]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
i386: Fix k*shift* intrinsics [PR93673]
[gcc.git] / gcc / ChangeLog
1 2020-02-13 Jakub Jelinek <jakub@redhat.com>
2
3 PR target/93673
4 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
5 use const_0_to_255_operand predicate instead of immediate_operand.
6 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
7 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
8 vgf2p8affineinvqb_<mode><mask_name>,
9 vgf2p8affineqb_<mode><mask_name>): Drop mode from
10 const_0_to_255_operand predicated operands.
11
12 2020-02-12 Jeff Law <law@redhat.com>
13
14 * config/h8300/h8300.md (comparison shortening peepholes): Use
15 a mode iterator to merge the HImode and SImode peepholes.
16
17 2020-02-12 Jakub Jelinek <jakub@redhat.com>
18
19 PR middle-end/93663
20 * real.c (is_even): Make static. Function comment fix.
21 (is_halfway_below): Make static, don't assert R is not inf/nan,
22 instead return false for those. Small formatting fixes.
23
24 2020-02-12 Martin Sebor <msebor@redhat.com>
25
26 PR middle-end/93646
27 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
28 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
29 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
30 (strlen_check_and_optimize_call): Adjust callee name.
31
32 2020-02-12 Jeff Law <law@redhat.com>
33
34 * config/h8300/h8300.md (comparison shortening peepholes): Drop
35 (and (xor)) variant. Combine other two into single peephole.
36
37 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
38
39 PR rtl-optimization/93565
40 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
41
42 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
43
44 * config/aarch64/aarch64-simd.md
45 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
46 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
47 generating separate ADDV and zero_extend patterns.
48 * config/aarch64/iterators.md (VDQV_E): New iterator.
49
50 2020-02-12 Jeff Law <law@redhat.com>
51
52 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
53 expanders, splits, etc.
54 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
55 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
56 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
57 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
58 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
59 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
60 function prototype.
61 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
62
63 2020-02-12 Jakub Jelinek <jakub@redhat.com>
64
65 PR target/93670
66 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
67 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
68 TARGET_AVX512DQ from condition.
69 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
70 instead of <mask_mode512bit_condition> in condition. If
71 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
72 vextract*32x8.
73 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
74 from condition.
75
76 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
77
78 PR target/91052
79 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
80
81 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
82
83 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
84 where strlen is more legible.
85 (rs6000_builtin_vectorized_libmass): Ditto.
86 (rs6000_print_options_internal): Ditto.
87
88 2020-02-11 Martin Sebor <msebor@redhat.com>
89
90 PR tree-optimization/93683
91 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
92
93 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
94
95 * config/rs6000/predicates.md (cint34_operand): Rename the
96 -mprefixed-addr option to be -mprefixed.
97 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
98 the -mprefixed-addr option to be -mprefixed.
99 (OTHER_FUTURE_MASKS): Likewise.
100 (POWERPC_MASKS): Likewise.
101 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
102 the -mprefixed-addr option to be -mprefixed. Change error
103 messages to refer to -mprefixed.
104 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
105 -mprefixed.
106 (rs6000_legitimate_offset_address_p): Likewise.
107 (rs6000_mode_dependent_address): Likewise.
108 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
109 "-mprefixed" for target attributes and pragmas.
110 (address_to_insn_form): Rename the -mprefixed-addr option to be
111 -mprefixed.
112 (rs6000_adjust_insn_length): Likewise.
113 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
114 -mprefixed-addr option to be -mprefixed.
115 (ASM_OUTPUT_OPCODE): Likewise.
116 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
117 -mprefixed-addr option to be -mprefixed.
118 * config/rs6000/rs6000.opt (-mprefixed): Rename the
119 -mprefixed-addr option to be prefixed. Change the option from
120 being undocumented to being documented.
121 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
122 -mprefixed option. Update the -mpcrel documentation to mention
123 -mprefixed.
124
125 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
126
127 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
128 including FIRST_PSEUDO_REGISTER - 1.
129 * ira-color.c (print_hard_reg_set): Ditto.
130
131 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
132
133 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
134 (USTERNOP_QUALIFIERS): New define.
135 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
136 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
137 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
138 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
139 * config/arm/arm_neon.h (vusdot_s32): New.
140 (vusdot_lane_s32): New.
141 (vusdotq_lane_s32): New.
142 (vsudot_lane_s32): New.
143 (vsudotq_lane_s32): New.
144 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
145 * config/arm/iterators.md (DOTPROD_I8MM): New.
146 (sup, opsuffix): Add <us/su>.
147 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
148 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
149
150 2020-02-11 Richard Biener <rguenther@suse.de>
151
152 PR tree-optimization/93661
153 PR tree-optimization/93662
154 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
155 tree_to_poly_int64.
156 * tree-sra.c (get_access_for_expr): Likewise.
157
158 2020-02-10 Jakub Jelinek <jakub@redhat.com>
159
160 PR target/93637
161 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
162 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
163 Change condition from TARGET_AVX2 to TARGET_AVX.
164
165 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
166
167 PR other/93641
168 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
169 argument of strncmp.
170
171 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
172
173 Try to generate zero-based comparisons.
174 * config/cris/cris.c (cris_reduce_compare): New function.
175 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
176 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
177 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
178
179 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
180
181 PR target/91913
182 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
183 in Thumb state and also as a destination in Arm state. Add T16
184 variants.
185
186 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
187
188 * md.texi (Define Subst): Match closing paren in example.
189
190 2020-02-10 Jakub Jelinek <jakub@redhat.com>
191
192 PR target/58218
193 PR other/93641
194 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
195 arguments of strncmp.
196
197 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
198
199 PR ipa/93203
200 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
201 but different source value.
202 (adjust_callers_for_value_intersection): New function.
203 (gather_edges_for_value): Adjust order of callers to let a
204 non-self-recursive caller be the first element.
205 (self_recursive_pass_through_p): Add a new parameter "simple", and
206 check generalized self-recursive pass-through jump function.
207 (self_recursive_agg_pass_through_p): Likewise.
208 (find_more_scalar_values_for_callers_subset): Compute value from
209 pass-through jump function for self-recursive.
210 (intersect_with_plats): Cleanup previous implementation code for value
211 itersection with self-recursive call edge.
212 (intersect_with_agg_replacements): Likewise.
213 (intersect_aggregates_with_edge): Deduce value from pass-through jump
214 function for self-recursive call edge. Cleanup previous implementation
215 code for value intersection with self-recursive call edge.
216 (decide_whether_version_node): Remove dead callers and adjust order
217 to let a non-self-recursive caller be the first element.
218
219 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
220
221 * recog.c: Move pass_split_before_sched2 code in front of
222 pass_split_before_regstack.
223 (pass_data_split_before_sched2): Rename pass to split3 from split4.
224 (pass_data_split_before_regstack): Rename pass to split4 from split3.
225 (rest_of_handle_split_before_sched2): Remove.
226 (pass_split_before_sched2::execute): Unconditionally call
227 split_all_insns.
228 (enable_split_before_sched2): New function.
229 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
230 (pass_split_before_regstack::gate): Ditto.
231 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
232 Update name check for renamed split4 pass.
233 * config/sh/sh.c (register_sh_passes): Update pass insertion
234 point for renamed split4 pass.
235
236 2020-02-09 Jakub Jelinek <jakub@redhat.com>
237
238 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
239 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
240 copying them around between host and target.
241
242 2020-02-08 Andrew Pinski <apinski@marvell.com>
243
244 PR target/91927
245 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
246 STRICT_ALIGNMENT also.
247
248 2020-02-08 Jim Wilson <jimw@sifive.com>
249
250 PR target/93532
251 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
252
253 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
254 Jakub Jelinek <jakub@redhat.com>
255
256 PR target/65782
257 * config/i386/i386.h (CALL_USED_REGISTERS): Make
258 xmm16-xmm31 call-used even in 64-bit ms-abi.
259
260 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
261
262 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
263 (simd_ummla, simd_usmmla): Likewise.
264 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
265 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
266 (vusmmlaq_s32): New.
267
268 2020-02-07 Richard Biener <rguenther@suse.de>
269
270 PR middle-end/93519
271 * tree-inline.c (fold_marked_statements): Do a PRE walk,
272 skipping unreachable regions.
273 (optimize_inline_calls): Skip folding stmts when we didn't
274 inline.
275
276 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
277
278 PR target/85667
279 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
280 Don't return aggregates with only SFmode and DFmode in SSE
281 register.
282 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
283
284 2020-02-07 Jakub Jelinek <jakub@redhat.com>
285
286 PR target/93122
287 * config/rs6000/rs6000-logue.c
288 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
289 if it fails, move rs into end_addr and retry. Add
290 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
291 the insn pattern doesn't describe well what exactly happens to
292 dwarf2cfi.c.
293
294 PR target/93594
295 * config/i386/predicates.md (avx_identity_operand): Remove.
296 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
297 (avx_<castmode><avxsizesuffix>_<castmode>,
298 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
299 a VEC_CONCAT of the operand and UNSPEC_CAST.
300 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
301 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
302 UNSPEC_CAST.
303
304 PR target/93611
305 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
306 recog_data.insn if distance_non_agu_define changed it.
307
308 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
309
310 PR target/93569
311 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
312 we only had X-FORM (reg+reg) addressing for vectors. Also before
313 ISA 3.0, we only had X-FORM addressing for scalars in the
314 traditional Altivec registers.
315
316 2020-02-06 <zhongyunde@huawei.com>
317 Vladimir Makarov <vmakarov@redhat.com>
318
319 PR rtl-optimization/93561
320 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
321 hard register range.
322
323 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
324
325 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
326 attribute.
327
328 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
329
330 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
331 where the low and the high 32 bits are equal to each other specially,
332 with an rldimi instruction.
333
334 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
335
336 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
337
338 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
339
340 * config/arm/arm-tables.opt: Regenerate.
341
342 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
343
344 PR target/87763
345 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
346 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
347 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
348
349 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
350
351 PR rtl-optimization/87763
352 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
353
354 2020-02-06 Delia Burduv <delia.burduv@arm.com>
355
356 * config/aarch64/aarch64-simd-builtins.def
357 (bfmlaq): New built-in function.
358 (bfmlalb): New built-in function.
359 (bfmlalt): New built-in function.
360 (bfmlalb_lane): New built-in function.
361 (bfmlalt_lane): New built-in function.
362 * config/aarch64/aarch64-simd.md
363 (aarch64_bfmmlaqv4sf): New pattern.
364 (aarch64_bfmlal<bt>v4sf): New pattern.
365 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
366 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
367 (vbfmlalbq_f32): New intrinsic.
368 (vbfmlaltq_f32): New intrinsic.
369 (vbfmlalbq_lane_f32): New intrinsic.
370 (vbfmlaltq_lane_f32): New intrinsic.
371 (vbfmlalbq_laneq_f32): New intrinsic.
372 (vbfmlaltq_laneq_f32): New intrinsic.
373 * config/aarch64/iterators.md (BF_MLA): New int iterator.
374 (bt): New int attribute.
375
376 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
377
378 * config/i386/i386.md (*pushtf): Emit "#" instead of
379 calling gcc_unreachable in insn output.
380 (*pushxf): Ditto.
381 (*pushdf): Ditto.
382 (*pushsf_rex64): Ditto for alternatives other than 1.
383 (*pushsf): Ditto for alternatives other than 1.
384
385 2020-02-06 Martin Liska <mliska@suse.cz>
386
387 PR gcov-profile/91971
388 PR gcov-profile/93466
389 * coverage.c (coverage_init): Revert mangling of
390 path into filename. It can lead to huge filename length.
391 Creation of subfolders seem more natural.
392
393 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
394
395 PR target/93300
396 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
397 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
398 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
399
400 2020-02-06 Jakub Jelinek <jakub@redhat.com>
401
402 PR target/93594
403 * config/i386/predicates.md (avx_identity_operand): New predicate.
404 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
405 define_insn_and_split.
406
407 PR libgomp/93515
408 * omp-low.c (use_pointer_for_field): For nested constructs, also
409 look for map clauses on target construct.
410 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
411 taskreg_nesting_level.
412
413 PR libgomp/93515
414 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
415 shared clause, call omp_notice_variable on outer context if any.
416
417 2020-02-05 Jason Merrill <jason@redhat.com>
418
419 PR c++/92003
420 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
421 non-zero address even if weak and not yet defined.
422
423 2020-02-05 Martin Sebor <msebor@redhat.com>
424
425 PR tree-optimization/92765
426 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
427 * tree-ssa-strlen.c (compute_string_length): Remove.
428 (determine_min_objsize): Remove.
429 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
430 Avoid using type size as the upper bound on string length.
431 (handle_builtin_string_cmp): Add an argument. Adjust.
432 (strlen_check_and_optimize_call): Pass additional argument to
433 handle_builtin_string_cmp.
434
435 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
436
437 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
438 (*pushdi2_rex64 peephole2): Unconditionally split after
439 epilogue_completed.
440 (*ashl<mode>3_doubleword): Ditto.
441 (*<shift_insn><mode>3_doubleword): Ditto.
442
443 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
444
445 PR target/93568
446 * config/rs6000/rs6000.c (get_vector_offset): Fix
447
448 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
449
450 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
451
452 2020-02-05 David Malcolm <dmalcolm@redhat.com>
453
454 * doc/analyzer.texi
455 (Special Functions for Debugging the Analyzer): Update description
456 of __analyzer_dump_exploded_nodes.
457
458 2020-02-05 Jakub Jelinek <jakub@redhat.com>
459
460 PR target/92190
461 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
462 include sets and not clobbers in the vzeroupper pattern.
463 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
464 the parallel has 17 (64-bit) or 9 (32-bit) elts.
465 (*avx_vzeroupper_1): New define_insn_and_split.
466
467 PR target/92190
468 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
469 don't run when !optimize.
470 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
471 when !optimize.
472
473 2020-02-05 Richard Biener <rguenther@suse.de>
474
475 PR middle-end/90648
476 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
477 checks before matching calls.
478
479 2020-02-05 Jakub Jelinek <jakub@redhat.com>
480
481 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
482 function comment typo.
483
484 PR middle-end/93555
485 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
486 simd_clone_create failed when i == 0, adjust clone->nargs by
487 clone->inbranch.
488
489 2020-02-05 Martin Liska <mliska@suse.cz>
490
491 PR c++/92717
492 * doc/invoke.texi: Document that one should
493 not combine ASLR and -fpch.
494
495 2020-02-04 Richard Biener <rguenther@suse.de>
496
497 PR tree-optimization/93538
498 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
499
500 2020-02-04 Richard Biener <rguenther@suse.de>
501
502 PR tree-optimization/91123
503 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
504 (vn_walk_cb_data::last_vuse): New member.
505 (vn_walk_cb_data::saved_operands): Likewsie.
506 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
507 (vn_walk_cb_data::push_partial_def): Use finish.
508 (vn_reference_lookup_2): Update last_vuse and use finish if
509 we've saved operands.
510 (vn_reference_lookup_3): Use finish and update calls to
511 push_partial_defs everywhere. When translating through
512 memcpy or aggregate copies save off operands and alias-set.
513 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
514 operation for redundant store removal.
515
516 2020-02-04 Richard Biener <rguenther@suse.de>
517
518 PR tree-optimization/92819
519 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
520 generating more stmts than before.
521
522 2020-02-04 Martin Liska <mliska@suse.cz>
523
524 * config/arm/arm.c (arm_gen_far_branch): Move the function
525 outside of selftests.
526
527 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
528
529 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
530 function to adjust PC-relative vector addresses.
531 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
532 handle vectors with PC-relative addresses.
533
534 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
535
536 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
537 reference.
538 (hard_reg_and_mode_to_addr_mask): Delete.
539 (rs6000_adjust_vec_address): If the original vector address
540 was REG+REG or REG+OFFSET and the element is not zero, do the add
541 of the elements in the original address before adding the offset
542 for the vector element. Use address_to_insn_form to validate the
543 address using the register being loaded, rather than guessing
544 whether the address is a DS-FORM or DQ-FORM address.
545
546 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
547
548 * config/rs6000/rs6000.c (get_vector_offset): New helper function
549 to calculate the offset in memory from the start of a vector of a
550 particular element. Add code to keep the element number in
551 bounds if the element number is variable.
552 (rs6000_adjust_vec_address): Move calculation of offset of the
553 vector element to get_vector_offset.
554 (rs6000_split_vec_extract_var): Do not do the initial AND of
555 element here, move the code to get_vector_offset.
556
557 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
558
559 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
560 gcc_asserts.
561
562 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
563
564 * config/rs6000/constraints.md: Improve documentation.
565
566 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
567
568 PR target/93548
569 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
570 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
571
572 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
573
574 * config.gcc: Remove "carrizo" support.
575 * config/gcn/gcn-opts.h (processor_type): Likewise.
576 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
577 * config/gcn/gcn.opt (gpu_type): Likewise.
578 * config/gcn/t-omp-device: Likewise.
579
580 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
581
582 PR target/91816
583 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
584 * config/arm/arm.c (arm_gen_far_branch): New function
585 arm_gen_far_branch.
586 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
587
588 2020-02-03 Julian Brown <julian@codesourcery.com>
589 Tobias Burnus <tobias@codesourcery.com>
590
591 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
592
593 2020-02-03 Jakub Jelinek <jakub@redhat.com>
594
595 PR target/93533
596 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
597 valid RTL to sum up the lowest and second lowest bytes of the popcnt
598 result.
599
600 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
601
602 PR rtl-optimization/91333
603 * ira-color.c (struct allocno_color_data): Add member
604 hard_reg_prefs.
605 (init_allocno_threads): Set the member up.
606 (bucket_allocno_compare_func): Add compare hard reg
607 prefs.
608
609 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
610
611 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
612
613 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
614 * config.in: Regenerated.
615 * configure: Regenerated.
616 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
617 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
618 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
619
620 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
621
622 * configure: Regenerate.
623
624 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
625
626 PR rtl-optimization/91333
627 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
628 reg preferences comparison up.
629
630 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
631
632 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
633 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
634 aarch64-sve-builtins-base.h.
635 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
636 aarch64-sve-builtins-base.cc.
637 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
638 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
639 (svcvtnt): Declare.
640 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
641 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
642 (svcvtnt): New functions.
643 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
644 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
645 (svcvtnt): New functions.
646 (svcvt): Add a form that converts f32 to bf16.
647 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
648 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
649 Declare.
650 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
651 Treat B as bfloat16_t.
652 (ternary_bfloat_lane_base): New class.
653 (ternary_bfloat_def): Likewise.
654 (ternary_bfloat): New shape.
655 (ternary_bfloat_lane_def): New class.
656 (ternary_bfloat_lane): New shape.
657 (ternary_bfloat_lanex2_def): New class.
658 (ternary_bfloat_lanex2): New shape.
659 (ternary_bfloat_opt_n_def): New class.
660 (ternary_bfloat_opt_n): New shape.
661 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
662 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
663 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
664 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
665 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
666 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
667 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
668 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
669 the pattern off the narrow mode instead of the wider one.
670 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
671 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
672 (sve_fp_op): Handle them.
673 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
674 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
675
676 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
677
678 * config/aarch64/arm_sve.h: Include arm_bf16.h.
679 * config/aarch64/aarch64-modes.def (BF): Move definition before
680 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
681 (SVE_MODES): Handle BF modes.
682 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
683 BF modes.
684 (aarch64_full_sve_mode): Likewise.
685 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
686 and VNx32BF.
687 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
688 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
689 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
690 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
691 new SVE BF modes.
692 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
693 type_class_index.
694 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
695 (TYPES_all_data): Add bf16.
696 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
697 (register_tuple_type): Increase buffer size.
698 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
699 (bf16): New type suffix.
700 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
701 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
702 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
703 Change type from all_data to all_arith.
704 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
705 (svminp): Likewise.
706
707 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
708 Matthew Malcomson <matthew.malcomson@arm.com>
709 Richard Sandiford <richard.sandiford@arm.com>
710
711 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
712 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
713 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
714 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
715 __ARM_FEATURE_MATMUL_FP64.
716 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
717 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
718 be disabled at the same time.
719 (f32mm): New extension.
720 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
721 (AARCH64_FL_F64MM): Bump to the next bit up.
722 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
723 (TARGET_SVE_F64MM): New macros.
724 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
725 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
726 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
727 (UNSPEC_ZIP2Q): New unspeccs.
728 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
729 (optab, sur, perm_insn): Handle the new unspecs.
730 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
731 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
732 TARGET_SVE_F64MM instead of separate tests.
733 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
734 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
735 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
736 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
737 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
738 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
739 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
740 (TYPES_s_signed): New macro.
741 (TYPES_s_integer): Use it.
742 (TYPES_d_float): New macro.
743 (TYPES_d_data): Use it.
744 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
745 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
746 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
747 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
748 (svmmla): New shape.
749 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
750 template parameters.
751 (ternary_resize2_lane_base): Likewise.
752 (ternary_resize2_base): New class.
753 (ternary_qq_lane_base): Likewise.
754 (ternary_intq_uintq_lane_def): Likewise.
755 (ternary_intq_uintq_lane): New shape.
756 (ternary_intq_uintq_opt_n_def): New class
757 (ternary_intq_uintq_opt_n): New shape.
758 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
759 (ternary_uintq_intq_def): New class.
760 (ternary_uintq_intq): New shape.
761 (ternary_uintq_intq_lane_def): New class.
762 (ternary_uintq_intq_lane): New shape.
763 (ternary_uintq_intq_opt_n_def): New class.
764 (ternary_uintq_intq_opt_n): New shape.
765 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
766 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
767 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
768 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
769 Generalize to...
770 (svdotprod_lane_impl): ...this new class.
771 (svmmla_impl, svusdot_impl): New classes.
772 (svdot_lane): Update to use svdotprod_lane_impl.
773 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
774 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
775 functions.
776 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
777 function, with no types defined.
778 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
779 AARCH64_FL_I8MM functions.
780 (svmmla): New AARCH64_FL_F32MM function.
781 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
782 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
783 AARCH64_FL_F64MM function.
784 (REQUIRED_EXTENSIONS):
785
786 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
787
788 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
789 alternative only.
790
791 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
792
793 * config/i386/i386.md (*movoi_internal_avx): Do not check for
794 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
795 (*movti_internal): Do not check for
796 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
797 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
798 just after check for TARGET_AVX.
799 (*movdf_internal): Ditto.
800 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
801 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
802 * config/i386/sse.md (mov<mode>_internal): Only check
803 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
804 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
805 (<sse>_andnot<mode>3<mask_name>): Move check for
806 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
807 (<code><mode>3<mask_name>): Ditto.
808 (*andnot<mode>3): Ditto.
809 (*andnottf3): Ditto.
810 (*<code><mode>3): Ditto.
811 (*<code>tf3): Ditto.
812 (*andnot<VI:mode>3): Remove
813 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
814 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
815 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
816 (sse4_1_blendv<ssemodesuffix>): Ditto.
817 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
818 Explain that tune applies to 128bit instructions only.
819
820 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
821
822 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
823 to definition of hsa_kernel_description. Parse assembly to find SGPR
824 and VGPR count of kernel and store in hsa_kernel_description.
825
826 2020-01-31 Tamar Christina <tamar.christina@arm.com>
827
828 PR rtl-optimization/91838
829 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
830 to truncate if allowed or reject combination.
831
832 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
833
834 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
835 (find_inv_vars_cb): Likewise.
836
837 2020-01-31 David Malcolm <dmalcolm@redhat.com>
838
839 * calls.c (special_function_p): Split out the check for DECL_NAME
840 being non-NULL and fndecl being extern at file scope into a
841 new maybe_special_function_p and call it. Drop check for fndecl
842 being non-NULL that was after a usage of DECL_NAME (fndecl).
843 * tree.h (maybe_special_function_p): New inline function.
844
845 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
846
847 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
848 (mask_gather_load<mode>): ... here, and zero-initialize the
849 destination.
850 (maskload<mode>di): Zero-initialize the destination.
851 * config/gcn/gcn.c:
852
853 2020-01-30 David Malcolm <dmalcolm@redhat.com>
854
855 PR analyzer/93356
856 * doc/analyzer.texi (Limitations): Note that constraints on
857 floating-point values are currently ignored.
858
859 2020-01-30 Jakub Jelinek <jakub@redhat.com>
860
861 PR lto/93384
862 * symtab.c (symtab_node::noninterposable_alias): If localalias
863 already exists, but is not usable, append numbers after it until
864 a unique name is found. Formatting fix.
865
866 PR middle-end/93505
867 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
868 rotate counts.
869
870 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
871
872 * config/gcn/gcn.c (print_operand): Handle LTGT.
873 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
874
875 2020-01-30 Richard Biener <rguenther@suse.de>
876
877 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
878 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
879
880 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
881
882 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
883 without a DECL in .data.rel.ro.local.
884
885 2020-01-30 Jakub Jelinek <jakub@redhat.com>
886
887 PR target/93494
888 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
889 returned.
890
891 PR target/91824
892 * config/i386/sse.md
893 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
894 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
895 any_extend code iterator instead of always zero_extend.
896 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
897 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
898 Use any_extend code iterator instead of always zero_extend.
899 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
900 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
901 Use any_extend code iterator instead of always zero_extend.
902 (*sse2_pmovmskb_ext): New define_insn.
903 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
904
905 PR target/91824
906 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
907 (*popcountsi2_zext_falsedep): New define_insn.
908
909 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
910
911 * config.in: Regenerated.
912 * configure: Regenerated.
913
914 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
915
916 PR bootstrap/93409
917 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
918 LLVM's assembler changed the default in version 9.
919
920 2020-01-24 Jeff Law <law@redhat.com>
921
922 PR tree-optimization/89689
923 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
924
925 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
926
927 Revert:
928
929 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
930
931 PR rtl-optimization/87763
932 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
933 simplification to handle subregs as well as bare regs.
934 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
935
936 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
937
938 PR target/93221
939 * ira.c (ira): Revert use of simplified LRA algorithm.
940
941 2020-01-29 Martin Jambor <mjambor@suse.cz>
942
943 PR tree-optimization/92706
944 * tree-sra.c (struct access): Fields first_link, last_link,
945 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
946 next_rhs_queued and grp_rhs_queued respectively, new fields
947 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
948 (struct assign_link): Field next renamed to next_rhs, new field
949 next_lhs. Updated comment.
950 (work_queue_head): Renamed to rhs_work_queue_head.
951 (lhs_work_queue_head): New variable.
952 (add_link_to_lhs): New function.
953 (relink_to_new_repr): Also relink LHS lists.
954 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
955 (add_access_to_lhs_work_queue): New function.
956 (pop_access_from_work_queue): Renamed to
957 pop_access_from_rhs_work_queue.
958 (pop_access_from_lhs_work_queue): New function.
959 (build_accesses_from_assign): Also add links to LHS lists and to LHS
960 work_queue.
961 (child_would_conflict_in_lacc): Renamed to
962 child_would_conflict_in_acc. Adjusted parameter names.
963 (create_artificial_child_access): New parameter set_grp_read, use it.
964 (subtree_mark_written_and_enqueue): Renamed to
965 subtree_mark_written_and_rhs_enqueue.
966 (propagate_subaccesses_across_link): Renamed to
967 propagate_subaccesses_from_rhs.
968 (propagate_subaccesses_from_lhs): New function.
969 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
970 RHSs.
971
972 2020-01-29 Martin Jambor <mjambor@suse.cz>
973
974 PR tree-optimization/92706
975 * tree-sra.c (struct access): Adjust comment of
976 grp_total_scalarization.
977 (find_access_in_subtree): Look for single children spanning an entire
978 access.
979 (scalarizable_type_p): Allow register accesses, adjust callers.
980 (completely_scalarize): Remove function.
981 (scalarize_elem): Likewise.
982 (create_total_scalarization_access): Likewise.
983 (sort_and_splice_var_accesses): Do not track total scalarization
984 flags.
985 (analyze_access_subtree): New parameter totally, adjust to new meaning
986 of grp_total_scalarization.
987 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
988 (can_totally_scalarize_forest_p): New function.
989 (create_total_scalarization_access): Likewise.
990 (create_total_access_and_reshape): Likewise.
991 (total_should_skip_creating_access): Likewise.
992 (totally_scalarize_subtree): Likewise.
993 (analyze_all_variable_accesses): Perform total scalarization after
994 subaccess propagation using the new functions above.
995 (initialize_constant_pool_replacements): Output initializers by
996 traversing the access tree.
997
998 2020-01-29 Martin Jambor <mjambor@suse.cz>
999
1000 * tree-sra.c (verify_sra_access_forest): New function.
1001 (verify_all_sra_access_forests): Likewise.
1002 (create_artificial_child_access): Set parent.
1003 (analyze_all_variable_accesses): Call the verifier.
1004
1005 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1006
1007 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1008 if called on indirect edge.
1009 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1010 speculative call if needed.
1011
1012 2020-01-29 Richard Biener <rguenther@suse.de>
1013
1014 PR tree-optimization/93428
1015 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1016 permutation when the load node is created.
1017 (vect_analyze_slp_instance): Re-use it here.
1018
1019 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1020
1021 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1022
1023 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1024
1025 PR rtl-optimization/93272
1026 * ira-lives.c (process_out_of_region_eh_regs): New function.
1027 (process_bb_node_lives): Call it.
1028
1029 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1030
1031 * coverage.c (read_counts_file): Make error message lowercase.
1032
1033 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1034
1035 * profile-count.c (profile_quality_display_names): Fix ordering.
1036
1037 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1038
1039 PR lto/93318
1040 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1041 hash only when edge is first within the sequence.
1042 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1043 (symbol_table::create_edge): Do not set target_prob.
1044 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1045 the call site hash.
1046 (cgraph_edge::make_speculative): Drop target_prob parameter.
1047 (cgraph_edge::speculative_call_info): Remove.
1048 (cgraph_edge::first_speculative_call_target): New member function.
1049 (update_call_stmt_hash_for_removing_direct_edge): New function.
1050 (cgraph_edge::resolve_speculation): Rewrite to new API.
1051 (cgraph_edge::speculative_call_for_target): New member function.
1052 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1053 multiple speculation targets.
1054 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1055 of profile.
1056 (verify_speculative_call): Verify that targets form an interval.
1057 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1058 (cgraph_edge::first_speculative_call_target): New member function.
1059 (cgraph_edge::next_speculative_call_target): New member function.
1060 (cgraph_edge::speculative_call_target_ref): New member function.
1061 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1062 (cgraph_edge): Remove target_prob.
1063 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1064 Fix handling of speculative calls.
1065 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1066 * ipa-fnsummary.c (analyze_function_body): Likewise.
1067 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1068 * ipa-profile.c (dump_histogram): Fix formating.
1069 (ipa_profile_generate_summary): Watch for overflows.
1070 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1071 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1072 (update_indirect_edges_after_inlining): Update to new API.
1073 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1074 profiles.
1075 * profile-count.h: (profile_probability::adjusted): New.
1076 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1077 updating of profile.
1078 * value-prof.c (gimple_ic_transform): Rename to ...
1079 (dump_ic_profile): ... this one; update dumping.
1080 (stream_in_histogram_value): Fix formating.
1081 (gimple_value_profile_transformations): Update.
1082
1083 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1084
1085 PR target/91461
1086 * config/i386/i386.md (*movoi_internal_avx): Remove
1087 TARGET_SSE_TYPELESS_STORES check.
1088 (*movti_internal): Prefer TARGET_AVX over
1089 TARGET_SSE_TYPELESS_STORES.
1090 (*movtf_internal): Likewise.
1091 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1092 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1093 from TARGET_SSE_TYPELESS_STORES.
1094
1095 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1096
1097 * diagnostic-core.h (warning_at): Rename overload to...
1098 (warning_meta): ...this.
1099 (emit_diagnostic_valist): Delete decl of overload taking
1100 diagnostic_metadata.
1101 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1102 (warning_at): Rename overload taking diagnostic_metadata to...
1103 (warning_meta): ...this.
1104
1105 2020-01-28 Richard Biener <rguenther@suse.de>
1106
1107 PR tree-optimization/93439
1108 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1109 * tree-cfg.c (move_sese_region_to_fn): ... here.
1110 (verify_types_in_gimple_reference): Verify used cliques are
1111 tracked.
1112
1113 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1114
1115 PR target/91399
1116 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1117 argument of a pointer to struct gcc_options and pass it to
1118 parse_mtune_ctrl_str.
1119 (ix86_function_specific_restore): Pass opts to
1120 set_ix86_tune_features.
1121 (ix86_option_override_internal): Likewise.
1122 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1123 gcc_options and use it for x_ix86_tune_ctrl_string.
1124
1125 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1126
1127 PR rtl-optimization/87763
1128 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1129 simplification to handle subregs as well as bare regs.
1130 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1131
1132 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1133
1134 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1135 for reduction chains that (now) include a call.
1136
1137 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1138
1139 PR tree-optimization/92822
1140 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1141 out the don't-care elements of a vector whose significant elements
1142 are duplicates, make the don't-care elements duplicates too.
1143
1144 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1145
1146 PR tree-optimization/93434
1147 * tree-predcom.c (split_data_refs_to_components): Record which
1148 components have had aliasing loads removed. Prevent store-store
1149 commoning for all such components.
1150
1151 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1152
1153 PR target/93418
1154 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1155 -1 or is_vshift is true, use new_vector with number of elts npatterns
1156 rather than new_unary_operation.
1157
1158 PR tree-optimization/93454
1159 * gimple-fold.c (fold_array_ctor_reference): Perform
1160 elt_size.to_uhwi () just once, instead of calling it in every
1161 iteration. Punt if that value is above size of the temporary
1162 buffer. Decrease third native_encode_expr argument when
1163 bufoff + elt_sz is above size of buf.
1164
1165 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1166
1167 * config/mips/mips.c (mips_declare_object_name)
1168 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1169
1170 2020-01-27 Martin Liska <mliska@suse.cz>
1171
1172 PR gcov-profile/93403
1173 * tree-profile.c (gimple_init_gcov_profiler): Generate
1174 both __gcov_indirect_call_profiler_v4 and
1175 __gcov_indirect_call_profiler_v4_atomic.
1176
1177 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1178
1179 PR target/92822
1180 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1181 expander.
1182 (@aarch64_split_simd_mov<mode>): Use it.
1183 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1184 Leave the vec_extract patterns to handle 2-element vectors.
1185 (aarch64_simd_mov_from_<mode>high): Likewise.
1186 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1187 (vec_extractv2dfv1df): Likewise.
1188
1189 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1190
1191 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1192 jump conditions for *compare_condjump<GPI:mode>.
1193
1194 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1195
1196 PR analyzer/93276
1197 * digraph.cc (test_edge::test_edge): Specify template for base
1198 class initializer.
1199
1200 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1201
1202 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1203
1204 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1205
1206 * config/arc/arc-protos.h (gen_mlo): Remove.
1207 (gen_mhi): Likewise.
1208 * config/arc/arc.c (AUX_MULHI): Define.
1209 (arc_must_save_reister): Special handling for r58/59.
1210 (arc_compute_frame_size): Consider mlo/mhi registers.
1211 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1212 paramter is true.
1213 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1214 mlo/mhi name selection.
1215 (arc_restore_callee_saves): Don't early restore blink when ISR.
1216 (arc_expand_prologue): Add mlo/mhi saving.
1217 (arc_expand_epilogue): Add mlo/mhi restoring.
1218 (gen_mlo): Remove.
1219 (gen_mhi): Remove.
1220 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1221 numbering when MUL64 option is used.
1222 (DWARF2_FRAME_REG_OUT): Define.
1223 * config/arc/arc.md (arc600_stall): New pattern.
1224 (VUNSPEC_ARC_ARC600_STALL): Define.
1225 (mulsi64): Use correct mlo/mhi registers.
1226 (mulsi_600): Clean it up.
1227 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1228 TARGET_BIG_ENDIAN.
1229 (mhi_operand): Likewise.
1230
1231 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1232 Petro Karashchenko <petro.karashchenko@ring.com>
1233
1234 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1235 attributes if needed.
1236 (prepare_move_operands): Generate special unspec instruction for
1237 direct access.
1238 (arc_isuncached_mem_p): Propagate uncached attribute to each
1239 structure member.
1240 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1241 (VUNSPEC_ARC_STDI): Likewise.
1242 (ALLI): New mode iterator.
1243 (mALLI): New mode attribute.
1244 (lddi): New instruction pattern.
1245 (stdi): Likewise.
1246 (stdidi_split): Split instruction for architectures which are not
1247 supporting ll64 option.
1248 (lddidi_split): Likewise.
1249
1250 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1251
1252 PR rtl-optimization/92989
1253 * lra-lives.c (process_bb_lives): Update the live-in set before
1254 processing additional clobbers.
1255
1256 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1257
1258 PR rtl-optimization/93170
1259 * cselib.c (cselib_invalidate_regno_val): New function, split out
1260 from...
1261 (cselib_invalidate_regno): ...here.
1262 (cselib_invalidated_by_call_p): New function.
1263 (cselib_process_insn): Iterate over all the hard-register entries in
1264 REG_VALUES and invalidate any that cross call-clobbered registers.
1265
1266 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1267
1268 * dojump.c (split_comparison): Use HONOR_NANS rather than
1269 HONOR_SNANS when splitting LTGT.
1270
1271 2020-01-27 Martin Liska <mliska@suse.cz>
1272
1273 PR driver/91220
1274 * opts.c (print_filtered_help): Exclude language-specific
1275 options from --help=common unless enabled in all FEs.
1276
1277 2020-01-27 Martin Liska <mliska@suse.cz>
1278
1279 * opts.c (print_help): Exclude params from
1280 all except --help=param.
1281
1282 2020-01-27 Martin Liska <mliska@suse.cz>
1283
1284 PR target/93274
1285 * config/i386/i386-features.c (make_resolver_func):
1286 Align the code with ppc64 target implementation.
1287 Do not generate a unique name for resolver function.
1288
1289 2020-01-27 Richard Biener <rguenther@suse.de>
1290
1291 PR tree-optimization/93397
1292 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1293 converted reduction chain SLP graph adjustment.
1294
1295 2020-01-26 Marek Polacek <polacek@redhat.com>
1296
1297 PR sanitizer/93436
1298 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1299 null DECL_NAME.
1300
1301 2020-01-26 Jason Merrill <jason@redhat.com>
1302
1303 PR c++/92601
1304 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1305 of complete types.
1306
1307 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1308
1309 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1310 (rx_setmem): Likewise.
1311
1312 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1313
1314 PR target/93412
1315 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1316 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1317 drop <di> from constraint of last operand.
1318
1319 PR target/93430
1320 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1321 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1322 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1323
1324 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1325
1326 PR ipa/93166
1327 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1328 check assertion.
1329
1330 2020-01-24 Jeff Law <law@redhat.com>
1331
1332 PR tree-optimization/92788
1333 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1334 not EDGE_ABNORMAL.
1335
1336 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1337
1338 PR target/93395
1339 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1340 *avx_vperm_broadcast_<mode>,
1341 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1342 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1343 Move before avx2_perm<mode>/avx512f_perm<mode>.
1344
1345 PR target/93376
1346 * simplify-rtx.c (simplify_const_unary_operation,
1347 simplify_const_binary_operation): Punt for mode precision above
1348 MAX_BITSIZE_MODE_ANY_INT.
1349
1350 2020-01-24 Andrew Pinski <apinski@marvell.com>
1351
1352 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1353 alu.shift_reg to 0.
1354
1355 2020-01-24 Jeff Law <law@redhat.com>
1356
1357 PR target/13721
1358 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1359 for REGs. Call output_operand_lossage to get more reasonable
1360 diagnostics.
1361
1362 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1363
1364 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1365 gcn_fp_compare_operator.
1366 (vec_cmpu<mode>di): Use gcn_compare_operator.
1367 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1368 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1369 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1370 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1371 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1372 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1373 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1374 gcn_fp_compare_operator.
1375 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1376 gcn_fp_compare_operator.
1377 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1378 gcn_fp_compare_operator.
1379 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1380 gcn_fp_compare_operator.
1381
1382 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1383
1384 * doc/install.texi (Cross-Compiler-Specific Options): Document
1385 `--with-toolexeclibdir' option.
1386
1387 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1388
1389 * target.def (flags_regnum): Also mention effect on delay slot filling.
1390 * doc/tm.texi: Regenerate.
1391
1392 2020-01-23 Jeff Law <law@redhat.com>
1393
1394 PR translation/90162
1395 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1396
1397 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1398
1399 PR target/92269
1400 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1401 profiling label
1402
1403 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1404
1405 PR rtl-optimization/93402
1406 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1407 USE insns.
1408
1409 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1410
1411 * config.in: Regenerated.
1412 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1413 for TARGET_LIBC_GNUSTACK.
1414 * configure: Regenerated.
1415 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1416 found to be 2.31 or greater.
1417
1418 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1419
1420 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1421 TARGET_SOFT_FLOAT.
1422 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1423 (mips_asm_file_end): New function. Delegate to
1424 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1425 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1426
1427 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1428
1429 PR target/93376
1430 * config/i386/i386-modes.def (POImode): New mode.
1431 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1432 * config/i386/i386.md (DPWI): New mode attribute.
1433 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1434 (QWI): Rename to...
1435 (QPWI): ... this. Use POI instead of OI for TImode.
1436 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1437 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1438 instead of <QWI>.
1439
1440 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1441
1442 PR target/93341
1443 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1444 unspec.
1445 (speculation_tracker_rev): New pattern.
1446 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1447 Use speculation_tracker_rev to track the inverse condition.
1448
1449 2020-01-23 Richard Biener <rguenther@suse.de>
1450
1451 PR tree-optimization/93381
1452 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1453 alias-set of the def as argument and record the first one.
1454 (vn_walk_cb_data::first_set): New member.
1455 (vn_reference_lookup_3): Pass the alias-set of the current def
1456 to push_partial_def. Fix alias-set used in the aggregate copy
1457 case.
1458 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1459 * real.c (clear_significand_below): Fix out-of-bound access.
1460
1461 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1462
1463 PR target/93346
1464 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1465 New define_insn patterns.
1466
1467 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1468
1469 * doc/sourcebuild.texi (check-function-bodies): Add an
1470 optional target/xfail selector.
1471
1472 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1473
1474 PR rtl-optimization/93124
1475 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1476 bare USE and CLOBBER insns.
1477
1478 2020-01-22 Andrew Pinski <apinski@marvell.com>
1479
1480 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1481
1482 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1483
1484 PR analyzer/93307
1485 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1486 diagnostic_manager into "ana" namespace.
1487 * selftest-run-tests.c (selftest::run_tests): Update for move of
1488 selftest::run_analyzer_selftests to
1489 ana::selftest::run_analyzer_selftests.
1490
1491 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1492
1493 * cfgexpand.c (union_stack_vars): Update the size.
1494
1495 2020-01-22 Richard Biener <rguenther@suse.de>
1496
1497 PR tree-optimization/93381
1498 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1499 throughout, handle all conversions the same.
1500
1501 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1502
1503 PR target/93335
1504 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1505 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1506 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1507 Call force_reg on high_in2 unconditionally.
1508
1509 2020-01-22 Martin Liska <mliska@suse.cz>
1510
1511 PR tree-optimization/92924
1512 * profile.c (compute_value_histograms): Divide
1513 all counter values.
1514
1515 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1516
1517 PR target/91298
1518 * output.h (assemble_name_resolve): Declare.
1519 * varasm.c (assemble_name_resolve): New function.
1520 (assemble_name): Use it.
1521 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1522
1523 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1524
1525 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1526 update_web_docs_git instead of update_web_docs_svn.
1527
1528 2020-01-21 Andrew Pinski <apinski@marvell.com>
1529
1530 PR target/9311
1531 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1532 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1533 (*tlsgd_small_<mode>): Likewise.
1534 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1535 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1536 register. Convert that register back to dest using convert_mode.
1537
1538 2020-01-21 Jim Wilson <jimw@sifive.com>
1539
1540 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1541 instead of XINT.
1542
1543 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1544 Uros Bizjak <ubizjak@gmail.com>
1545
1546 PR target/93319
1547 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1548 with ptr_mode.
1549 (legitimize_tls_address): Do GNU2 TLS address computation in
1550 ptr_mode and zero-extend result to Pmode.
1551 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1552 :P with :PTR and Pmode with ptr_mode.
1553 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1554 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1555 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1556
1557 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1558
1559 PR target/93333
1560 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1561 the last two operands are CONST_INT_P before using them as such.
1562
1563 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1564
1565 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1566 to get the integer element types.
1567
1568 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1569
1570 * config/aarch64/aarch64-sve-builtins.h
1571 (function_expander::convert_to_pmode): Declare.
1572 * config/aarch64/aarch64-sve-builtins.cc
1573 (function_expander::convert_to_pmode): New function.
1574 (function_expander::get_contiguous_base): Use it.
1575 (function_expander::prepare_gather_address_operands): Likewise.
1576 * config/aarch64/aarch64-sve-builtins-sve2.cc
1577 (svwhilerw_svwhilewr_impl::expand): Likewise.
1578
1579 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1580
1581 PR target/92424
1582 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1583 cfun->machine->label_is_assembled.
1584 (aarch64_print_patchable_function_entry): New.
1585 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1586 * config/aarch64/aarch64.h (struct machine_function): New field,
1587 label_is_assembled.
1588
1589 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1590
1591 PR ipa/93315
1592 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1593 NULL on exit.
1594
1595 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1596
1597 PR lto/93318
1598 * cgraph.c (cgraph_edge::resolve_speculation,
1599 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1600 call_stmt_site_hash.
1601
1602 2020-01-21 Martin Liska <mliska@suse.cz>
1603
1604 * config/rs6000/rs6000.c (common_mode_defined): Remove
1605 unused variable.
1606
1607 2020-01-21 Richard Biener <rguenther@suse.de>
1608
1609 PR tree-optimization/92328
1610 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1611 type when value-numbering same-sized store by inserting a
1612 VIEW_CONVERT_EXPR.
1613 (eliminate_dom_walker::eliminate_stmt): When eliminating
1614 a redundant store handle bit-reinterpretation of the same value.
1615
1616 2020-01-21 Andrew Pinski <apinski@marvel.com>
1617
1618 PR tree-opt/93321
1619 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1620 from ...
1621 (prepare_block_for_update): This. Use a worklist instead of
1622 recursing.
1623
1624 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1625
1626 * gcc/config/arm/arm.c (clear_operation_p):
1627 Initialise last_regno, skip first iteration
1628 based on the first_set value and use ints instead
1629 of the unnecessary HOST_WIDE_INTs.
1630
1631 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1632
1633 PR target/93073
1634 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1635 compare_mode other than SFmode or DFmode.
1636
1637 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1638
1639 PR target/93304
1640 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1641 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1642 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1643
1644 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1645
1646 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1647
1648 2020-01-20 Andrew Pinski <apinski@marvell.com>
1649
1650 PR middle-end/93242
1651 * targhooks.c (default_print_patchable_function_entry): Use
1652 output_asm_insn to emit the nop instruction.
1653
1654 2020-01-20 Fangrui Song <maskray@google.com>
1655
1656 PR middle-end/93194
1657 * targhooks.c (default_print_patchable_function_entry): Align to
1658 POINTER_SIZE.
1659
1660 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1661
1662 PR target/93319
1663 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1664 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1665 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1666 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1667 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1668 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1669 Remove the {q} suffix from lea.
1670 (*tls_dynamic_gnu2_call_64): Renamed to ...
1671 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1672 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1673 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1674 Pass Pmode to gen_tls_dynamic_gnu2_64.
1675
1676 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1677
1678 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1679
1680 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1681
1682 * config/aarch64/aarch64-sve-builtins-base.cc
1683 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1684
1685 2020-01-20 Richard Biener <rguenther@suse.de>
1686
1687 PR debug/92763
1688 * dwarf2out.c (prune_unused_types): Unconditionally mark
1689 called function DIEs.
1690
1691 2020-01-20 Martin Liska <mliska@suse.cz>
1692
1693 PR tree-optimization/93199
1694 * tree-eh.c (struct leh_state): Add
1695 new field outer_non_cleanup.
1696 (cleanup_is_dead_in): Pass leh_state instead
1697 of eh_region. Add a checking that state->outer_non_cleanup
1698 points to outer non-clean up region.
1699 (lower_try_finally): Record outer_non_cleanup
1700 for this_state.
1701 (lower_catch): Likewise.
1702 (lower_eh_filter): Likewise.
1703 (lower_eh_must_not_throw): Likewise.
1704 (lower_cleanup): Likewise.
1705
1706 2020-01-20 Richard Biener <rguenther@suse.de>
1707
1708 PR tree-optimization/93094
1709 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1710 (vect_transform_loop): Likewise.
1711 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1712 loop_vectorized_call to vect_transform_loop.
1713 * tree-vect-loop.c (vect_transform_loop): Pass down
1714 loop_vectorized_call to vect_loop_versioning.
1715 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1716 the earlier discovered loop_vectorized_call.
1717
1718 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1719
1720 * doc/contribute.texi: Update for SVN -> Git transition.
1721 * doc/install.texi: Likewise.
1722
1723 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1724
1725 PR lto/93318
1726 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1727 speculative targets.
1728 (verify_speculative_call): New function
1729 (cgraph_node::verify_node): Use it.
1730 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1731 speculations.
1732
1733 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1734
1735 PR lto/93318
1736 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1737 (cgraph_edge::make_direct): Remove all indirect targets.
1738 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1739 (cgraph_node::verify_node): Verify that only one call_stmt or
1740 lto_stmt_uid is set.
1741 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1742 lto_stmt_uid.
1743 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1744 (lto_output_ref): Simplify streaming of stmt.
1745 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1746
1747 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1748
1749 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1750 Mark parameter unused.
1751
1752 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1753
1754 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1755
1756 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1757
1758 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1759
1760 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1761
1762 * Makefile.in: Add coroutine-passes.o.
1763 * builtin-types.def (BT_CONST_SIZE): New.
1764 (BT_FN_BOOL_PTR): New.
1765 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1766 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1767 * coroutine-builtins.def: New file.
1768 * coroutine-passes.cc: New file.
1769 * function.h (struct GTY function): Add a bit to indicate that the
1770 function is a coroutine component.
1771 * internal-fn.c (expand_CO_FRAME): New.
1772 (expand_CO_YIELD): New.
1773 (expand_CO_SUSPN): New.
1774 (expand_CO_ACTOR): New.
1775 * internal-fn.def (CO_ACTOR): New.
1776 (CO_YIELD): New.
1777 (CO_SUSPN): New.
1778 (CO_FRAME): New.
1779 * passes.def: Add pass_coroutine_lower_builtins,
1780 pass_coroutine_early_expand_ifns.
1781 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1782 (make_pass_coroutine_early_expand_ifns): New.
1783 * doc/invoke.texi: Document the fcoroutines command line
1784 switch.
1785
1786 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1787
1788 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1789
1790 PR target/93312
1791 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1792 after checking the argument is a REG. Don't use REGNO (reg)
1793 again to set last_regno, reuse regno variable instead.
1794
1795 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1796
1797 * doc/analyzer.texi (Limitations): Add note about NaN.
1798
1799 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1800 Sudakshina Das <sudi.das@arm.com>
1801
1802 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1803 and valid immediate.
1804 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1805 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1806 * config/arm/constraints.md (Pg): New.
1807 * config/arm/predicates.md (long_shift_imm): New.
1808 (arm_reg_or_long_shift_imm): Likewise.
1809 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1810 (thumb2_lsll): Likewise.
1811 (thumb2_lsrl): New.
1812
1813 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1814 Sudakshina Das <sudi.das@arm.com>
1815
1816 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1817 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1818 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1819 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1820 * config/arm/thumb2.md (thumb2_asrl): New.
1821 (thumb2_lsll): Likewise.
1822
1823 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1824
1825 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1826 unused variable.
1827
1828 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1829
1830 * gdbinit.in (help-gcc-hooks): New command.
1831 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1832 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1833 documentation.
1834
1835 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1836
1837 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1838 correct target macro.
1839
1840 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1841
1842 * config/aarch64/aarch64-protos.h
1843 (aarch64_sve_ld1ro_operand_p): New.
1844 * config/aarch64/aarch64-sve-builtins-base.cc
1845 (class load_replicate): New.
1846 (class svld1ro_impl): New.
1847 (class svld1rq_impl): Change to inherit from load_replicate.
1848 (svld1ro): New sve intrinsic function base.
1849 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1850 New DEF_SVE_FUNCTION.
1851 * config/aarch64/aarch64-sve-builtins-base.h
1852 (svld1ro): New decl.
1853 * config/aarch64/aarch64-sve-builtins.cc
1854 (function_expander::add_mem_operand): Modify assert to allow
1855 OImode.
1856 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1857 pattern.
1858 * config/aarch64/aarch64.c
1859 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1860 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1861 (aarch64_sve_ld1ro_operand_p): New.
1862 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1863 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1864 * config/aarch64/predicates.md
1865 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1866
1867 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1868
1869 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1870 Introduce this ACLE specified predefined macro.
1871 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1872 (fp): Disabling this disables f64mm.
1873 (simd): Disabling this disables f64mm.
1874 (fp16): Disabling this disables f64mm.
1875 (sve): Disabling this disables f64mm.
1876 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1877 (AARCH64_ISA_F64MM): New.
1878 (TARGET_F64MM): New.
1879 * doc/invoke.texi (f64mm): Document new option.
1880
1881 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1882
1883 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1884 (neoversen1_tunings): Likewise.
1885
1886 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1887
1888 PR target/92692
1889 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1890 Add assert to ensure prolog has been emitted.
1891 (aarch64_split_atomic_op): Likewise.
1892 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1893 Use epilogue_completed rather than reload_completed.
1894 (aarch64_atomic_exchange<mode>): Likewise.
1895 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1896 (atomic_nand<mode>): Likewise.
1897 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1898 (atomic_fetch_nand<mode>): Likewise.
1899 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1900 (atomic_nand_fetch<mode>): Likewise.
1901
1902 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1903
1904 PR target/93133
1905 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1906 for FP modes.
1907 (REVERSE_CONDITION): Delete.
1908 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1909 (CCFP_CCFPE): Likewise.
1910 (e): New mode attribute.
1911 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1912 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1913 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1914 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1915 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1916 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1917 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1918 name of generator from gen_ccmpdi to gen_ccmpccdi.
1919 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1920 the previous comparison but aren't able to, use the new ccmp_rev
1921 patterns instead.
1922
1923 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1924
1925 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1926 than testing directly for INTEGER_CST.
1927 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1928
1929 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1930
1931 PR tree-optimization/93292
1932 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1933 get_vectype_for_scalar_type returns NULL.
1934
1935 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1936
1937 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1938 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1939
1940 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1941
1942 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1943 dump.
1944 * params.opt: (max-predicted-iterations): Set bounds.
1945 * predict.c (real_almost_one, real_br_prob_base,
1946 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1947 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1948 probabilities; do not truncate to reg_br_prob_bases.
1949 (estimate_loops_at_level): Pass max_cyclic_prob.
1950 (estimate_loops): Compute max_cyclic_prob.
1951 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1952 of back edge prob.
1953 * profile-count.c (profile_probability::to_sreal): New.
1954 * profile-count.h (class sreal): Move up in file.
1955 (profile_probability::to_sreal): Declare.
1956
1957 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1958
1959 * config/arm/arm.c
1960 (arm_invalid_conversion): New function for target hook.
1961 (arm_invalid_unary_op): New function for target hook.
1962 (arm_invalid_binary_op): New function for target hook.
1963
1964 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1965
1966 * config.gcc: Add arm_bf16.h.
1967 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1968 (arm_simd_builtin_std_type): Add BFmode.
1969 (arm_init_simd_builtin_types): Define element types for vector types.
1970 (arm_init_bf16_types): New function.
1971 (arm_init_builtins): Add arm_init_bf16_types function call.
1972 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1973 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1974 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1975 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1976 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1977 (arm_mangle_type): Add __bf16.
1978 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1979 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1980 arm_bf16_ptr_type_node.
1981 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1982 define_split between ARM registers.
1983 * config/arm/arm_bf16.h: New file.
1984 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1985 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1986 (VQXMOV): Add V8BF.
1987 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1988 * config/arm/vfp.md: Add BFmode to movhf patterns.
1989
1990 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1991 Andre Vieira <andre.simoesdiasvieira@arm.com>
1992
1993 * config/arm/arm-cpus.in (mve, mve_float): New features.
1994 (dsp, mve, mve.fp): New options.
1995 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1996 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1997 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1998
1999 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2000 Thomas Preud'homme <thomas.preudhomme@arm.com>
2001
2002 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2003 Armv8-M Mainline.
2004 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2005 error for using -mcmse when targeting Armv8.1-M Mainline.
2006
2007 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2008 Thomas Preud'homme <thomas.preudhomme@arm.com>
2009
2010 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2011 address in r4 when targeting Armv8.1-M Mainline.
2012 (nonsecure_call_value_internal): Likewise.
2013 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2014 a register match_operand again. Emit BLXNS when targeting
2015 Armv8.1-M Mainline.
2016 (nonsecure_call_value_reg_thumb2): Likewise.
2017
2018 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2019 Thomas Preud'homme <thomas.preudhomme@arm.com>
2020
2021 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2022 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2023 variable as true when floating-point ABI is not hard. Replace
2024 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2025 Generate VLSTM and VLLDM instruction respectively before and
2026 after a function call to cmse_nonsecure_call function.
2027 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2028 (VUNSPEC_VLLDM): Likewise.
2029 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2030 (lazy_load_multiple_insn): Likewise.
2031
2032 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2033 Thomas Preud'homme <thomas.preudhomme@arm.com>
2034
2035 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2036 (arm_emit_vfp_multi_reg_pop): Likewise.
2037 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2038 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2039 restore callee-saved VFP registers.
2040
2041 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2042 Thomas Preud'homme <thomas.preudhomme@arm.com>
2043
2044 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2045 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2046 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2047 callee-saved GPRs as well as clear ip register before doing a nonsecure
2048 call then restore callee-saved GPRs after it when targeting
2049 Armv8.1-M Mainline.
2050 (arm_reorg): Adapt to function rename.
2051
2052 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2053 Thomas Preud'homme <thomas.preudhomme@arm.com>
2054
2055 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2056 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2057 clear_vfp_multiple pattern based on a new vfp parameter.
2058 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2059 targeting Armv8.1-M Mainline.
2060 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2061 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2062 whether VFP registers are available before looking call_used_regs for a
2063 VFP register.
2064 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2065 of prototype of clear_operation_p.
2066 (clear_vfp_multiple_operation): New predicate.
2067 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2068 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2069
2070 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2071 Thomas Preud'homme <thomas.preudhomme@arm.com>
2072
2073 * config/arm/arm-protos.h (clear_operation_p): Declare.
2074 * config/arm/arm.c (clear_operation_p): New function.
2075 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2076 targeting Armv8.1-M Mainline or successor.
2077 (output_return_instruction): Only output APSR register clearing if
2078 Armv8.1-M Mainline instructions not available.
2079 (thumb_exit): Likewise.
2080 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2081 * config/arm/thumb2.md (clear_apsr): New define_insn.
2082 (clear_multiple): Likewise.
2083 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2084
2085 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2086 Thomas Preud'homme <thomas.preudhomme@arm.com>
2087
2088 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2089 (use_return_insn): Also return false for Armv8.1-M Mainline.
2090 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2091 Mainline instructions are available.
2092 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2093 when targeting Armv8.1-M Mainline Security Extensions.
2094 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2095 Mainline entry function.
2096 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2097 targeting Armv8.1-M Mainline or successor.
2098 (arm_expand_epilogue): Fix indentation of caller-saved register
2099 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2100 entry function.
2101 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2102 (FP_SYSREGS): Likewise.
2103 (enum vfp_sysregs_encoding): Define enum.
2104 (fp_sysreg_names): Declare.
2105 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2106 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2107 (pop_fpsysreg_insn): Likewise.
2108
2109 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2110 Thomas Preud'homme <thomas.preudhomme@arm.com>
2111
2112 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2113 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2114 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2115 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2116 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2117 (ARMv8_1m_main): New feature group.
2118 (armv8.1-m.main): New architecture.
2119 * config/arm/arm-tables.opt: Regenerate.
2120 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2121 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2122 (arm_options_perform_arch_sanity_checks): Error out when targeting
2123 Armv8.1-M Mainline Security Extensions.
2124 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2125
2126 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2127
2128 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2129 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2130 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2131 aarch64_bfdot_laneq): New.
2132 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2133 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2134 vbfdotq_laneq_f32): New.
2135 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2136 VBFMLA_W, VBF): New.
2137 (isquadop): Add V4BF, V8BF.
2138
2139 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2140
2141 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2142 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2143 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2144 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2145 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2146 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2147 usdot_laneq, sudot_lane,sudot_laneq): New.
2148 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2149 (aarch64_<sur>dot_lane): New.
2150 * config/aarch64/arm_neon.h (vusdot_s32): New.
2151 (vusdotq_s32): New.
2152 (vusdot_lane_s32): New.
2153 (vsudot_lane_s32): New.
2154 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2155 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2156
2157 2020-01-16 Martin Liska <mliska@suse.cz>
2158
2159 * value-prof.c (dump_histogram_value): Fix
2160 obvious spacing issue.
2161
2162 2020-01-16 Andrew Pinski <apinski@marvell.com>
2163
2164 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2165 !storage_order_barrier_p.
2166
2167 2020-01-16 Andrew Pinski <apinski@marvell.com>
2168
2169 * sched-int.h (_dep): Add unused bit-field field for the padding.
2170 * sched-deps.c (init_dep_1): Init unused field.
2171
2172 2020-01-16 Andrew Pinski <apinski@marvell.com>
2173
2174 * optabs.h (create_expand_operand): Initialize target field also.
2175
2176 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2177
2178 PR tree-optimization/92429
2179 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2180 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2181 control folding.
2182 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2183 tree.
2184
2185 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2186
2187 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2188 aarch64_sve_int_mode to each mode.
2189
2190 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2191
2192 * doc/analyzer.texi (Overview): Add note about
2193 -fdump-ipa-analyzer.
2194
2195 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2196
2197 PR tree-optimization/93231
2198 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2199 input_type is unsigned. Use tree_to_shwi for shift constant.
2200 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2201 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2202 inputs more efficiently.
2203
2204 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2205
2206 * config/i386/i386.md (*movsf_internal): Do not require
2207 SSE2 ISA for alternatives 14 and 15.
2208
2209 2020-01-15 Richard Biener <rguenther@suse.de>
2210
2211 PR middle-end/93273
2212 * tree-eh.c (sink_clobbers): If we already visited the destination
2213 block do not defer insertion.
2214 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2215 the purpose of defered insertion.
2216
2217 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2218
2219 * BASE-VER: Bump to 10.0.1.
2220
2221 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2222
2223 PR tree-optimization/93247
2224 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2225 type of the stmt that we're going to vectorize.
2226
2227 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2228
2229 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2230 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2231 type from the lhs.
2232
2233 2020-01-15 Martin Liska <mliska@suse.cz>
2234
2235 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2236 2 calls of streamer_read_hwi in a function call.
2237
2238 2020-01-15 Richard Biener <rguenther@suse.de>
2239
2240 * alias.c (record_alias_subset): Avoid redundant work when
2241 subset is already recorded.
2242
2243 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2244
2245 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2246 the analyzer options provide CWE identifiers.
2247
2248 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2249
2250 * tree-diagnostic-path.cc (path_summary::event_range::print):
2251 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2252 using get_pure_location.
2253
2254 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2255
2256 PR tree-optimization/93262
2257 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2258 perform head trimming only if the last argument is constant,
2259 either all ones, or larger or equal to head trim, in the latter
2260 case decrease the last argument by head_trim.
2261
2262 PR tree-optimization/93249
2263 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2264 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2265 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2266 perform head trim unless we can prove there are no '\0' chars
2267 from the source among the first head_trim chars.
2268
2269 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2270
2271 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2272
2273 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2274
2275 PR target/93009
2276 * config/i386/sse.md
2277 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2278 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2279 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2280 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2281 just a single alternative instead of two, make operands 1 and 2
2282 commutative.
2283
2284 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2285
2286 PR lto/91576
2287 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2288 TYPE_MODE.
2289
2290 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2291
2292 * Makefile.in (lang_opt_files): Add analyzer.opt.
2293 (ANALYZER_OBJS): New.
2294 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2295 tristate.o and ANALYZER_OBJS.
2296 (TEXI_GCCINT_FILES): Add analyzer.texi.
2297 * common.opt (-fanalyzer): New driver option.
2298 * config.in: Regenerate.
2299 * configure: Regenerate.
2300 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2301 (gccdepdir): Also create depdir for "analyzer" subdir.
2302 * digraph.cc: New file.
2303 * digraph.h: New file.
2304 * doc/analyzer.texi: New file.
2305 * doc/gccint.texi ("Static Analyzer") New menu item.
2306 (analyzer.texi): Include it.
2307 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2308 ("Warning Options"): Add static analysis warnings to the list.
2309 (-Wno-analyzer-double-fclose): New option.
2310 (-Wno-analyzer-double-free): New option.
2311 (-Wno-analyzer-exposure-through-output-file): New option.
2312 (-Wno-analyzer-file-leak): New option.
2313 (-Wno-analyzer-free-of-non-heap): New option.
2314 (-Wno-analyzer-malloc-leak): New option.
2315 (-Wno-analyzer-possible-null-argument): New option.
2316 (-Wno-analyzer-possible-null-dereference): New option.
2317 (-Wno-analyzer-null-argument): New option.
2318 (-Wno-analyzer-null-dereference): New option.
2319 (-Wno-analyzer-stale-setjmp-buffer): New option.
2320 (-Wno-analyzer-tainted-array-index): New option.
2321 (-Wno-analyzer-use-after-free): New option.
2322 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2323 (-Wno-analyzer-use-of-uninitialized-value): New option.
2324 (-Wanalyzer-too-complex): New option.
2325 (-fanalyzer-call-summaries): New warning.
2326 (-fanalyzer-checker=): New warning.
2327 (-fanalyzer-fine-grained): New warning.
2328 (-fno-analyzer-state-merge): New warning.
2329 (-fno-analyzer-state-purge): New warning.
2330 (-fanalyzer-transitivity): New warning.
2331 (-fanalyzer-verbose-edges): New warning.
2332 (-fanalyzer-verbose-state-changes): New warning.
2333 (-fanalyzer-verbosity=): New warning.
2334 (-fdump-analyzer): New warning.
2335 (-fdump-analyzer-callgraph): New warning.
2336 (-fdump-analyzer-exploded-graph): New warning.
2337 (-fdump-analyzer-exploded-nodes): New warning.
2338 (-fdump-analyzer-exploded-nodes-2): New warning.
2339 (-fdump-analyzer-exploded-nodes-3): New warning.
2340 (-fdump-analyzer-supergraph): New warning.
2341 * doc/sourcebuild.texi (dg-require-dot): New.
2342 (dg-check-dot): New.
2343 * gdbinit.in (break-on-saved-diagnostic): New command.
2344 * graphviz.cc: New file.
2345 * graphviz.h: New file.
2346 * ordered-hash-map-tests.cc: New file.
2347 * ordered-hash-map.h: New file.
2348 * passes.def (pass_analyzer): Add before
2349 pass_ipa_whole_program_visibility.
2350 * selftest-run-tests.c (selftest::run_tests): Call
2351 selftest::ordered_hash_map_tests_cc_tests.
2352 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2353 decl.
2354 * shortest-paths.h: New file.
2355 * timevar.def (TV_ANALYZER): New timevar.
2356 (TV_ANALYZER_SUPERGRAPH): Likewise.
2357 (TV_ANALYZER_STATE_PURGE): Likewise.
2358 (TV_ANALYZER_PLAN): Likewise.
2359 (TV_ANALYZER_SCC): Likewise.
2360 (TV_ANALYZER_WORKLIST): Likewise.
2361 (TV_ANALYZER_DUMP): Likewise.
2362 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2363 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2364 * tree-pass.h (make_pass_analyzer): New decl.
2365 * tristate.cc: New file.
2366 * tristate.h: New file.
2367
2368 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2369
2370 PR target/93254
2371 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2372 alternatives 9 and 10.
2373
2374 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2375
2376 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2377 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2378 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2379 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2380 (selftest::hash_map_tests_c_tests): Call it.
2381 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2382 New static constant, using the value of = H::empty_zero_p.
2383 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2384 from default_hash_traits <Value>.
2385 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2386 from Traits.
2387 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2388 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2389 calls to mark_empty with !Descriptor::empty_zero_p.
2390 (hash_table::empty_slow): Conditionalize the memset call with a
2391 check that Descriptor::empty_zero_p; otherwise, loop through the
2392 entries calling mark_empty on them.
2393 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2394 (pointer_hash::empty_zero_p): Likewise.
2395 (pair_hash::empty_zero_p): Likewise.
2396 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2397 Likewise.
2398 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2399 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2400 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2401 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2402 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2403 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2404 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2405 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2406 * tree-vectorizer.h
2407 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2408 Likewise.
2409
2410 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2411
2412 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2413 fix typo on return value.
2414
2415 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2416
2417 PR ipa/69678
2418 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2419 target_prob.
2420 (cgraph_edge::make_speculative): Add param for setting speculative_id
2421 and target_prob.
2422 (cgraph_edge::speculative_call_info): Update comments and find reference
2423 by speculative_id for multiple indirect targets.
2424 (cgraph_edge::resolve_speculation): Decrease the speculations
2425 for indirect edge, drop it's speculative if not direct target
2426 left. Update comments.
2427 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2428 (cgraph_node::dump): Print num_speculative_call_targets.
2429 (cgraph_node::verify_node): Don't report error if speculative
2430 edge not include statement.
2431 (cgraph_edge::num_speculative_call_targets_p): New function.
2432 * cgraph.h (int common_target_id): Remove.
2433 (int common_target_probability): Remove.
2434 (num_speculative_call_targets): New variable.
2435 (make_speculative): Add param for setting speculative_id.
2436 (cgraph_edge::num_speculative_call_targets_p): New declare.
2437 (target_prob): New variable.
2438 (speculative_id): New variable.
2439 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2440 call summaries for multiple speculative call targets.
2441 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2442 * ipa-profile.c (struct speculative_call_target): New struct.
2443 (class speculative_call_summary): New class.
2444 (class speculative_call_summaries): New class.
2445 (call_sums): New variable.
2446 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2447 (ipa_profile_write_edge_summary): New function.
2448 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2449 (ipa_profile_dump_all_summaries): New function.
2450 (ipa_profile_read_edge_summary): New function.
2451 (ipa_profile_read_summary_section): New function.
2452 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2453 (ipa_profile): Generate num_speculative_call_targets from
2454 profile summaries.
2455 * ipa-ref.h (speculative_id): New variable.
2456 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2457 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2458 common_target_probability. Stream out speculative_id and
2459 num_speculative_call_targets.
2460 (input_edge): Likewise.
2461 * predict.c (dump_prediction): Remove edges count assert to be
2462 precise.
2463 * symtab.c (symtab_node::create_reference): Init speculative_id.
2464 (symtab_node::clone_references): Clone speculative_id.
2465 (symtab_node::clone_referring): Clone speculative_id.
2466 (symtab_node::clone_reference): Clone speculative_id.
2467 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2468 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2469 if indirect call contains multiple speculative targets.
2470 * value-prof.h (check_ic_target): Remove.
2471 * value-prof.c (gimple_value_profile_transformations):
2472 Use void function gimple_ic_transform.
2473 * value-prof.c (gimple_ic_transform): Handle topn case.
2474 Fix comment typos. Change it to a void function.
2475
2476 2020-01-13 Andrew Pinski <apinski@marvell.com>
2477
2478 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2479 (octeontx2t98): New define.
2480 (octeontx2t96): New define.
2481 (octeontx2t93): New define.
2482 (octeontx2f95): New define.
2483 (octeontx2f95n): New define.
2484 (octeontx2f95mm): New define.
2485 * config/aarch64/aarch64-tune.md: Regenerate.
2486 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2487
2488 2020-01-13 Jason Merrill <jason@redhat.com>
2489
2490 PR c++/33799 - destroy return value if local cleanup throws.
2491 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2492
2493 2020-01-13 Martin Liska <mliska@suse.cz>
2494
2495 * ipa-cp.c (get_max_overall_size): Use newly
2496 renamed param param_ipa_cp_unit_growth.
2497 * params.opt: Remove legacy param name.
2498
2499 2020-01-13 Martin Sebor <msebor@redhat.com>
2500
2501 PR tree-optimization/93213
2502 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2503 stores to be eliminated.
2504
2505 2020-01-13 Martin Liska <mliska@suse.cz>
2506
2507 * opts.c (print_help): Do not print CL_PARAM
2508 and CL_WARNING for CL_OPTIMIZATION.
2509
2510 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2511
2512 PR driver/92757
2513 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2514 depending on optimization settings.
2515
2516 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2517
2518 PR tree-optimization/90838
2519 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2520 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2521 argument rather than to initialize temporary for targets that
2522 don't use the mode argument at all. Initialize ctzval to avoid
2523 warning at -O0.
2524
2525 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2526
2527 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2528 * tree-core.h: Document it.
2529 * gimplify.c (gimplify_omp_workshare): Set it.
2530 * omp-low.c (lower_omp_target): Use it.
2531 * tree-pretty-print.c (dump_omp_clause): Print it.
2532
2533 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2534 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2535
2536 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2537
2538 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2539 * common.opt (fdiagnostics-path-format=): New option.
2540 (diagnostic_path_format): New enum.
2541 (fdiagnostics-show-path-depths): New option.
2542 * coretypes.h (diagnostic_event_id_t): New forward decl.
2543 * diagnostic-color.c (color_dict): Add "path".
2544 * diagnostic-event-id.h: New file.
2545 * diagnostic-format-json.cc (json_from_expanded_location): Make
2546 non-static.
2547 (json_end_diagnostic): Call context->make_json_for_path if it
2548 exists and the diagnostic has a path.
2549 (diagnostic_output_format_init): Clear context->print_path.
2550 * diagnostic-path.h: New file.
2551 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2552 when printing a run of events in a diagnostic_path so that they
2553 all get the same color.
2554 (layout::m_diagnostic_path_p): New field.
2555 (layout::layout): Initialize it.
2556 (layout::print_any_labels): Don't colorize the label text for an
2557 event in a diagnostic_path.
2558 (gcc_rich_location::add_location_if_nearby): Add
2559 "restrict_to_current_line_spans" and "label" params. Pass the
2560 former to layout.maybe_add_location_range; pass the latter
2561 when calling add_range.
2562 * diagnostic.c: Include "diagnostic-path.h".
2563 (diagnostic_initialize): Initialize context->path_format and
2564 context->show_path_depths.
2565 (diagnostic_show_any_path): New function.
2566 (diagnostic_path::interprocedural_p): New function.
2567 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2568 (simple_diagnostic_path::num_events): New function.
2569 (simple_diagnostic_path::get_event): New function.
2570 (simple_diagnostic_path::add_event): New function.
2571 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2572 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2573 (debug): New overload taking a diagnostic_path *.
2574 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2575 * diagnostic.h (enum diagnostic_path_format): New enum.
2576 (json::value): New forward decl.
2577 (diagnostic_context::path_format): New field.
2578 (diagnostic_context::show_path_depths): New field.
2579 (diagnostic_context::print_path): New callback field.
2580 (diagnostic_context::make_json_for_path): New callback field.
2581 (diagnostic_show_any_path): New decl.
2582 (json_from_expanded_location): New decl.
2583 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2584 (-fdiagnostics-show-path-depths): New option.
2585 (-fdiagnostics-color): Add "path" to description of default
2586 GCC_COLORS; describe it.
2587 (-fdiagnostics-format=json): Document how diagnostic paths are
2588 represented in the JSON output format.
2589 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2590 Add optional params "restrict_to_current_line_spans" and "label".
2591 * opts.c (common_handle_option): Handle
2592 OPT_fdiagnostics_path_format_ and
2593 OPT_fdiagnostics_show_path_depths.
2594 * pretty-print.c: Include "diagnostic-event-id.h".
2595 (pp_format): Implement "%@" format code for printing
2596 diagnostic_event_id_t *.
2597 (selftest::test_pp_format): Add tests for "%@".
2598 * selftest-run-tests.c (selftest::run_tests): Call
2599 selftest::tree_diagnostic_path_cc_tests.
2600 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2601 * toplev.c (general_init): Initialize global_dc->path_format and
2602 global_dc->show_path_depths.
2603 * tree-diagnostic-path.cc: New file.
2604 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2605 non-static. Drop "diagnostic" param in favor of storing the
2606 original value of "where" and re-using it.
2607 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2608 maybe_unwind_expanded_macro_loc.
2609 (tree_diagnostics_defaults): Initialize context->print_path and
2610 context->make_json_for_path.
2611 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2612 decl.
2613 (default_tree_make_json_for_path): New decl.
2614 (maybe_unwind_expanded_macro_loc): New decl.
2615
2616 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2617
2618 PR tree-optimization/93210
2619 * fold-const.h (native_encode_initializer,
2620 can_native_interpret_type_p): Declare.
2621 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2622 simplify.
2623 (native_encode_initializer): New function, moved from dwarf2out.c.
2624 Adjust to native_encode_expr compatible arguments, including dry-run
2625 and partial extraction modes. Don't handle STRING_CST.
2626 (can_native_interpret_type_p): No longer static.
2627 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2628 offset / BITS_PER_UNIT fits into int and don't call it if
2629 can_native_interpret_type_p fails. If suboff is NULL and for
2630 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2631 native_encode_initializer.
2632 (fold_const_aggregate_ref_1): Formatting fix.
2633 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2634 (tree_add_const_value_attribute): Adjust caller.
2635
2636 PR tree-optimization/90838
2637 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2638 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2639 CTZ_DEFINED_VALUE_AT_ZERO.
2640
2641 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2642
2643 PR inline-asm/93027
2644 * lra-constraints.c (match_reload): Permit input operands have the
2645 same mode as output while other input operands have a different
2646 mode.
2647
2648 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2649
2650 PR tree-optimization/90838
2651 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2652 (check_ctz_string): Likewise.
2653 (optimize_count_trailing_zeroes): Likewise.
2654 (simplify_count_trailing_zeroes): Likewise.
2655 (pass_forwprop::execute): Try ctz simplification.
2656 * match.pd: Add matching for ctz idioms.
2657
2658 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2659
2660 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2661 for target hook.
2662 (aarch64_invalid_unary_op): New function for target hook.
2663 (aarch64_invalid_binary_op): New function for target hook.
2664
2665 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2666
2667 * config.gcc: Add arm_bf16.h.
2668 * config/aarch64/aarch64-builtins.c
2669 (aarch64_simd_builtin_std_type): Add BFmode.
2670 (aarch64_init_simd_builtin_types): Define element types for vector
2671 types.
2672 (aarch64_init_bf16_types): New function.
2673 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2674 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2675 modes.
2676 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2677 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2678 patterns.
2679 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2680 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2681 * config/aarch64/aarch64.c
2682 (aarch64_classify_vector_mode): Add support for BF types.
2683 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2684 (aarch64_vq_mode): Add support for BF types.
2685 (aarch64_simd_container_mode): Add support for BF types.
2686 (aarch64_mangle_type): Add support for BF scalar type.
2687 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2688 * config/aarch64/arm_bf16.h: New file.
2689 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2690 * config/aarch64/iterators.md: Add BF types to mode attributes.
2691 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2692
2693 2020-01-10 Jason Merrill <jason@redhat.com>
2694
2695 PR c++/93173 - incorrect tree sharing.
2696 * gimplify.c (copy_if_shared): No longer static.
2697 * gimplify.h: Declare it.
2698
2699 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2700
2701 * doc/invoke.texi (-msve-vector-bits=): Document that
2702 -msve-vector-bits=128 now generates VL-specific code for
2703 little-endian targets.
2704 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2705 build_vector_type_for_mode to construct the data vector types.
2706 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2707 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2708 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2709 for 128-bit vectors.
2710
2711 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2712
2713 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2714 invocation.
2715
2716 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2717
2718 * config/aarch64/aarch64-builtins.c
2719 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2720 rather than checking the number of elements and the element mode.
2721
2722 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2723
2724 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2725 get_related_vectype_for_scalar_type rather than build_vector_type
2726 to create the index type for a conditional reduction.
2727
2728 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2729
2730 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2731 for any type of gather or scatter, including strided accesses.
2732
2733 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2734
2735 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2736 comment.
2737
2738 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2739
2740 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2741 get_dr_vinfo_offset
2742 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2743 parameter and its use to reset DR_OFFSET's.
2744 (vect_transform_loop): Remove orig_drs_init argument.
2745 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2746 member of dr_vec_info rather than the offset of the associated
2747 data_reference's innermost_loop_behavior.
2748 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2749 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2750 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2751 get_dr_vinfo_offset.
2752 (vectorizable_store): Likewise.
2753 (vectorizable_load): Likewise.
2754
2755 2020-01-10 Richard Biener <rguenther@suse.de>
2756
2757 * gimple-ssa-store-merging
2758 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2759
2760 2020-01-10 Martin Liska <mliska@suse.cz>
2761
2762 PR ipa/93217
2763 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2764 encapsulation that was there before r280040.
2765
2766 2020-01-10 Richard Biener <rguenther@suse.de>
2767
2768 PR middle-end/93199
2769 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2770 sequences to avoid walking them again for secondary opportunities.
2771 (pass_lower_eh_dispatch::execute): Instead actually insert
2772 them here.
2773
2774 2020-01-10 Richard Biener <rguenther@suse.de>
2775
2776 PR middle-end/93199
2777 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2778 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2779 avoid quadraticness.
2780
2781 2020-01-10 Martin Jambor <mjambor@suse.cz>
2782
2783 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2784 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2785 to get param_ipa_sra_max_replacements.
2786 (param_splitting_across_edge): Pass the caller to
2787 pull_accesses_from_callee.
2788
2789 2020-01-10 Martin Jambor <mjambor@suse.cz>
2790
2791 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2792 * ipa-cp.c (max_new_size): Removed.
2793 (orig_overall_size): New variable.
2794 (get_max_overall_size): New function.
2795 (estimate_local_effects): Use it. Adjust dump.
2796 (decide_about_value): Likewise.
2797 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2798 orig_overall_size. Adjust dump.
2799 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2800
2801 2020-01-10 Martin Jambor <mjambor@suse.cz>
2802
2803 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2804 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2805 instead of param_ipa_max_agg_items.
2806 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2807 optimization info for the callee.
2808
2809 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2810
2811 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2812 markers if debug_inline_points is false.
2813
2814 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2815
2816 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2817 extra_objs.
2818 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2819 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2820 aarch64-sve-builtins-sve2.h.
2821 (aarch64-sve-builtins-sve2.o): New rule.
2822 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2823 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2824 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2825 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2826 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2827 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2828 TARGET_SVE2_SM4.
2829 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2830 instructions that are handled here.
2831 (@cond_asrd<mode>): Generalize to...
2832 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2833 (*cond_asrd<mode>_2): Generalize to...
2834 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2835 (*cond_asrd<mode>_z): Generalize to...
2836 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2837 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2838 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2839 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2840 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2841 pattern.
2842 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2843 (@aarch64_scatter_stnt<mode>): Likewise.
2844 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2845 (@aarch64_mul_lane_<mode>): Likewise.
2846 (@aarch64_sve_suqadd<mode>_const): Likewise.
2847 (*<sur>h<addsub><mode>): Generalize to...
2848 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2849 new pattern.
2850 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2851 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2852 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2853 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2854 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2855 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2856 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2857 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2858 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2859 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2860 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2861 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2862 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2863 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2864 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2865 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2866 (@aarch64_sve2_xar<mode>): Likewise.
2867 (@aarch64_sve2_bcax<mode>): Likewise.
2868 (*aarch64_sve2_eor3<mode>): Rename to...
2869 (@aarch64_sve2_eor3<mode>): ...this.
2870 (@aarch64_sve2_bsl<mode>): New expander.
2871 (@aarch64_sve2_nbsl<mode>): Likewise.
2872 (@aarch64_sve2_bsl1n<mode>): Likewise.
2873 (@aarch64_sve2_bsl2n<mode>): Likewise.
2874 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2875 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2876 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2877 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2878 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2879 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2880 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2881 (<su>mull<bt><Vwide>): Generalize to...
2882 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2883 pattern.
2884 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2885 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2886 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2887 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2888 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2889 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2890 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2891 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2892 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2893 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2894 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2895 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2896 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2897 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2898 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2899 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2900 (<SHRNB:r>shrnb<mode>): Generalize to...
2901 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2902 new pattern.
2903 (<SHRNT:r>shrnt<mode>): Generalize to...
2904 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2905 new pattern.
2906 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2907 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2908 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2909 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2910 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2911 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2912 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2913 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2914 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2915 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2916 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2917 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2918 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2919 (@aarch64_sve2_cvtnt<mode>): Likewise.
2920 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2921 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2922 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2923 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2924 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2925 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2926 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2927 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2928 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2929 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2930 (@aarch64_sve2_pmul<mode>): Likewise.
2931 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2932 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2933 (@aarch64_sve2_tbl2<mode>): Likewise.
2934 (@aarch64_sve2_tbx<mode>): Likewise.
2935 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2936 (@aarch64_sve2_histcnt<mode>): Likewise.
2937 (@aarch64_sve2_histseg<mode>): Likewise.
2938 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2939 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2940 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2941 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2942 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2943 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2944 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2945 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2946 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2947 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2948 (SVE2_PMULL_PAIR_I): New mode iterators.
2949 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2950 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2951 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2952 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2953 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2954 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2955 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2956 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2957 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2958 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2959 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2960 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2961 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2962 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2963 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2964 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2965 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2966 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2967 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2968 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2969 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2970 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2971 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2972 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2973 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2974 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2975 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2976 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2977 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2978 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2979 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2980 further down file.
2981 (VNARROW, Ventype): New mode attributes.
2982 (Vewtype): Handle VNx2DI. Fix typo in comment.
2983 (VDOUBLE): New mode attribute.
2984 (sve_lane_con): Handle VNx8HI.
2985 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2986 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2987 (sve_int_op, sve_int_op_rev): Handle the above codes.
2988 (sve_pred_int_rhs2_operand): Likewise.
2989 (MULLBT, SHRNB, SHRNT): Delete.
2990 (SVE_INT_SHIFT_IMM): New int iterator.
2991 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2992 and UNSPEC_WHILEHS for TARGET_SVE2.
2993 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2994 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2995 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2996 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2997 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2998 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2999 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3000 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3001 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3002 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3003 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3004 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3005 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3006 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3007 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3008 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3009 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3010 (optab): Handle the new unspecs.
3011 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3012 and UNSPEC_RSHRNT.
3013 (lr): Handle the new unspecs.
3014 (bt): Delete.
3015 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3016 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3017 (sve_int_qsub_op): New int attributes.
3018 (sve_fp_op, rot): Handle the new unspecs.
3019 * config/aarch64/aarch64-sve-builtins.h
3020 (function_resolver::require_matching_pointer_type): Declare.
3021 (function_resolver::resolve_unary): Add an optional boolean argument.
3022 (function_resolver::finish_opt_n_resolution): Add an optional
3023 type_suffix_index argument.
3024 (gimple_folder::redirect_call): Declare.
3025 (gimple_expander::prepare_gather_address_operands): Add an optional
3026 bool parameter.
3027 * config/aarch64/aarch64-sve-builtins.cc: Include
3028 aarch64-sve-builtins-sve2.h.
3029 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3030 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3031 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3032 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3033 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3034 (TYPES_s_unsigned): Likewise.
3035 (TYPES_s_integer): Use TYPES_s_unsigned.
3036 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3037 (TYPES_sd_integer): Use them.
3038 (TYPES_d_unsigned): New macro.
3039 (TYPES_d_integer): Use it.
3040 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3041 (TYPES_cvt_narrow): Likewise.
3042 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3043 (preds_mx): New variable.
3044 (function_builder::add_overloaded_function): Allow the new feature
3045 set to be more restrictive than the original one.
3046 (function_resolver::infer_pointer_type): Remove qualifiers from
3047 the pointer type before printing it.
3048 (function_resolver::require_matching_pointer_type): New function.
3049 (function_resolver::resolve_sv_displacement): Handle functions
3050 that don't support 32-bit vector indices or svint32_t vector offsets.
3051 (function_resolver::finish_opt_n_resolution): Take the inferred type
3052 as a separate argument.
3053 (function_resolver::resolve_unary): Optionally treat all forms in
3054 the same way as normal merging functions.
3055 (gimple_folder::redirect_call): New function.
3056 (function_expander::prepare_gather_address_operands): Add an argument
3057 that says whether scaled forms are available. If they aren't,
3058 handle scaling of vector indices and don't add the extension and
3059 scaling operands.
3060 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3061 fall back to using cond_* instead.
3062 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3063 Split out the member variables into...
3064 (rtx_code_function_base): ...this new base class.
3065 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3066 (unspec_based_function): Split out the member variables into...
3067 (unspec_based_function_base): ...this new base class.
3068 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3069 (unspec_based_function_exact_insn): New class.
3070 (unspec_based_add_function, unspec_based_add_lane_function)
3071 (unspec_based_lane_function, unspec_based_pred_function)
3072 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3073 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3074 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3075 typedefs.
3076 (unspec_based_fused_function): New class.
3077 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3078 (unspec_based_fused_lane_function): New class.
3079 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3080 typedefs.
3081 (CODE_FOR_MODE1): New macro.
3082 (fixed_insn_function): New class.
3083 (while_comparison): Likewise.
3084 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3085 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3086 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3087 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3088 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3089 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3090 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3091 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3092 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3093 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3094 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3095 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3096 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3097 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3098 Also add an initial argument for unary_convert_narrowt, regardless
3099 of the predication type.
3100 (build_32_64): Allow loads and stores to specify MODE_none.
3101 (build_sv_index64, build_sv_uint_offset): New functions.
3102 (long_type_suffix): New function.
3103 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3104 (binary_imm_long_base, load_gather_sv_base): Likewise.
3105 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3106 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3107 (unary_narrowb_base, unary_narrowt_base): Likewise.
3108 (binary_long_lane_def, binary_long_lane): New shape.
3109 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3110 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3111 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3112 (binary_to_uint_def, binary_to_uint): Likewise.
3113 (binary_wide_def, binary_wide): Likewise.
3114 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3115 (compare_def, compare): Likewise.
3116 (compare_ptr_def, compare_ptr): Likewise.
3117 (load_ext_gather_index_restricted_def,
3118 load_ext_gather_index_restricted): Likewise.
3119 (load_ext_gather_offset_restricted_def,
3120 load_ext_gather_offset_restricted): Likewise.
3121 (load_gather_sv_def): Inherit from load_gather_sv_base.
3122 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3123 (shift_left_imm_def, shift_left_imm): Likewise.
3124 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3125 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3126 (store_scatter_index_restricted_def,
3127 store_scatter_index_restricted): Likewise.
3128 (store_scatter_offset_restricted_def,
3129 store_scatter_offset_restricted): Likewise.
3130 (tbl_tuple_def, tbl_tuple): Likewise.
3131 (ternary_long_lane_def, ternary_long_lane): Likewise.
3132 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3133 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3134 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3135 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3136 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3137 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3138 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3139 (ternary_uint_def, ternary_uint): Likewise.
3140 (unary_convert): Fix typo in comment.
3141 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3142 (unary_long_def, unary_long): Likewise.
3143 (unary_narrowb_def, unary_narrowb): Likewise.
3144 (unary_narrowt_def, unary_narrowt): Likewise.
3145 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3146 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3147 (unary_to_int_def, unary_to_int): Likewise.
3148 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3149 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3150 (svasrd_impl): Delete.
3151 (svcadd_impl::expand): Handle integer operations too.
3152 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3153 new functions to derive the unspec numbers.
3154 (svmla_svmls_lane_impl): Replace with...
3155 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3156 integer operations too.
3157 (svwhile_impl): Rename to...
3158 (svwhilelx_impl): ...this and inherit from while_comparison.
3159 (svasrd): Use unspec_based_function.
3160 (svmla_lane): Use svmla_lane_impl.
3161 (svmls_lane): Use svmls_lane_impl.
3162 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3163 (svwhilele, svwhilelt): Use svwhilelx_impl.
3164 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3165 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3166 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3167 * config/aarch64/aarch64-sve-builtins.def: Include
3168 aarch64-sve-builtins-sve2.def.
3169
3170 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3171
3172 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3173 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3174 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3175 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3176 immediates as well as vector ones.
3177 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3178 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3179 (aarch64_sve_qsub_immediate): Update calls accordingly.
3180
3181 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3182
3183 * config/aarch64/aarch64-sve2.md: Add banner comments.
3184 (<su>mulh<r>s<mode>3): Move further up file.
3185 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3186 (*aarch64_sve2_sra<mode>): Move further down file.
3187 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3188
3189 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3190
3191 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3192 and UNSPEC_WHILEWR.
3193 (while_optab_cmp): Handle them.
3194 * config/aarch64/aarch64-sve.md
3195 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3196 and add a "@" marker.
3197 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3198 instead of gen_aarch64_sve2_while_ptest.
3199 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3200
3201 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3202
3203 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3204 (UNSPEC_WHILELE): ...this.
3205 (UNSPEC_WHILE_LO): Rename to...
3206 (UNSPEC_WHILELO): ...this.
3207 (UNSPEC_WHILE_LS): Rename to...
3208 (UNSPEC_WHILELS): ...this.
3209 (UNSPEC_WHILE_LT): Rename to...
3210 (UNSPEC_WHILELT): ...this.
3211 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3212 (cmp_op, while_optab_cmp): Likewise.
3213 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3214 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3215 (svwhilelt): Likewise.
3216
3217 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3218
3219 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3220 (unary_to_uint): Define.
3221 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3222 (unary_count): Rename to...
3223 (unary_to_uint_def, unary_to_uint): ...this.
3224 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3225
3226 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3227
3228 * config/aarch64/aarch64-sve-builtins-functions.h
3229 (code_for_mode_function): New class.
3230 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3231 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3232 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3233 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3234 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3235
3236 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3237
3238 * config/aarch64/iterators.md (addsub): New code attribute.
3239 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3240 Re-express as...
3241 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3242 in the asm string and attributes. Fix indentation.
3243 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3244 Re-express as...
3245 (@aarch64_sve_<optab><mode>): ...this.
3246 * config/aarch64/aarch64-sve-builtins.h
3247 (function_expander::expand_signed_unpred_op): Delete.
3248 * config/aarch64/aarch64-sve-builtins.cc
3249 (function_expander::expand_signed_unpred_op): Likewise.
3250 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3251 try using code_for_aarch64_sve instead.
3252 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3253 (svqsub_impl): Likewise.
3254 (svqadd, svqsub): Use rtx_code_function instead.
3255
3256 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3257
3258 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3259 (HADDSUB, sur, addsub): Remove them.
3260
3261 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3262
3263 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3264 functions the same way, rather than singling out those that
3265 aren't mapped directly to optabs.
3266
3267 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3268
3269 * target.def (compatible_vector_types_p): New target hook.
3270 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3271 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3272 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3273 * doc/tm.texi: Regenerate.
3274 * gimple-expr.c: Include target.h.
3275 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3276 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3277 function.
3278 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3279 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3280 Use the original predicate if it already has a suitable type.
3281
3282 2020-01-09 Martin Jambor <mjambor@suse.cz>
3283
3284 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3285 resolve_speculation and redirect_call_stmt_to_callee static. Change
3286 return type of set_call_stmt to cgraph_edge *.
3287 * auto-profile.c (afdo_indirect_call): Adjust call to
3288 redirect_call_stmt_to_callee.
3289 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3290 make the this pointer explicit, adjust self-recursive calls and the
3291 call top make_direct. Return the resulting edge.
3292 (cgraph_edge::remove): Make this pointer explicit.
3293 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3294 (cgraph_edge::make_direct): Likewise, adjust call to
3295 resolve_speculation.
3296 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3297 call to set_call_stmt.
3298 (cgraph_update_edges_for_call_stmt_node): Update call to
3299 set_call_stmt and remove.
3300 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3301 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3302 (cgraph_node::create_edge_including_clones): Moved "first" definition
3303 of edge to the block where it was used. Adjusted calls to
3304 set_call_stmt.
3305 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3306 cgraph_edge::remove.
3307 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3308 make_direct and redirect_call_stmt_to_callee.
3309 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3310 resolve_speculation and make_direct.
3311 * ipa-inline-transform.c (inline_transform): Adjust call to
3312 redirect_call_stmt_to_callee.
3313 (check_speculations_1):: Adjust call to resolve_speculation.
3314 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3315 resolve-speculation.
3316 (inline_small_functions): Adjust call to resolve_speculation.
3317 (ipa_inline): Likewise.
3318 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3319 make_direct.
3320 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3321 safe with regards to edge removal, adjust calls to
3322 redirect_call_stmt_to_callee.
3323 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3324 and redirect_call_stmt_to_callee.
3325 * multiple_target.c (create_dispatcher_calls): Adjust call to
3326 redirect_call_stmt_to_callee
3327 (redirect_to_specific_clone): Likewise.
3328 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3329 Adjust calls to cgraph_edge::remove.
3330 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3331 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3332 (expand_call_inline): Adjust call to cgraph_edge::remove.
3333
3334 2020-01-09 Martin Liska <mliska@suse.cz>
3335
3336 * params.opt: Set Optimization for
3337 param_max_speculative_devirt_maydefs.
3338
3339 2020-01-09 Martin Sebor <msebor@redhat.com>
3340
3341 PR middle-end/93200
3342 PR fortran/92956
3343 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3344
3345 2020-01-09 Martin Liska <mliska@suse.cz>
3346
3347 * auto-profile.c (auto_profile): Use opt_for_fn
3348 for a parameter.
3349 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3350 (propagate_vals_across_arith_jfunc): Likewise.
3351 (hint_time_bonus): Likewise.
3352 (incorporate_penalties): Likewise.
3353 (good_cloning_opportunity_p): Likewise.
3354 (perform_estimation_of_a_value): Likewise.
3355 (estimate_local_effects): Likewise.
3356 (ipcp_propagate_stage): Likewise.
3357 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3358 (set_switch_stmt_execution_predicate): Likewise.
3359 (analyze_function_body): Likewise.
3360 * ipa-inline-analysis.c (offline_size): Likewise.
3361 * ipa-inline.c (early_inliner): Likewise.
3362 * ipa-prop.c (ipa_analyze_node): Likewise.
3363 (ipcp_transform_function): Likewise.
3364 * ipa-sra.c (process_scan_results): Likewise.
3365 (ipa_sra_summarize_function): Likewise.
3366 * params.opt: Rename ipcp-unit-growth to
3367 ipa-cp-unit-growth. Add Optimization for various
3368 IPA-related parameters.
3369
3370 2020-01-09 Richard Biener <rguenther@suse.de>
3371
3372 PR middle-end/93054
3373 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3374
3375 2020-01-09 Richard Biener <rguenther@suse.de>
3376
3377 PR tree-optimization/93040
3378 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3379
3380 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3381
3382 * common/config/avr/avr-common.c (avr_option_optimization_table)
3383 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3384
3385 2020-01-09 Martin Liska <mliska@suse.cz>
3386
3387 * cgraphclones.c (symbol_table::materialize_all_clones):
3388 Use cgraph_node::dump_name.
3389
3390 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3391
3392 PR inline-asm/93202
3393 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3394 output_operand_lossage instead of gcc_unreachable.
3395 * doc/md.texi (riscv f constraint): Fix typo.
3396
3397 PR target/93141
3398 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3399 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3400 CONST_SCALAR_INT_P instead of CONST_INT_P.
3401 (*subv<mode>4_1): Rename to ...
3402 (subv<mode>4_1): ... this.
3403 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3404 define_insn_and_split patterns.
3405 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3406 patterns.
3407
3408 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3409
3410 * vec.c (class selftest::count_dtor): New class.
3411 (selftest::test_auto_delete_vec): New test.
3412 (selftest::vec_c_tests): Call it.
3413 * vec.h (class auto_delete_vec): New class template.
3414 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3415
3416 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3417
3418 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3419
3420 2020-01-08 Jim Wilson <jimw@sifive.com>
3421
3422 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3423 use of TLS_MODEL_LOCAL_EXEC when not pic.
3424
3425 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3426
3427 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3428 memory leak.
3429
3430 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3431
3432 PR target/93187
3433 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3434 *stack_protect_set_3 peephole2): Also check that the second
3435 insns source is general_operand.
3436
3437 PR target/93174
3438 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3439 predicate for output operand instead of register_operand.
3440 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3441 memory destination and non-memory operands[2].
3442
3443 2020-01-08 Martin Liska <mliska@suse.cz>
3444
3445 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3446 ::dump_asm_name instead of (::name or ::asm_name).
3447 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3448 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3449 (analyze_functions): Likewise.
3450 (expand_all_functions): Likewise.
3451 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3452 (propagate_bits_across_jump_function): Likewise.
3453 (dump_profile_updates): Likewise.
3454 (ipcp_store_bits_results): Likewise.
3455 (ipcp_store_vr_results): Likewise.
3456 * ipa-devirt.c (dump_targets): Likewise.
3457 * ipa-fnsummary.c (analyze_function_body): Likewise.
3458 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3459 (process_hsa_functions): Likewise.
3460 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3461 (set_alias_uids): Likewise.
3462 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3463 * ipa-inline.c (recursive_inlining): Likewise.
3464 (inline_to_all_callers_1): Likewise.
3465 (ipa_inline): Likewise.
3466 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3467 (ipa_propagate_frequency): Likewise.
3468 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3469 (remove_described_reference): Likewise.
3470 * ipa-pure-const.c (worse_state): Likewise.
3471 (check_retval_uses): Likewise.
3472 (analyze_function): Likewise.
3473 (propagate_pure_const): Likewise.
3474 (propagate_nothrow): Likewise.
3475 (dump_malloc_lattice): Likewise.
3476 (propagate_malloc): Likewise.
3477 (pass_local_pure_const::execute): Likewise.
3478 * ipa-visibility.c (optimize_weakref): Likewise.
3479 (function_and_variable_visibility): Likewise.
3480 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3481 (ipa_discover_variable_flags): Likewise.
3482 * lto-streamer-out.c (output_function): Likewise.
3483 (output_constructor): Likewise.
3484 * tree-inline.c (copy_bb): Likewise.
3485 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3486 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3487
3488 2020-01-08 Richard Biener <rguenther@suse.de>
3489
3490 PR middle-end/93199
3491 * tree-eh.c (sink_clobbers): Update virtual operands for
3492 the first and last stmt only. Add a dry-run capability.
3493 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3494 after CFG manipulations and in RPO order to catch all
3495 secondary opportunities reliably.
3496
3497 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3498
3499 PR target/93182
3500 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3501
3502 2019-01-08 Richard Biener <rguenther@suse.de>
3503
3504 PR middle-end/93199
3505 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3506 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3507 virtual operand, also updating SSA use.
3508 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3509 Update stmt after resetting virtual operand.
3510 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3511 * gimple-iterator.c (gsi_remove): When not removing the stmt
3512 permanently do not delink immediate uses or mark the stmt modified.
3513
3514 2020-01-08 Martin Liska <mliska@suse.cz>
3515
3516 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3517 (ipa_call_context::estimate_size_and_time): Likewise.
3518 (inline_analyze_function): Likewise.
3519
3520 2020-01-08 Martin Liska <mliska@suse.cz>
3521
3522 * cgraph.c (cgraph_node::dump): Use systematically
3523 dump_asm_name.
3524
3525 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3526
3527 Add -nodevicespecs option for avr.
3528
3529 PR target/93182
3530 * config/avr/avr.opt (-nodevicespecs): New driver option.
3531 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3532 "-specs=device-specs/..." if that option is not set.
3533 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3534
3535 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3536
3537 Implement 64-bit double functions for avr.
3538
3539 PR target/92055
3540 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3541 --with-double-comparison.
3542 * doc/install.texi: Document them.
3543 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3544 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3545 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3546 * doc/invoke.texi (AVR Built-in Macros): Document them.
3547 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3548 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3549 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3550
3551 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3552
3553 PR target/93188
3554 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3555 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3556 when only building rm-profile multilibs.
3557
3558 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3559
3560 PR ipa/93084
3561 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3562 lattice for a value to check.
3563 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3564 finite propagation in self-recursive scc.
3565
3566 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3567
3568 * ipa-inline.c (caller_growth_limits): Restore the AND.
3569
3570 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3571
3572 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3573 (VEC_ALLREG_ALT): New iterator.
3574 (VEC_ALLREG_INT_MODE): New iterator.
3575 (VCMP_MODE): New iterator.
3576 (VCMP_MODE_INT): New iterator.
3577 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3578 (vec_cmp<u>v64qidi): New define_expand.
3579 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3580 (vec_cmpu<mode>di_exec): New define_expand.
3581 (vec_cmp<u>v64qidi_exec): New define_expand.
3582 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3583 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3584 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3585 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3586 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3587 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3588 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3589 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3590 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3591 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3592 this.
3593 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3594 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3595
3596 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3597
3598 * config/gcn/constraints.md (DA): Update description and match.
3599 (DB): Likewise.
3600 (Db): New constraint.
3601 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3602 parameter.
3603 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3604 Implement 'Db' mixed immediate type.
3605 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3606 (addcv64si3_dup<exec_vcc>): Delete.
3607 (subcv64si3<exec_vcc>): Rework constraints.
3608 (addv64di3): Rework constraints.
3609 (addv64di3_exec): Rework constraints.
3610 (subv64di3): Rework constraints.
3611 (addv64di3_dup): Delete.
3612 (addv64di3_dup_exec): Delete.
3613 (addv64di3_zext): Rework constraints.
3614 (addv64di3_zext_exec): Rework constraints.
3615 (addv64di3_zext_dup): Rework constraints.
3616 (addv64di3_zext_dup_exec): Rework constraints.
3617 (addv64di3_zext_dup2): Rework constraints.
3618 (addv64di3_zext_dup2_exec): Rework constraints.
3619 (addv64di3_sext_dup2): Rework constraints.
3620 (addv64di3_sext_dup2_exec): Rework constraints.
3621
3622 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3623
3624 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3625 existing target checks.
3626
3627 2020-01-07 Richard Biener <rguenther@suse.de>
3628
3629 * doc/install.texi: Bump minimal supported MPC version.
3630
3631 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3632
3633 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3634 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3635 * langhooks.c: Include stor-layout.h.
3636 (lhd_simulate_enum_decl): New function.
3637 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3638 handle_arm_sve_h for the LTO frontend.
3639 (register_vector_type): Cope with null returns from pushdecl.
3640
3641 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3642
3643 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3644 (aarch64_sve::nvectors_if_data_type): Replace with...
3645 (aarch64_sve::builtin_type_p): ...this.
3646 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3647 (find_vector_type): Delete.
3648 (add_sve_type_attribute): New function.
3649 (lookup_sve_type_attribute): Likewise.
3650 (register_builtin_types): Add an "SVE type" attribute to each type.
3651 (register_tuple_type): Likewise.
3652 (svbool_type_p, nvectors_if_data_type): Delete.
3653 (mangle_builtin_type): Use lookup_sve_type_attribute.
3654 (builtin_type_p): Likewise. Add an overload that returns the
3655 number of constituent vector and predicate registers.
3656 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3657 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3658 instead of aarch64_sve_argument_p.
3659 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3660 (aarch64_pass_by_reference): Likewise.
3661 (aarch64_function_value_1): Likewise.
3662 (aarch64_return_in_memory): Likewise.
3663 (aarch64_layout_arg): Likewise.
3664
3665 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3666
3667 PR tree-optimization/93156
3668 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3669 least significant bit is always clear.
3670
3671 PR tree-optimization/93118
3672 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3673 simplifier with two intermediate conversions.
3674
3675 2020-01-07 Martin Liska <mliska@suse.cz>
3676
3677 * params.opt: Add Optimization for various parameters.
3678
3679 2020-01-07 Martin Liska <mliska@suse.cz>
3680
3681 PR ipa/83411
3682 * doc/extend.texi: Explain cloning for target_clone
3683 attribute.
3684
3685 2020-01-07 Martin Liska <mliska@suse.cz>
3686
3687 PR tree-optimization/92860
3688 * common.opt: Make in Optimization option
3689 as it is affected by -O0, which is an Optimization
3690 option.
3691 * tree-inline.c (tree_inlinable_function_p):
3692 Use opt_for_fn for warn_inline.
3693 (expand_call_inline): Likewise.
3694
3695 2020-01-07 Martin Liska <mliska@suse.cz>
3696
3697 PR tree-optimization/92860
3698 * common.opt: Make flag_ree as optimization
3699 attribute.
3700
3701 2020-01-07 Martin Liska <mliska@suse.cz>
3702
3703 PR optimization/92860
3704 * params.opt: Mark param_min_crossjump_insns with Optimization
3705 keyword.
3706
3707 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3708
3709 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3710 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3711
3712 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3713
3714 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3715 helper function to return the valid addressing formats for a given
3716 hard register and mode.
3717 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3718
3719 * config/rs6000/constraints.md (Q constraint): Update
3720 documentation.
3721 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3722 documentation.
3723
3724 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3725 Use 'Q' for doing vector extract from memory.
3726 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3727 memory.
3728 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3729 doing vector extract from memory.
3730 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3731 extract from memory.
3732
3733 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3734 for the offset being 34-bits when -mcpu=future is used.
3735
3736 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3737
3738 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3739 instead of cmpib_comparison_operator in cmpib patterns.
3740 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3741 of cmpib_comparison_operator. Revise comment.
3742
3743 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3744
3745 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3746 in an IFN_DIV_POW2 node to be equal.
3747
3748 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3749
3750 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3751 (vect_check_scalar_mask): ...this.
3752 (vectorizable_store, vectorizable_load): Update call accordingly.
3753 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3754 argument in calls to conditional internal functions.
3755
3756 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3757
3758 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3759 '0' matching inputs.
3760 (subv64di3_exec): Likewise.
3761
3762 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3763
3764 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3765 * doc/md.texi (movstr): Likewise.
3766
3767 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3768
3769 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3770 clobber.
3771
3772 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3773
3774 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3775 Depend on...
3776 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3777 to a temporary file and use move-if-change to update the real
3778 file where necessary.
3779
3780 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3781
3782 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3783 rather than Upa for CPY /M.
3784
3785 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3786
3787 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3788 immediate.
3789
3790 2020-01-06 Martin Liska <mliska@suse.cz>
3791
3792 PR tree-optimization/92860
3793 * params.opt: Mark param_max_combine_insns with Optimization
3794 keyword.
3795
3796 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3797
3798 PR target/93141
3799 * config/i386/i386.md (SWIDWI): New mode iterator.
3800 (DWI, dwi): Add TImode variants.
3801 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3802 <general_hilo_operand> instead of <general_operand>. Use
3803 CONST_SCALAR_INT_P instead of CONST_INT_P.
3804 (*addv<mode>4_1): Rename to ...
3805 (addv<mode>4_1): ... this.
3806 (QWI): New mode attribute.
3807 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3808 define_insn_and_split patterns.
3809 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3810 patterns.
3811 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3812 <general_hilo_operand> instead of <general_operand>.
3813 (*addcarry<mode>_1): New define_insn.
3814 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3815
3816 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3817
3818 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3819 Use "call" instead of "set".
3820
3821 2020-01-03 Martin Jambor <mjambor@suse.cz>
3822
3823 PR ipa/92917
3824 * ipa-cp.c (print_all_lattices): Skip functions without info.
3825
3826 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3827
3828 PR target/93089
3829 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3830 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3831 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3832 for 'e' simd clones.
3833
3834 PR target/93089
3835 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3836 entry.
3837 (mprefer-vector-width=): Add Save.
3838 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3839 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3840 (ix86_debug_options, ix86_function_specific_print): Adjust
3841 ix86_target_string callers.
3842 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3843 (ix86_valid_target_attribute_tree): Likewise.
3844 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3845 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3846 ix86_target_string caller.
3847
3848 PR target/93110
3849 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3850 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3851 instead of gen_int_shift_amount + convert_modes.
3852
3853 PR rtl-optimization/93088
3854 * loop-iv.c (find_single_def_src): Punt after looking through
3855 128 reg copies for regs with single definitions. Move definitions
3856 to first uses.
3857
3858 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3859
3860 * config/arm/arm-c.c (arm_cpu_builtins): Define
3861 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3862 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3863 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3864 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3865 * config/arm/arm-tables.opt: Regenerated.
3866 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3867 arm_arch_i8mm and arm_arch_bf16 when enabled.
3868 * config/arm/arm.h (TARGET_I8MM): New macro.
3869 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3870 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3871 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3872 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3873 (v8_6_a_simd_variants): New.
3874 (v8_*_a_simd_variants): Add i8mm and bf16.
3875 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3876
3877 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3878
3879 PR ipa/93087
3880 * predict.c (compute_function_frequency): Don't call
3881 warn_function_cold on functions that already have cold attribute.
3882
3883 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3884
3885 PR target/67834
3886 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3887 COMDAT group function labels in .data.rel.ro.local section.
3888 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3889
3890 PR target/93111
3891 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3892 comparison_operator in B and S integer comparisons. Likewise, use
3893 ordered_comparison_operator instead of cmpib_comparison_operator in
3894 cmpib patterns.
3895 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3896
3897 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3898
3899 Update copyright years.
3900
3901 * gcc.c (process_command): Update copyright notice dates.
3902 * gcov-dump.c (print_version): Ditto.
3903 * gcov.c (print_version): Ditto.
3904 * gcov-tool.c (print_version): Ditto.
3905 * gengtype.c (create_file): Ditto.
3906 * doc/cpp.texi: Bump @copying's copyright year.
3907 * doc/cppinternals.texi: Ditto.
3908 * doc/gcc.texi: Ditto.
3909 * doc/gccint.texi: Ditto.
3910 * doc/gcov.texi: Ditto.
3911 * doc/install.texi: Ditto.
3912 * doc/invoke.texi: Ditto.
3913
3914 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3915
3916 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3917 summary.
3918
3919 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3920
3921 PR tree-optimization/93098
3922 * match.pd (popcount): For shift amounts, use integer_onep
3923 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3924 tests. Make sure that precision is power of two larger than or equal
3925 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3926 instead of ULL suffixed constants. Formatting fixes.
3927 \f
3928 Copyright (C) 2020 Free Software Foundation, Inc.
3929
3930 Copying and distribution of this file, with or without modification,
3931 are permitted in any medium without royalty provided the copyright
3932 notice and this notice are preserved.
This page took 0.19869 seconds and 6 git commands to generate.