lanesoi<mode>)
19422 (vec_store_lanesoi<mode>): Move ...
19423 * config/arm/vec-common.md: here.
19425 2021-05-24 liuhongt <hongtao.liu@intel.com>
19428 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
19429 stmt with GIMPLE_NOP when lhs doesn't exist.
19431 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
19434 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
19435 New instruction pattern.
19436 (*push<VI_32:mode>2): Ditto.
19437 (push splitter for SSE registers): New splitter.
19439 2021-05-23 Andrew Pinski <apinski@marvell.com>
19441 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
19443 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
19445 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
19446 * config/rs6000/fusion.md: Regenerate file.
19448 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19450 * config/rs6000/genfusion.pl (gen_addadd): New function.
19451 * config/rs6000/fusion.md: Regenerate file.
19452 * config/rs6000/rs6000-cpus.def: Add
19453 OPTION_MASK_P10_FUSION_2ADD to masks.
19454 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19455 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
19456 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
19458 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19460 PR middle-end/99928
19461 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
19462 * gimplify.c (enum gimplify_omp_var_data): Fix up
19463 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
19464 (omp_lastprivate_for_combined_outer_constructs): If combined target
19465 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
19466 GOVD_MAP | GOVD_SEEN.
19467 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
19468 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
19469 (gimplify_adjust_omp_clauses): For firstprivate clauses with
19470 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
19471 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
19472 let it be replaced by implicit map clause.
19474 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19476 PR middle-end/99928
19477 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
19479 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
19480 (gimplify_omp_for): Likewise.
19482 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19484 PR middle-end/90115
19485 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
19486 'external' in blocks.
19488 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19490 PR middle-end/90115
19491 * flag-types.h (enum openacc_privatization): New.
19492 * params.opt (-param=openacc-privatization): New.
19493 * doc/invoke.texi (openacc-privatization): Document it.
19494 * omp-general.h (get_openacc_privatization_dump_flags): New
19496 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
19497 * omp-offload.c (execute_oacc_device_lower)
19498 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
19499 * target.def (goacc.adjust_private_decl): Add 'location_t'
19501 * doc/tm.texi: Regenerate.
19502 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
19503 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
19504 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
19505 Likewise. Preserve it for...
19506 (nvptx_goacc_expand_var_decl): ... use here.
19508 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19510 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
19513 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19515 PR middle-end/90115
19516 * omp-low.c (oacc_privatization_candidate_p): New function.
19517 (oacc_privatization_scan_clause_chain)
19518 (oacc_privatization_scan_decl_chain): Use it. Also
19519 'gcc_checking_assert' that we're not seeing duplicates.
19521 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19523 PR middle-end/90115
19524 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
19527 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19529 PR middle-end/90115
19530 * omp-offload.c (execute_oacc_device_lower): Explain.
19532 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19534 PR middle-end/90115
19535 * omp-offload.c (execute_oacc_device_lower)
19536 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
19538 * internal-fn.c (expand_UNIQUE): Don't expect
19539 'IFN_UNIQUE_OACC_PRIVATE'.
19541 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19543 PR middle-end/90115
19544 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
19546 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19548 PR middle-end/90115
19549 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
19550 (nvptx_goacc_expand_var_decl): Tighten.
19552 2021-05-21 Julian Brown <julian@codesourcery.com>
19553 Chung-Lin Tang <cltang@codesourcery.com>
19554 Thomas Schwinge <thomas@codesourcery.com>
19556 PR middle-end/90115
19557 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
19558 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
19559 * doc/tm.texi: Regenerate.
19560 * expr.c (expand_expr_real_1): Expand decls using the
19561 expand_var_decl OpenACC hook if defined.
19562 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
19563 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
19564 * omp-low.c (omp_context): Add oacc_privatization_candidates
19566 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
19568 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
19569 private marker's gimple call arguments, and pass it to
19570 lower_oacc_reductions.
19571 (oacc_privatization_scan_clause_chain)
19572 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
19574 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
19575 * omp-offload.c (convert.h): Include.
19576 (oacc_loop_xform_head_tail): Treat private-variable markers like
19577 fork/join when transforming head/tail sequences.
19578 (struct var_decl_rewrite_info): Add struct.
19579 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
19580 (execute_oacc_device_lower): Support rewriting gang-private
19581 variables using target hook, and fix up addr_expr and var_decl
19583 * target.def (adjust_private_decl, expand_var_decl): New hooks.
19584 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
19586 (gcn_goacc_adjust_private_decl): ...this.
19587 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
19589 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
19590 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
19591 definition using gcn_goacc_adjust_gangprivate_decl...
19592 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
19593 gcn_goacc_adjust_private_decl.
19594 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
19595 (gang_private_shared_size): New global variable.
19596 (gang_private_shared_align): Likewise.
19597 (gang_private_shared_sym): Likewise.
19598 (gang_private_shared_hmap): Likewise.
19599 (nvptx_option_override): Initialize these.
19600 (nvptx_file_end): Output gang_private_shared_sym.
19601 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
19603 (nvptx_set_current_function): Clear gang_private_shared_hmap.
19604 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
19605 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
19607 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
19609 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
19611 2021-05-21 Richard Biener <rguenther@suse.de>
19612 H.J. Lu <hjl.tools@gmail.com>
19614 PR middle-end/90773
19615 * expr.c (expand_constructor): Elide expand_constructor if
19616 move by pieces is preferred.
19618 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19620 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
19621 Take a flag and mode value as arguments.
19622 (aarch64_modifies_global_state_p): Likewise.
19623 (aarch64_reads_global_state_p): Likewise.
19624 (aarch64_could_trap_p): Likewise.
19625 (aarch64_get_attributes): Likewise.
19626 (aarch64_init_simd_builtins): Adjust callsite of above.
19627 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
19628 function attributes to apply to builtins.
19629 (aarch64_init_crc32_builtins): Likewise.
19630 (aarch64_init_builtin_rsqrt): Likewise.
19632 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19634 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
19635 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
19636 (gen_2logical): Use new fusion types.
19637 * config/rs6000/fusion.md: Regenerate.
19639 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19642 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
19643 Handle V4QI and V2HI modes.
19644 (ix86_expand_sse_movcc): Ditto.
19645 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
19646 New instruction pattern.
19647 (*eq<VI_32:mode>3): Ditto.
19648 (*gt<VI_32:mode>3): Ditto.
19649 (*xop_pcmov_<VI_32:mode>): Ditto.
19650 (mmx_pblendvb32): Ditto.
19651 (mmx_pblendvb64): Rename from mmx_pblendvb.
19652 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
19653 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
19654 (vcond<VI_32:mode><VI_32:mode>): Ditto.
19655 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
19656 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
19658 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19660 PR tree-optimization/94589
19661 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
19662 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
19663 not the same as cmp3) treat <= the same as < and >= the same as >.
19664 Don't require that cond2_phi_edge is true edge, instead take
19665 false/true edges into account based on cmp1/cmp2 comparison kinds.
19667 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19670 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
19671 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
19672 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
19673 using SMAXMIN_MMXMODEI mode iterator.
19674 (*<smaxmin:code>v4qi3): New insn pattern.
19675 (*<smaxmin:code>v2hi3): Ditto.
19676 (SMAXMIN_VI_32): New mode iterator.
19677 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
19678 (UMAXMIN_MMXMODEI): New mode iterator.
19679 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
19680 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
19681 using UMAXMIN_MMXMODEI mode iterator.
19682 (*<umaxmin:code>v4qi3): New insn pattern.
19683 (*<umaxmin:code>v2hi3): Ditto.
19684 (UMAXMIN_VI_32): New mode iterator.
19685 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
19686 (abs<VI_32:mode>2): New insn pattern.
19687 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
19688 * config/i386/sse.md: ... here.
19690 2021-05-20 Clement Chigot <clement.chigot@atos.net>
19691 David Edelsohn <dje.gcc@gmail.com>
19693 * collect2.c (scan_prog_file): Issue non-fatal warning for
19696 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
19698 * doc/invoke.texi (-Wno-c++11-extensions)
19699 (-Wno-c++14-extensions, -Wno-c++17-extensions)
19700 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
19703 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
19705 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
19706 * config/darwin.c (darwin_override_options): Likewise.
19707 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
19708 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
19709 (DWARF2_FRAME_REG_OUT): Likewise.
19710 * config/mips/mips.c (mips_output_filename): Likewise.
19711 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
19713 (rs6000_dbx_register_number): Likewise.
19714 * dbxout.c: Include flags.h.
19715 * dwarf2cfi.c (cfi_label_required_p): Likewise.
19716 (dwarf2out_do_frame): Likewise.
19717 * except.c: Include flags.h.
19718 * final.c (dwarf2_debug_info_emitted_p): Likewise.
19719 (final_scan_insn_1): Likewise.
19720 * flags.h (dwarf_debuginfo_p): New function declaration.
19721 * opts.c (dwarf_debuginfo_p): New function definition.
19722 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
19723 * toplev.c (process_options): Likewise.
19725 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
19727 * common.opt: Change type to support bitmasks.
19728 * flag-types.h (enum debug_info_type): Rename enumerator constants.
19729 (NO_DEBUG): New bitmask.
19730 (DBX_DEBUG): Likewise.
19731 (DWARF2_DEBUG): Likewise.
19732 (XCOFF_DEBUG): Likewise.
19733 (VMS_DEBUG): Likewise.
19734 (VMS_AND_DWARF2_DEBUG): Likewise.
19735 * flags.h (debug_set_to_format): New function declaration.
19736 (debug_set_count): Likewise.
19737 (debug_set_names): Likewise.
19738 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
19739 (debug_set_to_format): New function definition.
19740 (debug_set_count): Likewise.
19741 (debug_set_names): Likewise.
19742 (set_debug_level): Update access to debug_type_names.
19743 * toplev.c: Likewise.
19745 2021-05-20 Martin Sebor <msebor@redhat.com>
19747 PR middle-end/100684
19748 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
19750 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
19753 * config/i386/i386.md (isa): Remove x64_bmi.
19754 (enabled): Remove x64_bmi.
19755 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
19756 Remove general register alternative.
19757 (*andnot<VI_32:mode>3): Ditto.
19758 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
19759 (*<any_logic:code><VI_32:mode>3): Ditto.
19761 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
19763 * config/arm/arm.c: Include head files tree-vectorizer.h and
19766 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
19769 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
19770 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
19771 (<s>mulv4hi3_highpart): New expander.
19772 (*<s>mulv2hi3_highpart): New insn pattern.
19773 (<s>mulv2hi3_higpart): New expander.
19774 (*<any_shift:insn>v2hi3): New insn pattern.
19775 (<any_shift:insn>v2hi3): New expander.
19776 * config/i386/sse.md (smulhrsv2hi3): New expander.
19777 (*smulhrsv2hi3): New insn pattern.
19779 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
19781 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
19783 * params.opt (vect-inner-loop-cost-factor): New.
19784 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
19785 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
19786 tree-vectorizer.h and its required ones.
19787 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
19788 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
19789 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
19790 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
19791 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
19792 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19794 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
19795 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
19796 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
19798 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
19799 Torbjörn Svensson <torbjorn.svensson@st.com>
19802 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
19804 2021-05-20 Jakub Jelinek <jakub@redhat.com>
19806 PR middle-end/99928
19807 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
19808 explicit linear clause when combined with target, make it map(tofrom:)
19809 instead of no clause or firstprivate.
19811 2021-05-20 Jakub Jelinek <jakub@redhat.com>
19813 PR tree-optimization/94589
19814 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
19815 of integral conversions.
19817 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
19819 * gimple-range.cc (fur_source::get_operand): New.
19820 (gimple_range_fold): Delete.
19821 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
19822 (fold_using_range::range_of_range_op): Move from gimple_ranger.
19823 (fold_using_range::range_of_address): Ditto.
19824 (fold_using_range::range_of_phi): Ditto.
19825 (fold_using_range::range_of_call): Ditto.
19826 (fold_using_range::range_of_builtin_ubsan_call): Move from
19827 range_of_builtin_ubsan_call.
19828 (fold_using_range::range_of_builtin_call): Move from
19829 range_of_builtin_call.
19830 (gimple_ranger::range_of_builtin_call): Delete.
19831 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
19832 (gimple_ranger::fold_range_internal): New.
19833 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
19834 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
19835 gimple_ranger. Improve ranges of SSA_NAMES when possible.
19836 * gimple-range.h (gimple_ranger): Remove various range_of routines.
19837 (class fur_source): New.
19838 (class fold_using_range): New.
19839 (fur_source::fur_source): New.
19841 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
19842 instead of range_of_builtin_call.
19844 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
19846 * doc/cpp.texi (Common Predefined Macros): Update documentation
19847 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
19849 2021-05-19 Alex Coplan <alex.coplan@arm.com>
19852 * config/arm/arm.md (nonsecure_call_internal): Always ensure
19853 callee's address is in a register.
19855 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
19857 * common/config/riscv/riscv-common.c
19858 (riscv_subset_list::parsing_subset_version): Properly parse the letter
19860 (riscv_subset_list::parse_std_ext,
19861 riscv_subset_list::parse_multiletter_ext): To handle errors generated
19862 in riscv_subset_list::parsing_subset_version.
19864 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19866 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
19867 type attribute in patterns generating XTN(2).
19869 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19871 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
19872 Remove as duplicate of...
19873 (aarch64_xtn<mode>): This.
19874 (aarch64_xtn2<mode>_le): Move position in file.
19875 (aarch64_xtn2<mode>_be): Move position in file.
19876 (aarch64_xtn2<mode>): Move position in file.
19877 (vec_pack_trunc_<mode>): Define as an expander.
19879 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19881 * config/aarch64/aarch64-simd-builtins.def: Split builtin
19882 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
19883 separate scalar and vector generators.
19884 * config/aarch64/aarch64-simd.md
19885 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
19887 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
19888 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
19889 * config/aarch64/iterators.md: Define SD_HSDI iterator.
19891 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
19893 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
19895 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
19897 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19899 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
19900 Implement as an expand emitting a big/little endian
19901 instruction pattern.
19902 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
19903 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
19905 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19907 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
19908 Implement as an expand emitting a big/little endian
19909 instruction pattern.
19910 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
19911 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
19912 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
19913 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
19915 2021-05-19 Richard Biener <rguenther@suse.de>
19917 PR middle-end/100672
19918 * fold-const.c (fold_negate_expr_1): Use element_precision.
19919 (negate_expr_p): Likewise.
19921 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
19923 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
19924 (pred_load): New int attribute.
19925 * config/aarch64/aarch64-sve.md
19926 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
19927 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
19928 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
19929 code_for_aarch64_load.
19931 2021-05-19 Richard Biener <rguenther@suse.de>
19933 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
19934 sure TARGET_MEM_REF bases are expanded as memory.
19935 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
19936 Do not mark TARGET_MEM_REF bases addressable.
19937 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
19938 TARGET_MEM_REF bases as never rewritable.
19939 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
19940 walk TARGET_MEM_REF bases as address-takens.
19941 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
19943 2021-05-19 Richard Biener <rguenther@suse.de>
19945 * builtins.c (get_object_alignment_1): Strip outer
19947 * tree-dfa.c (get_ref_base_and_extent): Handle outer
19948 WITH_SIZE_EXPR for size processing and process the
19950 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
19951 outer WITH_SIZE_EXPR.
19952 (ao_ref_base_alias_ptr_type): Likewise.
19953 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
19954 and handle that accordingly, stripping it for the
19955 core alias workers.
19956 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
19957 looking through it instead of returning NULL.
19959 2021-05-19 Jakub Jelinek <jakub@redhat.com>
19961 PR middle-end/100576
19962 * builtins.c (check_read_access): Convert bound to size_type_node if
19965 2021-05-19 Richard Biener <rguenther@suse.de>
19967 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
19968 (verify_types_in_gimple_reference): ... here. Sanitize.
19969 (verify_gimple_call): Verify references in LHS and arguments.
19970 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
19972 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
19974 * config/i386/i386.h (VALID_INT_MODE_P):
19975 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
19976 * config/i386/i386.md (isa): Add x64_bmi.
19977 (enabled): Handle x64_bmi.
19978 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
19979 Add alternative using 64bit general registers.
19980 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
19982 2021-05-19 Jakub Jelinek <jakub@redhat.com>
19984 PR middle-end/99928
19985 * tree.h (OMP_MASTER_COMBINED): Define.
19986 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
19987 handling for outer combined/composite constructs to a loop.
19988 Handle lastprivate on combined target.
19989 (gimplify_expr): Formatting fix.
19991 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
19993 * passes.def: Add sink_code pass before store_merging.
19994 * tree-ssa-sink.c (pass_sink_code:clone): New.
19996 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
19998 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
19999 rs6000_special_adjust_field_align_p.
20000 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
20001 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
20002 Remove ABI warning.
20003 (rs6000_function_arg): Likewise.
20004 * config/rs6000/rs6000-protos.h
20005 (rs6000_special_adjust_field_align_p): Remove prototype.
20006 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
20008 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
20009 rs6000_special_adjust_field_align_p.
20011 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
20014 * config/i386/i386.h (VALID_SSE2_REG_MODE):
20015 Add V4QI and V2HI modes.
20016 (VALID_INT_MODE_P): Ditto.
20017 * config/i386/mmx.md (VI_32): New mode iterator.
20018 (mmxvecsize): Handle V4QI and V2HI.
20020 (mov<VI_32:mode>): New expander.
20021 (*mov<mode>_internal): New insn pattern.
20022 (movmisalign<VI_32:mode>): New expander.
20023 (neg<VI_32:mode>): New expander.
20024 (<plusminus:insn><VI_32:mode>3): New expander.
20025 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
20026 (mulv2hi3): New expander.
20027 (*mulv2hi3): New insn pattern.
20028 (one_cmpl<VI_32:mode>2): New expander.
20029 (*andnot<VI_32:mode>3): New insn pattern.
20030 (<any_logic:code><VI_32:mode>3): New expander.
20031 (*<any_logic:code><VI_32:mode>3): New insn pattern.
20033 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
20035 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
20036 Fix a mode mismatch with operand 1.
20038 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
20041 * config/i386/i386-expand.c (split_double_mode): Return
20042 temporary register when simplify_gen_subreg fails with
20043 the high half od the paradoxical subreg.
20045 2021-05-18 Richard Biener <rguenther@suse.de>
20047 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
20048 and honor it when expanding.
20049 (expand_used_vars_for_block): Pass through forced_stack_var.
20050 (expand_used_vars): Likewise.
20051 (discover_nonconstant_array_refs_r): Set bits in
20052 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
20053 (avoid_type_punning_on_regs): Likewise.
20054 (discover_nonconstant_array_refs): Likewise.
20055 (pass_expand::execute): Create and pass down forced_stack_var
20056 bitmap. For parameters and returns temporarily set
20057 TREE_ADDRESSABLE when expand_function_start.
20059 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
20061 * doc/sourcebuild.texi: Document 'dg-note'.
20063 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
20066 * configure: Regenerate.
20067 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
20069 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
20071 * gimple.h (is_gimple_omp_oacc): Tighten.
20072 * omp-low.c (check_omp_nesting_restrictions): Adjust.
20074 2021-05-18 Richard Biener <rguenther@suse.de>
20076 * tree-ssa-operands.c (mark_address_taken): Simplify.
20078 2021-05-18 Martin Liska <mliska@suse.cz>
20080 * config/gcn/mkoffload.c (STR): Redefine.
20081 * config/i386/intelmic-mkoffload.c (STR): Likewise.
20082 * config/nvptx/mkoffload.c (STR): Likewise.
20084 2021-05-18 Martin Liska <mliska@suse.cz>
20086 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
20087 Use startswith function instead of strncmp.
20088 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
20089 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
20090 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
20091 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
20092 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
20093 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20094 * config/arm/arm.c (arm_file_start): Likewise.
20095 (arm_valid_target_attribute_rec): Likewise.
20096 (thumb1_md_asm_adjust): Likewise.
20097 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
20098 * config/avr/avr.c (STR_PREFIX_P): Likewise.
20099 (avr_set_current_function): Likewise.
20100 (avr_handle_addr_attribute): Likewise.
20101 (avr_asm_output_aligned_decl_common): Likewise.
20102 (avr_asm_named_section): Likewise.
20103 (avr_section_type_flags): Likewise.
20104 (avr_asm_select_section): Likewise.
20105 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
20106 (c6x_section_type_flags): Likewise.
20107 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
20108 (darwin_objc_declare_unresolved_class_reference): Likewise.
20109 (darwin_objc_declare_class_definition): Likewise.
20110 * config/darwin.c (indirect_data): Likewise.
20111 (darwin_encode_section_info): Likewise.
20112 (darwin_objc2_section): Likewise.
20113 (darwin_objc1_section): Likewise.
20114 (machopic_select_section): Likewise.
20115 (darwin_globalize_label): Likewise.
20116 (darwin_label_is_anonymous_local_objc_name): Likewise.
20117 (darwin_asm_named_section): Likewise.
20118 (darwin_asm_output_dwarf_offset): Likewise.
20119 * config/frv/frv.c (frv_string_begins_with): Likewise.
20120 (frv_in_small_data_p): Likewise.
20121 * config/gcn/mkoffload.c (STR): Likewise.
20123 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
20124 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
20125 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
20126 (ix86_md_asm_adjust): Likewise.
20127 * config/i386/intelmic-mkoffload.c (STR): Likewise.
20128 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
20129 (i386_pe_file_end): Likewise.
20130 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
20131 (ia64_section_type_flags): Likewise.
20132 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
20133 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
20134 (mips16_stub_function_p): Likewise.
20135 (mips_function_rodata_section): Likewise.
20136 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
20137 (msp430_function_section): Likewise.
20138 (msp430_section_type_flags): Likewise.
20139 (msp430_expand_helper): Likewise.
20140 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
20141 (nios2_valid_target_attribute_rec): Likewise.
20142 * config/nvptx/mkoffload.c (process): Likewise.
20144 * config/pa/som.h: Likewise.
20145 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
20146 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
20147 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
20148 (rs6000_inner_target_options): Likewise.
20149 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
20150 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
20151 * config/vax/vax.c (vax_output_int_move): Likewise.
20152 * config/vms/vms-ld.c (startswith): Likewise.
20153 (process_args): Likewise.
20155 * config/vms/vms.c: Likewise.
20157 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20159 PR rtl-optimization/100590
20160 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
20161 they are NONJUMP_INSN_P.
20163 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20166 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
20167 DECL_ASSEMBLER_NAME on the fn_decl.
20169 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20171 PR tree-optimization/94589
20172 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
20173 phi result used in (res & ~1) == 0 comparison as res >= 0 as
20174 res == 2 would be UB with -ffinite-math-only.
20176 2021-05-18 Martin Liska <mliska@suse.cz>
20178 * Makefile.in: genversion.o should depend on DATESTAMP.
20180 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
20182 * config/arc/simdext.md (negv2si2): Remove round bracket.
20184 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
20186 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
20187 _Bool as macro expanding to _Bool.
20189 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
20192 * tree.c (build_reference_type_for_mode)
20193 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
20195 (build_reference_type, build_pointer_type): Invoke
20196 build_*_type_for_mode with VOIDmode.
20198 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
20200 PR tree-optimization/100512
20201 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
20202 and non-zero pointer ranges as invariant.
20203 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
20204 processing from here.
20206 2021-05-17 Tom de Vries <tdevries@suse.de>
20209 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
20210 * config/nvptx/nvptx.c (nvptx_output_barrier)
20211 (nvptx_output_atomic_insn): New function.
20212 (nvptx_print_operand): Add support for 'B'.
20213 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
20216 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
20218 PR tree-optimization/100349
20219 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
20222 2021-05-17 Tamar Christina <tamar.christina@arm.com>
20224 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
20225 (host_detect_local_cpu): Use it.
20227 2021-05-17 Martin Liska <mliska@suse.cz>
20229 * doc/invoke.texi: Add 2 missing dots.
20231 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
20233 PR bootstrap/100552
20234 * configure.ac: Replace pattern substitution with call to sed.
20235 * configure: Regenerate.
20237 2021-05-17 Richard Biener <rguenther@suse.de>
20239 PR middle-end/100582
20240 * tree.c (array_at_struct_end_p): Get to the base of the
20241 reference before looking for the underlying decl.
20243 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
20245 * genoutput.c (validate_insn_alternatives) Make "wrong number of
20246 alternatives" message more specific, and remove assumption on where
20249 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20251 * config/arm/iterators.md (V16): New iterator.
20252 (VH_cvtto): New iterator.
20253 (v_cmp_result): Added V4HF and V8HF support.
20254 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
20255 (vcond<mode><mode>): Likewise.
20256 (vcond_mask_<mode><v_cmp_result>): Likewise.
20257 (vcond<VH_cvtto><mode>): New expander.
20259 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20261 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
20263 * config/arm/arm.c (arm_expand_vector_compare): Add support for
20265 (arm_expand_vcond): Likewise.
20266 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
20267 VCMPEQQ_N_S, VCMPNEQ_N_S.
20268 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
20269 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
20270 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
20271 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
20272 (@mve_vpselq_<supf><mode>): Likewise.
20273 (@mve_vpselq_f<mode>"): Likewise.
20274 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
20275 and move to vec-common.md.
20276 (vec_cmpu<mode><mode>): Likewise.
20277 (vcond<mode><mode>): Likewise.
20278 (vcond<V_cvtto><mode>): Likewise.
20279 (vcondu<mode><v_cmp_result>): Likewise.
20280 (vcond_mask_<mode><v_cmp_result>): Likewise.
20281 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
20282 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
20283 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
20284 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
20285 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
20286 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
20288 (vec_cmpu<mode><mode>): Likewise.
20289 (vcond<mode><mode>): Likewise.
20290 (vcond<V_cvtto><mode>): Likewise.
20291 (vcondu<mode><v_cmp_result>): Likewise.
20292 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
20295 2021-05-17 liuhongt <hongtao.liu@intel.com>
20298 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
20299 gsi_insert_seq_before instead.
20301 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20303 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
20304 (arm_sat_ok): ...this.
20306 2021-05-17 Martin Liska <mliska@suse.cz>
20308 * lto-wrapper.c (merge_flto_options): Factor out a new function.
20309 (merge_and_complain): Use it.
20310 (run_gcc): Merge also linker command line -flto=foo argument
20313 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
20315 * config/arm/arm.h (CPP_SPEC): Remove error message about
20316 -mlittle-endian/-mbig-endian conflict.
20318 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20320 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
20321 __ROP_PROTECT__ if -mrop-protect is selected.
20323 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20325 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
20326 rop_hash_save_offset and rop_hash_size.
20327 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
20328 rop_hash_size and rop_hash_save_offset.
20329 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
20330 (rs6000_emit_prologue): Emit hashst[p] in prologue.
20331 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
20332 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
20334 (hashst): New define_insn.
20335 (hashchk): Likewise.
20337 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20339 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20340 Disable shrink wrap when inserting ROP-protect instructions.
20341 * config/rs6000/rs6000.opt (mrop-protect): New option.
20342 (mprivileged): Likewise.
20343 * doc/invoke.texi: Document mrop-protect and mprivileged.
20345 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
20347 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
20350 2021-05-15 Martin Jambor <mjambor@suse.cz>
20353 2021-05-13 Martin Jambor <mjambor@suse.cz>
20355 PR tree-optimization/100453
20356 * tree-sra.c (sra_modify_assign): All const base accesses do not
20357 need refreshing, not just those from decl_pool.
20358 (sra_modify_assign): Do not refresh into a const base decl.
20360 2021-05-15 Jakub Jelinek <jakub@redhat.com>
20362 PR rtl-optimization/100342
20363 * regcprop.c (copy_value): When copying a source reg in a wider
20364 mode than it has recorded for the value, adjust recorded destination
20365 mode too or punt if !REG_CAN_CHANGE_MODE_P.
20367 2021-05-14 Jason Merrill <jason@redhat.com>
20369 * intl.h: Add comments.
20371 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20373 * config/aarch64/aarch64-simd.md
20374 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
20375 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
20376 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
20377 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
20378 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
20379 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
20380 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
20381 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
20382 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
20384 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20387 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
20388 boolean logic equivalent.
20389 (vtst_s16): Likewise.
20390 (vtst_s32): Likewise.
20391 (vtst_u8): Likewise.
20392 (vtst_u16): Likewise.
20393 (vtst_u32): Likewise.
20394 (vtst_p8): Likewise.
20395 (vtst_p16): Likewise.
20396 (vtstq_s8): Likewise.
20397 (vtstq_s16): Likewise.
20398 (vtstq_s32): Likewise.
20399 (vtstq_u8): Likewise.
20400 (vtstq_u16): Likewise.
20401 (vtstq_u32): Likewise.
20402 (vtstq_p8): Likewise.
20403 (vtstq_p16): Likewise.
20404 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
20405 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
20407 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20409 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
20410 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
20411 (aarch64_sqdmlsl2<mode>): Delete.
20412 (aarch64_sqdmlal2_lane<mode>): Merge this...
20413 (aarch64_sqdmlsl2_lane<mode>): ... And this...
20414 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
20415 (aarch64_sqdmlal2_laneq<mode>): Merge this...
20416 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
20417 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
20418 (aarch64_sqdmlal2_n<mode>): Merge this...
20419 (aarch64_sqdmlsl2_n<mode>): ... And this...
20420 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
20422 2021-05-13 Martin Sebor <msebor@redhat.com>
20424 PR middle-end/100574
20425 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
20426 all null arguments.
20428 2021-05-13 Martin Sebor <msebor@redhat.com>
20430 PR tree-optimization/93100
20431 PR middle-end/98583
20432 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
20433 don't modify referenced objects.
20435 2021-05-13 Martin Jambor <mjambor@suse.cz>
20437 PR tree-optimization/100453
20438 * tree-sra.c (sra_modify_assign): All const base accesses do not
20439 need refreshing, not just those from decl_pool.
20440 (sra_modify_assign): Do not refresh into a const base decl.
20442 2021-05-13 Martin Liska <mliska@suse.cz>
20444 * tree-ssa-dom.c: Remove m_simplifier.
20446 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
20449 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
20450 canonicalize DImode inequality comparisons against the
20451 maximum integral value.
20453 2021-05-13 Jakub Jelinek <jakub@redhat.com>
20455 PR tree-optimization/98856
20456 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
20457 Expect V2DI and V4DI arithmetic right shifts to be emulated.
20458 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
20460 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
20461 expand_vec_perm_2perm_pblendv): New functions.
20462 (ix86_expand_vec_perm_const_1): Use them.
20463 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
20464 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
20465 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
20466 (ashrv4di3): New define_expand.
20467 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
20468 and !TARGET_AVX512VL expansion.
20470 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
20473 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
20474 sizes < 16 to a register when constructing vpcmov pattern.
20475 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
20477 2021-05-13 Martin Liska <mliska@suse.cz>
20479 * gcov-io.c (gcov_write_block): Remove.
20480 (gcov_write_words): Likewise.
20481 (gcov_read_words): Re-implement using gcov_read_bytes.
20482 (gcov_allocate): Remove.
20483 (GCOV_BLOCK_SIZE): Likewise.
20484 (struct gcov_var): Remove most of the fields.
20485 (gcov_position): Implement with ftell.
20486 (gcov_rewrite): Remove setting of start and offset fields.
20487 (from_file): Re-format.
20488 (gcov_open): Remove setbuf call. It should not be needed.
20489 (gcov_close): Remove internal buffer handling.
20490 (gcov_magic): Use __builtin_bswap32.
20491 (gcov_write_counter): Use directly gcov_write_unsigned.
20492 (gcov_write_string): Use direct fwrite and do not round
20494 (gcov_seek): Use directly fseek.
20495 (gcov_write_tag): Use gcov_write_unsigned directly.
20496 (gcov_write_length): Likewise.
20497 (gcov_write_tag_length): Likewise.
20498 (gcov_read_bytes): Use directly fread.
20499 (gcov_read_unsigned): Use gcov_read_words.
20500 (gcov_read_counter): Likewise.
20501 (gcov_read_string): Use gcov_read_bytes.
20502 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
20503 that size is not in bytes, but words (4B).
20504 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
20505 (GCOV_TAG_ARCS_LENGTH): Likewise.
20506 (GCOV_TAG_ARCS_NUM): Likewise.
20507 (GCOV_TAG_COUNTER_LENGTH): Likewise.
20508 (GCOV_TAG_COUNTER_NUM): Likewise.
20509 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
20511 2021-05-13 liuhongt <hongtao.liu@intel.com>
20514 * config/i386/sse.md (ssedoublevecmode): Add attribute for
20515 V64QI/V32HI/V16SI/V4DI.
20516 (ssehalfvecmode): Add attribute for V2DI/V2DF.
20517 (*vec_concatv4si_0): Extend to VI124_128.
20518 (*vec_concat<mode>_0): New pre-reload splitter.
20519 * config/i386/predicates.md (movq_parallel): New predicate.
20521 2021-05-13 Alexandre Oliva <oliva@adacore.com>
20523 * targhooks.c (default_zero_call_used_regs): Retry using
20524 successfully-zeroed registers as sources.
20526 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
20528 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
20530 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
20533 * gimple-range.cc (range_of_builtin_call): Skip out on
20534 processing __builtin_clz when varying.
20536 2021-05-12 Tom de Vries <tdevries@suse.de>
20539 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
20540 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
20541 to ptx_version_option.
20542 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
20543 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
20544 (define_insn "nvptx_vote_ballot"): Use sync variant for
20546 * config/nvptx/nvptx.opt (ptx_version): Add enum.
20547 (mptx): Add option.
20548 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
20550 2021-05-12 Richard Biener <rguenther@suse.de>
20552 PR tree-optimization/100566
20553 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
20554 allow_back for all edge queries.
20556 2021-05-12 liuhongt <hongtao.liu@intel.com>
20559 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
20560 splitters for pblendvb of NOT mask register.
20562 2021-05-12 Richard Biener <rguenther@suse.de>
20564 PR tree-optimization/100519
20565 * tree-ssa-reassoc.c (can_associate_p): Split into...
20566 (can_associate_op_p): ... this
20567 (can_associate_type_p): ... and this.
20568 (is_reassociable_op): Call can_associate_op_p.
20569 (break_up_subtract_bb): Call the appropriate predicates.
20570 (reassociate_bb): Likewise.
20572 2021-05-12 Martin Liska <mliska@suse.cz>
20574 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
20575 (run_gcc): Use -flto argument detection for merged
20578 2021-05-12 Martin Liska <mliska@suse.cz>
20580 * lto-wrapper.c (print_lto_docs_link): New function.
20581 (run_gcc): Print warning about missing job server detection
20582 after we know NR of partitions. Do the same for -flto{,=1}.
20583 * opts.c (get_option_html_page): Support -flto option.
20585 2021-05-12 Martin Liska <mliska@suse.cz>
20587 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
20589 (append_option): Remove.
20590 (find_option): Rework to use the vector type.
20591 (remove_option): Remove.
20592 (merge_and_complain): Use vectors for cl_decoded_option data
20594 (append_compiler_options): Likewise.
20595 (append_diag_options): Likewise.
20596 (append_linker_options): Likewise.
20597 (append_offload_options): Likewise.
20598 (compile_offload_image): Likewise.
20599 (compile_images_for_offload_targets): Likewise.
20600 (find_and_merge_options): Likewise.
20601 (run_gcc): Likewise.
20603 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
20606 * dwarf2out.c (dwarf2out_finish): Set
20607 have_multiple_function_sections with multi-range text_section.
20609 2021-05-12 Martin Liska <mliska@suse.cz>
20611 PR bootstrap/100560
20612 * Makefile.in: Remove version.h from linker command line.
20614 2021-05-12 Richard Biener <rguenther@suse.de>
20616 PR middle-end/100547
20617 * rtl.h (rtvec_alloc): Make argument size_t.
20618 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
20620 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20622 PR middle-end/100508
20623 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
20624 type, don't reuse DECL_RTL if it has different mode, instead force
20625 creation of a new DEBUG_EXPR.
20627 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20628 Marc Glisse <marc.glisse@inria.fr>
20630 PR tree-optimization/94589
20631 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
20632 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
20634 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
20637 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
20638 * config/i386/mmx.md (MMXMODE124): New mode iterator.
20640 (mmxintvecmode): New mode attribute.
20641 (mmxintvecmodelower): Ditto.
20642 (*mmx_maskcmpv2sf3_comm): New insn pattern.
20643 (*mmx_maskcmpv2sf3): Ditto.
20644 (vec_cmpv2sfv2si): New expander.
20645 (vcond<V2FI:mode>v2si): Ditto.
20646 (mmx_vlendvps): New insn pattern.
20647 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
20648 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
20649 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
20651 2021-05-11 Martin Sebor <msebor@redhat.com>
20653 PR middle-end/21433
20654 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
20656 2021-05-11 Richard Biener <rguenther@suse.de>
20658 * gimple-fold.c (gimple_fold_call): Do not call
20659 maybe_fold_reference on call arguments or the static chain.
20660 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
20663 2021-05-11 Martin Liska <mliska@suse.cz>
20665 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
20666 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
20667 (DEF_HSAIL_SAT_BUILTIN): Likewise.
20668 (DEF_HSAIL_INTR_BUILTIN): Likewise.
20669 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
20670 * doc/frontends.texi: Remove BRIG.
20671 * doc/install.texi: Likewise.
20672 * doc/invoke.texi: Likewise.
20673 * doc/standards.texi: Likewise.
20674 * brig-builtins.def: Removed.
20675 * brig/ChangeLog: Removed.
20676 * brig/Make-lang.in: Removed.
20677 * brig/brig-builtins.h: Removed.
20678 * brig/brig-c.h: Removed.
20679 * brig/brig-lang.c: Removed.
20680 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
20681 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
20682 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
20683 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
20684 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
20685 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
20686 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
20687 * brig/brigfrontend/brig-comment-handler.cc: Removed.
20688 * brig/brigfrontend/brig-control-handler.cc: Removed.
20689 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
20690 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
20691 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
20692 * brig/brigfrontend/brig-function-handler.cc: Removed.
20693 * brig/brigfrontend/brig-function.cc: Removed.
20694 * brig/brigfrontend/brig-function.h: Removed.
20695 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
20696 * brig/brigfrontend/brig-label-handler.cc: Removed.
20697 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
20698 * brig/brigfrontend/brig-machine.c: Removed.
20699 * brig/brigfrontend/brig-machine.h: Removed.
20700 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
20701 * brig/brigfrontend/brig-module-handler.cc: Removed.
20702 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
20703 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
20704 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
20705 * brig/brigfrontend/brig-to-generic.cc: Removed.
20706 * brig/brigfrontend/brig-to-generic.h: Removed.
20707 * brig/brigfrontend/brig-util.cc: Removed.
20708 * brig/brigfrontend/brig-util.h: Removed.
20709 * brig/brigfrontend/brig-variable-handler.cc: Removed.
20710 * brig/brigfrontend/hsa-brig-format.h: Removed.
20711 * brig/brigfrontend/phsa.h: Removed.
20712 * brig/brigspec.c: Removed.
20713 * brig/config-lang.in: Removed.
20714 * brig/gccbrig.texi: Removed.
20715 * brig/lang-specs.h: Removed.
20716 * brig/lang.opt: Removed.
20718 2021-05-11 Richard Biener <rguenther@suse.de>
20721 * ipa-param-manipulation.c
20722 (ipa_param_body_adjustments::modify_call_stmt): Avoid
20723 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
20724 via gimple_call_lhs_ptr.
20726 2021-05-11 Alex Coplan <alex.coplan@arm.com>
20729 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
20730 Avoid emitting CFA adjusts on the sp if we have the fp.
20732 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
20734 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
20735 (VMULD): New iterator.
20736 (VCOND): Handle V4HF and V8HF.
20737 (VCONQ): Fix entry for V2SF.
20738 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
20739 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
20740 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
20741 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
20742 Use a 128-bit vector mode for the indexed operand. Use stype for
20743 the scheduling type.
20745 2021-05-11 Richard Biener <rguenther@suse.de>
20747 * gimple-fold.c (maybe_fold_reference): Only return
20748 is_gimple_min_invariant values.
20750 2021-05-11 Richard Biener <rguenther@suse.de>
20752 PR middle-end/100509
20753 * gimple-fold.c (fold_gimple_assign): Only call
20754 get_symbol_constant_value on register type symbols.
20756 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
20757 Joe Ramsay <joe.ramsay@arm.com>
20760 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
20761 (__arm_vcmpneq): Remove duplicate definition.
20762 (__arm_vstrwq_scatter_offset_p): Likewise.
20763 (__arm_vmaxq_x): Likewise.
20764 (__arm_vmlsdavaq): Likewise.
20765 (__arm_vmlsdavaxq): Likewise.
20766 (__arm_vmlsdavq_p): Likewise.
20767 (__arm_vmlsdavxq_p): Likewise.
20768 (__arm_vrmlaldavhaq): Likewise.
20769 (__arm_vstrbq_p): Likewise.
20770 (__arm_vstrbq_scatter_offset): Likewise.
20771 (__arm_vstrbq_scatter_offset_p): Likewise.
20772 (__arm_vstrdq_scatter_offset): Likewise.
20773 (__arm_vstrdq_scatter_offset_p): Likewise.
20774 (__arm_vstrdq_scatter_shifted_offset): Likewise.
20775 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
20777 2021-05-11 Jakub Jelinek <jakub@redhat.com>
20779 PR middle-end/100471
20780 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
20781 is 0, bypass the reduction loop including
20782 GOMP_taskgroup_reduction_unregister call.
20784 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20786 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
20787 costing_for_scalar.
20788 (rs6000_density_test): Early return if costing_for_scalar is true.
20789 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
20791 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20793 * doc/tm.texi: Regenerated.
20794 * target.def (init_cost): Add new parameter costing_for_scalar.
20795 * targhooks.c (default_init_cost): Adjust for new parameter.
20796 * targhooks.h (default_init_cost): Likewise.
20797 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
20798 (vect_compute_single_scalar_iteration_cost): Likewise.
20799 (vect_analyze_loop_2): Likewise.
20800 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
20801 (vect_bb_vectorization_profitable_p): Likewise.
20802 * tree-vectorizer.h (init_cost): Likewise.
20803 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
20804 * config/i386/i386.c (ix86_init_cost): Likewise.
20805 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
20807 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20809 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
20810 vect_nonmem and moved into...
20811 (struct rs6000_cost_data): ...here.
20812 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
20813 (rs6000_add_stmt_cost): Likewise.
20814 (rs6000_finish_cost): Likewise.
20816 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
20818 * range-op.cc (get_bool_state): Adjust head comment.
20819 (operator_not_equal::op1_range): Fix comment.
20820 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
20822 2021-05-10 Martin Sebor <msebor@redhat.com>
20824 PR middle-end/100425
20825 PR middle-end/100510
20826 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
20827 (pass_walloca::xlimit_certain_p): ...to this.
20828 (pass_walloca::gate): Execute for any kind of handled warning.
20829 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
20830 warnings when xlimit_certain_p is set.
20832 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
20834 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
20835 Return ALTIVEC_REGS if that is best_class.
20836 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
20838 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20840 * config/arm/arm.h (CPP_SPEC): Remove error message about
20843 2021-05-10 Martin Jambor <mjambor@suse.cz>
20845 * ipa-prop.h (IPA_NODE_REF): Removed.
20846 (IPA_NODE_REF_GET_CREATE): Likewise.
20847 (IPA_EDGE_REF): Likewise.
20848 (IPA_EDGE_REF_GET_CREATE): Likewise.
20849 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
20850 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
20851 use of ipa_node_params_sum.
20852 (ipcp_versionable_function_p): Likewise.
20853 (push_node_to_stack): Likewise.
20854 (pop_node_from_stack): Likewise.
20855 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
20856 direct use of ipa_node_params_sum.
20857 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
20858 ipa_node_params_sum.
20859 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
20861 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
20862 use of ipa_node_params_sum.
20863 (self_recursively_generated_p): Likewise.
20864 (propagate_scalar_across_jump_function): Likewise.
20865 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
20866 direct use of ipa_edge_args_sum, moved the lookup after the early
20867 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
20868 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
20869 direct uses of ipa_node_params_sum.
20870 (propagate_vr_across_jump_function): Likewise.
20871 (propagate_aggregate_lattice): Likewise.
20872 (propagate_aggs_across_jump_function): Likewise.
20873 (propagate_constants_across_call): Likewise, also replaced
20874 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
20875 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
20876 of ipa_node_params_sum.
20877 (estimate_local_effects): Likewise.
20878 (add_all_node_vals_to_toposort): Likewise.
20879 (propagate_constants_topo): Likewise.
20880 (ipcp_propagate_stage): Likewise.
20881 (ipcp_discover_new_direct_edges): Likewise.
20882 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
20883 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
20884 (get_info_about_necessary_edges): Likewise.
20885 (want_remove_some_param_p): Likewise.
20886 (create_specialized_node): Likewise.
20887 (self_recursive_pass_through_p): Likewise.
20888 (self_recursive_agg_pass_through_p): Likewise.
20889 (find_more_scalar_values_for_callers_subset): Likewise and also
20890 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
20891 case replacing two of those with a single query.
20892 (find_more_contexts_for_caller_subset): Likewise for the
20893 ipa_polymorphic_call_context overload.
20894 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
20895 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
20896 ipa_node_params_sum.
20897 (find_aggregate_values_for_callers_subset): Likewise, also reusing
20898 results of ipa_edge_args_sum->get.
20899 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
20900 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
20901 direct use of ipa_edge_args_sum.
20902 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
20903 summary query after the early exit and reused the result later.
20904 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
20905 ipa_node_params_sum.
20906 (decide_whether_version_node): Likewise. Removed re-querying for
20907 summaries after cloning.
20908 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
20909 ipa_node_params_sum.
20910 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
20912 (identify_dead_nodes): Likewise.
20913 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
20914 ipa_node_params_sum.
20915 (ipcp_store_vr_results): Likewise.
20916 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
20917 (ipa_fn_summary_t::duplicate): Likewise.
20918 (analyze_function_body): Likewise.
20919 (estimate_calls_size_and_time): Likewise.
20920 (ipa_cached_call_context::duplicate_from): Likewise.
20921 (ipa_call_context::equal_to): Likewise.
20922 (remap_edge_params): Likewise.
20923 (ipa_merge_fn_summary_after_inlining): Likewise.
20924 (inline_read_section): Likewise.
20925 * ipa-icf.c (sem_function::param_used_p): Likewise.
20926 * ipa-modref.c (compute_parm_map): Likewise.
20927 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
20929 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
20930 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
20932 * ipa-profile.c (check_argument_count): Likewise.
20933 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
20934 with a direct use of ipa_node_params_sum.
20935 (ipa_initialize_node_params): Likewise.
20936 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
20937 direct use of ipa_edge_args_sum and reused the query result.
20938 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
20939 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
20940 direct use of ipa_edge_args_sum.
20941 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
20942 ipa_node_params_sum and reused the result of the query.
20943 (ipa_analyze_node): Likewise.
20944 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
20945 of ipa_node_params_sum.
20946 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
20947 direct uses of ipa_edge_args_sum.
20948 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
20949 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
20950 direct use of ipa_edge_args_sum. Removed superficial re-querying the
20952 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
20953 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
20955 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
20956 direct use of ipa_edge_args_sum.
20957 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
20958 use of ipa_node_params_sum.
20959 (ipa_print_node_params): Likewise.
20960 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
20961 direct uses of ipa_edge_args_sum.
20962 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
20964 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
20965 ipa_node_params_sum.
20966 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
20967 scopes where it is used.
20969 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
20971 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
20972 <case E_V2SImode>: Force op_true to register.
20974 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20976 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
20977 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
20978 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
20980 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
20981 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
20982 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
20983 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
20984 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
20985 (mve_vcmpneq_n_f<mode>): Remove.
20986 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
20987 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
20988 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
20990 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20992 * config/arm/iterators.md (MVE_COMPARISONS): New.
20994 (mve_cmp_type): New.
20995 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
20997 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
20998 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
20999 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
21000 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
21001 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
21002 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
21003 (mve_vcmpneq_n_<mode>): Remove.
21005 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21007 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
21008 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
21009 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
21012 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21014 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
21015 (vcmpneq_n_u): Likewise.
21016 (vcmpeqq_u,): Likewise.
21017 (vcmpeqq_n_u): Likewise.
21018 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
21019 VCMPEQQ_N_U and VCMPNEQ_N_U.
21020 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
21021 (mve_vcmpeqq_n): Likewise.
21022 (mve_vcmpeqq): Likewise.
21023 (mve_vcmpneq_n): Likewise.
21025 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21027 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
21028 the 's' version of the builtin.
21030 2021-05-10 Richard Biener <rguenther@suse.de>
21032 PR tree-optimization/100492
21033 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
21034 Find nothing when the loop contains an irreducible region.
21036 2021-05-10 Richard Biener <rguenther@suse.de>
21038 PR middle-end/100464
21040 * gimple-fold.c (canonicalize_constructor_val): Do not set
21043 2021-05-10 Richard Biener <rguenther@suse.de>
21045 PR tree-optimization/100434
21046 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
21048 (dse_optimize_stmt): Handle call LHS by dropping the
21049 LHS or the whole call if it doesn't have other
21051 (pass_dse::execute): Adjust.
21053 2021-05-10 Martin Liska <mliska@suse.cz>
21055 * Makefile.in: Add missing genversion rule.
21057 2021-05-10 Alex Coplan <alex.coplan@arm.com>
21060 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
21061 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
21063 2021-05-10 Martin Liska <mliska@suse.cz>
21065 * builtins.c (is_builtin_name): Use startswith
21066 function instead of strncmp.
21067 * collect2.c (main): Likewise.
21068 (has_lto_section): Likewise.
21069 (scan_libraries): Likewise.
21070 * coverage.c (coverage_checksum_string): Likewise.
21071 (coverage_init): Likewise.
21072 * dwarf2out.c (is_cxx): Likewise.
21073 (gen_compile_unit_die): Likewise.
21074 * gcc-ar.c (main): Likewise.
21075 * gcc.c (init_spec): Likewise.
21076 (read_specs): Likewise.
21077 (execute): Likewise.
21078 (check_live_switch): Likewise.
21079 * genattrtab.c (write_attr_case): Likewise.
21080 (IS_ATTR_GROUP): Likewise.
21081 * gencfn-macros.c (main): Likewise.
21082 * gengtype.c (type_for_name): Likewise.
21083 (gen_rtx_next): Likewise.
21084 (get_file_langdir): Likewise.
21085 (write_local): Likewise.
21086 * genmatch.c (get_operator): Likewise.
21087 (get_operand_type): Likewise.
21088 (expr::gen_transform): Likewise.
21089 * genoutput.c (validate_optab_operands): Likewise.
21090 * incpath.c (add_sysroot_to_chain): Likewise.
21091 * langhooks.c (lang_GNU_C): Likewise.
21092 (lang_GNU_CXX): Likewise.
21093 (lang_GNU_Fortran): Likewise.
21094 (lang_GNU_OBJC): Likewise.
21095 * lto-wrapper.c (run_gcc): Likewise.
21096 * omp-general.c (omp_max_simt_vf): Likewise.
21097 * omp-low.c (omp_runtime_api_call): Likewise.
21098 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
21099 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
21100 * real.c (real_from_string): Likewise.
21101 * selftest.c (assert_str_startswith): Likewise.
21102 * timevar.c (timer::validate_phases): Likewise.
21103 * tree.c (get_file_function_name): Likewise.
21104 * ubsan.c (ubsan_use_new_style_p): Likewise.
21105 * varasm.c (default_function_rodata_section): Likewise.
21106 (incorporeal_function_p): Likewise.
21107 (default_section_type_flags): Likewise.
21108 * system.h (startswith): Define startswith.
21110 2021-05-10 Martin Liska <mliska@suse.cz>
21112 * bitmap.h (class auto_bitmap): Remove
21113 __cplusplus >= 201103.
21114 * config/aarch64/aarch64.c: Likewise.
21115 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
21117 * sbitmap.h: Likewise.
21119 2021-05-10 Martin Liska <mliska@suse.cz>
21121 * Makefile.in: Rename gcov-iov to genversion and depend
21122 on version.h (instead of gcov-iov.h).
21123 * gcov-io.h: Include version.h instread of gcov-iov.h.
21124 * gengtype-state.c (read_state_version): Likewise.
21125 * gcov-iov.c: Moved to...
21126 * genversion.c: ...here.
21127 * lto-streamer.h (LTO_major_version): Define it with
21129 * version.c: Removed.
21130 * version.h: Removed.
21132 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21134 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
21135 * config/arc/simdext.md (VCT): Add predicates for iterator
21138 (voptab): Likewise.
21139 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
21140 (<voptab>v2si3): New patterns.
21142 (reduc_plus_scal_v4hi): Likewise.
21143 (reduc_plus_scal_v2si): Likewise.
21144 (vec_duplicatev2si): Likewise.
21145 (vec_duplicatev4hi): Likewise.
21147 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21149 * config/arc/simdext.md: Format and cleanup file.
21151 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21153 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
21154 only when munaligned-access option is on.
21155 (movmisalign<mode>): Likewise.
21157 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21159 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
21161 * config/arc/arc.c (arc_reorg): Remove underscore from string.
21163 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21165 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21166 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
21167 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
21168 (norm_f): Likewise.
21171 (clzsi2): Use fls instruction when available.
21172 (arc_clzsi2): Likewise.
21174 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21176 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
21178 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21180 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
21182 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
21184 PR middle-end/100467
21185 * toplev.c (compile_file): Call insn_locations_init before
21186 targetm.asm_out.code_end.
21188 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21191 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21193 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21195 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21196 Andrew Stubbs <amd@codesourcery.com>
21199 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
21200 emit_move_insn operands.
21202 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21204 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
21205 location for the outgoing edges of an empty block.
21206 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
21207 type by means of the get_array_descr_info langhook, if it is set and
21208 returns true. Remove obsolete code dealing with unnamed subtypes.
21210 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21212 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
21213 (sbr_vector): Renamed from ssa_block_cache.
21214 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
21215 (ssa_block_ranges::~ssa_block_ranges): Remove.
21216 (sbr_vector::set_bb_range): Use varying and undefined cached values.
21217 (ssa_block_ranges::set_bb_varying): Remove.
21218 (sbr_vector::get_bb_range): Adjust assert.
21219 (sbr_vector::bb_range_p): Adjust assert.
21220 (~block_range_cache): No freeing loop required.
21221 (block_range_cache::get_block_ranges): Remove.
21222 (block_range_cache::set_bb_range): Inline get_block_ranges.
21223 (block_range_cache::set_bb_varying): Remove.
21224 * gimple-range-cache.h (set_bb_varying): Remove prototype.
21225 * value-range.h (irange_allocator::get_memory): New.
21227 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21229 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
21230 dominator tree is available and requested.
21231 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
21232 (ranger_cache::fill_block_cache): Don't search dom tree here either.
21233 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
21235 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21237 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
21238 only PHI nodes better.
21240 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21242 * gimple-range-edge.h (gimple_outgoing_range): Rename from
21244 (gcond_edge_range): Export prototype.
21245 * gimple-range-edge.cc (gcond_edge_range): New.
21246 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
21247 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
21249 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21251 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
21252 default range into a temp and allocate only what is needed.
21254 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21256 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
21258 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21260 * gimple-range.h (gimple_range_global): Pick up parameter initial
21261 values, and use-before defined locals are UNDEFINED.
21263 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21265 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
21267 * tree.h (reverse_storage_order_for_component_p): Return false if
21268 the type is a pointer.
21270 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21272 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21274 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21277 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
21278 Handle V8QI, V4HI and V2SI modes.
21279 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
21280 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
21281 * config/i386/i386.md (unspec): ... here.
21283 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
21284 Tom de Vries <tdevries@suse.de>
21286 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
21287 a truth_value_p reduction variable is nonintegral.
21289 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21292 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
21293 Return false for mode sizes < 16.
21295 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21298 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
21300 2021-05-06 Martin Jambor <mjambor@suse.cz>
21302 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
21303 when there is no function summary.
21304 (ipa_sra_summarize_function): produce edge summaries even when
21307 2021-05-06 Tom Tromey <tom@tromey.com>
21309 * godump.c (string_hash_eq): Remove.
21310 (go_finish): Use htab_eq_string.
21312 2021-05-06 Tom Tromey <tom@tromey.com>
21314 * gengtype-state.c (read_state): Use htab_eq_string.
21315 (string_eq): Remove.
21317 2021-05-06 Tom Tromey <tom@tromey.com>
21319 * gensupport.c (htab_eq_string): Remove.
21321 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
21324 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
21325 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
21326 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
21327 set_ignored_loc callbacks.
21328 * debug.c (do_nothing_debug_hooks): Likewise.
21329 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21330 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
21331 (in_text_section_p, last_text_label, last_cold_label,
21332 switch_text_ranges, switch_cold_ranges): New data items.
21333 (dwarf2out_note_section_used): Remove.
21334 (dwarf2out_begin_prologue): Set fde->ignored_debug and
21336 (mark_ignored_debug_section): New helper function.
21337 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
21338 mark_ignored_debug_section.
21339 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
21340 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
21341 (size_of_aranges): Adjust formula for multi-part text ranges size.
21342 (output_aranges): Output multi-part text ranges.
21343 (dwarf2out_set_ignored_loc): New callback function.
21344 (dwarf2out_finish): Output multi-part text ranges.
21345 (dwarf2out_c_finalize): Clear new data items.
21346 * final.c (final_start_function_1): Call set_ignored_loc callback.
21347 (final_scan_insn_1): Likewise.
21348 * ggc-page.c (gt_ggc_mx): New helper function.
21349 * stringpool.c (gt_pch_nx): Likewise.
21351 2021-05-06 Richard Biener <rguenther@suse.de>
21353 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
21354 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
21355 (TV_TREE_INTO_SSA): New.
21356 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
21357 (rewrite_blocks): Likewise.
21358 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
21360 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21362 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
21363 abs_replacement, xor_replacement,
21364 cond_removal_in_popcount_clz_ctz_pattern,
21365 replace_phi_edge_with_variable): Change type of phi argument from
21366 gimple * to gphi *.
21368 2021-05-06 Richard Biener <rguenther@suse.de>
21370 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
21371 Output an opt-info message.
21372 (do_split_loop_on_cond): Likewise.
21373 (tree_ssa_split_loops): Update SSA form here.
21375 2021-05-06 Richard Biener <rguenther@suse.de>
21377 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
21378 return variable removal.
21380 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
21382 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
21383 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
21384 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
21386 * config/s390/s390.c (s390_const_operand_ok): Remove unused
21389 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21391 PR tree-optimization/94589
21392 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
21393 spaceship_replacement.
21394 (cond_only_block_p, spaceship_replacement): New functions.
21396 2021-05-06 Richard Biener <rguenther@suse.de>
21399 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
21400 dealing with a debug use and only query existing addresses
21402 (lower_emutls_1): Avoid splitting out addresses for debug
21403 stmts, reset the debug stmt when we fail to find existing
21405 (lower_emutls_phi_arg): Set wi.stmt.
21407 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
21410 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
21411 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
21412 (stack_protect_test): Use cbranch helper.
21414 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21417 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21418 always return the establisher frame for __builtin_frame_address (0).
21420 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
21423 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
21424 (ix86_cpu_model_var): Likewise.
21425 (ix86_cpu_features2_type_node): Likewise.
21426 (ix86_cpu_features2_var): Likewise.
21427 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
21430 2021-05-05 Martin Sebor <msebor@redhat.com>
21432 * passes.def (pass_warn_printf): Run after SSA.
21434 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21436 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
21437 * config/arm/predicates.md (minus_one_operand): New predicate.
21439 2021-05-05 Jeff Law <jlaw@tachyum.com>
21441 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
21443 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21445 PR rtl-optimization/100263
21446 * postreload.c (move2add_valid_value_p): Ensure register can
21449 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21451 PR rtl-optimization/100411
21452 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
21453 and beginning of function markers.
21455 2021-05-05 Jeff Law <jlaw@tachyum.com>
21457 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
21458 * config/cr16/cr16.c (notice_update_cc): Remove.
21459 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
21461 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
21464 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
21465 Handle V8QI, V4HI and V2SI modes.
21466 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
21467 (ix86_build_signbit_mask): Ditto.
21468 * config/i386/mmx.md (MMXMODE14): New mode iterator.
21469 (<smaxmin:code><MMXMODE14:mode>3): New expander.
21470 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
21471 (<umaxmin:code><MMXMODE24:mode>3): New expander.
21472 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
21473 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
21474 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21475 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21476 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21477 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21479 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21481 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
21482 not expand the VALUE_EXPR of variables put in the non-local frame.
21483 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
21484 to be ignored for debug info, ensure its variable offsets are not.
21486 2021-05-05 Richard Biener <rguenther@suse.de>
21488 PR tree-optimization/79333
21489 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
21490 Fold stmt following SSA edges.
21492 2021-05-05 Richard Biener <rguenther@suse.de>
21494 PR middle-end/100394
21495 * calls.c (expand_call): Preserve possibly throwing calls.
21496 * cfgexpand.c (expand_call_stmt): When a call can throw signal
21497 RTL expansion there are side-effects.
21498 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
21499 mark all possibly throwing stmts necessary unless we can elide
21501 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
21502 -fdelete-dead-exceptions.
21503 * tree.h (DECL_PURE_P): Add note about exceptions.
21505 2021-05-05 Alexandre Oliva <oliva@adacore.com>
21507 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
21510 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
21512 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
21513 get_fnname_from_decl for name of thunk.
21514 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
21515 and ASM_OUTPUT_LABEL.
21516 (rs6000_xcoff_declare_function_name): Use assemble_name and
21518 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
21519 (rs6000_xcoff_encode_section_info): Don't add mapping class
21520 for aliases. Always add [DS] mapping class to primary
21522 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
21524 2021-05-04 Martin Sebor <msebor@redhat.com>
21526 PR middle-end/100307
21527 * builtins.c (compute_objsize_r): Clear base0 for pointers.
21529 2021-05-04 Jeff Law <jlaw@tachyum.com>
21531 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
21533 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
21535 * caller-save.c: Remove CC0.
21536 * cfgcleanup.c: Remove CC0.
21537 * cfgrtl.c: Remove CC0.
21538 * combine.c: Remove CC0.
21539 * compare-elim.c: Remove CC0.
21540 * conditions.h: Remove CC0.
21541 * config/h8300/h8300.h: Remove CC0.
21542 * config/h8300/h8300-protos.h: Remove CC0.
21543 * config/h8300/peepholes.md: Remove CC0.
21544 * config/i386/x86-tune-sched.c: Remove CC0.
21545 * config/m68k/m68k.c: Remove CC0.
21546 * config/rl78/rl78.c: Remove CC0.
21547 * config/sparc/sparc.c: Remove CC0.
21548 * config/xtensa/xtensa.c: Remove CC0.
21549 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
21550 RTL where that is used as a placeholder only.
21551 * cprop.c: Remove CC0.
21552 * cse.c: Remove CC0.
21553 * cselib.c: Remove CC0.
21554 * df-problems.c: Remove CC0.
21555 * df-scan.c: Remove CC0.
21556 * doc/md.texi: Remove CC0. Adjust an example.
21557 * doc/rtl.texi: Remove CC0. Adjust an example.
21558 * doc/tm.texi: Regenerate.
21559 * doc/tm.texi.in: Remove CC0.
21560 * emit-rtl.c: Remove CC0.
21561 * final.c: Remove CC0.
21562 * fwprop.c: Remove CC0.
21563 * gcse-common.c: Remove CC0.
21564 * gcse.c: Remove CC0.
21565 * genattrtab.c: Remove CC0.
21566 * genconfig.c: Remove CC0.
21567 * genemit.c: Remove CC0.
21568 * genextract.c: Remove CC0.
21569 * gengenrtl.c: Remove CC0.
21570 * genrecog.c: Remove CC0.
21571 * haifa-sched.c: Remove CC0.
21572 * ifcvt.c: Remove CC0.
21573 * ira-costs.c: Remove CC0.
21574 * ira.c: Remove CC0.
21575 * jump.c: Remove CC0.
21576 * loop-invariant.c: Remove CC0.
21577 * lra-constraints.c: Remove CC0.
21578 * lra-eliminations.c: Remove CC0.
21579 * optabs.c: Remove CC0.
21580 * postreload-gcse.c: Remove CC0.
21581 * postreload.c: Remove CC0.
21582 * print-rtl.c: Remove CC0.
21583 * read-rtl-function.c: Remove CC0.
21584 * reg-notes.def: Remove CC0.
21585 * reg-stack.c: Remove CC0.
21586 * reginfo.c: Remove CC0.
21587 * regrename.c: Remove CC0.
21588 * reload.c: Remove CC0.
21589 * reload1.c: Remove CC0.
21590 * reorg.c: Remove CC0.
21591 * resource.c: Remove CC0.
21592 * rtl.c: Remove CC0.
21593 * rtl.def: Remove CC0.
21594 * rtl.h: Remove CC0.
21595 * rtlanal.c: Remove CC0.
21596 * sched-deps.c: Remove CC0.
21597 * sched-rgn.c: Remove CC0.
21598 * shrink-wrap.c: Remove CC0.
21599 * simplify-rtx.c: Remove CC0.
21600 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
21601 CC_STATUS_MDEP, and CC_STATUS.
21602 * target.def: Remove CC0.
21603 * valtrack.c: Remove CC0.
21604 * var-tracking.c: Remove CC0.
21606 2021-05-04 Richard Biener <rguenther@suse.de>
21608 PR tree-optimization/100414
21609 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
21611 (tree_ssa_phiopt_worker): But unconditionally here.
21613 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
21615 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
21616 && and || with floating-point and complex arguments.
21618 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21620 * tree-inline.c (insert_debug_decl_map): Delete.
21621 (copy_debug_stmt): Minor tweak.
21622 (setup_one_parameter): Do not use a variable if the value is either
21623 a read-only DECL or a non-addressable local variable in the caller.
21624 In this case, insert the debug-only variable in the map manually.
21625 (expand_call_inline): Do not generate a CLOBBER for these values.
21626 * tree-inline.h (debug_map): Minor tweak.
21628 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21630 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
21631 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
21633 2021-05-04 Richard Biener <rguenther@suse.de>
21635 PR tree-optimization/100329
21636 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
21638 (insert_stmt_after): Assert we're not running into asm goto.
21640 2021-05-04 Richard Biener <rguenther@suse.de>
21642 PR tree-optimization/100398
21643 * tree-ssa-dse.c (pass_dse::execute): Preserve control
21646 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21648 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
21650 2021-05-04 Alexandre Oliva <oliva@adacore.com>
21652 * builtins.c (try_store_by_multiple_pieces): New.
21653 (expand_builtin_memset_args): Use it. If target_char_cast
21654 fails, proceed as for non-constant val. Pass len's ctz to...
21655 * expr.c (clear_storage_hints): ... this. Try store by
21656 multiple pieces after setmem.
21657 (clear_storage): Adjust.
21658 * expr.h (clear_storage_hints): Likewise.
21659 (try_store_by_multiple_pieces): Declare.
21660 * passes.def: Replace the last copy_prop with ccp.
21662 2021-05-03 Tom de Vries <tdevries@suse.de>
21665 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
21668 2021-05-03 Richard Biener <rguenther@suse.de>
21670 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
21672 2021-05-03 Richard Biener <rguenther@suse.de>
21674 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
21675 (dse_dom_walker): Remove.
21676 (dse_dom_walker::dse_optimize_stmt): Rename...
21677 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
21678 (dse_dom_walker::before_dom_children): Inline ...
21679 (pass_dse::execute): ... here. Perform a reverse program
21682 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
21685 * configure: Regenerated.
21687 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
21690 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
21691 (s390_md_asm_adjust): Handle hard registers.
21693 2021-05-03 Jakub Jelinek <jakub@redhat.com>
21695 PR tree-optimization/100382
21696 * tree-ssa-dse.c: Include tree-eh.h.
21697 (dse_dom_walker::before_dom_children): Don't remove stmts if
21698 stmt_unremovable_because_of_non_call_eh_p is true.
21700 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
21702 * varasm.c (compute_reloc_for_var): Split out from...
21703 (get_variable_section): Use it.
21704 * output.h (compute_reloc_for_var): Declare.
21705 * config/rs6000/rs6000-protos.h
21706 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
21708 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
21709 Don't append storage mapping class to symbol.
21710 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
21711 Don't convert TLS BSS to common.
21712 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
21713 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
21715 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
21717 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
21718 If align is 0 from TLS class, use the same rules as varasm.c
21719 If not common, switch to BSS section manually.
21720 If common, emit appropriate comm or lcomm directive.
21721 (rs6000_xcoff_encode_section_info): Add logic to append all
21722 storage mapping classes.
21723 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
21724 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
21725 rs6000_xcoff_asm_output_aligned_decl_common.
21726 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
21727 rs6000_xcoff_asm_output_aligned_decl_common.
21728 (ASM_OUTPUT_TLS_COMMON): Use
21729 rs6000_xcoff_asm_output_aligned_decl_common.
21731 2021-05-02 Jakub Jelinek <jakub@redhat.com>
21734 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
21735 as first argument of pseudo_node_t constructors.
21737 2021-05-02 Jakub Jelinek <jakub@redhat.com>
21740 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
21742 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
21744 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
21745 (gt_pch_nx (int_range<1> *&)): New.
21746 (gt_ggc_mx (int_range<1> *&)): New.
21747 * value-range.h (class irange): Add GTY support for
21750 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
21752 * doc/options.texi (Negative): Change either or to both and.
21754 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21756 * config/aarch64/aarch64-simd-builtins.def: Add
21757 float_ml[as][q]_laneq builtin generator macros.
21758 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
21759 (aarch64_float_mla_laneq<mode>): Define.
21760 (aarch64_float_mls_laneq<mode>): Define.
21761 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
21762 instead of GCC vector extensions.
21763 (vmlaq_laneq_f32): Likewise.
21764 (vmls_laneq_f32): Likewise.
21765 (vmlsq_laneq_f32): Likewise.
21767 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21769 * config/aarch64/aarch64-simd-builtins.def: Add
21770 float_ml[as]_lane builtin generator macros.
21771 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
21773 (mul_lane<mode>3): This, and re-order arguments.
21774 (aarch64_float_mla_lane<mode>): Define.
21775 (aarch64_float_mls_lane<mode>): Define.
21776 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
21777 instead of GCC vector extensions.
21778 (vmlaq_lane_f32): Likewise.
21779 (vmls_lane_f32): Likewise.
21780 (vmlsq_lane_f32): Likewise.
21782 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21784 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
21785 builtin generator macros.
21786 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
21788 (aarch64_float_mls<mode>): Define.
21789 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
21790 instead of relying on GCC vector extensions.
21791 (vmla_f64): Likewise.
21792 (vmlaq_f32): Likewise.
21793 (vmlaq_f64): Likewise.
21794 (vmls_f32): Likewise.
21795 (vmls_f64): Likewise.
21796 (vmlsq_f32): Likewise.
21797 (vmlsq_f64): Likewise.
21798 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
21800 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21802 * config/aarch64/aarch64-simd-builtins.def: Add
21803 float_ml[as]_n_builtin generator macros.
21804 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
21806 (mul_n<mode>3): This, and re-order arguments.
21807 (aarch64_float_mla_n<mode>): Define.
21808 (aarch64_float_mls_n<mode>): Define.
21809 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
21810 instead of inline asm.
21811 (vmlaq_n_f32): Likewise.
21812 (vmls_n_f32): Likewise.
21813 (vmlsq_n_f32): Likewise.
21815 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
21817 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
21818 builtin generator macros.
21819 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
21820 (aarch64_pmull_hiv16qi_insn): Define.
21821 (aarch64_pmull_hiv16qi): Define.
21822 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
21823 instead of inline asm.
21824 (vmull_p8): Likewise.
21826 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
21828 * config/avr/avr.md: Adjust peepholes to match and
21829 generate parallels with clobber of REG_CC.
21830 (mov<mode>_insn): Rename to mov<mode>_insn_split.
21831 (*mov<mode>_insn): Rename to mov<mode>_insn.
21833 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
21835 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
21836 for VAR_DECLs if -fdata-sections enabled.
21838 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
21840 PR bootstrap/100327
21841 * config/rs6000/rs6000.c
21842 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
21843 (rs6000_libgcc_floating_mode_supported_p): New target hook.
21845 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
21847 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
21849 (class back_threader_registry): ...to here...
21850 (class back_threader_profitability): ...and here...
21851 (thread_jumps::thread_through_all_blocks): Remove argument.
21852 (back_threader_registry::back_threader_registry): New.
21853 (back_threader_registry::~back_threader_registry): New.
21854 (back_threader_registry::thread_through_all_blocks): New.
21855 (thread_jumps::profitable_jump_thread_path): Move from here...
21856 (back_threader_profitability::profitable_path_p): ...to here.
21857 (thread_jumps::find_taken_edge): New.
21858 (thread_jumps::convert_and_register_current_path): Move...
21859 (back_threader_registry::register_path): ...to here.
21860 (thread_jumps::register_jump_thread_path_if_profitable): Move...
21861 (thread_jumps::maybe_register_path): ...to here.
21862 (thread_jumps::handle_phi): Call find_taken_edge and
21863 maybe_register_path.
21864 (thread_jumps::handle_assignment): Same.
21865 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
21866 tree argument to handle_phi and handle_assignment.
21867 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
21868 set of m_speed_p and m_max_threaded_paths.
21869 (pass_thread_jumps::execute): Remove second argument from
21870 find_jump_threads_backwards.
21871 (pass_early_thread_jumps::execute): Same.
21873 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
21875 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
21876 (class dom_opt_dom_walker): Initialize some class variables.
21877 (pass_dominator::execute): Pass evrp_range_analyzer and
21878 dom_jump_threader_simplifier to dom_opt_dom_walker.
21879 Adjust for some functions moving into classes.
21880 (simplify_stmt_for_jump_threading): Adjust and move to...
21881 (jump_threader_simplifier::simplify): ...here.
21882 (dom_opt_dom_walker::before_dom_children): Adjust for
21883 m_evrp_range_analyzer.
21884 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
21885 (test_for_singularity): Place in dom_opt_dom_walker class.
21886 (dom_opt_dom_walker::optimize_stmt): The argument
21887 evrp_range_analyzer is now a class field.
21888 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
21889 (thread_jumps::thread_through_all_blocks): New.
21890 (thread_jumps::convert_and_register_current_path): Use m_registry.
21891 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
21892 being in the threader class.
21893 (pass_early_thread_jumps::execute): Same.
21894 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
21895 (jump_threader::jump_threader): ...here.
21896 (threadedge_finalize_values): Move...
21897 (jump_threader::~jump_threader): ...here.
21898 (jump_threader::remove_jump_threads_including): New.
21899 (jump_threader::thread_through_all_blocks): New.
21900 (record_temporary_equivalences_from_phis): Move...
21901 (jump_threader::record_temporary_equivalences_from_phis): ...here.
21902 (record_temporary_equivalences_from_stmts_at_dest): Move...
21903 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
21905 (simplify_control_stmt_condition_1): Move to jump_threader class.
21906 (simplify_control_stmt_condition): Move...
21907 (jump_threader::simplify_control_stmt_condition): ...here.
21908 (thread_around_empty_blocks): Move...
21909 (jump_threader::thread_around_empty_blocks): ...here.
21910 (thread_through_normal_block): Move...
21911 (jump_threader::thread_through_normal_block): ...here.
21912 (thread_across_edge): Move...
21913 (jump_threader::thread_across_edge): ...here.
21914 (thread_outgoing_edges): Move...
21915 (jump_threader::thread_outgoing_edges): ...here.
21916 * tree-ssa-threadedge.h: Move externally facing functings...
21917 (class jump_threader): ...here...
21918 (class jump_threader_simplifier): ...and here.
21919 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
21920 (jump_thread_path_allocator::jump_thread_path_allocator): New.
21921 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
21922 (jump_thread_path_allocator::allocate_thread_edge): New.
21923 (jump_thread_path_allocator::allocate_thread_path): New.
21924 (jump_thread_path_registry::jump_thread_path_registry): New.
21925 (jump_thread_path_registry::~jump_thread_path_registry): New.
21926 (jump_thread_path_registry::allocate_thread_edge): New.
21927 (jump_thread_path_registry::allocate_thread_path): New.
21928 (dump_jump_thread_path): Make extern.
21929 (debug (const vec<jump_thread_edge *> &path)): New.
21930 (struct removed_edges): Move to tree-ssa-threadupdate.h.
21931 (struct thread_stats_d): Remove.
21932 (remove_ctrl_stmt_and_useless_edges): Make static.
21933 (lookup_redirection_data): Move...
21934 (jump_thread_path_registry::lookup_redirection_data): ...here.
21935 (ssa_redirect_edges): Make static.
21936 (thread_block_1): Move...
21937 (jump_thread_path_registry::thread_block_1): ...here.
21938 (thread_block): Move...
21939 (jump_thread_path_registry::thread_block): ...here.
21940 (thread_through_loop_header): Move...
21941 (jump_thread_path_registry::thread_through_loop_header): ...here.
21942 (mark_threaded_blocks): Move...
21943 (jump_thread_path_registry::mark_threaded_blocks): ...here.
21944 (debug_path): Move...
21945 (jump_thread_path_registry::debug_path): ...here.
21946 (debug_all_paths): Move...
21947 (jump_thread_path_registry::dump): ..here.
21948 (rewire_first_differing_edge): Move...
21949 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
21950 (adjust_paths_after_duplication): Move...
21951 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
21952 (duplicate_thread_path): Move...
21953 (jump_thread_path_registry::duplicate_thread_path): ..here.
21954 (remove_jump_threads_including): Move...
21955 (jump_thread_path_registry::remove_jump_threads_including): ...here.
21956 (thread_through_all_blocks): Move to...
21957 (jump_thread_path_registry::thread_through_all_blocks): ...here.
21958 (delete_jump_thread_path): Remove.
21959 (register_jump_thread): Move...
21960 (jump_thread_path_registry::register_jump_thread): ...here.
21961 * tree-ssa-threadupdate.h: Move externally facing functions...
21962 (class jump_thread_path_allocator): ...here...
21963 (class jump_thread_path_registry): ...and here.
21964 (thread_through_all_blocks): Remove.
21965 (struct removed_edges): New.
21966 (register_jump_thread): Remove.
21967 (remove_jump_threads_including): Remove.
21968 (delete_jump_thread_path): Remove.
21969 (remove_ctrl_stmt_and_useless_edges): Remove.
21970 (free_dom_edge_info): New prototype.
21971 * tree-vrp.c: Remove x_vr_values hack.
21972 (class vrp_jump_threader_simplifier): New.
21973 (vrp_jump_threader_simplifier::simplify): New.
21974 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
21975 Remove m_dummy_cond.
21976 Instantiate m_simplifier and m_threader.
21977 (vrp_jump_threader::thread_through_all_blocks): New.
21978 (vrp_jump_threader::simplify_stmt): Remove.
21979 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
21980 Remove x_vr_values hack.
21981 (execute_vrp): Adjust for thread_through_all_blocks being in a
21984 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
21986 * genflags.c (gen_insn): Print failed expansion string.
21988 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
21990 * expr.c (alignment_for_piecewise_move): Call mode_for_size
21991 without limit to MAX_FIXED_MODE_SIZE.
21993 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
21995 PR middle-end/90773
21996 * builtins.c (builtin_memset_gen_str): Don't use return from
21997 simplify_gen_subreg.
21999 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
22002 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
22003 (*addsi3_carry_zext_0r): Ditto.
22004 (*sub<mode>3_carry_0): Ditto.
22005 (*subsi3_carry_zext_0r): Ditto.
22006 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
22008 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
22009 Also consider ix86_carry_flag_unset_operator to calculate
22010 the cost of adc/sbb insn.
22012 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
22014 PR rtl-optimization/100225
22015 PR rtl-optimization/84878
22016 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
22017 where we have an instruction which touches (writes) any hard
22018 register from df->regular_block_artificial_uses set.
22019 Allow not-single-set instruction only right before basic block
22022 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
22024 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
22026 2021-04-30 LevyHsu <admin@levyhsu.com>
22028 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
22029 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
22030 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
22031 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
22033 2021-04-29 Alexandre Oliva <oliva@adacore.com>
22035 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
22037 2021-04-29 Alexandre Oliva <oliva@adacore.com>
22039 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
22040 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
22041 * config/i386/i386.c: Adjust.
22042 * config/i386/i386.md: Adjust.
22043 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
22044 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22045 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22046 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22047 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22048 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22049 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22050 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22051 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22052 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22053 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
22055 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22057 * config/i386/i386-expand.c (ix86_expand_int_compare):
22058 Swap operands of GTU and LEU comparison to emit carry flag comparison.
22059 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
22060 predicate to allow more combine opportunities with memory operands.
22061 (*sub<mode>3_carry_0): Ditto.
22063 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
22065 PR rtl-optimization/100303
22066 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
22067 boolean that indicates whether the use will only be used in
22068 debug instructions. Treat it in the same way that existing
22069 cross-EBB debug references would be handled if so.
22070 (function_info::make_uses_available): Likewise.
22071 * rtl-ssa/functions.h (function_info::make_uses_available): Update
22072 prototype accordingly.
22073 (function_info::make_uses_available): Likewise.
22074 * fwprop.c (try_fwprop_subst): Update call accordingly.
22076 2021-04-29 Jeff Law <jlaw@tachyum.com>
22078 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
22081 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22082 Richard Biener <rguenther@suse.de>
22085 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
22086 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
22087 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
22088 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
22089 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
22090 to PURE_ARGS category.
22091 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
22092 Handle PURE_ARGS category.
22093 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
22095 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
22097 * configure.ac: Check for the presence of sys/locking.h header and
22098 for whether _LK_LOCK is supported by _locking.
22099 * configure: Regenerate.
22100 * config.in: Likewise.
22101 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
22102 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
22103 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
22105 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22107 * config/i386/predicates.md (fcmov_comparison_operator):
22108 Do not check for trivial FP comparison operator.
22109 <case GEU, case LTU>: Allow CCGZmode.
22110 <case GTU, case LEU>: Do not allow CCCmode.
22111 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
22112 (ix86_carry_flag_operator): Match only LTU and UNLT code.
22113 Do not check for trivial FP comparison operator. Allow CCGZmode.
22115 2021-04-29 Tom de Vries <tdevries@suse.de>
22117 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
22118 fd->loop.step by either step or orig_step.
22120 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
22122 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
22123 (load_got_register): Do the PIC dance here.
22124 (sparc_legitimize_tls_address): Simplify.
22125 (sparc_emit_probe_stack_range): Likewise.
22126 (sparc32_initialize_trampoline): Likewise.
22127 (sparc64_initialize_trampoline): Likewise.
22128 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
22129 (probe_stack_range<P:mode>): Likewise.
22130 (flush<P:mode>): Likewise.
22131 (tgd_hi22<P:mode>): Likewise.
22132 (tgd_lo10<P:mode>): Likewise.
22133 (tgd_add<P:mode>): Likewise.
22134 (tgd_call<P:mode>): Likewise.
22135 (tldm_hi22<P:mode>): Likewise.
22136 (tldm_lo10<P:mode>): Likewise.
22137 (tldm_add<P:mode>): Likewise.
22138 (tldm_call<P:mode>): Likewise.
22139 (tldo_hix22<P:mode>): Likewise.
22140 (tldo_lox10<P:mode>): Likewise.
22141 (tldo_add<P:mode>): Likewise.
22142 (tie_hi22<P:mode>): Likewise.
22143 (tie_lo10<P:mode>): Likewise.
22144 (tie_add<P:mode>): Likewise.
22145 (tle_hix22<P:mode>): Likewise.
22146 (tle_lox10<P:mode>): Likewise.
22147 (stack_protect_setsi): Rename to...
22148 (stack_protect_set32): ...this.
22149 (stack_protect_setdi): Rename to...
22150 (stack_protect_set64): ...this.
22151 (stack_protect_set): Adjust calls to above.
22152 (stack_protect_testsi): Rename to...
22153 (stack_protect_test32): ...this.
22154 (stack_protect_testdi): Rename to...
22155 (stack_protect_test64): ...this.
22156 (stack_protect_test): Adjust calls to above.
22158 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
22160 PR middle-end/90773
22161 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
22162 (builtin_strncpy_read_str): Likewise.
22163 (builtin_memset_read_str): Add an argument for the previous RTL
22164 information and generate the new RTL from the previous RTL info.
22165 (builtin_memset_gen_str): Likewise.
22166 * builtins.h (builtin_strncpy_read_str): Update the prototype.
22167 (builtin_memset_read_str): Likewise.
22168 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
22169 returns true, round up size and alignment to the widest integer
22170 mode for maximum size.
22171 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
22172 and pass it to m_constfn.
22173 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
22174 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
22175 initialize m_push. Initialize m_overlap_op_by_pieces with
22176 targetm.overlap_op_by_pieces_p ().
22177 (op_by_pieces_d::run): Pass the previous RTL information to
22178 pieces_addr::adjust and generate overlapping operations if
22179 m_overlap_op_by_pieces is true.
22181 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
22183 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
22185 (can_store_by_pieces): Use by_pieces_constfn on constfun.
22186 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
22187 for op_by_pieces_d change.
22188 (clear_by_pieces_1): Add a dummy argument.
22189 (clear_by_pieces): Updated for op_by_pieces_d change.
22190 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
22191 (string_cst_read_str): Add a dummy argument.
22192 * expr.h (by_pieces_constfn): Add a dummy argument.
22193 (by_pieces_prev): New.
22194 * target.def (overlap_op_by_pieces_p): New target hook.
22195 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
22196 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
22197 * doc/tm.texi: Regenerated.
22199 2021-04-29 Richard Biener <rguenther@suse.de>
22201 PR tree-optimization/100253
22202 * tree-vect-stmts.c (vectorizable_load): Do not assume
22203 element alignment when DR_MISALIGNMENT is -1.
22204 (vectorizable_store): Likewise.
22206 2021-04-29 Jakub Jelinek <jakub@redhat.com>
22209 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
22210 absu_hwi instead of abs_hwi.
22212 2021-04-29 Richard Biener <rguenther@suse.de>
22214 PR middle-end/38474
22215 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
22216 forwarding when indirect forwarding through ESCAPED
22219 2021-04-29 Tom de Vries <tdevries@suse.de>
22222 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
22223 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
22224 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
22225 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
22227 2021-04-29 Richard Biener <rguenther@suse.de>
22229 PR tree-optimization/99912
22230 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
22231 (dse_dom_walker::todo): Likewise.
22232 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
22234 (dse_dom_walker::before_dom_children): Remove trivially
22235 dead SSA defs and schedule CFG cleanup if we removed all
22237 (pass_dse::execute): Get TODO as computed by the DOM walker
22238 and return it. Wipe dominator info earlier.
22240 2021-04-29 Richard Biener <rguenther@suse.de>
22243 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
22244 Track blocks to cleanup EH in new m_need_eh_cleanup.
22245 (ipcp_modif_dom_walker::cleanup_eh): New.
22246 (ipcp_transform_function): Release dominator info before
22249 2021-04-29 Martin Sebor <msebor@redhat.com>
22251 PR middle-end/100250
22252 * attribs.c (attr_access::array_as_string): Avoid dereferencing
22253 a pointer when it's null.
22255 2021-04-29 Martin Sebor <msebor@redhat.com>
22257 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
22258 * ipa-free-lang-data.cc: New file.
22259 * tree.c: Move pass free_lang_data to file above.
22260 (build_array_type_1): Declare extern.
22261 * tree.h (build_array_type_1): Declare.
22263 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22265 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
22266 make consistent with updated RTL pattern.
22267 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
22268 Implement using ss_truncate and us_truncate rather than
22270 * config/aarch64/iterators.md: Remove redundant unspecs and
22271 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
22273 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22275 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
22276 attributes consistent with those defined in arm_neon.h.
22278 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22280 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
22281 attributes consistent with those defined in arm_neon.h.
22283 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22285 * config/aarch64/aarch64-simd-builtins.def: Add
22286 float_trunc_rodd builtin generator macros.
22287 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
22289 (aarch64_float_trunc_rodd_lo_v2sf): Define.
22290 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
22291 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
22292 (aarch64_float_trunc_rodd_hi_v4sf): Define.
22293 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
22294 instead of inline asm.
22295 (vcvtx_high_f32_f64): Likewise.
22296 (vcvtxd_f32_f64): Likewise.
22297 * config/aarch64/iterators.md: Add FCVTXN unspec.
22299 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22301 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
22303 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
22305 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
22306 instead of inline asm.
22307 (vqtbx1_u8): Likewise.
22308 (vqtbx1_p8): Likewise.
22309 (vqtbx1q_s8): Likewise.
22310 (vqtbx1q_u8): Likewise.
22311 (vqtbx1q_p8): Likewise.
22312 (vtbx2_s8): Likewise.
22313 (vtbx2_u8): Likewise.
22314 (vtbx2_p8): Likewise.
22316 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22318 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
22320 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
22321 instead of inline asm.
22322 (vqtbl1_s8): Likewise.
22323 (vqtbl1_u8): Likewise.
22324 (vqtbl1q_p8): Likewise.
22325 (vqtbl1q_s8): Likewise.
22326 (vqtbl1q_u8): Likewise.
22327 (vtbl1_s8): Likewise.
22328 (vtbl1_u8): Likewise.
22329 (vtbl1_p8): Likewise.
22330 (vtbl2_s8): Likewise.
22331 (vtbl2_u8): Likewise.
22332 (vtbl2_p8): Likewise.
22334 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22336 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
22337 ssri_n buitin generator macro.
22338 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
22339 instead of inline asm.
22340 (vsri_n_p16): Likewise.
22341 (vsri_n_p64): Likewise.
22342 (vsriq_n_p8): Likewise.
22343 (vsriq_n_p16): Likewise.
22344 (vsriq_n_p64): Likewise.
22346 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22348 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
22349 iterator for polynomial ssli_n builtin generator macro.
22350 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
22351 instead of inline asm.
22352 (vsli_n_p16): Likewise.
22353 (vsliq_n_p8): Likewise.
22354 (vsliq_n_p16): Likewise.
22355 * config/aarch64/iterators.md: Define VALLP mode iterator.
22357 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22359 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
22360 iterator to generate [su]adalp RTL builtins.
22361 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
22362 [su]adalp RTL pattern.
22363 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
22364 instead of inline asm.
22365 (vpadal_u32): Likewise.
22367 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22369 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
22370 builtin generator macros.
22371 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
22373 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
22374 instead of inline asm.
22375 (vpaddl_s16): Likewise.
22376 (vpaddl_s32): Likewise.
22377 (vpaddl_u8): Likewise.
22378 (vpaddl_u16): Likewise.
22379 (vpaddl_u32): Likewise.
22380 (vpaddlq_s8): Likewise.
22381 (vpaddlq_s16): Likewise.
22382 (vpaddlq_s32): Likewise.
22383 (vpaddlq_u8): Likewise.
22384 (vpaddlq_u16): Likewise.
22385 (vpaddlq_u32): Liwewise.
22386 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
22387 appropriate attributes.
22389 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22391 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
22392 for aarch64_addp<mode> builtin macro generator.
22393 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
22394 aarch64_addp<mode> RTL pattern.
22395 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
22396 instead of inline asm.
22397 (vpaddq_s16): Likewise.
22398 (vpaddq_s32): Likewise.
22399 (vpaddq_s64): Likewise.
22400 (vpaddq_u8): Likewise.
22401 (vpaddq_u16): Likewise.
22402 (vpaddq_u32): Likewise.
22403 (vpaddq_u64): Likewise.
22405 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22407 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
22408 builtin generator macros.
22409 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
22411 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
22412 instead of inline asm.
22413 (vqdmulh_n_s32): Likewise.
22414 (vqdmulhq_n_s16): Likewise.
22415 (vqdmulhq_n_s32): Likewise.
22416 (vqrdmulh_n_s16): Likewise.
22417 (vqrdmulh_n_s32): Likewise.
22418 (vqrdmulhq_n_s16): Likewise.
22419 (vqrdmulhq_n_s32): Likewise.
22421 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
22423 * doc/install.texi (--enable-offload-defaulted): Document.
22425 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
22427 * config/avr/avr-dimode.md: Turn existing patterns into
22428 define_insn_and_split style patterns where the splitter
22429 adds a clobber of the condition code register. Drop "cc"
22430 attribute. Add new patterns to match output of
22432 * config/avr/avr-fixed.md: Likewise.
22433 * config/avr/avr.c (cc_reg_rtx): New.
22434 (avr_parallel_insn_from_insns): Adjust insn count
22435 for removal of set of cc0.
22436 (avr_is_casesi_sequence): Likewise.
22437 (avr_casei_sequence_check_operands): Likewise.
22438 (avr_optimize_casesi): Likewise. Also insert
22439 new insns after jump_insn.
22440 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
22441 for removal of set of cc0.
22442 (avr_init_expanders): Initialize cc_reg_rtx.
22443 (avr_regno_reg_class): Handle REG_CC.
22444 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
22445 (avr_notice_update_cc): Remove function.
22446 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
22447 (compare_condition): Adjust for PARALLEL with
22449 (out_shift_with_cnt): Likewise.
22450 (ashlhi3_out): Likewise.
22451 (ashrhi3_out): Likewise.
22452 (lshrhi3_out): Likewise.
22453 (avr_class_max_nregs): Return single reg for REG_CC.
22454 (avr_compare_pattern): Check for REG_CC instead
22456 (avr_reorg_remove_redundant_compare): Likewise.
22457 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
22458 (avr_hard_regno_nregs): Return single reg for REG_CC.
22459 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
22460 (avr_md_asm_adjust): Clobber REG_CC.
22461 (TARGET_HARD_REGNO_NREGS): Define.
22462 (TARGET_CLASS_MAX_NREGS): Define.
22463 (TARGET_MD_ASM_ADJUST): Define.
22464 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
22466 (enum reg_class): Add CC_REG class.
22467 (NOTICE_UPDATE_CC): Remove.
22468 (CC_OVERFLOW_UNUSABLE): Remove.
22469 (CC_NO_CARRY): Remove.
22470 * config/avr/avr.md: Turn existing patterns into
22471 define_insn_and_split style patterns where the splitter
22472 adds a clobber of the condition code register. Drop "cc"
22473 attribute. Add new patterns to match output of
22475 (sez): Remove unused pattern.
22477 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
22480 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
22483 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
22486 * config/aarch64/constraints.md (Utq): Require the address to
22487 be valid for both the element mode and for V2DImode.
22489 2021-04-28 Jakub Jelinek <jakub@redhat.com>
22490 Tobias Burnus <tobias@codesourcery.com>
22492 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
22493 * gcc.c (process_command): New variable.
22494 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
22495 set it if -foffload is defaulted.
22496 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
22497 (compile_offload_image): If OFFLOAD_DEFAULTED and
22498 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
22499 if corresponding mkoffload can't be found.
22500 (compile_images_for_offload_targets): Likewise. Free and clear
22501 offload_names if no valid offload is found.
22502 * config.in: Regenerate.
22503 * configure: Regenerate.
22505 2021-04-28 Richard Biener <rguenther@suse.de>
22507 PR tree-optimization/100292
22508 * tree-vect-generic.c (expand_vector_condition): Do not fold
22511 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22513 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
22514 * config/rs6000/aix64.opt (m64): New.
22517 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22519 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
22520 (index_term_p): Handle ASHIFT too.
22522 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22524 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
22525 (sync_lock_test_and_set<mode>): Adjust accordingly.
22526 (sync_lock_release<mode>): Likewise.
22528 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22530 * config/vax/vax-protos.h (adjacent_operands_p): Remove
22532 * config/vax/vax.c (adjacent_operands_p): Remove.
22534 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
22536 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
22537 through to the non-conditional execution case if getting the
22538 condition for conditional execution has failed.
22540 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22542 PR middle-end/100284
22543 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
22544 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
22545 than asserting on it.
22547 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22549 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
22550 with TARGET_AIX_OS.
22552 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22555 * calls.c (precompute_register_parameters): Additionally test
22556 targetm.precompute_tls_p to pre-compute argument.
22557 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
22558 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
22559 * target.def (precompute_tls_p): New.
22560 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
22561 * doc/tm.texi: Regenerated.
22563 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22566 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
22567 back to HOST_WIDE_INT.
22569 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22572 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
22573 memory alignment for the outer mode.
22575 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
22577 PR middle-end/90773
22578 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
22579 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
22581 2021-04-27 Alex Coplan <alex.coplan@arm.com>
22584 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
22585 with negative immediates: ensure we expand cbranchsi4_scratch
22586 correctly and ensure we satisfy its constraints.
22587 * config/arm/sync.md
22588 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
22589 attempt to tie two output operands together with constraints;
22590 collapse two alternatives.
22591 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
22592 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
22594 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22597 * config/aarch64/predicates.md (aarch64_sub_immediate,
22598 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
22599 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
22600 * config/aarch64/aarch64.c (aarch64_print_operand,
22601 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
22603 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22605 PR tree-optimization/100239
22606 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
22607 permutations with all indices from the first zero element as vec_shl.
22609 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22611 PR rtl-optimization/100254
22612 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
22613 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
22615 2021-04-27 Richard Biener <rguenther@suse.de>
22617 PR tree-optimization/99912
22618 * passes.def: Add comment about new TODO_remove_unused_locals.
22619 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
22622 2021-04-27 Richard Biener <rguenther@suse.de>
22624 PR tree-optimization/99912
22625 * passes.def (pass_all_optimizations): Add pass_dse before
22626 the first pass_dce, move the first pass_dse before the
22627 pass_dce following pass_pre.
22629 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22631 PR tree-optimization/95527
22632 * generic-match-head.c: Include tm.h.
22633 * gimple-match-head.c: Include tm.h.
22634 * match.pd (CLZ == INTEGER_CST): Don't use
22635 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
22636 if clz == CFN_CLZ. Add missing val declaration.
22637 (CTZ cmp CST): New simplifications.
22639 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22641 PR tree-optimization/96696
22642 * expr.c (expand_expr_divmod): New function.
22643 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
22644 divisions. Formatting fixes.
22645 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
22648 2021-04-27 Martin Jambor <mjambor@suse.cz>
22651 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
22652 If removing a call statement LHS SSA name, release it.
22654 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
22657 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
22658 is valid before including it in the mask.
22660 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22663 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
22666 2021-04-27 Richard Biener <rguenther@suse.de>
22668 PR tree-optimization/100051
22669 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
22670 disambiguator based on access size vs. decl size.
22672 2021-04-27 Richard Biener <rguenther@suse.de>
22674 PR tree-optimization/100278
22675 * tree-ssa-pre.c (compute_avail): Give up when we cannot
22676 adjust TBAA beacuse of mismatching bases.
22678 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22681 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
22682 For any_rotate define_insn_split and following splitters, use
22683 SWI iterator instead of SWI48.
22685 2021-04-27 Richard Biener <rguenther@suse.de>
22687 PR tree-optimization/99776
22688 * match.pd (bit_field_ref (ctor)): Relax element extract
22689 type compatibility checks.
22691 2021-04-27 Cui,Lili <lili.cui@intel.com>
22693 * common/config/i386/i386-common.c (processor_names):
22694 Sync processor_names with processor_type.
22695 * config/i386/i386-options.c (processor_cost_table):
22696 Sync processor_cost_table with processor_type.
22698 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22700 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
22701 (irange::set): Call irange_set_1bit_anti_range for handling all
22702 1-bit ranges. Fall through on ~[MIN,MAX].
22704 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22706 * value-range.cc (irange::legacy_num_pairs): Remove.
22707 (irange::invert): Change gcc_assert to gcc_checking_assert.
22708 * value-range.h (irange::num_pairs): Adjust for a cached
22709 num_pairs(). Also, rename all gcc_assert's to
22710 gcc_checking_assert's.
22712 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22714 * value-range.cc (irange::operator=): Set m_kind.
22715 (irange::copy_to_legacy): Handle varying and undefined sources
22716 as a legacy copy since they can be easily copied.
22717 (irange::irange_set): Set m_kind.
22718 (irange::irange_set_anti_range): Same.
22719 (irange::set): Rename normalize_min_max to normalize_kind.
22720 (irange::verify_range): Adjust for multi-ranges having the
22722 (irange::irange_union): Set m_kind.
22723 (irange::irange_intersect): Same.
22724 (irange::invert): Same.
22725 * value-range.h (irange::kind): Always return m_kind.
22726 (irange::varying_p): Rename to...
22727 (irange::varying_comptaible_p): ...this.
22728 (irange::undefined_p): Only look at m_kind.
22729 (irange::irange): Always set VR_UNDEFINED if applicable.
22730 (irange::set_undefined): Always set VR_UNDEFINED.
22731 (irange::set_varying): Always set m_kind to VR_VARYING.
22732 (irange::normalize_min_max): Rename to...
22733 (irange::normalize_kind): ...this.
22735 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22737 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
22738 Adjust for constant_p including varying_p.
22739 * tree-vrp.c (vrp_prop::finalize): Same.
22740 (determine_value_range): Same.
22741 * vr-values.c (vr_values::range_of_expr): Same.
22742 * value-range.cc (irange::symbolic_p): Do not check varying_p.
22743 (irange::constant_p): Same.
22745 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22747 * value-range.cc (irange::legacy_lower_bound): Replace
22748 !undefined_p check with num_ranges > 0.
22749 (irange::legacy_upper_bound): Same.
22750 * value-range.h (irange::type): Same.
22751 (irange::lower_bound): Same.
22752 (irange::upper_bound): Same.
22754 2021-04-26 Richard Biener <rguenther@suse.de>
22756 PR tree-optimization/99956
22757 * gimple-loop-interchange.cc (compute_access_stride):
22758 Try instantiating the access in a shallower loop nest
22759 if instantiating failed.
22760 (compute_access_strides): Pass adjustable loop_nest
22761 to compute_access_stride.
22763 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
22765 * doc/sourcebuild.texi (arm_cmse_hw): Document.
22767 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22769 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
22771 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
22772 Nathan Sidwell <nathan@codesourcery.com>
22773 Tom de Vries <vries@codesourcery.com>
22774 Julian Brown <julian@codesourcery.com>
22775 Kwok Cheung Yeung <kcy@codesourcery.com>
22777 * omp-offload.c (oacc_validate_dims): Implement
22778 '-Wopenacc-parallelism'.
22779 * doc/invoke.texi (-Wopenacc-parallelism): Document.
22781 2021-04-26 Richard Biener <rguenther@suse.de>
22783 * tree-cfg.h (gimplify_build1): Remove.
22784 (gimplify_build2): Likewise.
22785 (gimplify_build3): Likewise.
22786 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
22787 (gimplify_build2): Likewise.
22788 (gimplify_build3): Likewise.
22789 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
22791 (gimplify_build2): Likewise.
22792 (gimplify_build3): Likewise.
22793 (tree_vec_extract): Use resimplify with following SSA edges.
22794 (expand_vector_parallel): Avoid passing NULL size/bitpos
22795 to tree_vec_extract.
22796 * expr.c (store_constructor): Deal with zero-element CTORs.
22797 * match.pd (bit_field_ref <vector CTOR>): Make sure to
22798 produce vector constants when possible.
22800 2021-04-26 Richard Biener <rguenther@suse.de>
22802 * tree-complex.c: Include gimple-fold.h.
22803 (expand_complex_addition): Use gimple_build.
22804 (expand_complex_multiplication_components): Likewise.
22805 (expand_complex_multiplication): Likewise.
22806 (expand_complex_div_straight): Likewise.
22807 (expand_complex_div_wide): Likewise.
22808 (expand_complex_division): Likewise.
22809 (expand_complex_conjugate): Likewise.
22810 (expand_complex_comparison): Likewise.
22812 2021-04-26 Richard Biener <rguenther@suse.de>
22814 * tree-ssa-phiopt.c (two_value_replacement): Remove use
22815 of legacy gimplify_buildN API.
22817 2021-04-26 Richard Biener <rguenther@suse.de>
22819 PR tree-optimization/99473
22820 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
22823 2021-04-26 Richard Biener <rguenther@suse.de>
22825 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
22826 Use replace_call_with_value.
22828 2021-04-26 Richard Biener <rguenther@suse.de>
22830 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
22831 (update_gimple_call): Likewise.
22832 (update_call_from_tree): Likewise.
22833 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
22834 (valid_gimple_call_p): Likewise.
22835 (move_ssa_defining_stmt_for_defs): Likewise.
22836 (finish_update_gimple_call): Likewise.
22837 (update_gimple_call): Likewise.
22838 (update_call_from_tree): Likewise.
22839 (propagate_tree_value_into_stmt): Use replace_call_with_value.
22840 * gimple-fold.h (update_gimple_call): Declare.
22841 * gimple-fold.c (valid_gimple_rhs_p): Move here from
22842 tree-ssa-propagate.c.
22843 (update_gimple_call): Likewise.
22844 (valid_gimple_call_p): Likewise.
22845 (finish_update_gimple_call): Likewise, and simplify.
22846 (gimplify_and_update_call_from_tree): Implement
22847 update_call_from_tree functionality, avoid excessive
22848 push/pop_gimplify_context.
22849 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
22850 (gimple_fold_call): Likewise.
22851 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
22852 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
22853 (pass_fold_builtins::execute): Likewise.
22854 (optimize_stack_restore): Use replace_call_with_value.
22855 * tree-cfg.c (fold_loop_internal_call): Likewise.
22856 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
22857 only gimplify_and_update_call_from_tree.
22858 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
22859 (handle_builtin_strchr): Likewise.
22860 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
22862 2021-04-26 Jakub Jelinek <jakub@redhat.com>
22865 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
22866 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
22869 2021-04-25 liuhongt <hongtao.liu@intel.com>
22872 * config/i386/i386-builtin.def (BDESC): Change the icode of
22873 the following builtins to CODE_FOR_nothing.
22874 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
22875 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
22876 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
22877 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
22878 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
22879 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
22880 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
22881 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
22882 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
22883 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
22884 (sse2_eq<mode>3): Ditto.
22885 (sse4_1_eqv2di3): Ditto.
22886 (sse2_gt<mode>3): Rename to ..
22887 (*sse2_gt<mode>3): .. this.
22889 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22892 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22895 * config/darwin.c (darwin_binds_local_p): Assume that any
22896 public symbol might be interposed for PIC code. Update function
22897 header comment to reflect current Darwin capability.
22899 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22902 * config/darwin.c (darwin_binds_local_p): Assume that any
22903 public symbol might be interposed for PIC code. Update function
22904 header comment to reflect current Darwin capability.
22906 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
22908 * doc/sourcebuild.texi: Document no-opts and any-opts target
22911 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
22913 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
22915 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
22917 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
22919 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
22922 * config/i386/i386-options.c (ix86_option_override_internal):
22923 Error out when -m96bit-long-double is used with 64bit targets.
22924 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
22926 2021-04-23 Martin Liska <mliska@suse.cz>
22928 * lto-wrapper.c: Remove FIXME about usage of
22929 hardware_concurrency. The function is not on par with
22932 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
22935 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
22936 Copy operand 3 to operand 4. Use sse_reg_operand
22937 as operand 3 predicate.
22938 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
22939 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
22940 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
22941 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
22942 Copy operand 1 to operand 0.
22943 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
22944 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
22945 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
22947 2021-04-23 Alex Coplan <alex.coplan@arm.com>
22949 PR rtl-optimization/100230
22950 * early-remat.c (early_remat::sort_candidates): Use delete[]
22951 instead of delete for array allocated with new[].
22953 2021-04-23 Richard Biener <rguenther@suse.de>
22955 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
22956 (capture_info::capture_info): Likewise.
22957 (capture_info::walk_match): Likewise.
22958 (expr::gen_transform): Likewise.
22959 (dt_simplify::gen_1): Likewise.
22960 * gimple-match-head.c (maybe_resimplify_conditional_op):
22961 Remove VEC_COND_EXPR special-casing.
22962 (gimple_simplify): Likewise.
22963 * gimple.c (gimple_could_trap_p_1): Adjust.
22964 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
22965 to participate in PRE.
22967 2021-04-23 Richard Biener <rguenther@suse.de>
22969 * cfganal.c (connect_infinite_loops_to_exit): First call
22970 add_noreturn_fake_exit_edges.
22971 * ipa-sra.c (process_scan_results): Do not call the now redundant
22972 add_noreturn_fake_exit_edges.
22973 * predict.c (tree_estimate_probability): Likewise.
22974 (rebuild_frequencies): Likewise.
22975 * store-motion.c (one_store_motion_pass): Likewise.
22977 2021-04-23 Richard Biener <rguenther@suse.de>
22979 PR tree-optimization/100222
22980 * predict.c (pass_profile::execute): Remove redundant call to
22981 mark_irreducible_loops.
22982 (report_predictor_hitrates): Likewise.
22984 2021-04-23 Richard Biener <rguenther@suse.de>
22986 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
22987 valid_gimple_rhs_p by instead gimplifying to one.
22989 2021-04-23 Richard Biener <rguenther@suse.de>
22991 PR tree-optimization/99971
22992 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
22993 Always use TBAA for loads.
22995 2021-04-23 liuhongt <hongtao.liu@intel.com>
22998 * config/i386/i386-options.c (ix86_option_override_internal):
22999 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
23000 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
23001 by target attribute.
23003 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
23005 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
23007 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
23009 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
23011 * config.gcc (powerpc-ibm-aix6.*): Remove.
23012 * config/rs6000/aix61.h: Delete.
23014 2021-04-22 Martin Liska <mliska@suse.cz>
23016 PR testsuite/100159
23017 PR testsuite/100192
23018 * builtins.c (expand_builtin): Fix typos and missing comments.
23019 * dwarf2out.c (gen_subprogram_die): Likewise.
23020 (gen_struct_or_union_type_die): Likewise.
23022 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
23025 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
23026 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
23028 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
23030 * config/i386/darwin.h (TARGET_64BIT): Remove definition
23031 based on TARGET_ISA_64BIT.
23032 (TARGET_64BIT_P): Remove definition based on
23033 TARGET_ISA_64BIT_P().
23035 2021-04-21 Martin Liska <mliska@suse.cz>
23038 2021-04-21 Martin Liska <mliska@suse.cz>
23040 * lto-wrapper.c (cpuset_popcount): Remove.
23041 (init_num_threads): Remove and use hardware_concurrency.
23043 2021-04-21 Martin Liska <mliska@suse.cz>
23046 * main.c (main): Call toplev::finalize in CHECKING_P mode.
23047 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
23048 when incremental LTO linking happens.
23050 2021-04-21 Martin Liska <mliska@suse.cz>
23052 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
23053 makeserver cannot be detected, then use -flto=N fallback.
23055 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
23057 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
23058 default to yes for aarch64-linux-gnu.
23059 * configure: Regenerate.
23061 2021-04-21 Martin Liska <mliska@suse.cz>
23063 * lto-wrapper.c (cpuset_popcount): Remove.
23064 (init_num_threads): Remove and use hardware_concurrency.
23066 2021-04-21 Martin Liska <mliska@suse.cz>
23068 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
23069 which remains to be '(... || 0)' and clang complains about it.
23070 * dwarf2out.c (AT_vms_delta): Declare conditionally.
23071 (add_AT_vms_delta): Likewise.
23072 * tree.c (fld_simplified_type): Use rather more common pattern
23073 for disabling of something (#if 0).
23074 (get_tree_code_name): Likewise.
23075 (verify_type_variant): Likewise.
23077 2021-04-21 Martin Liska <mliska@suse.cz>
23079 * config/i386/i386-expand.c (decide_alignment): Use newly named
23080 macro TARGET_CPU_P.
23081 * config/i386/i386.c (ix86_decompose_address): Likewise.
23082 (ix86_address_cost): Likewise.
23083 (ix86_lea_outperforms): Likewise.
23084 (ix86_avoid_lea_for_addr): Likewise.
23085 (ix86_add_stmt_cost): Likewise.
23086 * config/i386/i386.h (TARGET_*): Remove.
23087 (TARGET_CPU_P): New macro.
23088 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
23089 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
23090 (swap_top_of_ready_list): Likewise.
23091 (ix86_atom_sched_reorder): Likewise.
23092 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
23093 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
23095 2021-04-21 Martin Liska <mliska@suse.cz>
23097 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
23099 (SET_TARGET_NO_SAHF): Likewise.
23100 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
23101 (SET_TARGET_PREFETCH_SSE): Likewise.
23102 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
23103 (SET_TARGET_NO_TUNE): Likewise.
23104 (TARGET_EXPLICIT_NO_80387_P): Likewise.
23105 (SET_TARGET_NO_80387): Likewise.
23107 * config/i386/i386.h (TARGET_*): Remove.
23108 * opth-gen.awk: Generate new used macros.
23110 2021-04-21 Martin Liska <mliska@suse.cz>
23112 * config/i386/i386.h (PTA_*): Remove.
23113 (enum pta_flag): New.
23114 (DEF_PTA): Generate PTA_* values from i386-isa.def.
23115 * config/i386/i386-isa.def: New file.
23117 2021-04-21 Alex Coplan <alex.coplan@arm.com>
23120 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
23121 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
23122 jump table targets.
23124 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
23126 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
23127 x86_64-*-* targets.
23128 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
23130 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
23131 (ix86_handle_option): Handle -mmwait.
23132 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
23133 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
23134 __builtin_ia32_monitor and __builtin_ia32_mwait.
23135 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
23136 (ix86_valid_target_attribute_inner_p): Likewise.
23137 (ix86_option_override_internal): Enable mwait/monitor
23138 instructions for -msse3.
23139 * config/i386/i386.h (TARGET_MWAIT): New.
23140 (TARGET_MWAIT_P): Likewise.
23141 * config/i386/i386.opt: Add -mmwait.
23142 * config/i386/mwaitintrin.h: New file.
23143 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
23144 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
23146 (@sse3_monitor_<mode>): Likewise.
23147 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
23148 * doc/extend.texi: Document mwait target attribute.
23149 * doc/invoke.texi: Document -mmwait.
23151 2021-04-21 Martin Liska <mliska@suse.cz>
23153 * config/i386/i386-options.c (DEF_ENUM): Remove it.
23154 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
23155 * config/i386/stringop.def (DEF_ENUM): Likewise.
23157 2021-04-21 Martin Liska <mliska@suse.cz>
23159 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
23160 of print_generic_expr.
23162 2021-04-21 Jakub Jelinek <jakub@redhat.com>
23164 PR rtl-optimization/100148
23165 * cprop.c (constprop_register): Use next_nondebug_insn instead of
23168 2021-04-21 Martin Liska <mliska@suse.cz>
23171 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
23172 free_dominance_info calls.
23174 2021-04-21 Richard Biener <rguenther@suse.de>
23176 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
23177 parameter (and assume it to be false).
23178 (fold_gimple_assign): Adjust, remove all callers of
23179 maybe_fold_reference calling it with is_lhs true.
23180 (gimple_fold_call): Likewise.
23181 (fold_stmt_1): Likewise.
23183 2021-04-21 Richard Biener <rguenther@suse.de>
23185 * fold-const.c (pedantic_non_lvalue_loc): Remove.
23186 (fold_binary_loc): Adjust.
23187 (fold_ternary_loc): Likewise.
23189 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
23191 PR middle-end/100130
23192 * varasm.c (get_block_for_decl): Make sure that any use of the
23193 retain attribute matches the section's retain flag.
23194 (switch_to_section): Check for retain mismatches even when
23195 changing sections, but do not warn if the given decl is the
23196 section's named.decl.
23197 (output_object_block): Pass the first decl in the block (if any)
23198 to switch_to_section.
23200 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
23202 * config/i386/i386-c.c (ix86_target_macros_internal): Define
23203 __CRC32__ for -mcrc32.
23204 * config/i386/i386-options.c (ix86_option_override_internal):
23205 Enable crc32 instruction for -msse4.2.
23206 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
23208 (sse4_2_crc32di): Likewise.
23209 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
23212 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
23215 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
23218 2021-04-20 Martin Liska <mliska@suse.cz>
23220 * doc/invoke.texi: Fix typo.
23221 * params.opt: Likewise.
23223 2021-04-20 Martin Liska <mliska@suse.cz>
23225 * doc/invoke.texi: Document new param.
23227 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
23229 PR tree-optimization/100081
23230 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
23231 rather than gori_compute_cache.
23232 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
23233 (range_def_chain::m_logical_depth): New member.
23234 (range_def_chain::range_def_chain): Initialize m_logical_depth.
23235 (range_def_chain::get_def_chain): Don't build defchains through more
23236 than LOGICAL_LIMIT logical expressions.
23237 * params.opt (param_ranger_logical_depth): New.
23239 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
23242 * config/arm/arm.c (arm_configure_build_target): Do not strip
23243 extended FPU/SIMD feature bits from the target ISA when -mfpu
23244 is specified (partial revert of r11-8168).
23246 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
23248 * params.opt (-param=openacc-kernels=): Add.
23249 * omp-oacc-kernels-decompose.cc
23250 (pass_omp_oacc_kernels_decompose::gate): Use it.
23251 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
23252 (--param): ... here, 'openacc-kernels'.
23254 2021-04-19 Martin Liska <mliska@suse.cz>
23257 * gengtype.c (finish_root_table): Align function arguments
23258 in between declaration and definition.
23260 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
23262 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
23263 frames larger than the SEH maximum frame size.
23265 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
23267 PR rtl-optimization/99927
23268 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
23269 is dead, just drop it.
23271 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23274 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
23275 * doc/tm.texi: Regenerate.
23276 * doc/tm.texi.in (D language and ABI): Add @hook for
23277 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
23279 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23281 * config/darwin-d.c (darwin_d_handle_target_object_format): New
23283 (darwin_d_register_target_info): New function.
23284 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23285 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
23287 (dragonfly_d_register_target_info): New function.
23288 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23289 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
23291 (freebsd_d_register_target_info): New function.
23292 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23293 * config/glibc-d.c (glibc_d_handle_target_object_format): New
23295 (glibc_d_register_target_info): New function.
23296 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23297 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
23299 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
23300 as handler for objectFormat key.
23301 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
23303 (winnt_d_register_target_info): New function.
23304 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23305 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
23307 (netbsd_d_register_target_info): New function.
23308 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23309 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
23311 (openbsd_d_register_target_info): New function.
23312 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23313 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
23314 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
23315 handler for objectFormat key.
23316 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
23318 (rs6000_d_register_target_info): Add
23319 rs6000_d_handle_target_object_format as handler for objectFormat key.
23320 * config/sol2-d.c (solaris_d_handle_target_object_format): New
23322 (solaris_d_register_target_info): New function.
23323 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23325 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23328 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
23329 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
23331 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
23332 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
23333 the old and new alignment after applying MIN/MAX to it is different.
23335 2021-04-16 Tamar Christina <tamar.christina@arm.com>
23338 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
23339 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
23341 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
23343 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
23345 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
23346 this section and its subsections.
23348 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23351 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
23352 define_insn patterns.
23354 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23356 PR rtl-optimization/98689
23357 * reg-notes.def (UNTYPED_CALL): New note.
23358 * combine.c (distribute_notes): Handle it.
23359 * emit-rtl.c (try_split): Likewise.
23360 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
23361 that calls with the note implicitly set all return value registers.
23362 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
23365 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23367 PR rtl-optimization/99596
23368 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
23369 register accesses for const calls. Assume that pure functions
23370 can only read from global registers. Ignore cases in which
23371 the stack pointer has been marked global.
23373 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23376 * tree-vect-loop.c (vect_transform_loop): Don't remove just
23377 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
23378 them by their last argument.
23380 2021-04-15 Martin Liska <mliska@suse.cz>
23382 * doc/invoke.texi: Other params don't use it, remove it.
23384 2021-04-15 Richard Biener <rguenther@suse.de>
23386 * gimple-builder.h: Add deprecation note.
23388 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23391 * attribs.h (restrict_type_identity_attributes_to): Declare.
23392 * attribs.c (restrict_type_identity_attributes_to): New function.
23394 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23397 * attribs.h (affects_type_identity_attributes): Declare.
23398 * attribs.c (remove_attributes_matching): New function.
23399 (affects_type_identity_attributes): Likewise.
23401 2021-04-15 Jakub Jelinek <jakub@redhat.com>
23404 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
23405 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
23406 ZERO_EXTEND, SIGN_EXTEND or AND.
23408 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23410 PR rtl-optimization/99929
23411 * rtl.h (same_vector_encodings_p): New function.
23412 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
23413 * cselib.c (rtx_equal_for_cselib_1): Likewise.
23414 * jump.c (rtx_renumbered_equal_p): Likewise.
23415 * lra-constraints.c (operands_match_p): Likewise.
23416 * reload.c (operands_match_p): Likewise.
23417 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
23419 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23421 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
23422 more information about variable-length CONST_VECTORs.
23424 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
23426 PR rtl-optimization/100066
23427 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
23428 ordered modes when choosing splitting mode for hard reg.
23430 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23433 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
23435 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
23437 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
23439 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
23440 for mask operand types.
23441 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
23442 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
23444 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
23446 * config/s390/s390.c (s390_const_operand_ok): Check the new
23447 operand types and generate a list of valid values.
23449 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23451 * doc/tm.texi: Regenerate.
23452 * doc/tm.texi.in (D language and ABI): Add @hook for
23453 TARGET_D_REGISTER_OS_TARGET_INFO.
23455 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23457 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
23459 (aarch64_d_register_target_info): New function.
23460 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
23462 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
23464 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
23465 (arm_d_register_target_info): New function.
23466 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
23467 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23468 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
23469 (ix86_d_register_target_info): New function.
23470 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
23471 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23472 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
23473 (mips_d_register_target_info): New function.
23474 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
23475 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23476 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
23477 (pa_d_register_target_info): New function.
23478 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
23479 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23480 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
23482 (riscv_d_register_target_info): New function.
23483 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
23484 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23485 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
23487 (rs6000_d_register_target_info): New function.
23488 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
23490 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23491 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
23492 (s390_d_register_target_info): New function.
23493 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
23494 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23495 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
23497 (sparc_d_register_target_info): New function.
23498 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
23499 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23500 * doc/tm.texi: Regenerate.
23501 * doc/tm.texi.in (D language and ABI): Add @hook for
23502 TARGET_D_REGISTER_CPU_TARGET_INFO.
23504 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23506 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
23507 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
23508 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
23509 * doc/tm.texi: Regenerate.
23510 * doc/tm.texi.in (D language and ABI): Add @hook for
23511 TARGET_D_HAS_STDCALL_CONVENTION.
23513 2021-04-14 Richard Biener <rguenther@suse.de>
23515 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
23516 VEC_COND_EXPRs have a gimple_val condition.
23517 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
23518 can no longer have a GENERIC condition.
23520 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
23523 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
23524 from the isa_delta when -mfpu has been used.
23525 (arm_options_perform_arch_sanity_checks): It's the architecture that
23526 lacks an FPU not the processor.
23528 2021-04-13 Richard Biener <rguenther@suse.de>
23530 PR tree-optimization/100053
23531 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
23532 not use optimistic dominance queries for backedges to validate
23534 (dominated_by_p_w_unex): Add parameter to ignore executable
23535 state on backedges.
23536 (rpo_elim::eliminate_avail): Adjust.
23538 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23541 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
23542 *aarch64_bfxilsi_extrdi): New define_insn patterns.
23544 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23547 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
23548 outermode, return NULL if the result doesn't encode back to the
23549 original byte sequence.
23550 (simplify_gen_subreg): Don't create SUBREGs from constants to
23551 MODE_COMPOSITE_P outermode.
23553 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23555 PR rtl-optimization/99905
23556 * combine.c (expand_compound_operation): If pos + len > modewidth,
23557 perform the right shift by pos in inner_mode and then convert to mode,
23558 instead of trying to simplify a shift of rtx with inner_mode by pos
23559 as if it was a shift in mode.
23561 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23564 * combine.c (simplify_and_const_int_1): Don't optimize varop
23565 away if it has side-effects.
23567 2021-04-12 Martin Liska <mliska@suse.cz>
23569 * doc/extend.texi: Escape @smallexample content.
23571 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
23573 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
23574 alternative in order to load a DFP zero.
23576 2021-04-12 Martin Liska <mliska@suse.cz>
23578 * doc/extend.texi: Be more precise in documentation
23579 of symver attribute.
23581 2021-04-12 Martin Liska <mliska@suse.cz>
23584 * gimplify.c (gimplify_expr): Right now, we unpoison all
23585 variables before a goto <dest>. We should not do it if we are
23588 2021-04-12 Cui,Lili <lili.cui@intel.com>
23590 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
23592 * common/config/i386/i386-common.c (processor_names): Add
23594 (processor_alias_table): Add rocketlake.
23595 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
23596 INTEL_COREI7_ROCKETLAKE.
23597 * config.gcc: Add -march=rocketlake.
23598 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23600 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
23601 (processor_cost_table): Add rocketlake cost.
23602 * config/i386/i386.h (ix86_size_cost) : Define
23604 (processor_type) : Add PROCESSOR_ROCKETLAKE.
23605 (PTA_ROCKETLAKE): Ditto.
23606 * doc/extend.texi: Add rocketlake.
23607 * doc/invoke.texi: Add rocketlake.
23609 2021-04-12 Cui,Lili <lili.cui@intel.com>
23611 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
23612 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
23613 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
23614 * doc/invoke.texi: Change alderlake ISA list.
23616 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
23618 PR middle-end/98088
23619 * omp-expand.c (expand_oacc_collapse_init): Update condition in
23622 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
23625 * config/i386/serializeintrin.h (_serialize): Defined as macro.
23627 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23630 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
23631 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
23633 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23635 PR middle-end/99989
23636 * gimple-ssa-warn-alloca.c
23637 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
23638 0 with integer precision unconditionally.
23640 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23642 PR rtl-optimization/98601
23643 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
23644 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
23645 unaligned_mems handle VOIDmode like BLKmode.
23647 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
23650 * tree.c (free_lang_data_in_decl): Do not release body of
23651 declare_variant_alt.
23653 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23655 * config/aarch64/aarch64.c (aarch64_option_restore): If the
23656 architecture was specified explicitly and the tuning wasn't,
23657 tune for the architecture rather than the configured default CPU.
23659 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23661 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
23662 as the temporary register.
23664 2021-04-09 Martin Liska <mliska@suse.cz>
23666 * doc/extend.texi: Move non-target attributes on the top level.
23668 2021-04-09 Martin Liska <mliska@suse.cz>
23670 * doc/invoke.texi: Document minimum and maximum value of the
23671 argument for both supported compression algorithms.
23673 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
23675 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
23676 TLS BSS before TLS data.
23677 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
23679 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23681 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
23683 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23685 * match.pd: Extend vec_cond folds to handle shifts.
23687 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
23689 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
23692 2021-04-08 Alex Coplan <alex.coplan@arm.com>
23695 * config/arm/iterators.md (MVE_vecs): New.
23696 (V_elem): Also handle V2DF.
23697 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
23698 (*mve_vdup<mode>): ... this. Remove second alternative since
23699 vec_duplicate of const_int is not canonical RTL, and we don't
23700 want to match symbol_refs.
23701 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
23703 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
23705 * fold-const.c (fold_single_bit_test): Fix typo.
23706 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
23709 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
23711 PR tree-optimization/97513
23712 * tree-vect-slp.c (vect_add_slp_permutation): New function,
23714 (vectorizable_slp_permutation): ...here. Detect cases in which
23715 all VEC_PERM_EXPRs are guaranteed to have the same stepped
23716 permute vector and only generate one permute vector for that case.
23717 Extend that case to handle variable-length vectors.
23719 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
23721 PR tree-optimization/99873
23722 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
23723 (vect_build_slp_instance): Don't split store groups that could
23724 use IFN_STORE_LANES.
23726 2021-04-07 Jakub Jelinek <jakub@redhat.com>
23729 * varasm.c (output_constant_pool_contents): Don't strip name encoding
23730 from XSTR (desc->sym, 0) or from label before passing those to
23733 2021-04-07 Richard Biener <rguenther@suse.de>
23735 PR tree-optimization/99954
23736 * tree-loop-distribution.c: Include tree-affine.h.
23737 (generate_memcpy_builtin): Try using tree-affine to prove
23739 (loop_distribution::classify_builtin_ldst): Always classify
23742 2021-04-07 Richard Biener <rguenther@suse.de>
23744 PR tree-optimization/99947
23745 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
23746 steps vector to avoid pushing elements from the reallocated
23749 2021-04-07 Richard Biener <rguenther@suse.de>
23751 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
23752 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
23754 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
23756 (debug_vn_reference_ops): New.
23758 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
23760 PR tree-optimization/98736
23761 * tree-loop-distribution.c
23762 * (loop_distribution::bb_top_order_init):
23763 Compute RPO with programing order preserved by calling function
23764 rev_post_order_and_mark_dfs_back_seme.
23766 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
23769 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
23770 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
23772 (process_bb_lives): Don't update biggest mode of hard reg for
23773 implicit in multi-register group. Use the new functions for
23774 updating dead_set and unused_set by register notes.
23776 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
23778 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
23781 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
23783 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
23784 (skylake_memset): Likewise.
23785 (skylake_cost): Change CLEAR_RATIO to 17.
23786 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
23787 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
23788 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
23790 2021-04-06 Richard Biener <rguenther@suse.de>
23792 PR tree-optimization/99880
23793 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
23794 set vectorized defs of relevant PHIs.
23796 2021-04-06 Richard Biener <rguenther@suse.de>
23798 PR tree-optimization/99924
23799 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
23800 nodes w/o scalar stmts as visited.
23802 2021-04-06 Alex Coplan <alex.coplan@arm.com>
23805 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
23806 PCS for [su]fix_optab.
23808 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
23810 * config/darwin.c (machopic_legitimize_pic_address): Check
23811 that the current pic register is one of the hard reg set
23812 before setting liveness.
23814 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
23816 * config/darwin.c (machopic_legitimize_pic_address): Fix
23817 whitespace, remove unused code.
23819 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23821 PR tree-optimization/99882
23822 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
23825 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23827 PR rtl-optimization/99863
23828 * dse.c (replace_read): Drop regs_live argument. Instead of
23829 regs_live, use store_insn->fixed_regs_live if non-NULL,
23830 otherwise punt if insns sequence clobbers or sets any hard
23833 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23836 * targhooks.h (default_print_patchable_function_entry_1): Declare.
23837 * targhooks.c (default_print_patchable_function_entry_1): New function,
23838 copied from default_print_patchable_function_entry with an added flags
23840 (default_print_patchable_function_entry): Rewritten into a small
23841 wrapper around default_print_patchable_function_entry_1.
23842 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23844 (rs6000_print_patchable_function_entry): New function.
23846 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
23848 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
23850 2021-04-01 Jason Merrill <jason@redhat.com>
23853 * common.opt: Document v15 and v16.
23855 2021-04-01 Richard Biener <rguenther@suse.de>
23857 PR tree-optimization/99863
23858 * gimplify.c (gimplify_init_constructor): Recompute vector
23861 2021-04-01 Jakub Jelinek <jakub@redhat.com>
23863 * doc/extend.texi (symver attribute): Fix up syntax errors
23866 2021-04-01 Jakub Jelinek <jakub@redhat.com>
23868 PR tree-optimization/96573
23869 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
23870 also pointer types.
23872 2021-04-01 Richard Biener <rguenther@suse.de>
23874 PR tree-optimization/99856
23875 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
23876 precision to vector element precision.
23878 2021-04-01 Martin Jambor <mjambor@suse.cz>
23880 PR tree-optimization/97009
23881 * tree-sra.c (access_or_its_child_written): New function.
23882 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
23885 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
23888 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
23890 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
23893 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
23894 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
23895 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
23896 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
23897 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
23898 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
23900 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
23901 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
23902 (define_attr "prefixed"): Update initializer.
23904 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23907 * dwarf2out.c (debug_ranges_dwo_section): New variable.
23908 (DW_RANGES_IDX_SKELETON): Define.
23909 (struct dw_ranges): Add begin_entry and end_entry members.
23910 (DEBUG_DWO_RNGLISTS_SECTION): Define.
23911 (add_ranges_num): Adjust r initializer for addition of *_entry
23913 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
23914 set idx to DW_RANGES_IDX_SKELETON.
23915 (use_distinct_base_address_for_range): New function.
23916 (index_rnglists): Don't set r->idx if it is equal to
23917 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
23918 r->end_entry for -gsplit-dwarf if those will be needed by
23920 (output_rnglists): Add DWO argument. If true, switch to
23921 debug_ranges_dwo_section rather than debug_ranges_section.
23922 Adjust l1/l2 label indexes. Only output the offset table when
23923 dwo is true and don't include in there the skeleton range
23924 entry if present. For -gsplit-dwarf, skip ranges that belong
23925 to the other rnglists section. Change return type from void
23926 to bool and return true if there are any range entries for
23927 the other section. For dwarf_split_debug_info use
23928 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
23929 entries instead of DW_RLE_start_end, DW_RLE_start_length and
23930 DW_RLE_base_address. Use use_distinct_base_address_for_range.
23931 (init_sections_and_labels): Initialize debug_ranges_dwo_section
23932 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
23933 and range_base_label indexes.
23934 (dwarf2out_finish): Call index_rnglists earlier before finalizing
23935 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
23936 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
23937 with different dwo arguments.
23938 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
23940 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23942 PR tree-optimization/98268
23943 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
23944 recompute_tree_invariant_for_addr_expr after successfully
23945 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
23947 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23949 PR tree-optimization/99726
23950 * tree-data-ref.c (create_intersect_range_checks_index): Bail
23951 out if there is more than one access function SCEV for the loop
23954 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23956 PR rtl-optimization/97141
23957 PR rtl-optimization/98726
23958 * emit-rtl.c (valid_for_const_vector_p): Return true for
23960 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
23961 poly_wide_int instead of a wide_int.
23962 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
23964 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
23965 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
23966 false for CONST_VECTORs that cannot be forced to memory.
23967 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
23968 is too complex to force to memory, build it up from individual
23971 2021-03-31 Jan Hubicka <jh@suse.cz>
23974 * cgraph.c (cgraph_node::release_body): Fix overactive check.
23976 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
23979 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
23982 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
23984 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
23985 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
23987 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
23988 "rep movsb/stosb" only for known sizes.
23989 * config/i386/i386-options.c (processor_cost_table): Use Ice
23990 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
23991 Rapids and Alder Lake.
23992 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
23993 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
23994 (icelake_memset): Likewise.
23995 (icelake_cost): Likewise.
23996 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
23999 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
24002 * config/aarch64/aarch64.c
24003 (aarch64_vectorize_preferred_vector_alignment): Query the size
24004 of the provided SVE vector; do not assume that all SVE vectors
24005 have the same size.
24007 2021-03-31 Jan Hubicka <jh@suse.cz>
24010 * cgraph.c (cgraph_node::release_body): Remove all callers and
24012 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
24013 * cgraphunit.c (cgraph_node::expand): And here.
24015 2021-03-31 Martin Liska <mliska@suse.cz>
24017 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
24018 and one negated condition.
24020 2021-03-31 Jakub Jelinek <jakub@redhat.com>
24021 Richard Sandiford <richard.sandiford@arm.com>
24024 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
24025 constraints on operands[2] and similarly 0 and rk constraints
24026 on operands[1] corresponding to that.
24028 2021-03-31 Jakub Jelinek <jakub@redhat.com>
24031 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
24032 linker doesn't support DWARF sections new in DWARF5.
24033 * config/i386/i386-options.c (ix86_option_override_internal): Default
24034 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
24036 * config.in: Regenerated.
24037 * configure: Regenerated.
24039 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24042 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
24043 available issue_info before using it.
24045 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24048 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
24051 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
24054 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
24055 (altivec_lvsl_reg_<mode>): ... this.
24056 (altivec_lvsr_reg): Change to ...
24057 (altivec_lvsr_reg_<mode>): ... this.
24058 * config/rs6000/predicates.md (vec_set_index_operand): New.
24059 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24060 Enable 32bit variable vec_insert for all TARGET_VSX.
24061 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
24062 Enable 32bit variable vec_insert for p9 and above.
24063 (rs6000_expand_vector_set_var_p8): Rename to ...
24064 (rs6000_expand_vector_set_var_p7): ... this.
24065 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
24067 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
24068 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
24069 gen_altivec_lvsr_reg_di.
24071 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
24074 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
24075 (__rdtscp): Likewise.
24077 2021-03-30 Tamar Christina <tamar.christina@arm.com>
24079 PR tree-optimization/99825
24080 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
24081 Reject non-mult 2 lanes.
24083 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
24086 * config/arm/arm.c (arm_file_start): Fix emission of
24087 Tag_ABI_VFP_args attribute.
24089 2021-03-30 Richard Biener <rguenther@suse.de>
24091 PR tree-optimization/99824
24092 * stor-layout.c (set_min_and_max_values_for_integral_type):
24093 Assert the precision is within the bounds of
24094 WIDE_INT_MAX_PRECISION.
24095 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
24096 the outermost component ref only to lower the access size
24097 and initialize that from the access type.
24099 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
24102 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
24103 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
24105 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
24107 * config/aarch64/aarch64.md
24108 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
24109 attribute to disambiguate between SIMD and FP variants of the
24112 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
24114 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
24115 (analyze_ssa_name_flags): Fix typo in comment.
24117 2021-03-29 Alex Coplan <alex.coplan@arm.com>
24120 * config/aarch64/aarch64-sve-builtins.cc
24121 (function_builder::add_function): Add placeholder_p argument, use
24122 placeholder decls if this is set.
24123 (function_builder::add_unique_function): Instead of conditionally adding
24124 direct overloads, unconditionally add either a direct overload or a
24126 (function_builder::add_overloaded_function): Set placeholder_p if we're
24127 using C++ overloads. Use the obstack for string storage instead
24128 of relying on the tree nodes.
24129 (function_builder::add_overloaded_functions): Don't return early for
24130 m_direct_overloads: we need to add placeholders.
24131 * config/aarch64/aarch64-sve-builtins.h
24132 (function_builder::add_function): Add placeholder_p argument.
24134 2021-03-29 Richard Biener <rguenther@suse.de>
24136 PR tree-optimization/99807
24137 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
24138 assert below VEC_PERM handling.
24140 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24143 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
24144 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
24145 matching const_int 0.
24146 (move_lo_quad_internal_be_<mode>): Likewise.
24147 (move_lo_quad_<mode>): Update for the above.
24148 * config/aarch64/iterators.md (VQ_2E): Delete.
24150 2021-03-29 Jakub Jelinek <jakub@redhat.com>
24152 PR tree-optimization/99777
24153 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
24154 types other than scalar integral types.
24156 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
24158 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
24159 XCOFF TLS reloc decorations.
24161 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
24163 * doc/analyzer.texi (Analyzer Internals): Update link to
24164 "A Memory Model for Static Analysis of C Programs".
24166 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
24168 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
24169 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
24171 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
24172 (rs6000_special_round_type_align): Recursively check innermost first
24175 2021-03-26 Jakub Jelinek <jakub@redhat.com>
24178 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
24179 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
24180 assignment with drap_reg active, queue reg save for hfp with offset 0
24181 and flush queued reg saves. When handling a push with rule18,
24182 defer queueing reg save for hfp and just assert the offset is 0.
24183 (scan_trace): Assert that fde->rule18 is false.
24185 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
24188 * ira-costs.c (record_reg_classes): Put case with
24189 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
24190 * ira.c (ira_setup_alts): Ditto.
24191 * lra-constraints.c (process_alt_operands): Ditto.
24192 * recog.c (asm_operand_ok): Ditto.
24193 * reload.c (find_reloads): Ditto.
24195 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24197 * config/aarch64/aarch64-protos.h
24198 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
24199 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
24200 * config/aarch64/aarch64.c (generic_addrcost_table): Update
24201 accordingly, using the same costs as for post_modify.
24202 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
24203 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
24204 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
24205 (a64fx_addrcost_table): Likewise.
24206 (neoversev1_addrcost_table): New.
24207 (neoversev1_tunings): Use neoversev1_addrcost_table.
24208 (aarch64_address_cost): Use the new post_modify costs for CImode
24211 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24213 * config/aarch64/aarch64.opt
24214 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
24215 * doc/invoke.texi: Document it.
24216 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
24217 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
24218 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
24219 (aarch64_vec_issue_info): New structures.
24220 (cpu_vector_cost): Write comments above the variables rather
24222 (cpu_vector_cost::issue_info): New member variable.
24223 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
24224 and tree-ssa-loop-niter.h.
24225 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
24226 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
24227 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
24228 (thunderx3t110_vector_cost): Initialize issue_info to null.
24229 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
24230 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
24231 (neoversev1_vector_cost): Use them.
24232 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
24233 (aarch64_vector_costs::saw_sve_only_op): New member variable.
24234 (aarch64_vector_costs::num_vector_iterations): Likewise.
24235 (aarch64_vector_costs::scalar_ops): Likewise.
24236 (aarch64_vector_costs::advsimd_ops): Likewise.
24237 (aarch64_vector_costs::sve_ops): Likewise.
24238 (aarch64_vector_costs::seen_loads): Likewise.
24239 (aarch64_simd_vec_costs_for_flags): New function.
24240 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
24241 Count the number of predicate operations required by SVE WHILE
24243 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
24244 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
24245 (aarch64_count_ops): Likewise.
24246 (aarch64_add_stmt_cost): Record whether see an SVE operation
24247 that cannot currently be implementing using Advanced SIMD.
24248 Record issue information about the scalar, Advanced SIMD
24249 and (where relevant) SVE versions of a loop.
24250 (aarch64_vec_op_count::dump): New function.
24251 (aarch64_sve_op_count::dump): Likewise.
24252 (aarch64_estimate_min_cycles_per_iter): Likewise.
24253 (aarch64_adjust_body_cost): If issue information is available,
24254 try to compare the issue rates of the various loop implementations
24255 and increase or decrease the vector body cost accordingly.
24257 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24259 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
24260 Assume a zero cost for induction phis.
24262 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24264 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
24266 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
24267 vector comparisons.
24269 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24271 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
24273 (aarch64_add_stmt_cost): Call it.
24275 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24277 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
24278 New tuning parameter.
24279 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
24280 (aarch64_estimated_sve_vq): New function.
24281 (aarch64_vector_costs::analyzed_vinfo): New member variable.
24282 (aarch64_vector_costs::is_loop): Likewise.
24283 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
24284 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
24285 (aarch64_record_potential_advsimd_unrolling): New function.
24286 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
24287 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
24288 aarch64_analyze_bb_vinfo on the first use of a costs structure.
24289 Detect whether we're vectorizing a loop for SVE that might be
24290 completely unrolled if it used Advanced SIMD instead.
24291 (aarch64_adjust_body_cost_for_latency): New function.
24292 (aarch64_finish_cost): Call it.
24294 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24296 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
24297 (aarch64_init_cost): New function.
24298 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
24299 the default unsigned[3].
24300 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
24301 (TARGET_VECTORIZE_INIT_COST): Override.
24302 (TARGET_VECTORIZE_FINISH_COST): Likewise.
24303 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
24305 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24307 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
24308 (neoversev1_sve_vector_cost): New cost structures.
24309 (neoversev1_vector_cost): Likewise.
24310 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
24312 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24314 * config/aarch64/aarch64-protos.h
24315 (sve_vec_cost::scatter_store_elt_cost): New member variable.
24316 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24317 accordingly, taking the cost from the cost of a scalar_store.
24318 (a64fx_sve_vector_cost): Likewise.
24319 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
24321 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24323 * config/aarch64/aarch64-protos.h
24324 (simd_vec_cost::store_elt_extra_cost): New member variable.
24325 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24326 accordingly, using the vec_to_scalar cost for the new field.
24327 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24328 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24329 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24330 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24331 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24332 (thunderx3t110_advsimd_vector_cost): Likewise.
24333 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
24335 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24337 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
24338 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
24339 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
24340 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24341 accordingly, using zero for the new costs.
24342 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24343 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24344 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24345 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24346 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24347 (thunderx3t110_advsimd_vector_cost): Likewise.
24348 (aarch64_ld234_st234_vectors): New function.
24349 (aarch64_adjust_stmt_cost): Likewise.
24350 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
24351 the new vector costs.
24353 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24355 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
24356 derived class of simd_vec_cost. Add information about CLAST[AB]
24357 and FADDA instructions.
24358 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24359 accordingly, using the vec_to_scalar costs for the new fields.
24360 (a64fx_sve_vector_cost): Likewise.
24361 (aarch64_reduc_type): New function.
24362 (aarch64_sve_in_loop_reduction_latency): Likewise.
24363 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
24364 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
24365 that occur in the loop body.
24366 (aarch64_add_stmt_cost): Update call accordingly.
24368 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24370 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
24372 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
24373 above the fields rather than to the right.
24374 (simd_vec_cost::reduc_i8_cost): New member variable.
24375 (simd_vec_cost::reduc_i16_cost): Likewise.
24376 (simd_vec_cost::reduc_i32_cost): Likewise.
24377 (simd_vec_cost::reduc_i64_cost): Likewise.
24378 (simd_vec_cost::reduc_f16_cost): Likewise.
24379 (simd_vec_cost::reduc_f32_cost): Likewise.
24380 (simd_vec_cost::reduc_f64_cost): Likewise.
24381 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24382 accordingly, using the vec_to_scalar_cost for the new fields.
24383 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24384 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24385 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24386 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24387 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24388 (thunderx3t110_advsimd_vector_cost): Likewise.
24389 (aarch64_use_new_vector_costs_p): New function.
24390 (aarch64_simd_vec_costs): New function, split out from...
24391 (aarch64_builtin_vectorization_cost): ...here.
24392 (aarch64_is_reduction): New function.
24393 (aarch64_detect_vector_stmt_subtype): Likewise.
24394 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
24395 using the new vector costs.
24397 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24400 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
24401 TLS declarations as public.
24403 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24405 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
24406 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
24407 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
24408 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
24409 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
24410 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
24411 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
24412 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
24413 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
24415 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24418 * config.gcc (*-*-cygwin*): Add winnt-d.o
24419 (*-*-mingw*): Likewise.
24420 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
24421 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
24422 * config/i386/t-cygming: Add winnt-d.o.
24423 * config/i386/winnt-d.c: New file.
24425 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24427 * config/freebsd-d.c: Include memmodel.h.
24429 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24432 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
24433 * config/t-openbsd: Add openbsd-d.o.
24434 * config/openbsd-d.c: New file.
24436 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
24438 PR tree-optimization/96974
24439 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
24440 with graceful exit.
24442 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24445 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24449 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24450 always_inline in system headers.
24452 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
24454 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
24456 2021-03-25 Jakub Jelinek <jakub@redhat.com>
24459 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
24460 * fold-const.c (operand_compare::operand_equal_p): Don't compare
24461 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
24463 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24467 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24468 always_inline in system headers.
24470 2021-03-25 Richard Biener <rguenther@suse.de>
24472 PR tree-optimization/99746
24473 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
24474 the scalar stmt as patterned. Instead set up required things
24477 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
24479 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
24482 2021-03-24 Martin Liska <mliska@suse.cz>
24485 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
24487 * config/i386/i386-options.c (ix86_option_override_internal):
24488 Add run-time assert.
24490 2021-03-24 Martin Jambor <mjambor@suse.cz>
24493 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
24494 parameters with unknown type.
24495 (ipacp_value_safe_for_type): New function.
24496 (propagate_vals_across_arith_jfunc): Verify that the constant type
24497 can be used for a type of the formal parameter.
24498 (propagate_vals_across_ancestor): Likewise.
24499 (propagate_scalar_across_jump_function): Likewise. Pass the type
24500 also to propagate_vals_across_ancestor.
24502 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
24505 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
24507 (movmisalign<mode>_mve_load): Likewise.
24509 2021-03-24 Jakub Jelinek <jakub@redhat.com>
24512 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
24513 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
24515 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24517 * doc/sourcebuild.texi (sysconf): New effective target.
24519 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24521 * config/i386/predicates.md (reg_or_const_vec_operand): New.
24522 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
24523 the now *-prefixed insn_and_split, turn the splitter const vec
24524 into an input for the insn, making it an ignored immediate for
24525 non-split cases, and loaded into the scratch register
24528 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
24531 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
24532 Use define_relaxed_memory_constraint for them.
24534 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
24537 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
24538 colon to the diagnostic message.
24540 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
24542 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
24544 (try_fwprop_subst_note): Use set_info instead of insn_info.
24545 (try_fwprop_subst_pattern): Likewise.
24546 (try_fwprop_subst_notes): Likewise.
24547 (try_fwprop_subst): Likewise.
24548 (forward_propagate_subreg): Likewise.
24549 (forward_propagate_and_simplify): Likewise.
24550 (forward_propagate_into): Likewise.
24551 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
24553 (set_info::single_nondebug_insn_use): Likewise.
24554 (set_info::single_phi_use): Likewise.
24555 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
24557 (set_info::single_nondebug_insn_use): Likewise.
24558 (set_info::single_phi_use): Likewise.
24560 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
24562 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
24564 2021-03-23 Jakub Jelinek <jakub@redhat.com>
24567 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
24568 expand_mult to perform an unsigned rather than a signed
24571 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
24574 * config/i386/cpuid.h (__cpuid): Add __volatile__.
24575 (__cpuid_count): Likewise.
24577 2021-03-23 Richard Biener <rguenther@suse.de>
24579 PR tree-optimization/99721
24580 * tree-vect-slp.c (vect_slp_analyze_node_operations):
24581 Make sure we can schedule the node.
24583 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24585 * config/riscv/riscv.c (riscv_subword): Take endianness into
24586 account when calculating the byte offset.
24588 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24590 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
24591 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
24592 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
24593 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
24594 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
24595 new predicate "subreg_lowpart_operator"
24597 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24599 * config/riscv/riscv.c (riscv_swap_instruction): New function
24600 to byteswap an SImode rtx containing an instruction.
24601 (riscv_trampoline_init): Byteswap the generated instructions
24604 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24606 * common/config/riscv/riscv-common.c
24607 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
24608 * config.gcc (riscv32be-*, riscv64be-*): Set
24609 TARGET_BIG_ENDIAN_DEFAULT to 1.
24610 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
24611 depending on default endianness.
24612 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24613 * config/riscv/linux.h (LINK_SPEC): Likewise.
24614 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
24615 default endianness.
24616 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
24618 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24620 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
24621 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24622 * config/riscv/linux.h (LINK_SPEC): Likewise.
24623 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
24625 (BYTES_BIG_ENDIAN): Handle big endian.
24626 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
24627 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
24629 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
24631 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
24633 * regcprop.c (find_oldest_value_reg): Ask target whether
24634 different mode is fine for replacement register.
24636 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
24638 PR tree-optimization/99296
24639 * value-range.cc (irange::irange_set_1bit_anti_range): New.
24640 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
24641 * value-range.h (irange::irange_set_1bit_anti_range): New.
24643 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
24646 * config/aarch64/constraints.md (UtQ): Use
24647 define_relaxed_memory_constraint for it.
24648 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
24649 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24650 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
24651 (have_relaxed_memory_constraints): New static var.
24652 (relaxed_memory_start, relaxed_memory_end): Ditto.
24653 (add_constraint): Add arg is_relaxed_memory. Check name for
24654 relaxed memory. Set up is_relaxed_memory in constraint_data and
24655 have_relaxed_memory_constraints. Adjust calls.
24656 (choose_enum_order): Process relaxed memory.
24657 (write_tm_preds_h): Ditto.
24658 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24659 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24660 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
24661 * ira-lives.c (single_reg_class): Use
24662 insn_extra_relaxed_memory_constraint.
24663 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
24664 * lra-constraints.c (valid_address_p): Use
24665 insn_extra_relaxed_memory_constraint instead of other memory
24667 (process_alt_operands): Process CT_RELAXED_MEMORY.
24668 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
24669 * recog.c (asm_operand_ok, preprocess_constraints): Process
24671 * reload.c (find_reloads): Ditto.
24672 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
24673 * stmt.c (parse_input_constraint): Use
24674 insn_extra_relaxed_memory_constraint.
24676 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
24679 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
24682 2021-03-22 Alex Coplan <alex.coplan@arm.com>
24685 * config/arm/arm-protos.h (neon_make_constant): Add generate
24686 argument to guard emitting insns, default to true.
24687 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
24688 CONST_VECTORs which neon_make_constant can't handle.
24689 (neon_vdup_constant): Add generate argument, avoid emitting
24690 insns if it's not set.
24691 (neon_make_constant): Plumb new generate argument through.
24692 * config/arm/constraints.md (Ui): New. Use it...
24693 * config/arm/mve.md (*mve_mov<mode>): ... here.
24694 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
24695 synthesize constants.
24697 2021-03-22 Richard Biener <rguenther@suse.de>
24699 * debug.h: Add deprecation warning.
24701 2021-03-22 Richard Biener <rguenther@suse.de>
24703 PR tree-optimization/99694
24704 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
24707 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
24710 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
24711 after type checking.
24713 2021-03-22 Jakub Jelinek <jakub@redhat.com>
24717 * dwarf2out.c (get_full_len): Use get_precision rather than
24719 (add_const_value_attribute): Make sure add_AT_wide argument has
24720 precision prec rather than some very wide one.
24722 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
24724 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
24725 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
24726 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
24727 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
24728 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
24729 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
24730 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
24731 *fix_trunc<mode>si2_internal): Fix empty split condition.
24732 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
24733 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
24734 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
24735 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
24737 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
24740 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
24741 Convert idx to DImode.
24742 (rs6000_expand_vector_set_var_p8): Likewise.
24744 2021-03-21 Jakub Jelinek <jakub@redhat.com>
24747 * dwarf2out.c (insert_float): Change return type from void to
24748 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
24749 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
24752 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
24755 * config/i386/i386.c (construct_container): Check cfun != NULL
24756 before accessing silent_p.
24758 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
24760 * asan.c: Fix typos in comments.
24762 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
24764 PR rtl-optimization/99680
24765 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
24766 (process_address_1): Check empty constraint before using
24769 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
24771 * config/rs6000/rs6000.c (power10_cost): New.
24772 (rs6000_option_override_internal): Set Power10 costs.
24773 (rs6000_issue_rate): Set Power10 issue rate.
24774 * config/rs6000/power10.md: Rewrite for Power10.
24776 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
24779 * lra-constraints.c (process_address_1): Don't use unknown
24780 constraint for address constraint.
24782 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
24785 * config.gcc (powerpc-*-darwin8): Delete the reference to
24786 the now removed darwin8.h.
24788 2021-03-19 Olivier Hainque <hainque@adacore.com>
24791 * config/vxworksae.h (VX_CPU_PREFIX): Define.
24793 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
24795 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
24797 2021-03-19 Tamar Christina <tamar.christina@arm.com>
24799 PR tree-optimization/99656
24800 * tree-vect-slp-patterns.c (linear_loads_p,
24801 complex_add_pattern::matches, is_eq_or_top,
24802 vect_validate_multiplication, complex_mul_pattern::matches,
24803 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
24804 * tree-vectorizer.h: (complex_load_perm_t): Removed.
24805 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
24806 complex_load_perm_t.
24808 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
24811 * config/i386/i386-options.c (ix86_init_machine_status): Set
24813 * config/i386/i386.c (init_cumulative_args): Set silent_p to
24815 (construct_container): Return early for return and argument
24816 errors if silent_p is true.
24817 * config/i386/i386.h (machine_function): Add silent_p.
24819 2021-03-19 Jakub Jelinek <jakub@redhat.com>
24822 * config/arm/constraints.md (Ds): New constraint.
24823 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
24824 constraint instead of w,Dm.
24826 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
24828 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
24831 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
24833 PR middle-end/99641
24834 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
24835 array type, do the computation of the current position in sizetype.
24837 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
24840 * lra-constraints.c (process_address_1): Use lookup_constraint
24841 only for a single constraint.
24843 2021-03-18 Martin Sebor <msebor@redhat.com>
24845 PR middle-end/99502
24846 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
24847 (inbounds_memaccess_p): ...to this. Check the ending offset of
24848 the accessed member.
24850 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
24852 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
24853 %> quote markers to error messages.
24854 (gcn_goacc_validate_dims): Likewise.
24855 (gcn_conditional_register_usage): Remove exclaimation mark from error
24857 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
24859 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
24861 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
24864 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
24865 Kito Cheng <kito.cheng@sifive.com>
24867 * config/riscv/riscv.c (riscv_block_move_straight): Change type
24868 to unsigned HOST_WIDE_INT for parameter and local variable with
24869 HOST_WIDE_INT type.
24870 (riscv_adjust_block_mem): Ditto.
24871 (riscv_block_move_loop): Ditto.
24872 (riscv_expand_block_move): Ditto.
24874 2021-03-18 Nick Clifton <nickc@redhat.com>
24876 * config/v850/v850.c (construct_restore_jr): Increase static
24878 (construct_save_jarl): Likewise.
24879 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
24881 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24883 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
24884 (aarch64_override_options_internal): Use it.
24885 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
24888 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
24890 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
24891 error message format issues.
24892 (nios2_option_override): Likewise.
24893 (nios2_expand_fpu_builtin): Likewise.
24894 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
24895 truncation warning.
24896 (nios2_expand_custom_builtin): More error message format fixes.
24897 (nios2_expand_rdwrctl_builtin): Likewise.
24898 (nios2_expand_rdprs_builtin): Likewise.
24899 (nios2_expand_eni_builtin): Likewise.
24900 (nios2_expand_builtin): Likewise.
24901 (nios2_register_custom_code): Likewise.
24902 (nios2_valid_target_attribute_rec): Likewise.
24903 (nios2_add_insn_asm): Fix uninitialized variable warning.
24905 2021-03-17 Jan Hubicka <jh@suse.cz>
24907 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
24908 of gather to match reality.
24909 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
24911 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24913 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
24914 to compare against CC_REG rather than NE.
24916 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
24919 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
24920 inline assembly statements.
24921 (ix86_print_operand): Update 'P' handling for -fno-plt.
24923 2021-03-17 Tamar Christina <tamar.christina@arm.com>
24926 * config/aarch64/aarch64.c
24927 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
24929 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
24932 * config/rs6000/predicates.md (branch_comparison_operator): Allow
24933 ordered and unordered for CCFPmode, if flag_finite_math_only.
24935 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24938 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
24939 rather than ASHIFT.
24940 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
24942 2021-03-16 Martin Liska <mliska@suse.cz>
24945 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
24946 cl_optimization_compare function.
24948 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
24950 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
24953 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24956 * config/i386/i386.h (struct machine_function): Add
24957 has_explicit_vzeroupper bitfield.
24958 * config/i386/i386-expand.c (ix86_expand_builtin): Set
24959 cfun->machine->has_explicit_vzeroupper when expanding
24960 IX86_BUILTIN_VZEROUPPER.
24961 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
24962 Do the mode switching only when TARGET_VZEROUPPER, expensive
24963 optimizations turned on and not optimizing for size.
24964 (pass_insert_vzeroupper::gate): Enable even when
24965 cfun->machine->has_explicit_vzeroupper is set.
24967 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24970 * config/aarch64/aarch64.c
24971 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
24972 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
24973 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
24975 2021-03-15 Richard Biener <rguenther@suse.de>
24977 PR tree-optimization/98834
24978 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
24979 subsetting by truncating the access size.
24981 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
24983 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
24984 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
24987 2021-03-15 Martin Liska <mliska@suse.cz>
24989 * spellcheck.c: Add missing comma in initialization.
24991 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
24993 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
24994 alternative 2 and alternative 1 with alternative 3 using
24995 YW register constraint.
24996 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
24997 using YW register constraint.
24998 (*vec_extractv16qi_zext): Ditto.
24999 (*vec_extractv4si): Merge alternatives 4 and 5
25000 using Yw register constraint.
25001 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
25003 2021-03-13 Martin Sebor <msebor@redhat.com>
25005 PR tree-optimization/99489
25006 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
25007 is not a call statement.
25009 2021-03-13 Jakub Jelinek <jakub@redhat.com>
25011 PR tree-optimization/99544
25012 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
25013 if for vector types multiplication can't be done in type's mode.
25015 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
25018 * config/sparc/constraints.md (w): Rename to...
25019 (W): ... this and ditch previous implementation.
25020 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
25021 (*movdf_insn_sp64): Likewise.
25022 (*mov<VM64:mode>_insn_sp64): Likewise.
25023 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
25025 (atomic_compare_and_swap_leon3_1): Likewise.
25026 (*atomic_compare_and_swapdi_v8plus): Likewise.
25027 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
25028 architecture and add missing address validity check during LRA.
25030 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
25033 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
25034 occuring for assumed-size arrays in use_device_{ptr,addr}.
25036 2021-03-12 Jakub Jelinek <jakub@redhat.com>
25039 * config/i386/constraints.md (YW): New internal constraint.
25040 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
25041 (*<sse2_avx2>_<insn><mode>3<mask_name>,
25042 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
25043 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
25045 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
25046 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
25047 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
25048 into one, use Yw instead of former x,v.
25049 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
25050 the last alternative.
25051 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
25052 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
25053 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
25054 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
25055 into one, use <v_Yw> instead of former x,v.
25056 (avx2_interleave_highv32qi<mask_name>,
25057 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
25058 constraints. Add && <mask_avx512bw_condition> to condition.
25059 (avx2_interleave_lowv32qi<mask_name>,
25060 vec_interleave_lowv16qi<mask_name>,
25061 avx2_interleave_highv16hi<mask_name>,
25062 vec_interleave_highv8hi<mask_name>,
25063 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
25064 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
25065 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
25066 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
25067 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
25068 Yw instead of v in constraints.
25069 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
25070 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
25071 instead of Yv in constraints.
25072 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
25073 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
25074 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
25075 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
25076 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
25078 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
25079 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
25080 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
25081 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
25082 two, one with just x, another isa avx512vl with v.
25084 2021-03-12 Martin Liska <mliska@suse.cz>
25086 * doc/invoke.texi: Add missing param documentation.
25088 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25091 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
25092 analyzer/trimmed-graph.o.
25093 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
25094 feasibility checking to reflect new implementation.
25095 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
25097 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
25099 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25101 * digraph.cc (selftest::test_shortest_paths): Update
25102 shortest_paths init for new param. Add test of
25103 SPS_TO_GIVEN_TARGET.
25104 * shortest-paths.h (enum shortest_path_sense): New.
25105 (shortest_paths::shortest_paths): Add "sense" param.
25106 Update for renamings. Generalize to use "sense" param.
25107 (shortest_paths::get_shortest_path): Rename param.
25108 (shortest_paths::m_sense): New field.
25109 (shortest_paths::m_prev): Rename...
25110 (shortest_paths::m_best_edge): ...to this.
25111 (shortest_paths::get_shortest_path): Update for renamings.
25112 Conditionalize flipping of path on sense of traversal.
25114 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25116 * digraph.cc (selftest::test_shortest_paths): Add test coverage
25117 for paths from B and C.
25118 * shortest-paths.h (shortest_paths::shortest_paths): Handle
25119 unreachable nodes, rather than asserting.
25121 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
25124 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
25125 xcoff_tbss_section_name.
25126 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
25127 * xcoffout.c (xcoff_tbss_section_name): Delete.
25128 * xcoffout.h (xcoff_tbss_section_name): Delete.
25130 2021-03-11 Richard Biener <rguenther@suse.de>
25132 PR tree-optimization/99523
25133 * tree-cfg.c (dump_function_to_file): Dump SSA names
25134 w/o identifier to the decls section as well, not only those
25135 without a VAR_DECL.
25137 2021-03-11 Jakub Jelinek <jakub@redhat.com>
25140 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
25141 function calls with lhs fail if the lhs don't have compatible types.
25143 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
25145 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
25146 Change FRAME_POINTER_REGNUM to correspond to a new faked
25147 register faked_fp, part of GENNONACR_REGS like faked_ap.
25148 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
25149 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
25150 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
25151 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
25152 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
25154 (CRIS_REAL_FP_REGNUM): New constant.
25155 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
25156 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
25157 (cris_initial_elimination_offset): Handle elimination changes
25158 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
25159 and add one from FRAME_POINTER_REGNUM to
25160 HARD_FRAME_POINTER_REGNUM.
25161 (cris_expand_prologue, cris_expand_epilogue): Emit code for
25162 hard_frame_pointer_rtx instead of frame_pointer_rtx.
25164 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
25167 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
25168 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
25170 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
25173 * lra-constraints.c (process_address_1): Don't check unknown
25174 constraint, use X for empty constraint.
25176 2021-03-10 Alex Coplan <alex.coplan@arm.com>
25178 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
25179 Fix typo in comment describing "is_ha" argument.
25181 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
25183 * doc/sourcebuild.texi: Document LRA target selector.
25185 2021-03-10 David Malcolm <dmalcolm@redhat.com>
25187 * doc/ux.texi: Add subsection contrasting interactive versus
25188 batch usage of GCC.
25190 2021-03-10 Joel Hutton <joel.hutton@arm.com>
25193 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
25195 (vectorizable_load): Fix gather load mask check condition.
25197 2021-03-10 Richard Biener <rguenther@suse.de>
25199 PR tree-optimization/99510
25200 * tree.c (check_aligned_type): Check that the candidate
25201 has TYPE_USER_ALIGN set instead of matching with the
25204 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
25206 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
25207 float and vector integer modes only if the mode is not larger.
25209 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
25211 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
25213 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25215 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
25217 * ira-lives.c (single_reg_class): Ditto.
25219 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
25221 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
25222 the architecture-specific rtems.h.
25223 (aarch64-*-rtems*): Likewise.
25224 (arm*-*-rtems*): Likewise.
25225 (epiphany-*-rtems*): Likewise.
25226 (riscv*-*-rtems*): Likewise.
25228 2021-03-09 Jakub Jelinek <jakub@redhat.com>
25230 PR tree-optimization/99305
25231 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
25232 before integer_all_onesp instead of vice versa.
25234 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
25236 * common/config/arm/arm-common.c (arm_config_default): Change type
25237 of 'i' to unsigned.
25239 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25242 * lra-constraints.c (process_address_1): Process constraint 'g'
25243 separately and digital constraints containing more one digit.
25245 2021-03-09 Nick Clifton <nickc@redhat.com>
25247 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
25248 (DWARF"_DEBUGGING_INFO): Define.
25250 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
25253 * calls.c (initialize_argument_information): When the argument
25254 is passed by reference, do not make a copy in a thunk only if
25255 the argument is already in memory. Remove redundant test for
25256 the case of callee copy.
25258 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25261 * lra-constraints.c (process_address_1): Process 0..9 constraints
25262 in process_address_1.
25264 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
25266 * config/s390/s390.c (struct s390_processor processor_table):
25267 Binutils name string must not be empty.
25269 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
25271 * config/arc/arc.c (arc_attr_type): Remove function.
25273 2021-03-09 Martin Liska <mliska@suse.cz>
25276 * config/i386/i386-options.c (ix86_option_override_internal):
25277 Set isa_flags for OPTS argument and not for the global
25280 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25282 * config/rs6000/predicates.md (ds_form_mem_operand): Check
25285 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25288 * config/rs6000/predicates.md (ds_form_mem_operand) New
25290 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
25291 ds_form_mem_operand in ld/lwa patterns.
25292 * config/rs6000/fusion.md: Regenerate file.
25294 2021-03-08 Martin Sebor <msebor@redhat.com>
25296 PR middle-end/98266
25297 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
25298 (array_bounds_checker::check_array_bounds): Call it.
25300 2021-03-08 Martin Sebor <msebor@redhat.com>
25302 PR middle-end/97631
25303 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
25304 (handle_builtin_stxncpy_strncat): Rename locals. Determine
25305 destination size from allocation calls. Issue a more appropriate
25307 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
25308 (handle_builtin_memset): Same.
25310 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
25313 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
25314 to ensure we do not have an Altivec style address.
25315 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
25316 an Altivec style address.
25317 (*vsx_le_perm_store_<mode>): Likewise.
25318 (splitters after *vsx_le_perm_store_<mode>): Likewise.
25319 (vsx_load_<mode>): Disable special expander if passed an Altivec
25321 (vsx_store_<mode>): Likewise.
25323 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25326 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
25327 (aarch64_simd_shift_imm_vec_hi): Likewise.
25328 (aarch64_simd_shift_imm_vec_si): Likewise.
25329 (aarch64_simd_shift_imm_vec_di): Likewise.
25330 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
25331 predicate from above.
25332 (aarch64_shrn<mode>_insn_be): Likewise.
25333 (aarch64_rshrn<mode>_insn_le): Likewise.
25334 (aarch64_rshrn<mode>_insn_be): Likewise.
25335 (aarch64_shrn2<mode>_insn_le): Likewise.
25336 (aarch64_shrn2<mode>_insn_be): Likewise.
25337 (aarch64_rshrn2<mode>_insn_le): Likewise.
25338 (aarch64_rshrn2<mode>_insn_be): Likewise.
25340 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
25343 * lra-constraints.c (skip_contraint_modifiers): New function.
25344 (process_address_1): Use it before lookup_constraint call.
25346 2021-03-08 Martin Liska <mliska@suse.cz>
25349 * config/i386/i386-options.c (ix86_option_override_internal):
25350 Enable UINTR and HRESET for -march that supports it.
25352 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
25354 * config/s390/s390.c (f_constraint_p): New function.
25355 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
25356 (TARGET_MD_ASM_ADJUST): Likewise.
25358 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
25361 * tree-nested.c (convert_local_reference_stmt): Avoid calling
25362 lookup_field_for_decl for Fortran module (= namespace context).
25364 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
25366 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
25367 comparison with arithmetic right shift.
25368 (s390_expand_vcond): No need for a force_reg anymore.
25369 s390_vec_compare will do it.
25370 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
25371 immediate operands.
25373 2021-03-07 Jakub Jelinek <jakub@redhat.com>
25376 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
25377 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
25379 * config/i386/sse.md (v_Yw): New define_mode_attr.
25380 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
25381 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
25383 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
25384 xYw in constraints.
25386 2021-03-06 Julian Brown <julian@codesourcery.com>
25388 * tree-pretty-print.c (dump_generic_node): Emit non-generic
25389 address space info for aggregates.
25391 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
25393 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
25395 2021-03-05 Jakub Jelinek <jakub@redhat.com>
25397 PR middle-end/99322
25398 * tree-cfg.c (bb_to_omp_idx): New variable.
25399 (execute_build_cfg): Release the bb_to_omp_idx vector after
25400 cleanup_tree_cfg returns.
25401 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
25402 for bb_to_omp_idx being a vec<int> instead of pointer to array
25404 (make_edges): Remove bb_to_omp_idx local variable, don't pass
25405 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
25406 vec<int> instead of pointer to array of ints and don't free/release
25408 (remove_bb): When removing a bb and placing forced label somewhere
25409 else, ensure it is put into the same OpenMP region during cfg
25410 pass if possible or to entry successor as fallback. Unregister
25411 bb from bb_to_omp_idx.
25413 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
25416 * lra-constraints.c (process_address_1): Skip decomposing address
25417 for asm insn operand with unknown constraint.
25419 2021-03-05 Martin Jambor <mjambor@suse.cz>
25422 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
25423 corresponding speculative edges if we are about to resolve
25424 sepculation. Make edge direct (and so resolve speculations) before
25425 removing it from call_site_hash.
25426 (cgraph_edge::make_direct): Relax the initial assert to allow calling
25427 the function on speculative direct edges.
25429 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
25431 PR rtl-optimization/99376
25432 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
25433 of low-order zero bits is too large, set the result to 0 directly.
25435 2021-03-04 Jakub Jelinek <jakub@redhat.com>
25437 PR middle-end/93235
25438 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
25439 SUBREG and a SUBREG to op_mode can't be created.
25441 2021-03-04 Alex Coplan <alex.coplan@arm.com>
25444 * config/aarch64/aarch64-sve-builtins.cc
25445 (function_resolver::require_vector_type): Handle error_mark_node.
25447 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
25449 * cfgexpand.c (expand_asm_loc): Pass new parameter.
25450 (expand_asm_stmt): Likewise.
25451 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
25453 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25454 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
25455 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
25456 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
25457 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
25458 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
25459 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
25460 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
25461 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
25462 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
25463 * doc/tm.texi (md_asm_adjust): Likewise.
25464 * target.def (md_asm_adjust): Likewise.
25466 2021-03-04 Richard Biener <rguenther@suse.de>
25468 PR middle-end/97855
25469 * tree-pretty-print.c: Poison pp_printf.
25470 (dump_decl_name): Avoid use of pp_printf.
25471 (dump_block_node): Likewise.
25472 (dump_generic_node): Likewise.
25474 2021-03-04 Martin Sebor <msebor@redhat.com>
25476 PR middle-end/96963
25477 PR middle-end/94655
25478 * builtins.c (handle_array_ref): New helper.
25479 (handle_mem_ref): New helper.
25480 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
25481 into new helper functions. Correct a workaround for vectorized
25484 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
25486 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
25487 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
25488 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
25489 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
25490 attribute for Power10.
25491 * config/rs6000/mma.md (*movoo): Likewise.
25492 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
25493 (define_mode_attr bits): Add DD/TD modes.
25494 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
25495 store_conditionalpti): Update size attribute for Power10.
25497 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25500 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
25501 -Wuninitialized, -Wmaybe-uninitialized.
25502 (wide-int.o-warn): Likewise.
25504 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
25506 * common/config/arm/arm-common.c: Include configargs.h.
25507 (arm_config_default): New function.
25508 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
25509 processors that do not support Thumb. Take into account the
25510 --with-mode configuration setting for selecting the default.
25511 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
25512 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
25514 2021-03-03 Martin Liska <mliska@suse.cz>
25516 PR gcov-profile/97461
25517 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
25519 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
25522 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25523 point back the hard frame pointer to its default location when the
25524 frame is larger than SEH_MAX_FRAME_SIZE.
25526 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25529 * config/i386/predicates.md (logic_operator): New define_predicate.
25530 * config/i386/i386.md (mov + mem using comm arith peephole2):
25531 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
25532 and the inner mode is [QH]Imode.
25534 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25537 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
25538 (new_loc_list): Clear end_entry.
25539 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
25540 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
25542 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
25543 initialize also end_entry.
25545 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25548 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
25549 partitions, if in non-layout mode after reorder_blocks also move
25550 affected blocks to ensure a single partition transition.
25552 2021-03-03 Jason Merrill <jason@redhat.com>
25555 * cgraphunit.c (process_function_and_variable_attributes): Don't
25556 warn about flatten on an alias if the target also has it.
25557 * cgraph.h (symtab_node::get_alias_target_tree): New.
25559 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
25561 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
25562 period to symbol name.
25563 (tls_get_addr_internal<mode>): Same.
25565 2021-03-02 David Malcolm <dmalcolm@redhat.com>
25568 * diagnostic-show-locus.c
25569 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
25572 2021-03-02 Martin Sebor <msebor@redhat.com>
25574 PR middle-end/99276
25575 * builtins.c (warn_for_access): Remove stray warning text.
25577 2021-03-02 Martin Sebor <msebor@redhat.com>
25579 PR middle-end/99295
25580 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
25583 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25586 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
25587 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
25588 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
25589 DW_MACRO_define_strx and DW_MACRO_undef_strx.
25590 (save_macinfo_strings): Use DW_MACRO_*_str* even with
25591 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
25592 DW_MACRO_undef_strx.
25594 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25596 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
25598 (BT_FN_V8HI_V8HI_UINT): Likewise.
25599 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
25600 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
25601 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
25602 New builtin definitions.
25603 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
25604 vector extension version.
25605 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
25606 available with current -march level.
25607 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
25608 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
25609 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
25610 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
25611 (vec_extend_to_fp32_lo): Likewise.
25612 (vec_round_from_fp32): Likewise.
25613 (vec_convert_to_fp16): Likewise.
25614 (vec_convert_from_fp16): Likewise.
25615 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
25616 (vclfnls_v8hi): Likewise.
25617 (vcrnfs_v8hi): Likewise.
25618 (vcfn_v8hi): Likewise.
25619 (vcnf_v8hi): Likewise.
25621 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25623 * common/config/s390/s390-common.c (processor_flags_table): New entry.
25624 * config.gcc: Enable arch14 for --with-arch and --with-tune.
25625 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
25626 arch14 for unknown CPU models.
25627 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
25628 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
25629 (s390_get_sched_attrmask): Likewise.
25630 (s390_get_unit_mask): Likewise.
25631 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
25632 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
25633 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
25634 (TARGET_NNPA_P): New macro definitions.
25635 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
25636 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
25638 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25640 PR middle-end/95757
25641 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
25642 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
25643 != 1 comparisons if name is lhs of a comparison.
25645 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
25649 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
25650 * config/darwin.c (darwin_should_restore_cfa_state): New.
25651 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
25652 * doc/tm.texi: Regenerated.
25653 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
25654 * dwarf2cfi.c (connect_traces): If the target requests, restore
25655 the CFA expression after a DW_CFA_restore.
25656 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
25658 2021-03-01 Martin Liska <mliska@suse.cz>
25661 * optc-save-gen.awk: Add 4 more exceptions.
25663 2021-03-01 Nathan Sidwell <nathan@acm.org>
25666 * tree.h (TYPE_ALIGN_RAW): New accessor.
25667 (TYPE_ALIGN): Use it.
25669 2021-03-01 Jan Hubicka <jh@suse.cz>
25672 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
25674 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
25677 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25678 point the hard frame pointer to the SSE register save area instead
25679 of the general register save area. Perform only minimal adjustment
25680 for small frames if it is initially not correctly aligned.
25681 (ix86_expand_prologue): Remove early saves for a SEH target.
25682 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
25684 2021-02-28 Jakub Jelinek <jakub@redhat.com>
25687 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
25688 typo - referneced -> referenced.
25689 * tree.c (component_ref_size): Fix comment typo -
25690 refernce -> reference.
25691 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
25692 traling -> trailing.
25693 (aliasing_component_refs_p): Fix comment typos -
25694 refernce -> reference and refernece -> reference and
25695 traling -> trailing.
25696 (nonoverlapping_refs_since_match_p): Fix comment typo -
25697 referneces -> references.
25698 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
25699 referneces -> references.
25701 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
25703 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
25704 diagnostic message to avoid use of a contraction and format
25707 2021-02-27 Jakub Jelinek <jakub@redhat.com>
25710 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
25711 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
25712 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
25713 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
25716 2021-02-26 Richard Biener <rguenther@suse.de>
25718 PR middle-end/99281
25719 * expr.c (store_field): For calls with return-slot optimization
25720 and addressable return type expand the store directly.
25722 2021-02-26 Richard Biener <rguenther@suse.de>
25725 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
25727 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
25730 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
25733 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
25735 * config.gcc: Add rs6000-pcrel-opt.o.
25736 * config/rs6000/rs6000-pcrel-opt.c: New file.
25737 * config/rs6000/pcrel-opt.md: New file.
25738 * config/rs6000/predicates.md: Add d_form_memory predicate.
25739 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
25740 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
25741 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
25742 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
25743 and make_pass_pcrel_opt().
25744 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
25745 (rs6000_option_override_internal): Add pcrel-opt.
25746 (rs6000_delegitimize_address): Support pcrel-opt.
25747 (rs6000_opt_masks): Add pcrel-opt.
25748 (pcrel_opt_valid_mem_p): New function.
25749 (reg_to_non_prefixed): Make global.
25750 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
25751 (output_pcrel_opt_reloc): New function.
25752 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
25753 (pcrel_extern_addr): Set loads_extern_addr.
25754 Add include for pcrel-opt.md.
25755 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
25756 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
25759 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
25762 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
25763 If TARGET_64BIT and dest is SUBREG, we check the width, if it
25764 equal to SImode, we use SImode operation, just like what we are
25767 2021-02-26 Marek Polacek <polacek@redhat.com>
25769 * builtins.c (warn_for_access): Fix typos.
25771 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
25773 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
25774 mark in front of the immediate quantity.
25775 (<optab>_rolsi3_uxtw): Likewise.
25777 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
25780 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
25781 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
25782 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
25783 address and disable when the FPCXT is not available.
25784 (nonsecure_call_value_reg_thumb2): Likewise.
25786 2021-02-25 Nathan Sidwell <nathan@acm.org>
25789 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
25791 2021-02-25 Tamar Christina <tamar.christina@arm.com>
25793 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
25795 2021-02-25 Richard Biener <rguenther@suse.de>
25797 PR tree-optimization/99253
25798 * tree-vect-loop.c (check_reduction_path): First compute
25799 code, then verify out-of-loop uses.
25801 2021-02-25 Jakub Jelinek <jakub@redhat.com>
25804 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
25806 2021-02-25 Jakub Jelinek <jakub@redhat.com>
25808 PR tree-optimization/80635
25809 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
25810 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
25811 has mode precision.
25813 2021-02-25 Richard Biener <rguenther@suse.de>
25815 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
25816 load_map population.
25817 (vect_match_slp_patterns_2): Revert part of last change.
25818 (vect_analyze_slp): Do not interleave optimize_load_redistribution
25819 with pattern detection but do it afterwards. Dump the
25820 whole SLP graph after pattern recognition and load
25821 redistribution optimization finished.
25823 2021-02-24 Jakub Jelinek <jakub@redhat.com>
25826 * omp-low.c (struct omp_context): Add teams_nested_p and
25827 nonteams_nested_p members.
25828 (scan_omp_target): Diagnose teams nested inside of target with other
25829 directives strictly nested inside of the same target.
25830 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
25831 ctx->nonteams_nested_p as needed.
25833 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
25835 PR inline-asm/99123
25836 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
25838 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
25840 * config/cris/cris.c (cris_expand_prologue): Set
25841 current_function_static_stack_size, if flag_stack_usage_info.
25843 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
25845 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
25846 (rs6000_final_prescan_insn): Adjust.
25847 (rs6000_asm_output_opcode): Likewise.
25849 2021-02-24 Martin Sebor <msebor@redhat.com>
25851 PR middle-end/97172
25852 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
25853 from function arguments.
25855 2021-02-24 Tamar Christina <tamar.christina@arm.com>
25857 PR tree-optimization/99220
25858 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
25859 node from cache when it's about to be deleted.
25861 2021-02-24 Jakub Jelinek <jakub@redhat.com>
25863 PR tree-optimization/99225
25864 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
25865 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
25866 build_int_cst (..., 1). Formatting fixes.
25868 2021-02-24 Tamar Christina <tamar.christina@arm.com>
25870 PR tree-optimization/99149
25871 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
25873 (vect_slp_reset_pattern): Remove.
25874 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
25875 (complex_mul_pattern::build, complex_fma_pattern::build,
25876 complex_fms_pattern::build): Fix ref counts.
25877 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
25878 when node is being deleted.
25879 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
25880 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
25882 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
25884 2021-02-24 Matthias Klose <doko@ubuntu.com>
25887 2020-12-07 Matthias Klose <doko@ubuntu.com>
25889 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
25890 and ENABLE_RTL_FLAG_CHECKING.
25892 2021-02-24 Richard Biener <rguenther@suse.de>
25895 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
25897 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
25899 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
25900 (vsx_assemble_pair): ...to this.
25901 (*mma_assemble_pair): Rename from this...
25902 (*vsx_assemble_pair): ...to this.
25903 (mma_disassemble_pair): Rename from this...
25904 (vsx_disassemble_pair): ...to this.
25905 (*mma_disassemble_pair): Rename from this...
25906 (*vsx_disassemble_pair): ...to this.
25907 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
25908 BU_COMPAT): New macros.
25909 (mma_assemble_pair): Rename from this...
25910 (vsx_assemble_pair): ...to this.
25911 (mma_disassemble_pair): Rename from this...
25912 (vsx_disassemble_pair): ...to this.
25913 (mma_assemble_pair): New compatibility built-in.
25914 (mma_disassemble_pair): Likewise.
25915 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
25916 (RS6000_BUILTIN_COMPAT): Define.
25917 (bdesc_compat): New.
25918 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
25919 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
25920 and VSX_BUILTIN_ASSEMBLE_PAIR.
25921 (rs6000_init_builtins): Register compatibility built-ins.
25922 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
25923 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
25924 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
25925 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
25926 (__builtin_vsx_assemble_pair): ...to this.
25927 (__builtin_mma_disassemble_pair): Rename from this...
25928 (__builtin_vsx_disassemble_pair): ...to this.
25930 2021-02-23 Martin Liska <mliska@suse.cz>
25933 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
25934 with different alignment. That leads to an invalid red zone
25935 size allocated in runtime.
25937 2021-02-23 Jakub Jelinek <jakub@redhat.com>
25939 PR tree-optimization/99204
25940 * fold-const.c (fold_read_from_constant_string): Check that
25941 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
25943 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
25944 Kewen Lin <linkw@gcc.gnu.org>
25946 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
25947 (rotl<mode>3_insert_3): ...this.
25948 (plus_ior_xor): New code_iterator.
25949 (define_split for GPR rl*imi): New splitter.
25950 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
25951 for integer merging.
25953 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25955 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
25957 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
25958 into a register when the above is enabled.
25959 * config/aarch64/aarch64.c (neoversev1_tunings):
25960 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
25961 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
25963 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
25965 * config/cris/cris.c (cris_print_operand) <'T'>: Change
25966 valid operand from is now an addi mult-value to shift-value.
25967 * config/cris/cris.md (*addi): Change expression of scaled
25968 operand from mult to ashift.
25969 * config/cris/cris.md (*addi_reload): New insn_and_split.
25971 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
25974 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
25975 hook_bool_const_tree_hwi_hwi_const_tree_true.
25976 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
25978 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
25980 PR rtl-optimization/98791
25981 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
25982 for unordered modes.
25984 2021-02-22 Martin Liska <mliska@suse.cz>
25986 * tree-inline.c (inline_forbidden_p): Set
25987 inline_forbidden_reason.
25989 2021-02-22 Richard Biener <rguenther@suse.de>
25991 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
25994 2021-02-22 Richard Biener <rguenther@suse.de>
25996 PR tree-optimization/99165
25997 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
25998 Accumulate changed to ret.
26000 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
26003 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
26005 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
26007 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
26010 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
26012 (trunctf<DFP_ALL:mode>2): Likewise.
26013 (trunctdtf2_vr): Likewise.
26014 (trunctdtf2): Likewise.
26015 (extend<DFP_ALL:mode>tf2_vr): Likewise.
26016 (extend<DFP_ALL:mode>tf2): Likewise.
26017 (extendtftd2_vr): Likewise.
26018 (extendtftd2): Likewise.
26020 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
26022 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
26023 add memory alternative.
26024 (tf_to_fprx2): New pattern.
26026 2021-02-19 Martin Sebor <msebor@redhat.com>
26029 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
26030 (attr_access::free_lang_data): Remove a spurious test.
26032 2021-02-19 Nathan Sidwell <nathan@acm.org>
26034 * doc/invoke.texi (flang-info-module-read): Document.
26036 2021-02-19 Martin Liska <mliska@suse.cz>
26038 PR translation/99167
26039 * params.opt: Fix typo.
26041 2021-02-19 Richard Biener <rguenther@suse.de>
26043 PR middle-end/99122
26044 * tree-inline.c (inline_forbidden_p): Do not inline functions
26045 with VLA arguments or return value.
26047 2021-02-19 Jakub Jelinek <jakub@redhat.com>
26050 * config/arm/arm.md (*stack_protect_combined_set_insn,
26051 *stack_protect_combined_test_insn): If force_const_mem result
26052 is not valid general operand, force its address into the destination
26055 2021-02-19 Jakub Jelinek <jakub@redhat.com>
26058 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
26059 pad or non-local label, put FORCED_LABELs from bb b after that label
26060 rather than before it.
26062 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
26065 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
26066 expand_vector_broadcast' to emit the vec_duplicate operand.
26068 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
26070 PR rtl-optimization/96264
26071 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
26074 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
26077 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
26078 looking up the retain attribute.
26079 (resolve_unique_section): Likewise.
26080 (get_variable_section): Likewise.
26081 (switch_to_section): Likewise. Warn when a symbol without the
26082 retain attribute and a symbol with the retain attribute are
26083 placed in the section with the same name, instead of the used
26085 * doc/extend.texi: Document the "retain" attribute.
26087 2021-02-18 Nathan Sidwell <nathan@acm.org>
26090 * doc/invoke.texi (flang-info-include-translate): Document header
26093 2021-02-18 Richard Biener <rguenther@suse.de>
26095 PR middle-end/99122
26096 * ipa-fnsummary.c (analyze_function_body): Set
26097 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
26098 * tree-inline.c (insert_init_debug_bind): Pass NULL for
26099 error_mark_node values.
26100 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
26102 (setup_one_parameter): Delay force_value_to_type until when
26105 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
26107 PR tree-optimization/99142
26108 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
26110 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26112 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
26113 wide_int_bitmask::wide_int_bitmask (uint64_t),
26114 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
26115 wide_int_bitmask::operator ~ () const,
26116 wide_int_bitmask::operator | (wide_int_bitmask) const,
26117 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
26119 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
26120 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
26121 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
26122 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
26123 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
26124 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
26125 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
26126 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
26127 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
26128 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
26129 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
26130 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
26131 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
26132 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
26133 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
26134 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
26135 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
26136 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
26137 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
26138 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
26139 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
26140 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
26141 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
26142 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
26144 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26146 PR middle-end/99109
26147 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
26148 (build_printable_array_type): ... this. Add nelts argument. For
26149 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
26150 nelts, call build_array_type_nelts.
26151 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
26152 instead of build_zero_elt_array_type and build_array_type_nelts.
26154 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26157 * config/i386/i386.c (distance_non_agu_define): Don't call
26158 extract_insn_cached here.
26159 (ix86_lea_outperforms): Save and restore recog_data around call
26160 to distance_non_agu_define and distance_agu_use.
26161 (ix86_ok_to_clobber_flags): Remove.
26162 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
26163 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
26164 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
26165 into define_insn. Move the splitting to define_peephole2 and
26166 check there using peep2_regno_dead_p if FLAGS_REG is dead.
26168 2021-02-17 Julian Brown <julian@codesourcery.com>
26170 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
26173 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
26176 * config/mips/mips.c (mips_symbol_insns): Do not use
26177 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
26179 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
26181 PR inline-asm/98096
26182 * stmt.c (resolve_operand_name_1): Take inout operands into account
26183 for access to labels by names.
26184 * doc/extend.texi: Describe counting operands for accessing labels.
26186 2021-02-16 Richard Biener <rguenther@suse.de>
26188 PR tree-optimization/38474
26189 * tree-ssa-structalias.c (variable_info::address_taken): New.
26190 (new_var_info): Initialize address_taken.
26191 (process_constraint): Set address_taken.
26192 (solve_constraints): Use the new address_taken flag rather
26193 than is_reg_var for sorting variables.
26194 (dump_constraint): Dump the variable number if the name
26197 2021-02-16 Jakub Jelinek <jakub@redhat.com>
26200 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
26201 multiply by 4096 and for inbranch by 8192.
26202 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
26203 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
26205 2021-02-15 Maya Rashish <coypu@sdf.org>
26207 * config/aarch64/aarch64.c (aarch64_init_builtins):
26208 Call SUBTARGET_INIT_BUILTINS.
26210 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
26212 PR rtl-optimization/98872
26213 * init-regs.c (initialize_uninitialized_regs): Skip initialization
26214 if CONST0_RTX is NULL.
26216 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
26218 PR rtl-optimization/98863
26219 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
26220 (function_info::build_info): Turn into a declaration, moving the
26221 definition to internals.h.
26222 (function_info::bb_walker): Declare.
26223 (function_info::create_reg_use): Likewise.
26224 (function_info::calculate_potential_phi_regs): Take a build_info
26226 (function_info::place_phis, function_info::create_ebbs): Declare.
26227 (function_info::calculate_ebb_live_in_for_debug): Likewise.
26228 (function_info::populate_backedge_phis): Delete.
26229 (function_info::start_block, function_info::end_block): Declare.
26230 (function_info::populate_phi_inputs): Delete.
26231 (function_info::m_potential_phi_regs): Move information to build_info.
26232 * rtl-ssa/internals.h: New file.
26233 (function_info::bb_phi_info): New class.
26234 (function_info::build_info): Moved from functions.h.
26235 Add a constructor and destructor.
26236 (function_info::build_info::ebb_use): Delete.
26237 (function_info::build_info::ebb_def): Likewise.
26238 (function_info::build_info::bb_live_out): Likewise.
26239 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
26240 (function_info::build_info::potential_phi_regs): Likewise.
26241 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
26242 (function_info::build_info::ebb_def_regs): Likewise.
26243 (function_info::build_info::bb_phis): Likewise.
26244 (function_info::build_info::bb_mem_live_out): Likewise.
26245 (function_info::build_info::bb_to_rpo): Likewise.
26246 (function_info::build_info::def_stack): Likewise.
26247 (function_info::build_info::old_def_stack_limit): Likewise.
26248 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
26249 Remove the regno argument. Push the previous definition onto the
26250 definition stack where necessary.
26251 * rtl-ssa/accesses.cc: Include internals.h.
26252 * rtl-ssa/changes.cc: Likewise.
26253 * rtl-ssa/blocks.cc: Likewise.
26254 (function_info::build_info::build_info): Define.
26255 (function_info::build_info::~build_info): Likewise.
26256 (function_info::bb_walker): New class.
26257 (function_info::bb_walker::bb_walker): Define.
26258 (function_info::add_live_out_use): Convert a logarithmic-complexity
26259 test into a linear one. Allow the same definition to be passed
26261 (function_info::calculate_potential_phi_regs): Moved from
26262 functions.cc. Take a build_info parameter and store the
26263 information there instead.
26264 (function_info::place_phis): New function.
26265 (function_info::add_entry_block_defs): Update call to record_reg_def.
26266 (function_info::calculate_ebb_live_in_for_debug): New function.
26267 (function_info::add_phi_nodes): Use bb_phis to decide which
26268 registers need phi nodes and initialize ebb_def_regs accordingly.
26269 Do not add degenerate phis here.
26270 (function_info::add_artificial_accesses): Use create_reg_use.
26271 Assert that all definitions are listed in the DF LR sets.
26272 Update call to record_reg_def.
26273 (function_info::record_block_live_out): Record live-out register
26274 values in the phis of successor blocks. Use the live-out set
26275 when processing the last block in an EBB, instead of always
26276 using the live-in sets of successor blocks. AND the live sets
26277 with the set of registers that have been defined in the EBB,
26278 rather than with all potential phi registers. Cope correctly
26279 with branches back to the start of the current EBB.
26280 (function_info::start_block): New function.
26281 (function_info::end_block): Likewise.
26282 (function_info::populate_phi_inputs): Likewise.
26283 (function_info::create_ebbs): Likewise.
26284 (function_info::process_all_blocks): Rewrite into a multi-phase
26286 * rtl-ssa/functions.cc: Include internals.h.
26287 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
26288 (function_info::init_function_data): Remove caller.
26289 * rtl-ssa/insns.cc: Include internals.h
26290 (function_info::create_reg_use): New function. Lazily any
26291 degenerate phis needed by the linear RPO view.
26292 (function_info::record_use): Use create_reg_use. When processing
26293 debug uses, use potential_phi_regs and test it before checking
26294 whether the register is live on entry to the current EBB. Lazily
26295 calculate ebb_live_in_for_debug.
26296 (function_info::record_call_clobbers): Update call to record_reg_def.
26297 (function_info::record_def): Likewise.
26299 2021-02-15 Martin Liska <mliska@suse.cz>
26301 * toplev.c (init_asm_output): Free output of
26302 gen_command_line_string function.
26303 (process_options): Likewise.
26305 2021-02-15 Martin Liska <mliska@suse.cz>
26307 * params.opt: Add 2 missing Param keywords.
26309 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
26311 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
26313 2021-02-15 Jakub Jelinek <jakub@redhat.com>
26315 PR tree-optimization/99079
26316 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
26317 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
26318 require both type and TREE_TYPE (@1) to be integral types and either
26319 type having smaller or equal precision, or TREE_TYPE (@1) being
26320 unsigned type, or type being signed type. If TREE_TYPE (@1)
26321 doesn't have wrapping overflow, perform the subtraction of one in
26324 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
26325 Richard Biener <rguether@suse.de>
26328 * ipa-reference.c (ipa_init): Only conditinally initialize
26329 reference_vars_to_consider.
26330 (propagate): Conditionally deninitialize reference_vars_to_consider.
26331 (ipa_reference_write_optimization_summary): Sanity check that
26332 reference_vars_to_consider is not allocated.
26334 2021-02-13 Levy Hsu <admin@levyhsu.com>
26337 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
26338 extend parameter to get_si_mem_base_reg declaration.
26339 (get_si_mem_base_reg): Add extend parameter. Set it.
26340 (analyze): Pass extend arg to get_si_mem_base_reg.
26341 (transform): Likewise. Use it when rewriting mems.
26342 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
26343 loads and emit sign/zero extending load followed by subreg move.
26345 2021-02-13 Jim Wilson <jimw@sifive.com>
26348 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
26349 exit when !reload_completed. Only perform check for compressed reg
26350 if reload_completed.
26351 (riscv_rtx_costs): In MEM case, when optimizing for size and
26352 shorten memrefs, if not compressible, then increase cost.
26354 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26356 PR rtl-optimization/98439
26357 * recog.c (pass_split_before_regstack::gate): Enable even when
26358 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
26361 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26364 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
26365 swap of V2SImode elements in memory into DImode memory rotate by 32.
26367 2021-02-12 Martin Sebor <msebor@redhat.com>
26369 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
26371 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26373 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
26374 m_temp_obstack rather than m_obstack to allocate the temporary use.
26376 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26378 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
26379 as read-modify operations.
26381 2021-02-12 Richard Biener <rguenther@suse.de>
26383 PR middle-end/38474
26384 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
26385 fbi->aa_walk_budget is bigger than zero. Update
26386 fbi->aa_walk_budget.
26387 (param_change_prob): Likewise.
26388 * ipa-prop.c (detect_type_change_from_memory_writes):
26389 Properly account walk_aliased_vdefs.
26390 (parm_preserved_before_stmt_p): Canonicalize updates.
26391 (parm_ref_data_preserved_p): Likewise.
26392 (parm_ref_data_pass_through_p): Likewise.
26393 (determine_known_aggregate_parts): Account own alias queries.
26395 2021-02-12 Martin Liska <mliska@suse.cz>
26397 * opts-common.c (decode_cmdline_option): Release werror_arg.
26398 * opts.c (gen_producer_string): Release output of
26399 gen_command_line_string.
26401 2021-02-12 Richard Biener <rguenther@suse.de>
26403 PR tree-optimization/38474
26404 * params.opt (-param=max-store-chains-to-track=): New param.
26405 (-param=max-stores-to-track=): Likewise.
26406 * doc/invoke.texi (max-store-chains-to-track): Document.
26407 (max-stores-to-track): Likewise.
26408 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
26410 (pass_store_merging::m_n_stores): Likewise.
26411 (pass_store_merging::terminate_and_process_chain): Update
26412 m_n_stores and m_n_chains.
26413 (pass_store_merging::process_store): Likewise. Terminate
26414 oldest chains if the number of stores or chains get too large.
26415 (imm_store_chain_info::terminate_and_process_chain): Dump
26418 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
26420 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
26421 the cold section, emit a nop before the directive if the previous
26422 active instruction can throw.
26424 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
26427 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
26428 memory addresses that are legal for quad word accesses.
26430 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
26433 * config/arm/thumb2.md (*doloop_end_internal): Generate
26434 alternative sequence to handle long range branches.
26436 2021-02-11 Joel Hutton <joel.hutton@arm.com>
26438 PR tree-optimization/98772
26439 * optabs-tree.c (supportable_half_widening_operation): New function
26440 to check for supportable V8QI->V8HI widening patterns.
26441 * optabs-tree.h (supportable_half_widening_operation): New function.
26442 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
26443 to create promotion stmts for V8QI->V8HI widening patterns.
26444 (vectorizable_conversion): Add case for V8QI->V8HI.
26446 2021-02-11 Richard Biener <rguenther@suse.de>
26448 * sparseset.h (SPARSESET_ELT_BITS): Remove.
26449 (SPARSESET_ELT_TYPE): Use unsigned int.
26450 * fwprop.c: Do not include sparseset.h.
26452 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26455 * varasm.c (declare_weak): For -fsyntax-only, allow even
26456 TREE_ASM_WRITTEN function decls.
26458 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26461 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
26462 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
26463 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
26464 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
26465 calling simplify_gen_subreg on it.
26467 2021-02-10 Martin Liska <mliska@suse.cz>
26469 * config/nvptx/nvptx.c (nvptx_option_override): Use
26470 flag_patchable_function_entry instead of the removed
26471 function_entry_patch_area_size.
26473 2021-02-10 Martin Liska <mliska@suse.cz>
26475 PR tree-optimization/99002
26476 PR tree-optimization/99026
26477 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
26478 leak when adjacent cases are merged.
26479 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
26481 (make_pass_lower_switch): Remove trailing whitespace.
26482 * tree-switch-conversion.h (release_clusters): New.
26484 2021-02-10 Richard Biener <rguenther@suse.de>
26486 PR rtl-optimization/99054
26487 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
26488 (fixup_partitions): Adjust.
26489 (rtl_verify_edges): Likewise.
26491 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26493 PR middle-end/99007
26494 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
26495 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
26498 2021-02-10 Richard Biener <rguenther@suse.de>
26501 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
26504 2021-02-10 Richard Biener <rguenther@suse.de>
26506 PR tree-optimization/99024
26507 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
26508 clear loop->aux if it is associated with the destroyed loop_vinfo.
26510 2021-02-10 Martin Liska <mliska@suse.cz>
26512 PR tree-optimization/99002
26513 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
26516 2021-02-10 Martin Liska <mliska@suse.cz>
26519 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
26520 a reference exists.
26522 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26525 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
26526 at class scope for DWARF5+.
26528 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
26530 PR rtl-optimization/96015
26531 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
26532 (relax_delay_slots): When deleting a jump to the next active
26533 instruction over a barrier, first delete the barrier if the
26534 jump is the only way to reach the target label.
26536 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
26538 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
26539 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
26540 vector multiplies and vect.alu for SSRA.
26541 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
26542 vect.mul cost field.
26543 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
26544 * config/arm/arm.c: Likewise.
26546 2021-02-09 Richard Biener <rguenther@suse.de>
26548 PR tree-optimization/98863
26549 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
26550 * tree-ssa-sccvn.c (last_pushed_avail): New global.
26551 (rpo_elim::eliminate_push_avail): Chain pushed avails.
26552 (unwind_state::avail_top): Add.
26553 (do_unwind): Rewrite unwinding of avail entries.
26554 (do_rpo_vn): Initialize last_pushed_avail and
26555 avail_top of the undo state.
26557 2021-02-09 Jakub Jelinek <jakub@redhat.com>
26559 PR middle-end/99004
26560 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
26561 const char * to char * and free those pointers after use.
26563 2021-02-09 Richard Biener <rguenther@suse.de>
26565 PR tree-optimization/99017
26566 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
26567 zero vector cost entries.
26569 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
26571 PR middle-end/98974
26572 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
26573 parameter in vectorizable_condition.
26575 2021-02-08 Richard Biener <rguenther@suse.de>
26578 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
26580 2021-02-08 Martin Liska <mliska@suse.cz>
26583 * cfgexpand.c (pass_expand::execute): Parse per-function option
26584 flag_patchable_function_entry and use it.
26585 * common.opt: Remove function_entry_patch_area_size and
26586 function_entry_patch_area_start global variables.
26587 * opts.c (parse_and_check_patch_area): New function.
26588 (common_handle_option): Use it.
26589 * opts.h (parse_and_check_patch_area): New function.
26590 * toplev.c (process_options): Parse and use
26591 function_entry_patch_area_size.
26593 2021-02-08 Martin Sebor <msebor@redhat.com>
26595 * doc/extend.texi (attribute malloc): Correct typos.
26597 2021-02-05 Nathan Sidwell <nathan@acm.org>
26600 * gcc.c (driver::maybe_run_linker): Check for input file
26601 accessibility if not linking.
26603 2021-02-05 Richard Biener <rguenther@suse.de>
26605 PR tree-optimization/98855
26606 * tree-vectorizer.h (add_stmt_cost): New overload.
26607 * tree-vect-slp.c (li_cost_vec_cmp): New.
26608 (vect_bb_slp_scalar_cost): Cost individual loop regions
26609 separately. Account for the scalar instance root stmt.
26611 2021-02-05 Tom de Vries <tdevries@suse.de>
26614 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
26616 (bit_test_cluster::emit): Reuse location_t for newly created
26618 (switch_decision_tree::try_switch_expansion): Preserve
26620 * tree-switch-conversion.h: Change function signatures.
26622 2021-02-05 Jakub Jelinek <jakub@redhat.com>
26625 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
26626 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
26627 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
26628 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
26630 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26632 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
26633 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
26634 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
26635 (vget_high_f16): Reimplement using new builtin.
26636 (vget_high_f32): Likewise.
26637 (vget_high_f64): Likewise.
26638 (vget_high_p8): Likewise.
26639 (vget_high_p16): Likewise.
26640 (vget_high_p64): Likewise.
26641 (vget_high_s8): Likewise.
26642 (vget_high_s16): Likewise.
26643 (vget_high_s32): Likewise.
26644 (vget_high_s64): Likewise.
26645 (vget_high_u8): Likewise.
26646 (vget_high_u16): Likewise.
26647 (vget_high_u32): Likewise.
26648 (vget_high_u64): Likewise.
26650 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26652 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
26653 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
26654 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
26655 (vget_low_f16): Reimplement using new builtin.
26656 (vget_low_f32): Likewise.
26657 (vget_low_f64): Likewise.
26658 (vget_low_p8): Likewise.
26659 (vget_low_p16): Likewise.
26660 (vget_low_p64): Likewise.
26661 (vget_low_s8): Likewise.
26662 (vget_low_s16): Likewise.
26663 (vget_low_s32): Likewise.
26664 (vget_low_s64): Likewise.
26665 (vget_low_u8): Likewise.
26666 (vget_low_u16): Likewise.
26667 (vget_low_u32): Likewise.
26668 (vget_low_u64): Likewise.
26670 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
26672 * gcc.c (print_multilib_info): Check all required argument is provided
26675 2021-02-05 liuhongt <hongtao.liu@intel.com>
26678 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
26679 generate integer mask comparison for 128/256-bits vector when
26680 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
26681 delete redundant !maskcmp condition.
26682 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
26684 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
26685 condition directly to if (maskcmp), add extra check for
26686 cmpmode, it should be MODE_INT.
26687 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
26688 parameters op_true/op_false.
26689 (ix86_use_mask_cmp_p): New.
26691 2021-02-05 liuhongt <hongtao.liu@intel.com>
26694 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
26695 Remove m_GENERIC from ~list.
26696 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
26698 2021-02-04 David Malcolm <dmalcolm@redhat.com>
26701 * diagnostic-show-locus.c (compatible_locations_p): Require
26702 locations in the same macro map to be either both from the
26703 macro definition, or both from the macro arguments.
26705 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
26707 * config/aarch64/aarch64-simd-builtins.def: Add
26708 [su]mull_hi_lane[q] builtin generator macros.
26709 * config/aarch64/aarch64-simd.md
26710 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
26711 (aarch64_<su>mull_hi_lane<mode>): Define.
26712 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
26713 (aarch64_<su>mull_hi_laneq<mode>): Define.
26714 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
26715 builtin instead of inline asm.
26716 (vmull_high_lane_s32): Likewise.
26717 (vmull_high_lane_u16): Likewise.
26718 (vmull_high_lane_u32): Likewise.
26719 (vmull_high_laneq_s16): Likewise.
26720 (vmull_high_laneq_s32): Likewise.
26721 (vmull_high_laneq_u16): Likewise.
26722 (vmull_high_laneq_u32): Liekwise.
26724 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
26726 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
26727 builtin generator macros.
26728 * config/aarch64/aarch64-simd.md
26729 (aarch64_<su>mull_hi_n<mode>_insn): Define.
26730 (aarch64_<su>mull_hi_n<mode>): Define.
26731 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
26732 instead of inline asm.
26733 (vmull_high_n_s32): Likewise.
26734 (vmull_high_n_u16): Likewise.
26735 (vmull_high_n_u32): Likewise.
26737 2021-02-04 Richard Biener <rguenther@suse.de>
26739 PR tree-optimization/98855
26740 * tree-vect-loop.c (vectorizable_phi): Do not cost
26741 single-argument PHIs.
26742 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
26743 * tree-vect-stmts.c (vectorizable_bswap): Also perform
26744 costing for SLP operation.
26746 2021-02-04 Martin Liska <mliska@suse.cz>
26748 * doc/extend.texi: Mention -mprefer-vector-width in target
26751 2021-02-03 Martin Sebor <msebor@redhat.com>
26753 PR tree-optimization/98937
26754 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
26755 Flush pointer_query cache.
26757 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
26759 * config/rs6000/genfusion.pl (gen_2logical): Add missing
26760 fixes based on patch review.
26761 * config/rs6000/fusion.md: Regenerate file.
26763 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
26765 * config/rs6000/t-rs6000: Comment out auto generation of
26768 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
26770 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
26771 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
26772 (output_file_start): Add gfx908.
26773 * config/gcn/gcn.opt (gpu_type): Add gfx908.
26774 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
26775 (MULTILIB_DIRNAMES): Add gfx908.
26776 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
26777 (main): Recognize gfx908.
26778 * config/gcn/t-omp-device: Add gfx908.
26780 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26782 * config/aarch64/aarch64-simd-builtins.def: Add
26783 [su]mlsl_hi_lane[q] builtin macro generators.
26784 * config/aarch64/aarch64-simd.md
26785 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
26786 (aarch64_<su>mlsl_hi_lane<mode>): Define.
26787 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
26788 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
26789 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
26790 builtin instead of inline asm.
26791 (vmlsl_high_lane_s32): Likewise.
26792 (vmlsl_high_lane_u16): Likewise.
26793 (vmlsl_high_lane_u32): Likewise.
26794 (vmlsl_high_laneq_s16): Likewise.
26795 (vmlsl_high_laneq_s32): Likewise.
26796 (vmlsl_high_laneq_u16): Likewise.
26797 (vmlsl_high_laneq_u32): Likewise.
26798 (vmlal_high_laneq_u32): Likewise.
26800 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26802 * config/aarch64/aarch64-simd-builtins.def: Add
26803 [su]mlal_hi_lane[q] builtin generator macros.
26804 * config/aarch64/aarch64-simd.md
26805 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
26806 (aarch64_<su>mlal_hi_lane<mode>): Define.
26807 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
26808 (aarch64_<su>mlal_hi_laneq<mode>): Define.
26809 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
26810 builtin instead of inline asm.
26811 (vmlal_high_lane_s32): Likewise.
26812 (vmlal_high_lane_u16): Likewise.
26813 (vmlal_high_lane_u32): Likewise.
26814 (vmlal_high_laneq_s16): Likewise.
26815 (vmlal_high_laneq_s32): Likewise.
26816 (vmlal_high_laneq_u16): Likewise.
26817 (vmlal_high_laneq_u32): Likewise.
26819 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26821 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
26822 builtin generator macros.
26823 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
26825 (aarch64_<su>mlsl_hi_n<mode>): Define.
26826 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
26827 instead of inline asm.
26828 (vmlsl_high_n_s32): Likewise.
26829 (vmlsl_high_n_u16): Likewise.
26830 (vmlsl_high_n_u32): Likewise.
26832 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26834 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
26835 builtin generator macros.
26836 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
26838 (aarch64_<su>mlal_hi_n<mode>): Define.
26839 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
26840 instead of inline asm.
26841 (vmlal_high_n_s32): Likewise.
26842 (vmlal_high_n_u16): Likewise.
26843 (vmlal_high_n_u32): Likewise.
26845 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26847 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
26849 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
26851 (aarch64_<su>mlal_hi<mode>_insn): This.
26852 (aarch64_<su>mlal_hi<mode>): Define.
26853 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
26854 instead of inline asm.
26855 (vmlal_high_s16): Likewise.
26856 (vmlal_high_s32): Likewise.
26857 (vmlal_high_u8): Likewise.
26858 (vmlal_high_u16): Likewise.
26859 (vmlal_high_u32): Likewise.
26861 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
26863 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
26864 after calling alter_subreg() on a (mem).
26866 2021-02-03 Martin Liska <mliska@suse.cz>
26869 * lto-streamer-out.c (produce_lto_section): Fill up missing
26871 * lto-streamer.h (struct lto_section): Add _padding field.
26873 2021-02-03 Richard Biener <rguenther@suse.de>
26875 * lto-streamer.c (lto_get_section_name): Free temporary
26877 * tree-loop-distribution.c
26878 (loop_distribution::merge_dep_scc_partitions): Free edge data.
26880 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26882 PR middle-end/97487
26883 * ifcvt.c (noce_can_force_operand): New function.
26884 (noce_emit_move_insn): Use it.
26885 (noce_try_sign_mask): Likewise. Formatting fix.
26887 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26889 PR middle-end/97971
26890 * lra-constraints.c (process_alt_operands): For inline asm, don't call
26891 fatal_insn, but instead return false.
26893 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26895 PR tree-optimization/98287
26896 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
26899 2021-02-03 Tamar Christina <tamar.christina@arm.com>
26901 PR tree-optimization/98928
26902 * tree-vect-loop.c (vect_analyze_loop_2): Change
26903 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
26904 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
26905 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
26906 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
26908 2021-02-02 Richard Biener <rguenther@suse.de>
26910 * gimple-loop-interchange.cc (prepare_data_references):
26912 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
26913 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
26914 * tree-vect-stmts.c (vectorizable_condition): Do not
26916 (vectorizable_comparison): Likewise.
26918 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26920 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
26921 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
26922 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
26923 (vrsqrteq_u32): Likewise.
26925 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26927 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
26928 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
26929 (aarch64_sqxtun2<mode>_be): Likewise.
26930 (aarch64_sqxtun2<mode>): Likewise.
26931 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
26932 (vqmovun_high_s32): Likewise.
26933 (vqmovun_high_s64): Likewise.
26934 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
26936 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26938 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
26940 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
26942 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26944 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
26945 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
26946 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
26947 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
26948 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
26949 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
26950 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
26951 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
26953 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26955 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
26956 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
26957 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
26959 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26961 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
26962 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
26964 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26966 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
26967 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
26968 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
26969 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
26970 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
26971 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
26972 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
26973 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
26974 NONE builtin flags.
26976 2021-02-02 Jakub Jelinek <jakub@redhat.com>
26978 PR tree-optimization/98848
26979 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
26980 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
26982 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
26985 * expr.c: Check mode before calling store_expr.
26987 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
26989 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
26991 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
26992 instruction using expression ior.
26993 (mve_vornq_u<mode>): New expander.
26994 (mve_vornq_f<mode>): Use ior code instead of unspec.
26995 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
26997 2021-02-02 Alexandre Oliva <oliva@adacore.com>
26999 * tree-nested.c (convert_nonlocal_reference_op): Move
27000 current_function_decl restore after re-gimplification.
27001 (convert_local_reference_op): Likewise.
27003 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27005 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
27007 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
27009 (aarch64_rshrn<mode>_insn_be): Likewise.
27010 (aarch64_rshrn<mode>): Likewise.
27011 (aarch64_rshrn2<mode>_insn_le): Likewise.
27012 (aarch64_rshrn2<mode>_insn_be): Likewise.
27013 (aarch64_rshrn2<mode>): Likewise.
27014 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
27015 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
27017 (vrshrn_high_n_s32): Likewise.
27018 (vrshrn_high_n_s64): Likewise.
27019 (vrshrn_high_n_u16): Likewise.
27020 (vrshrn_high_n_u32): Likewise.
27021 (vrshrn_high_n_u64): Likewise.
27022 (vrshrn_n_s16): Likewise.
27023 (vrshrn_n_s32): Likewise.
27024 (vrshrn_n_s64): Likewise.
27025 (vrshrn_n_u16): Likewise.
27026 (vrshrn_n_u32): Likewise.
27027 (vrshrn_n_u64): Likewise.
27029 2021-02-01 Sergei Trofimovich <siarheit@google.com>
27031 PR tree-optimization/98499
27032 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
27033 conservatively and assume all possible side-effects.
27035 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27037 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
27038 vec_unpacku_hi_): Define builtins.
27039 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
27041 (vmovl_high_s16): Likewise.
27042 (vmovl_high_s32): Likewise.
27043 (vmovl_high_u8): Likewise.
27044 (vmovl_high_u16): Likewise.
27045 (vmovl_high_u32): Likewise.
27047 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27049 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
27051 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
27053 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
27055 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
27057 (vabdl_s16): Likewise.
27058 (vabdl_s32): Likewise.
27059 (vabdl_u8): Likewise.
27060 (vabdl_u16): Likewise.
27061 (vabdl_u32): Likewise.
27062 * config/aarch64/iterators.md (ABDL): New int iterator.
27063 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
27065 2021-02-01 Martin Sebor <msebor@redhat.com>
27067 * tree.h (BLOCK_VARS): Add comment.
27068 (BLOCK_SUBBLOCKS): Same.
27069 (BLOCK_SUPERCONTEXT): Same.
27070 (BLOCK_ABSTRACT_ORIGIN): Same.
27071 (inlined_function_outer_scope_p): Same.
27073 2021-02-01 Martin Sebor <msebor@redhat.com>
27075 PR middle-end/97172
27076 * attribs.c (attr_access::free_lang_data): Define new function.
27077 * attribs.h (attr_access::free_lang_data): Declare new function.
27079 2021-02-01 Richard Biener <rguenther@suse.de>
27081 * vec.h (auto_vec::auto_vec): Add memory stat parameters
27083 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
27085 2021-02-01 Tamar Christina <tamar.christina@arm.com>
27087 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
27088 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
27090 2021-02-01 Richard Biener <rguenther@suse.de>
27092 PR rtl-optimization/98863
27093 * config/i386/i386-features.c (convert_scalars_to_vector):
27094 Set DF_RD_PRUNE_DEAD_DEFS.
27096 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
27098 * system.h (SIZE_MAX): Define if not already defined.
27100 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
27102 * config/rs6000/genfusion.pl (gen_2logical): New function to
27103 generate patterns for logical-logical fusion.
27104 * config/rs6000/fusion.md: Regenerated patterns.
27105 * config/rs6000/rs6000-cpus.def: Add
27106 OPTION_MASK_P10_FUSION_2LOGICAL.
27107 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27108 Enable logical-logical fusion for p10.
27109 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
27111 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
27113 * config/rs6000/rs6000.opt: Add periods to new AIX options.
27115 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
27117 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
27118 (mabi=vec-default): New.
27119 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
27120 __EXTABI__ for AIX Vector extended ABI.
27121 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
27123 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
27125 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
27127 2021-01-30 Jakub Jelinek <jakub@redhat.com>
27129 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
27130 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
27132 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
27135 * lra-constraints.c (in_class_p): Don't narrow class only for REG
27138 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
27140 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
27141 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
27142 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
27144 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
27146 PR tree-optimization/98866
27147 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
27148 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
27149 (gori_map::m_maybe_invariant): Rename from all_outgoing.
27150 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
27151 (gori_map::is_export_p): Ditto.
27152 (gori_map::calculate_gori): Ditto.
27153 (gori_compute::set_range_invariant): New.
27154 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
27155 invariant for pointers evaluating to [1, +INF].
27157 2021-01-29 Richard Biener <rguenther@suse.de>
27159 PR rtl-optimization/98863
27160 * config/i386/i386-features.c (remove_partial_avx_dependency):
27161 Do not perform DF analysis.
27162 (pass_data_remove_partial_avx_dependency): Remove
27165 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27167 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
27168 builtin generator macros.
27169 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
27171 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
27172 instead of inline asm.
27173 (vmull_n_s32): Likewise.
27174 (vmull_n_u16): Likewise.
27175 (vmull_n_u32): Likewise.
27177 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27179 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
27181 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
27183 (aarch64_<sur>abdl2<mode>): ... This.
27184 (<sur>sadv16qi): Adjust use of above.
27185 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
27187 (vabdl_high_s16): Likewise.
27188 (vabdl_high_s32): Likewise.
27189 (vabdl_high_u8): Likewise.
27190 (vabdl_high_u16): Likewise.
27191 (vabdl_high_u32): Likewise.
27193 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27195 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
27197 (uabal2): Likewise.
27198 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
27200 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
27202 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
27204 (vabal_high_s16): Likewise.
27205 (vabal_high_s32): Likewise.
27206 (vabal_high_u8): Likewise.
27207 (vabal_high_u16): Likewise.
27208 (vabal_high_u32): Likewise.
27209 * config/aarch64/iterators.md (ABAL2): New mode iterator.
27210 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
27212 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27214 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
27217 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
27219 (aarch64_<sur>abal<mode>): ... This
27220 (<sur>sadv16qi): Adust use of the above.
27221 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
27223 (vabal_s16): Likewise.
27224 (vabal_s32): Likewise.
27225 (vabal_u8): Likewise.
27226 (vabal_u16): Likewise.
27227 (vabal_u32): Likewise.
27229 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27231 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
27233 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
27235 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
27237 (vaddlv_s16): Likewise.
27238 (vaddlv_u8): Likewise.
27239 (vaddlv_u16): Likewise.
27240 (vaddlvq_s8): Likewise.
27241 (vaddlvq_s16): Likewise.
27242 (vaddlvq_s32): Likewise.
27243 (vaddlvq_u8): Likewise.
27244 (vaddlvq_u16): Likewise.
27245 (vaddlvq_u32): Likewise.
27246 (vaddlv_s32): Likewise.
27247 (vaddlv_u32): Likewise.
27248 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
27249 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
27250 (Vwstype): New mode attribute.
27252 (VWIDE_S): Likewise.
27253 (USADDLV): New int iterator.
27254 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
27256 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27258 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
27259 builtin generator macros.
27260 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
27262 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
27263 instead of inline asm.
27264 (vmlsl_lane_s32): Likewise.
27265 (vmlsl_lane_u16): Likewise.
27266 (vmlsl_lane_u32): Likewise.
27267 (vmlsl_laneq_s16): Likewise.
27268 (vmlsl_laneq_s32): Likewise.
27269 (vmlsl_laneq_u16): Likewise.
27270 (vmlsl_laneq_u32): Likewise.
27272 2021-01-29 Richard Biener <rguenther@suse.de>
27274 * doc/invoke.texi (--param max-gcse-memory): Document unit
27276 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
27277 * params.opt (--param max-gcse-memory): Adjust default and
27278 document unit of size.
27280 2021-01-29 Richard Biener <rguenther@suse.de>
27282 PR rtl-optimization/98863
27283 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
27284 HOST_WIDE_INT for the memory estimate.
27286 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
27287 Richard Biener <rguenther@suse.de>
27289 PR tree-optimization/97627
27290 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
27291 Do not analyze fake edges.
27293 2021-01-29 Richard Biener <rguenther@suse.de>
27295 PR rtl-optimization/98144
27296 * df.h (df_mir_bb_info): Add con_visited member.
27297 * df-problems.c (df_mir_alloc): Initialize con_visited,
27298 do not fully populate IN and OUT.
27299 (df_mir_reset): Likewise.
27300 (df_mir_confluence_0): Set con_visited.
27301 (df_mir_confluence_n): Properly handle implicitely
27302 fully populated IN and OUT as designated by con_visited
27303 and update con_visited accordingly.
27305 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27308 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
27309 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
27310 && !TARGET_REALLY_IWMMXT to conditions.
27312 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27315 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
27318 2021-01-28 Marek Polacek <polacek@redhat.com>
27321 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
27322 the main variant, maybe reset it in its variants too.
27323 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
27324 (check_aligned_type): Check if TYPE_USER_ALIGN match.
27326 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
27329 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
27330 of constant zero for comparisons.
27332 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
27334 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
27335 support for mapping built-in function names for long double
27336 built-in functions if long double is IEEE 128-bit.
27338 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27340 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
27341 builtin generator macros.
27342 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
27344 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
27345 instead of inline asm.
27346 (vmlsl_n_s32): Likewise.
27347 (vmlsl_n_u16): Likewise.
27348 (vmlsl_n_u32): Likewise.
27350 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27352 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
27353 builtin generator macros.
27354 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
27356 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
27357 instead of inline asm.
27358 (vmlal_n_s32): Likewise.
27359 (vmlal_n_u16): Likewise.
27360 (vmlal_n_u32): Likewise.
27362 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27364 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
27366 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
27368 (aarch64_shrn2<mode>_insn_be): Likewise.
27369 (aarch64_shrn2<mode>): Likewise.
27370 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
27372 (vshrn_high_n_s32): Likewise.
27373 (vshrn_high_n_s64): Likewise.
27374 (vshrn_high_n_u16): Likewise.
27375 (vshrn_high_n_u32): Likewise.
27376 (vshrn_high_n_u64): Likewise.
27378 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27380 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
27382 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
27384 (aarch64_shrn<mode>_insn_be): Likewise.
27385 (aarch64_shrn<mode>): Likewise.
27386 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
27388 (vshrn_n_s32): Likewise.
27389 (vshrn_n_s64): Likewise.
27390 (vshrn_n_u16): Likewise.
27391 (vshrn_n_u32): Likewise.
27392 (vshrn_n_u64): Likewise.
27393 * config/aarch64/iterators.md (vn_mode): New mode attribute.
27395 2021-01-28 Richard Biener <rguenther@suse.de>
27397 PR rtl-optimization/80960
27398 * dse.c (check_mem_read_rtx): Call get_addr on the
27401 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
27402 David Edelsohn <dje.gcc@gmail.com>
27405 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27406 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
27408 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27410 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
27411 wrapper call rs6000_expand_vector_set_var for cleanup. Call
27412 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
27414 (rs6000_expand_vector_set_var): Delete.
27415 (rs6000_expand_vector_set_var_p9): Make static.
27416 (rs6000_expand_vector_set_var_p8): Make static.
27418 2021-01-28 Xing GUO <higuoxing@gmail.com>
27420 * common/config/riscv/riscv-common.c
27421 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
27422 when `p` extension exists.
27424 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
27426 PR rtl-optimization/97684
27427 * ira.c (ira): Call ira_set_pseudo_classes before
27428 update_equiv_regs when it is necessary.
27430 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27433 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
27434 %w0, %w1 and %2 instead of %0, %1 and %2.
27436 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
27438 * config/rs6000/genfusion.pl: New script to generate
27439 define_insn_and_split patterns so combine can arrange fused
27440 instructions next to each other.
27441 * config/rs6000/fusion.md: New file, generated fused instruction
27442 patterns for combine.
27443 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
27444 (non_update_memory_operand): New predicate.
27445 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
27446 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
27448 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
27450 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27451 Automatically set OPTION_MASK_P10_FUSION and
27452 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
27453 (rs600_opt_masks): Allow -mpower10-fusion
27454 in function attributes.
27455 (address_is_non_pfx_d_or_x): New function.
27456 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
27457 * config/rs6000/rs6000.md: Include fusion.md.
27458 * config/rs6000/rs6000.opt: Add -mpower10-fusion
27459 and -mpower10-fusion-ld-cmpi.
27460 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
27462 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27464 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
27465 builtin generator macros.
27466 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
27468 (aarch64_<su>mlal<mode>): This.
27469 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
27470 instead of inline asm.
27471 (vmlal_s16): Likewise.
27472 (vmlal_s32): Likewise.
27473 (vmlal_u8): Likewise.
27474 (vmlal_u16): Likewise.
27475 (vmlal_u32): Likewise.
27477 2021-01-27 Richard Biener <rguenther@suse.de>
27479 PR tree-optimization/98854
27480 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
27481 PHIs from scalars when the number of CTORs matches the
27482 number of children.
27484 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27486 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
27488 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
27490 (aarch64_mls_n<mode>): This.
27491 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
27493 (vmls_n_s32): Likewise.
27494 (vmls_n_u16): Likewise.
27495 (vmls_n_u32): Likewise.
27496 (vmlsq_n_s16): Likewise.
27497 (vmlsq_n_s32): Likewise.
27498 (vmlsq_n_u16): Likewise.
27499 (vmlsq_n_u32): Likewise.
27501 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27503 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
27505 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
27507 (vmls_s16): Likewise.
27508 (vmls_s32): Likewise.
27509 (vmls_u8): Likewise.
27510 (vmls_u16): Likewise.
27511 (vmls_u32): Likewise.
27512 (vmlsq_s8): Likewise.
27513 (vmlsq_s16): Likewise.
27514 (vmlsq_s32): Likewise.
27515 (vmlsq_u8): Likewise.
27516 (vmlsq_u16): Likewise.
27517 (vmlsq_u32): Likewise.
27519 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27521 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
27523 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
27525 (aarch64_mla_n<mode>): This.
27526 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
27528 (vmla_n_s32): Likewise.
27529 (vmla_n_u16): Likewise.
27530 (vmla_n_u32): Likewise.
27531 (vmlaq_n_s16): Likewise.
27532 (vmlaq_n_s32): Likewise.
27533 (vmlaq_n_u16): Likewise.
27534 (vmlaq_n_u32): Likewise.
27536 2021-01-27 liuhongt <hongtao.liu@intel.com>
27539 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
27540 (*sse2_eq<mode>3): Ditto.
27542 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27544 * tree-pass.h (PROP_trees): Rename to ...
27545 (PROP_gimple): ... this.
27546 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
27547 * passes.c (execute_function_dump, execute_function_todo,
27548 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
27549 * varpool.c (ctor_for_folding): Likewise.
27551 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27553 PR tree-optimization/97260
27554 * varpool.c: Include tree-pass.h.
27555 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
27556 non-TREE_SIDE_EFFECTS automatic variables.
27558 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
27560 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
27562 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
27563 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
27564 (gen_compile_unit_die): Recognise C++23.
27566 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27569 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
27572 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27575 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
27576 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
27577 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
27579 2021-01-26 Richard Biener <rguenther@suse.de>
27581 * gimple-pretty-print.c (dump_binary_rhs): Handle
27582 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
27584 2021-01-26 Richard Biener <rguenther@suse.de>
27586 PR middle-end/98726
27587 * tree.h (vector_cst_int_elt): Remove.
27588 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
27591 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
27593 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
27594 for V64DFmode min/max reductions.
27596 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27598 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
27599 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
27600 two halves, one with x and the other with const0_rtx, ordered
27601 depending on endianity.
27603 2021-01-26 Alexandre Oliva <oliva@adacore.com>
27605 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
27606 temporaries not seen in binding block, and not about to be
27607 added as gimple variables.
27609 2021-01-25 Martin Sebor <msebor@redhat.com>
27612 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
27614 2021-01-25 Martin Liska <mliska@suse.cz>
27616 * value-prof.c (get_nth_most_common_value): Use %s instead
27619 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27622 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
27623 readelf -wi is able to read the emitted .debug_info back.
27624 * configure: Regenerated.
27626 2021-01-25 Martin Liska <mliska@suse.cz>
27628 PR gcov-profile/98739
27629 * common.opt: Add missing sign symbol.
27630 * value-prof.c (get_nth_most_common_value): Restore handling
27631 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
27632 PROFILE_REPRODUCIBILITY_MULTITHREADED.
27634 2021-01-25 Richard Biener <rguenther@suse.de>
27636 PR middle-end/98807
27637 * tree.c (vector_element_bits): Always use precision of
27638 the element type for boolean vectors.
27640 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27642 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
27643 (ENDFILE_SPEC): Evaluate qnolinkcmds.
27645 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27647 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
27648 nostartfiles handling since this is already done by
27649 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
27650 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
27651 is already done by LINK_COMMAND_SPEC.
27652 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
27653 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
27656 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27659 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
27660 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
27661 value rather than host size_t.
27662 (fold_const_call): Change type of s2 from size_t to
27663 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
27664 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
27665 instead of s2 as last argument.
27667 2021-01-25 Tamar Christina <tamar.christina@arm.com>
27669 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
27670 VCMLA_OP, VCMUL_OP): New.
27671 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
27672 * config/arm/neon.md (cmul<conj_op><mode>3): New.
27673 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
27674 UNSPEC_VCMUL_CONJ): New.
27675 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
27676 cml<fcmac1><conj_op><mode>4): New.
27678 2021-01-23 Jakub Jelinek <jakub@redhat.com>
27681 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
27683 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
27685 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
27687 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
27689 (vmla_s16): Likewise.
27690 (vmla_s32): Likewise.
27691 (vmla_u8): Likewise.
27692 (vmla_u16): Likewise.
27693 (vmla_u32): Likewise.
27694 (vmlaq_s8): Likewise.
27695 (vmlaq_s16): Likewise.
27696 (vmlaq_s32): Likewise.
27697 (vmlaq_u8): Likewise.
27698 (vmlaq_u16): Likewise.
27699 (vmlaq_u32): Likewise.
27701 2021-01-22 David Malcolm <dmalcolm@redhat.com>
27703 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
27706 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27709 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
27710 filenames to emit, still emit the required 0 index directory and
27711 filename entries that match DW_AT_comp_dir and DW_AT_name of the
27714 2021-01-22 Marek Polacek <polacek@redhat.com>
27717 * doc/invoke.texi: Update C++ ABI Version 15 description.
27719 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27721 PR tree-optimization/98766
27722 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
27723 comparing against type size with param_avoid_fma_max_bits.
27725 2021-01-22 Richard Biener <rguenther@suse.de>
27727 PR middle-end/98793
27728 * tree.c (vector_element_bits): Key single-bit bool vector on
27729 integer mode rather than not vector mode.
27731 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
27734 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27735 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
27737 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
27738 to call different path for P8 and P9.
27739 (rs6000_expand_vector_set_var_p9): New function.
27740 (rs6000_expand_vector_set_var_p8): New function.
27742 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
27746 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27747 Ajdust variable index vec_insert from address dereference to
27748 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
27749 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27751 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
27753 2021-01-22 Martin Liska <mliska@suse.cz>
27755 PR gcov-profile/98739
27756 * profile.c (compute_value_histograms): Drop time profile for
27757 -fprofile-reproducible=multithreaded.
27759 2021-01-22 Nathan Sidwell <nathan@acm.org>
27761 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
27764 2021-01-22 Richard Biener <rguenther@suse.de>
27766 PR middle-end/98773
27767 * tree-data-ref.c (initalize_matrix_A): Revert previous
27768 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
27770 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27772 PR tree-optimization/90248
27773 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
27774 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
27776 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
27777 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
27779 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27781 PR tree-optimization/98255
27782 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
27783 extend index - low_bound from sizetype's precision rather than index
27785 (get_addr_base_and_unit_offset_1): Likewise.
27786 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
27787 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
27789 2021-01-22 Richard Biener <rguenther@suse.de>
27791 PR tree-optimization/98786
27792 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
27793 adding new uses of abnormals. Verify we deal with a conditional
27796 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27799 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
27801 2021-01-22 liuhongt <hongtao.liu@intel.com>
27805 * config/i386/sse.md (VI_128_256): New mode iterator.
27806 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
27807 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
27808 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
27809 define_insn_and_split to lower avx512 vector comparison to avx
27810 version when dest is vector.
27811 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
27812 define_insn_and_split for negating the comparison result.
27813 * config/i386/predicates.md (float_vector_all_ones_operand):
27815 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
27816 general NOT operator without UNSPEC_MASKOP.
27818 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
27820 PR rtl-optimization/98777
27821 * lra-int.h (lra_pmode_pseudo): New extern.
27822 * lra.c (lra_pmode_pseudo): New global.
27824 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
27826 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
27828 * fwprop.c (fwprop_propagation::classify_result): Allow
27829 (subreg (mem)) simplifications.
27831 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27833 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
27835 (aarch64_sqdmlal<mode>): ... This...
27836 (aarch64_sqdmlsl<mode>): ... And this.
27837 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
27838 (aarch64_sqdmlal_lane<mode>): ... This...
27839 (aarch64_sqdmlsl_lane<mode>): ... And this.
27840 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
27841 (aarch64_sqdmlsl_laneq<mode>): ... This...
27842 (aarch64_sqdmlal_laneq<mode>): ... And this.
27843 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
27844 (aarch64_sqdmlsl_n<mode>): ... This...
27845 (aarch64_sqdmlal_n<mode>): ... And this.
27846 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
27847 (aarch64_sqdmlal2<mode>_internal): ... This...
27848 (aarch64_sqdmlsl2<mode>_internal): ... And this.
27850 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
27852 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
27854 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
27857 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
27859 2021-01-21 liuhongt <hongtao.liu@intel.com>
27861 PR rtl-optimization/98694
27862 * regcprop.c (copy_value): If SRC had been assigned a mode
27863 narrower than the copy, we can't link DEST into the chain even
27864 they have same hard_regno_nregs(i.e. HImode/SImode in i386
27867 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27869 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
27870 Convert to define_insn_and_split. Split into simple move when moving
27873 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
27875 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
27876 Adjust comment. Simplify code.
27878 2021-01-20 Jakub Jelinek <jakub@redhat.com>
27881 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
27882 with DW_FORM_line_strp form.
27883 (prune_unused_types_update_strings): Don't add into debug_str_hash
27884 indirect strings with DW_FORM_line_strp form.
27885 (adjust_name_comp_dir): New function.
27886 (dwarf2out_finish): Call it on CU DIEs after resetting
27887 debug_line_str_hash.
27889 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
27891 PR rtl-optimization/98722
27892 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
27893 has no 3-op add insn to transform insns containing two pluses.
27895 2021-01-20 Richard Biener <rguenther@suse.de>
27897 * hwint.h (add_hwi): New function.
27898 (mul_hwi): Likewise.
27899 * tree-data-ref.c (initialize_matrix_A): Properly translate
27900 tree constants and avoid HOST_WIDE_INT_MIN.
27901 (lambda_matrix_row_add): Avoid undefined integer overflow
27902 and return true on such overflow.
27903 (lambda_matrix_right_hermite): Handle overflow from
27904 lambda_matrix_row_add gracefully. Simplify previous fix.
27905 (analyze_subscript_affine_affine): Likewise.
27907 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
27909 PR tree-optimization/96674
27910 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
27911 x >= y && y != XXX_MIN --> x > y - 1
27913 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
27915 PR tree-optimization/98535
27916 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
27917 If the high and low permutes are the same, remove the high permutes
27918 from the working set and only continue with the low ones.
27920 2021-01-20 Jakub Jelinek <jakub@redhat.com>
27922 PR tree-optimization/98721
27923 * builtins.c (access_ref::inform_access): Don't assume
27924 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
27925 object whenever allocfn is NULL, rather than only when DECL_P
27926 is true. Use %qE instead of %qD for that. Formatting fixes.
27928 2021-01-20 Richard Biener <rguenther@suse.de>
27930 PR tree-optimization/98758
27931 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
27932 (lambda_matrix_right_hermite): Avoid undefinedness with
27933 signed integer abs and multiplication.
27934 (analyze_subscript_affine_affine): Use lambda_int.
27936 2021-01-20 David Malcolm <dmalcolm@redhat.com>
27939 * dwarf2out.c (output_line_info): Rename static variable
27940 "generation", moving it out of the function to...
27941 (output_line_info_generation): New.
27942 (init_sections_and_labels): Likewise, renaming the variable to...
27943 (init_sections_and_labels_generation): New.
27944 (dwarf2out_c_finalize): Reset the new variables.
27946 2021-01-19 Martin Sebor <msebor@redhat.com>
27948 PR middle-end/98664
27949 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
27950 all functions, even if they're not declared artificial or inline.
27951 * tree.c (tree_inlined_location): Use macro expansion location
27952 only if scope traversal fails to expose one.
27954 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
27956 PR rtl-optimization/92294
27957 * alias.c (compare_base_symbol_refs): Take an extra parameter
27958 and add the distance between two symbols to it. Enshrine in
27959 comments that -1 means "either 0 or 1, but we can't tell
27960 which at compile time".
27961 (memrefs_conflict_p): Update call accordingly.
27962 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
27965 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27967 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
27968 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
27969 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
27970 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
27971 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
27972 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
27973 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
27974 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
27975 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
27976 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
27977 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
27978 sqmovun): Use NONE flags.
27980 2021-01-19 Richard Biener <rguenther@suse.de>
27983 * ipa-modref.c (analyze_stmt): Only record a summary for a
27986 2021-01-19 Richard Biener <rguenther@suse.de>
27988 PR middle-end/98638
27989 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
27991 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
27993 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
27994 built-in define __FIX_LEON3FT_TN0018.
27996 2021-01-19 Richard Biener <rguenther@suse.de>
27999 * tree-inline.c (tree_function_versioning): Set input_location
28000 to UNKNOWN_LOCATION throughout the function.
28002 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
28005 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
28007 2021-01-19 Martin Jambor <mjambor@suse.cz>
28010 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
28011 whether non-call exceptions allow removal of a statement.
28012 (isra_analyze_call): Pass the appropriate function to
28013 ssa_name_only_returned_p.
28015 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
28017 * config/riscv/arch-canonicalize (longext_sort): New function for
28018 sorting 'multi-letter'.
28019 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
28020 'alts'. The 'arch' may not be the first of 'alts'.
28021 (_expand_combination): Add underline for the 'ext' without '*'.
28022 This is because, a single-letter extension can always be treated well
28023 with a '_' prefix, but it cannot be separated out if it is appended
28026 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
28029 * ira.c (ira): Skip abnormal critical edge splitting.
28031 2021-01-18 Jakub Jelinek <jakub@redhat.com>
28033 PR tree-optimization/98727
28034 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
28035 second .MUL_OVERFLOW operand for signed multiplication with overflow
28036 checking if the second operand of multiplication is not constant.
28038 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
28040 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
28041 defaults to version 4.
28043 2021-01-18 David Malcolm <dmalcolm@redhat.com>
28045 * attribs.h (fndecl_dealloc_argno): New decl.
28046 * builtins.c (call_dealloc_argno): Split out second half of
28048 (fndecl_dealloc_argno): New.
28049 * doc/extend.texi (Common Function Attributes): Document the
28050 interaction between the analyzer and the malloc attribute.
28051 * doc/invoke.texi (Static Analyzer Options): Likewise.
28053 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
28055 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
28056 dwarf_version to 4.
28057 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
28059 2021-01-17 Martin Jambor <mjambor@suse.cz>
28062 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
28065 2021-01-17 Mark Wielaard <mark@klomp.org>
28067 * common.opt (gdwarf-): Init(5).
28068 * doc/invoke.texi (-gdwarf): Document default to 5.
28070 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
28072 * builtin-types.def
28073 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
28075 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
28076 ...this. Add extra argument.
28077 * gimplify.c (omp_default_clause): Ensure that event handle is
28078 firstprivate in a task region.
28079 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
28080 (gimplify_adjust_omp_clauses): Likewise.
28081 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
28082 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
28083 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
28084 if detach clause specified. Add detach argument when generating
28086 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
28088 (finish_taskreg_scan): Move field for variable containing the event
28089 handle to the front of the struct.
28090 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
28092 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
28093 OMP_CLAUSE_DETACH clause.
28094 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
28095 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
28096 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
28098 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
28100 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
28102 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28104 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
28105 multilib directory names. Use MULTILIB_REQUIRED instead of
28106 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
28107 -mcustom-fpu-cfg=fph2 multilib.
28109 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28111 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
28112 (nios2_init_fpu_configs): Provide register values for new
28113 -mcustom-fpu-cfg=fph2 option variant.
28114 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
28117 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28119 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
28120 custom instruction warnings.
28122 2021-01-16 Jakub Jelinek <jakub@redhat.com>
28124 PR tree-optimization/96669
28125 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
28127 2021-01-16 Jakub Jelinek <jakub@redhat.com>
28129 PR tree-optimization/96271
28130 * passes.def: Pass false argument to first two pass_cd_dce
28131 instances and true to last instance. Add comment that
28132 last instance rewrites no longer addressed locals.
28133 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
28135 (pass_cd_dce::set_pass_param): New method.
28136 (pass_cd_dce::execute): Return TODO_update_address_taken from
28137 last cd_dce instance.
28139 2021-01-15 Carl Love <cel@us.ibm.com>
28141 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
28143 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
28144 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
28145 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
28146 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
28147 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
28148 Add builtin define.
28149 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
28150 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
28151 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
28152 New overloaded definitions.
28153 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
28154 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
28155 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
28156 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
28157 P10V_BUILTIN_MULHU_V4SI]: Add case
28158 statement for builtins.
28159 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
28160 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
28161 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
28162 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
28163 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
28164 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
28165 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
28166 Add define_insn, mode is VIlong.
28167 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
28168 Add builtin descriptions.
28170 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
28172 * final.c (final_start_function_1): Reset force_source_line.
28174 2021-01-15 Jakub Jelinek <jakub@redhat.com>
28176 PR tree-optimization/96669
28177 * match.pd (((1 << A) & 1) != 0 -> A == 0,
28178 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
28179 possibly different power of two constants and to right shift too.
28181 2021-01-15 Jakub Jelinek <jakub@redhat.com>
28183 PR tree-optimization/96681
28184 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
28185 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
28186 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
28187 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
28189 2021-01-15 Alexandre Oliva <oliva@adacore.com>
28191 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
28193 2021-01-15 Tamar Christina <tamar.christina@arm.com>
28195 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
28196 cmul<conj_op><mode>3): New.
28197 * config/aarch64/iterators.md (UNSPEC_FCMUL,
28198 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
28199 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
28200 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
28201 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
28202 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
28203 (rot_op): Renamed to conj_op.
28204 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
28205 cmul<conj_op><mode>3): New.
28206 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
28207 cmul<conj_op><mode>3): New.
28209 2021-01-15 David Malcolm <dmalcolm@redhat.com>
28213 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28214 Escape the tempfile name when constructing the expected output.
28216 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28218 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
28220 (aarch64_<su>mlsl_hi<mode>): ... This.
28221 (aarch64_<su>mlsl_hi<mode>): Define.
28222 (*aarch64_<su>mlsl<mode): Rename to...
28223 (aarch64_<su>mlsl<mode): ... This.
28224 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
28225 smlsl_hi, umlsl_hi): Define builtins.
28226 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
28227 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
28228 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
28229 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
28231 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
28233 * config/i386/i386-c.c (ix86_target_macros):
28234 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
28236 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28239 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
28240 * Makefile.in (RTL_SSA_H): New variable.
28241 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
28242 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
28243 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
28245 * config/aarch64/aarch64-cc-fusion.cc: New file.
28247 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28249 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
28250 calling cancel_changes for changes that no longer exist.
28252 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28254 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
28255 (function_info::reg_defs): ...this.
28256 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
28257 (function_info::reg_defs): ...this.
28259 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28262 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28264 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28267 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28270 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28272 2021-01-15 Richard Biener <rguenther@suse.de>
28274 PR tree-optimization/96376
28275 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
28276 for VMAT_INVARIANT.
28278 2021-01-15 Martin Liska <mliska@suse.cz>
28280 * doc/install.texi: Document that some tests need pytest module.
28281 * doc/sourcebuild.texi: Likewise.
28283 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28286 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28288 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28290 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
28291 (mve_vshrq_n_u<mode>_imm): Likewise.
28292 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
28293 * config/arm/vec-common.md: ... here.
28295 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28297 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
28299 * config/arm/neon.md (vashl<mode>3): Delete.
28300 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
28301 (vasl<mode>3): New expander.
28303 2021-01-15 Richard Biener <rguenther@suse.de>
28305 PR tree-optimization/98685
28306 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
28307 of vector extern defs.
28309 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28312 * diagnostic.c (diagnostic_kind_text): Break out this array
28314 (diagnostic_build_prefix): ...here.
28315 (fancy_abort): Detect when diagnostic_initialize has not yet been
28316 called and fall back to a minimal implementation of printing the
28317 ICE, rather than segfaulting in internal_error.
28319 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28321 * diagnostic.c (diagnostic_initialize): Eliminate
28322 parseable_fixits_p in favor of initializing extra_output_kind from
28323 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28324 (convert_column_unit): New function, split out from...
28325 (diagnostic_converted_column): ...this.
28326 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
28327 Use them to call convert_column_unit on the column values.
28328 (diagnostic_report_diagnostic): Eliminate conditional on
28329 parseable_fixits_p in favor of a switch statement on
28330 extra_output_kind, passing the appropriate values to the new
28331 params of print_parseable_fixits.
28332 (selftest::test_print_parseable_fixits_none): Update for new
28333 params of print_parseable_fixits.
28334 (selftest::test_print_parseable_fixits_insert): Likewise.
28335 (selftest::test_print_parseable_fixits_remove): Likewise.
28336 (selftest::test_print_parseable_fixits_replace): Likewise.
28337 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28339 (selftest::diagnostic_c_tests): Call it.
28340 * diagnostic.h (enum diagnostics_extra_output_kind): New.
28341 (diagnostic_context::parseable_fixits_p): Delete field in favor
28343 (diagnostic_context::extra_output_kind): ...this new field.
28344 * doc/invoke.texi (Environment Variables): Add
28345 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28346 * opts.c (common_handle_option): Update handling of
28347 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
28350 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28352 * tree-vect-slp-patterns.c (class complex_operations_pattern,
28353 complex_operations_pattern::matches,
28354 complex_operations_pattern::recognize,
28355 complex_operations_pattern::build): New.
28356 (slp_patterns): Use it.
28358 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28360 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
28361 * optabs.def (cmls_optab, cmls_conj_optab): New.
28362 * doc/md.texi: Document them.
28363 * tree-vect-slp-patterns.c (class complex_fms_pattern,
28364 complex_fms_pattern::matches, complex_fms_pattern::recognize,
28365 complex_fms_pattern::build): New.
28367 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28369 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
28370 * optabs.def (cmla_optab, cmla_conj_optab): New.
28371 * doc/md.texi: Document them.
28372 * tree-vect-slp-patterns.c (vect_match_call_p,
28373 class complex_fma_pattern, vect_slp_reset_pattern,
28374 complex_fma_pattern::matches, complex_fma_pattern::recognize,
28375 complex_fma_pattern::build): New.
28377 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28379 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
28380 * optabs.def (cmul_optab, cmul_conj_optab): New.
28381 * doc/md.texi: Document them.
28382 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
28383 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
28384 vect_build_combine_node, class complex_mul_pattern,
28385 complex_mul_pattern::matches, complex_mul_pattern::recognize,
28386 complex_mul_pattern::build): New.
28388 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28390 * tree-vect-slp.c (optimize_load_redistribution_1): New.
28391 (optimize_load_redistribution, vect_is_slp_load_node): New.
28392 (vect_match_slp_patterns): Use it.
28394 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28396 * tree-vect-slp-patterns.c (complex_add_pattern::build):
28399 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
28401 * config/gcn/mkoffload.c (main): Create an offload image only in
28402 64-bit configurations.
28404 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
28407 * config/i386/i386-options.c (ix86_option_override_internal):
28408 Issue an error for -fcf-protection with CF_BRANCH when compiling
28409 for 32-bit non-TARGET_CMOV targets.
28411 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28414 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
28415 Remove declaration and initialization of shadow variable "ret".
28416 (ix86_option_override_internal): Remove delcaration of
28417 shadow variable "i". Redeclare shadowed variable to unsigned.
28418 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
28419 * config/i386/i386-builtins.c (get_builtin_code_for_version):
28420 Update for redeclaration.
28421 * config/i386/i386.h (pta_size): Ditto.
28423 2021-01-14 Richard Biener <rguenther@suse.de>
28425 PR tree-optimization/98674
28426 * tree-data-ref.c (base_supports_access_fn_components_p): New.
28427 (initialize_data_dependence_relation): For two bases without
28428 possible access fns resort to type size equality when determining
28429 shape compatibility.
28431 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28434 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
28435 <=, >= operators in vcle and vcge intrinsics respectively.
28436 * config/arm/arm_neon_builtins.def: Remove entry for
28439 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28442 * config/i386/i386-options.c (ix86_function_specific_save):
28443 Remove redundant assignment to opts->x_ix86_branch_cost.
28444 * config/i386/i386.c (ix86_prefetch_sse):
28445 Rename from x86_prefetch_sse. Update all uses.
28446 * config/i386/i386.h: Update for rename.
28447 * config/i386/i386-options.h: Ditto.
28449 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28452 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
28453 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
28454 Use Bm instead of m for non-avx. Add isa attribute.
28456 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28458 PR tree-optimization/96688
28459 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
28460 ~X can be simplified.
28462 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
28464 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
28465 IFN_LOAD_LANES results.
28467 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28469 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
28471 (aarch64_xtn<mode>): Likewise.
28472 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
28475 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28477 (vmovl_s16): Likewise.
28478 (vmovl_s32): Likewise.
28479 (vmovl_u8): Likewise.
28480 (vmovl_u16): Likewise.
28481 (vmovl_u32): Likewise.
28482 (vmovn_s16): Likewise.
28483 (vmovn_s32): Likewise.
28484 (vmovn_s64): Likewise.
28485 (vmovn_u16): Likewise.
28486 (vmovn_u32): Likewise.
28487 (vmovn_u64): Likewise.
28489 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28491 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
28493 (aarch64_<su>qxtn2<mode>_be): Likewise.
28494 (aarch64_<su>qxtn2<mode>): Likewise.
28495 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
28497 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
28498 (su): Handle ss_truncate and us_truncate.
28499 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
28501 (vqmovn_high_s32): Likewise.
28502 (vqmovn_high_s64): Likewise.
28503 (vqmovn_high_u16): Likewise.
28504 (vqmovn_high_u32): Likewise.
28505 (vqmovn_high_u64): Likewise.
28507 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28509 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
28511 (aarch64_xtn2<mode>_be): Likewise.
28512 (aarch64_xtn2<mode>): Likewise.
28513 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
28515 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
28517 (vmovn_high_s32): Likewise.
28518 (vmovn_high_s64): Likewise.
28519 (vmovn_high_u16): Likewise.
28520 (vmovn_high_u32): Likewise.
28521 (vmovn_high_u64): Likewise.
28523 2021-01-13 Stafford Horne <shorne@gmail.com>
28525 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
28527 2021-01-13 Stafford Horne <shorne@gmail.com>
28529 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
28531 2021-01-13 Stafford Horne <shorne@gmail.com>
28533 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
28534 define for __or1k_hard_float__.
28536 2021-01-13 Stafford Horne <shorne@gmail.com>
28538 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
28539 (PROFILE_HOOK): Define to call _mcount.
28540 (FUNCTION_PROFILER): Change from abort to no-op.
28542 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28544 PR tree-optimization/96691
28545 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
28546 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
28547 (~D ^ C) or (D ^ C) can be simplified.
28549 2021-01-13 Richard Biener <rguenther@suse.de>
28551 PR tree-optimization/92645
28552 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
28553 until after vector lowering.
28555 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28557 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
28559 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
28560 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
28562 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28564 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
28566 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
28567 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
28569 2021-01-13 Richard Biener <rguenther@suse.de>
28571 PR tree-optimization/92645
28572 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
28573 BIT_FIELD_REF argument.
28574 (vect_build_slp_tree_2): Record the desired vector type
28575 on the external vector def.
28576 (vectorizable_slp_permutation): Handle required punning
28577 of existing vector defs.
28579 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28581 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
28583 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28585 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
28587 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
28589 * config.gcc [$target == *-*-gnu*]: Enable
28590 'default_gnu_indirect_function'.
28592 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28595 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
28596 registers before calling targetm.vectorize.vec_perm_const, only after
28598 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
28599 two argument permutation when one operand is zero vector and only
28600 after that force operands into registers.
28601 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
28602 define_insn_and_split pattern.
28603 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
28604 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
28605 (*avx2_zero_extendv8hiv8si2_1): Likewise.
28606 (*avx512f_zero_extendv8siv8di2_1): Likewise.
28607 (*avx2_zero_extendv4siv4di2_1): Likewise.
28608 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
28610 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
28611 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
28612 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
28613 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
28614 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
28615 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
28617 2021-01-13 Martin Liska <mliska@suse.cz>
28619 PR tree-optimization/98455
28620 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
28621 Record also virtual PHIs.
28622 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
28625 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
28627 * doc/invoke.texi (C++ Modules): Fix typos.
28629 2021-01-13 Richard Biener <rguenther@suse.de>
28631 PR tree-optimization/98640
28632 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
28633 handle plus or minus from a truncated operand to be
28636 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28639 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
28640 define_insn_and_split patterns.
28641 (splitter after *btr<mode>_2): New splitter.
28643 2021-01-13 Martin Liska <mliska@suse.cz>
28646 * cgraphunit.c (analyze_functions): Remove dead code.
28648 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
28650 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
28651 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
28652 (a64fx_regmove_cost, a64fx_vector_cost): New.
28653 (a64fx_tunings): Use the new added cost tables.
28655 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28658 * config/i386/predicates.md (pmovzx_parallel): New predicate.
28659 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
28660 define_insn_and_split pattern.
28661 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
28662 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
28664 2021-01-13 Julian Brown <julian@codesourcery.com>
28666 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
28667 to fix v0 register.
28669 2021-01-13 Julian Brown <julian@codesourcery.com>
28671 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
28674 2021-01-13 Julian Brown <julian@codesourcery.com>
28676 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
28677 for reciprocal-approximation instructions.
28678 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
28679 refinement and division result.
28680 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
28682 2021-01-13 Julian Brown <julian@codesourcery.com>
28684 * config/gcn/gcn-valu.md (subdf): Rename to...
28687 2021-01-12 Martin Liska <mliska@suse.cz>
28689 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
28691 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
28693 * function-abi.h: Fix typo.
28695 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
28699 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
28700 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
28701 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
28702 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
28703 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
28704 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
28705 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
28706 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
28707 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
28708 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
28709 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
28710 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
28711 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
28712 (ARM_HAVE_V2DI_LDST): Likewise.
28713 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
28714 (*movmisalign<mode>_mve_load): New pattern.
28715 * config/arm/neon.md (movmisalign<mode>): Move to ...
28716 * config/arm/vec-common.md: ... here.
28718 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
28721 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
28722 of pattern 'plus (plus (hard reg, const), pseudo)'.
28724 2021-01-12 Richard Biener <rguenther@suse.de>
28726 PR tree-optimization/98550
28727 * tree-vect-slp.c (vect_record_max_nunits): Check whether
28728 the group size is a multiple of the vector element count.
28729 (vect_build_slp_tree_1): When we need to fail because
28730 the vector type choosen causes unrolling do so lazily
28731 without affecting matches only at the end to guide group splitting.
28733 2021-01-12 Martin Liska <mliska@suse.cz>
28736 * optc-save-gen.awk: Compare also n_target_save vars with
28739 2021-01-12 Martin Liska <mliska@suse.cz>
28741 * gcov.c (source_info::debug): New.
28742 (print_usage): Add --debug (-D) option.
28743 (process_args): Likewise.
28744 (generate_results): Call src->debug after
28745 accumulate_line_counts.
28746 (read_graph_file): Properly assign id for EXIT_BLOCK.
28747 * profile.c (branch_prob): Dump function body before it is
28750 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28752 PR tree-optimization/98629
28753 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
28754 unless returning non-zero.
28756 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28758 PR tree-optimization/95731
28759 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
28760 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
28761 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
28762 only after optimize_range_tests_var_bound.
28764 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28766 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
28767 * configure: Regenerated.
28769 2021-01-12 liuhongt <hongtao.liu@intel.com>
28772 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
28774 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
28777 2021-01-12 Alexandre Oliva <oliva@adacore.com>
28779 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
28781 (auto_end_imm_use_stmt_traverse): New struct.
28782 (FOR_EACH_IMM_USE_STMT): Use it.
28783 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
28785 * gimple-ssa-strength-reduction.c: ... here, ...
28786 * graphite-scop-detection.c: ... here, ...
28787 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
28788 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
28789 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
28790 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
28791 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
28792 * tree-vect-slp.c: ... and here, ...
28793 * doc/tree-ssa.texi: ... and the example here.
28795 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28797 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
28798 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
28799 (*sdiv_pow2<mode>3): New pattern.
28800 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
28801 Wrap the ASRD in an UNSPEC_PRED_X.
28802 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
28803 predicate with a constant PTRUE, if it isn't already.
28804 (*cond_<sve_int_op><mode>_z): Replace with...
28805 (*cond_<sve_int_op><mode>_any): ...this new pattern.
28807 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28809 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
28810 SVE_FULL_I to SVE_I.
28811 (*cond_bic<mode>_any): Likewise.
28813 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28815 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
28816 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
28819 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28821 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
28822 SVE_FULL_I to SVE_I.
28823 (*aarch64_cond_<su>abd<mode>_2): Likewise.
28824 (*aarch64_cond_<su>abd<mode>_any): Likewise.
28825 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
28826 for the max and min but not for the minus.
28827 (*aarch64_cond_<su>abd<mode>_3): New pattern.
28829 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28831 * config/aarch64/iterators.md (SVE_24I): New iterator.
28832 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
28833 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
28835 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28837 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
28838 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
28840 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
28841 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
28842 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
28843 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
28845 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28847 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
28848 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
28849 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
28852 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28854 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
28855 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
28856 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
28858 2021-01-11 Martin Liska <mliska@suse.cz>
28861 * symtab-clones.h (clone_info::release): Release
28862 symtab::m_clones with ggc_delete as it's a GGC memory.
28864 2021-01-11 Matthias Klose <doko@ubuntu.com>
28866 * Makefile.in (LINK_PROGRESS): Show the link target.
28868 2021-01-11 Richard Biener <rguenther@suse.de>
28870 PR tree-optimization/91403
28871 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
28872 single-element interleaving group size at 4096 elements.
28874 2021-01-11 Richard Biener <rguenther@suse.de>
28876 PR tree-optimization/98526
28877 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
28878 of the actual reduction op for the regular case.
28879 (vectorizable_reduction): Cost the stmts
28880 vect_transform_reduction produces here.
28882 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
28884 * tree-ssa-forwprop.c (simplify_vector_constructor): For
28885 big-endian, use UNPACK[_FLOAT]_HI.
28887 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28889 * tree-vect-slp-patterns.c (class complex_pattern,
28890 class complex_add_pattern): Add parameters to matches.
28891 (complex_add_pattern::build): Free memory.
28892 (complex_add_pattern::matches): Move validation end of match.
28893 (complex_add_pattern::recognize): Likewise.
28895 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28897 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
28899 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28901 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
28903 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28905 PR tree-optimization/95867
28906 * tree-ssa-math-opts.h: New header.
28907 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
28908 (powi_as_mults): No longer static. Use build_one_cst instead of
28909 build_real. Formatting fix.
28910 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
28911 (attempt_builtin_powi): Handle multiplication reassociation without
28912 powi_fndecl using powi_as_mults.
28913 (reassociate_bb): For integral types don't require
28914 -funsafe-math-optimizations to call attempt_builtin_powi.
28916 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28918 PR tree-optimization/95852
28919 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
28920 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
28921 allow in the bb any of the stmts in that vector, div_stmt and
28922 up to 3 cast stmts.
28923 (arith_cast_equal_p): New function.
28924 (arith_overflow_check_p): Add cast_stmt argument, handle signed
28925 multiply overflow checks.
28926 (match_arith_overflow): Adjust caller. Handle signed multiply
28929 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28931 PR tree-optimization/95852
28932 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
28933 (uaddsub_overflow_check_p): Renamed to ...
28934 (arith_overflow_check_p): ... this. Handle also multiplication
28935 with overflow check.
28936 (match_uaddsub_overflow): Renamed to ...
28937 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
28938 also multiplication with overflow check. Adjust function comment.
28939 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
28940 match_arith_overflow also for MULT_EXPR.
28942 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28944 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28945 __builtin_convertvector.
28946 (vmovl_s16): Likewise.
28947 (vmovl_s32): Likewise.
28948 (vmovl_u8): Likewise.
28949 (vmovl_u16): Likewise.
28950 (vmovl_u32): Likewise.
28951 (vmovn_s16): Likewise.
28952 (vmovn_s32): Likewise.
28953 (vmovn_s64): Likewise.
28954 (vmovn_u16): Likewise.
28955 (vmovn_u32): Likewise.
28956 (vmovn_u64): Likewise.
28958 2021-01-11 Martin Liska <mliska@suse.cz>
28960 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
28961 (if_chain::is_beneficial): Delete clusters
28962 (find_conditions): Make second argument of conditions_in_bbs a
28963 pointer so that we control over it's lifetime.
28964 (pass_if_to_switch::execute): Delete them.
28966 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
28968 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
28971 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28973 * config/vax/vax.md (cc): Remove mode attribute.
28974 (subst_<cc>, subst_f<cc>): Rename to...
28975 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
28976 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
28977 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
28978 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
28980 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28982 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
28983 `const_double_zero'.
28985 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28987 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
28988 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
28991 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28993 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
28995 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
28996 with `const_double_zero'.
28997 * doc/rtl.texi (Constant Expression Types): Document it.
28999 2021-01-09 Jakub Jelinek <jakub@redhat.com>
29002 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
29003 POINTER_DIFF_EXPR to be any integral type.
29005 2021-01-09 Jakub Jelinek <jakub@redhat.com>
29007 PR rtl-optimization/98603
29008 * function.c (instantiate_virtual_regs_in_insn): For asm goto
29009 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
29010 if any, set ASM_OPERANDS mode to VOIDmode and change
29011 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
29013 2021-01-09 Alexandre Oliva <oliva@gnu.org>
29016 * final.c (notice_source_line): Narrow down the condition to
29017 skip a line-0 marker.
29019 2021-01-08 Sergei Trofimovich <siarheit@google.com>
29021 * ipa-modref.c (merge_call_side_effects): Fix
29022 linebreak split by reordering two print calls.
29024 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
29026 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
29027 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
29028 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
29029 consistency, fix constraint.
29031 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
29033 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
29034 callables instead of mask values.
29035 (struct target_flag_set_p): New predicate.
29036 (s390_cpu_cpp_builtins_internal): Define or undefine
29037 __LONG_DOUBLE_VX__ macro.
29039 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
29042 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
29043 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
29046 2021-01-08 Richard Biener <rguenther@suse.de>
29048 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
29050 2021-01-08 Richard Biener <rguenther@suse.de>
29052 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
29053 (vect_build_slp_tree): On cache hit release the matched
29054 scalar stmts vector.
29055 * tree-vect-stmts.c (vectorizable_store): Properly free
29056 vec_oprnds before possibly gathering them again.
29058 2021-01-08 Richard Biener <rguenther@suse.de>
29060 PR tree-optimization/98544
29061 * tree-vect-slp.c (vect_optimize_slp): Always materialize
29062 permutes at a permute node.
29064 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
29067 * config/i386/i386.c (x86_function_profiler): Use R10 to call
29068 mcount in large model. Sorry for large model with PIC.
29070 2021-01-08 Jakub Jelinek <jakub@redhat.com>
29073 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
29074 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
29075 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
29076 TargetSave and initialize for variables with enum types.
29077 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
29078 mstack-protector-guard-symbol=): Add Save.
29079 * config/i386/i386-options.c (ix86_function_specific_save,
29080 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
29081 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
29082 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
29083 x_ix86_veclibabi_type.
29085 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
29087 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
29088 SVE_FULL_I to SVE_I.
29089 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
29091 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
29093 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
29094 SVE_FULL_I to SVE_I.
29095 (*cond_uxt<mode>_any): Likewise.
29097 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29099 * config/aarch64/iterators.md (Vwhalf): New iterator.
29100 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
29102 (aarch64_<sur>adalp<mode>): ... This. Make more
29104 (<sur>sadv16qi): Adjust callsite of the above.
29105 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
29107 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
29109 (vpadal_s16): Likewise.
29110 (vpadal_u8): Likewise.
29111 (vpadal_u16): Likewise.
29112 (vpadalq_s8): Likewise.
29113 (vpadalq_s16): Likewise.
29114 (vpadalq_s32): Likewise.
29115 (vpadalq_u8): Likewise.
29116 (vpadalq_u16): Likewise.
29117 (vpadalq_u32): Likewise.
29119 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29121 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
29123 (aarch64_<su>abd<mode>): ... This.
29124 (<sur>sadv16qi): Adjust callsite of the above.
29125 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
29127 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
29129 (vabd_s16): Likewise.
29130 (vabd_s32): Likewise.
29131 (vabd_u8): Likewise.
29132 (vabd_u16): Likewise.
29133 (vabd_u32): Likewise.
29134 (vabdq_s8): Likewise.
29135 (vabdq_s16): Likewise.
29136 (vabdq_s32): Likewise.
29137 (vabdq_u8): Likewise.
29138 (vabdq_u16): Likewise.
29139 (vabdq_u32): Likewise.
29141 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29143 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
29145 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
29146 (vaba_s16): Likewise.
29147 (vaba_s32): Likewise.
29148 (vaba_u8): Likewise.
29149 (vaba_u16): Likewise.
29150 (vaba_u32): Likewise.
29151 (vabaq_s8): Likewise.
29152 (vabaq_s16): Likewise.
29153 (vabaq_s32): Likewise.
29154 (vabaq_u8): Likewise.
29155 (vabaq_u16): Likewise.
29156 (vabaq_u32): Likewise.
29158 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29160 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
29161 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
29162 Change RTL pattern to match.
29164 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
29166 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
29167 * config/riscv/riscv-c.c (riscv-subset.h): New.
29168 (INCLUDE_STRING): Define.
29169 (riscv_cpu_cpp_builtins): Add new style architecture extension
29171 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
29172 (riscv_subset_list::end): New.
29173 (riscv_current_subset_list): New.
29175 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
29177 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
29178 Move to riscv-subset.h.
29179 (struct riscv_subset_t): Ditto.
29180 (class riscv_subset_list): Ditto.
29181 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
29182 from riscv-common.c.
29183 (struct riscv_subset_t): Ditto.
29184 (class riscv_subset_list): Ditto.
29185 * config/riscv/t-riscv ($(common_out_file)): Add file
29188 2021-01-07 Jakub Jelinek <jakub@redhat.com>
29191 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
29192 New define_insn patterns.
29194 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29196 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
29197 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
29198 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
29200 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29202 PR tree-optimization/98560
29203 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
29204 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
29206 (vec_cond_direct): Likewise argument 2.
29207 (vec_condu_direct, vec_condeq_direct): Delete.
29208 (expand_vect_cond_optab_fn): Rename to...
29209 (expand_vec_cond_optab_fn): ...this, replacing old macro.
29210 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
29211 (expand_vect_cond_mask_optab_fn): Rename to...
29212 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
29213 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
29215 (direct_vec_cond_optab_supported_p): Likewise.
29216 (direct_vec_condu_optab_supported_p): Delete.
29217 (direct_vec_condeq_optab_supported_p): Delete.
29218 * gimple-isel.cc: Include internal-fn.h.
29219 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
29222 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29224 PR tree-optimization/98560
29225 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
29226 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
29228 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
29230 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
29231 rotate_insn and optab code attributes.
29232 Update all uses to merged code attribute.
29233 * config/i386/sse.md: Update all uses to merged code attribute.
29234 * config/i386/mmx.md: Update all uses to merged code attribute.
29236 2021-01-07 Jakub Jelinek <jakub@redhat.com>
29238 PR tree-optimization/98568
29239 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
29240 (bswap_replace): Use it.
29242 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
29244 PR rtl-optimization/97978
29245 * lra-int.h (lra_hard_reg_split_p): New external.
29246 * lra.c (lra_hard_reg_split_p): New global.
29247 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
29248 * lra-assigns.c (lra_assign): Don't check allocation correctness
29249 after hard reg splitting.
29251 2021-01-06 Martin Sebor <msebor@redhat.com>
29254 * builtins.c (new_delete_mismatch_p): New overload.
29255 (new_delete_mismatch_p (tree, tree)): Call it.
29257 2021-01-06 Alexandre Oliva <oliva@adacore.com>
29259 * Makefile.in (T_GLIMITS_H): New.
29260 (stmp-int-hdrs): Depend on it, use it.
29261 * config/t-vxworks (T_GLIMITS_H): Override it.
29262 (vxw-glimits.h): New.
29264 2021-01-06 Richard Biener <rguenther@suse.de>
29266 PR tree-optimization/98513
29267 * value-range.cc (intersect_ranges): Compare the upper bounds
29268 for the expected relation.
29270 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
29273 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
29275 * doc/standards.texi (HSAIL): Remove section.
29277 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
29279 * configure: Re-generate.
29281 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29283 * doc/invoke.texi (-std=c++20): Adjust for the publication of
29284 ISO 14882:2020 standard.
29285 * doc/standards.texi: Likewise.
29287 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29289 PR tree-optimization/94802
29290 * expr.h (maybe_optimize_sub_cmp_0): Declare.
29291 * expr.c: Include tree-pretty-print.h and flags.h.
29292 (maybe_optimize_sub_cmp_0): New function.
29293 (do_store_flag): Use it.
29294 * cfgexpand.c (expand_gimple_cond): Likewise.
29296 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29298 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
29299 * rtlanal.c (simple_regno_set): Tweak description to clarify the
29302 2021-01-05 Richard Biener <rguenther@suse.de>
29304 PR tree-optimization/98516
29305 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
29306 lanes when materializing on a VEC_PERM node.
29307 (vectorizable_slp_permutation): Dump the permute properly.
29309 2021-01-05 Richard Biener <rguenther@suse.de>
29311 * tree-vect-slp.c (vect_slp_region): Move debug counter
29312 to cover individual subgraphs.
29314 2021-01-05 Richard Biener <rguenther@suse.de>
29316 PR tree-optimization/98428
29317 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
29318 vector lane extracts for loop vectorization.
29320 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29322 PR tree-optimization/98514
29323 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
29325 (operand_rank): Change type from hash_map<tree, long> to
29326 hash_map<tree, int64_t>.
29327 (phi_rank): Change return type from long to int64_t.
29328 (loop_carried_phi): Change block_rank variable type from long to
29330 (propagate_rank): Change return type, rank parameter type and
29331 op_rank variable type from long to int64_t.
29332 (find_operand_rank): Change return type from long to int64_t
29333 and change slot variable type from long * to int64_t *.
29334 (insert_operand_rank): Change rank parameter type from long to
29336 (get_rank): Change return type and rank variable type from long to
29337 int64_t. Use PRId64 instead of ld to print the rank.
29338 (init_reassoc): Change rank variable type from long to int64_t
29339 and adjust correspondingly bb_rank and operand_rank initialization.
29341 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29343 PR tree-optimization/96928
29344 * tree-ssa-phiopt.c (xor_replacement): New function.
29345 (tree_ssa_phiopt_worker): Call it.
29347 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29349 PR tree-optimization/96930
29350 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
29351 from narrower value which has the same type as 1 << B, perform
29352 the right shift on the narrower value followed by extension.
29354 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29356 PR tree-optimization/96239
29357 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
29359 (get_status_for_store_merging): Don't return BB_INVALID for blocks
29360 with potential bswap optimizable CONSTRUCTORs.
29361 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
29364 2021-01-05 Richard Biener <rguenther@suse.de>
29366 PR tree-optimization/98381
29367 * tree.c (vector_element_bits): Properly compute bool vector
29369 * tree-vect-loop.c (vectorizable_live_operation): Properly
29370 compute the last lane bit offset.
29372 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29375 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
29376 Clear the top 64 bytes of the input XMM register.
29377 (sse_cvttps2pi): Ditto.
29379 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29382 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
29384 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
29387 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
29390 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
29392 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
29393 (umaddsidi4_split): Likewise.
29395 2021-01-05 liuhongt <hongtao.liu@intel.com>
29398 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
29399 define_insn_and_split for zero_extend of subreg HI of pmovskb
29401 (*sse2_pmovskb_zexthisi): Add new combine splitters for
29402 zero_extend of not of subreg HI of pmovskb result.
29404 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29407 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
29409 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
29410 convert_memory_address to convert symbolic immediates to ptr_mode
29411 before forcing them to memory.
29413 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29415 PR rtl-optimization/97144
29416 * recog.c (constrain_operands): Initialize matching_operand
29417 for each alternative, rather than only doing it once.
29419 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29421 PR rtl-optimization/98403
29422 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
29423 why we don't remove call clobbers.
29424 (function_info::apply_changes_to_insn): Don't attempt to add
29425 call clobbers here.
29427 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29429 PR tree-optimization/98371
29430 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
29431 (vect_analyze_loop): If an epilogue loop appears to be cheaper
29432 than the main loop, re-analyze it as a main loop before adopting
29435 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29438 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
29439 * aclocal.m4, configure: Regenerate.
29440 * Makefile.in (NETLIBS): Define.
29441 (BACKEND): Remove $(CODYLIB).
29443 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29445 PR rtl-optimization/98334
29446 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
29447 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
29449 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
29451 * tree-inline.c (expand_call_inline): Restore input_location.
29452 Return result from recursive call.
29454 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29456 PR tree-optimization/95401
29457 * config/aarch64/aarch64-sve-builtins.cc
29458 (gimple_folder::load_store_cookie): Use bits rather than bytes
29459 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
29460 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
29461 * tree-vect-stmts.c (vectorizable_store): Likewise.
29462 (vectorizable_load): Likewise.
29464 2021-01-04 Richard Biener <rguenther@suse.de>
29466 PR tree-optimization/98308
29467 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
29470 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29472 PR tree-optimization/95771
29473 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
29474 with precision smaller than int's precision and types with precision
29475 twice as large as long long. Formatting fixes.
29477 2021-01-04 Richard Biener <rguenther@suse.de>
29479 PR tree-optimization/98464
29480 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
29481 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
29482 (process_bb): Adjust.
29484 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
29487 * doc/invoke.texi (-fsanitize=address): Fix wording describing
29488 clash with -fsanitize=hwaddress.
29490 2021-01-04 Richard Biener <rguenther@suse.de>
29492 PR tree-optimization/98282
29493 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
29494 invariants as VN_NARY.
29496 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29499 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
29500 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
29501 to handle zero operands.
29503 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29505 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
29506 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
29507 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
29510 2021-01-04 Richard Biener <rguenther@suse.de>
29512 PR tree-optimization/98393
29513 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
29514 when hitting the limit.
29516 2021-01-04 Richard Biener <rguenther@suse.de>
29518 PR tree-optimization/98291
29519 * tree-vect-loop.c (vectorizable_reduction): Bypass
29520 associativity check for SLP reductions with VF 1.
29522 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29524 PR tree-optimization/96782
29525 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
29527 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
29529 * collect-utils.c (collect_execute): Check dumppfx.
29530 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
29531 to collect_execute.
29532 (do_link): Add new parameter atsuffix.
29533 (main): Handle -dumpdir option. Skip one argument for
29534 -o, -isystem and -B options.
29535 * gcc.c (make_at_file): New helper function.
29536 (close_at_file): Use it.
29538 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29540 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
29541 Amend handling for LD64_VERSION fallback defaults.
29543 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29545 * config.gcc: Compute default version information
29546 from the configured target. Likewise defaults for
29548 * config/darwin10.h: Removed.
29549 * config/darwin12.h: Removed.
29550 * config/darwin9.h: Removed.
29551 * config/rs6000/darwin8.h: Removed.
29553 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29555 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
29557 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29559 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
29560 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
29562 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29564 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
29566 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
29568 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29570 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
29571 for the Darwin10 unwinder stub from here ...
29572 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
29574 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29576 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
29577 (ASM_DEBUG_SPEC):Only define if the assembler supports
29579 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
29580 (DARWIN_PREFER_DWARF): Define.
29581 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
29582 (DARWIN_PREFER_DWARF): Likewise
29583 (DSYMUTIL_SPEC): Likewise.
29584 (COLLECT_RUN_DSYMUTIL): Likewise.
29585 (ASM_DEBUG_SPEC): Likewise.
29586 (ASM_DEBUG_OPTION_SPEC): Likewise.
29588 2021-01-02 Jan Hubicka <jh@suse.cz>
29590 * cfg.c (free_block): ggc_free bb.
29592 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29594 * gcc.c (process_command): Update copyright notice dates.
29595 * gcov-dump.c (print_version): Ditto.
29596 * gcov.c (print_version): Ditto.
29597 * gcov-tool.c (print_version): Ditto.
29598 * gengtype.c (create_file): Ditto.
29599 * doc/cpp.texi: Bump @copying's copyright year.
29600 * doc/cppinternals.texi: Ditto.
29601 * doc/gcc.texi: Ditto.
29602 * doc/gccint.texi: Ditto.
29603 * doc/gcov.texi: Ditto.
29604 * doc/install.texi: Ditto.
29605 * doc/invoke.texi: Ditto.
29607 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29609 * ChangeLog-2020: Rotate ChangeLog. New file.
29612 Copyright (C) 2021 Free Software Foundation, Inc.
29614 Copying and distribution of this file, with or without modification,
29615 are permitted in any medium without royalty provided the copyright
29616 notice and this notice are preserved.