1 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
4 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
7 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
9 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
10 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
11 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
12 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
13 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
14 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
15 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
16 (V_bf_low, V_bf_cvt_m): New mode attributes.
17 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
18 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
19 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
20 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
21 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
23 2020-03-03 Jakub Jelinek <jakub@redhat.com>
25 PR tree-optimization/93582
26 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
27 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
28 members, initialize them in the constructor and if mask is non-NULL,
29 artificially push_partial_def {} for the portions of the mask that
31 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
32 val and return (void *)-1. Formatting fix.
33 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
35 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
36 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
38 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
40 (visit_stmt): Formatting fix.
42 2020-03-03 Richard Biener <rguenther@suse.de>
44 PR tree-optimization/93946
45 * alias.h (refs_same_for_tbaa_p): Declare.
46 * alias.c (refs_same_for_tbaa_p): New function.
47 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
49 * tree-ssa-scopedtables.h
50 (avail_exprs_stack::lookup_avail_expr): Add output argument
51 giving access to the hashtable entry.
52 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
54 * tree-ssa-dom.c: Include alias.h.
55 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
56 removing redundant store.
57 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
58 (ao_ref_init_from_vn_reference): Adjust prototype.
59 (vn_reference_lookup_pieces): Likewise.
60 (vn_reference_insert_pieces): Likewise.
61 * tree-ssa-sccvn.c: Track base alias set in addition to alias
63 (eliminate_dom_walker::eliminate_stmt): Also check base alias
64 set when removing redundant stores.
65 (visit_reference_op_store): Likewise.
66 * dse.c (record_store): Adjust valdity check for redundant
69 2020-03-03 Jakub Jelinek <jakub@redhat.com>
72 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
74 PR rtl-optimization/94002
75 * explow.c (plus_constant): Punt if cst has VOIDmode and
76 get_pool_mode is different from mode.
78 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
80 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
81 address has an offset which fits the scalling constraint for a
83 (legitimate_scaled_address_p): Update use
84 leigitimate_small_data_address_p.
85 (arc_print_operand): Likewise.
86 (arc_legitimate_address_p): Likewise.
87 (legitimate_small_data_address_p): Likewise.
89 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
91 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
92 (fnmasf4_fpu): Likewise.
94 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
96 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
99 (adddi3_i): Remove pattern.
100 (subdi3_i): Likewise.
102 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
104 * config/arc/arc.md (eh_return): Add length info.
106 2020-03-02 David Malcolm <dmalcolm@redhat.com>
108 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
110 2020-03-02 David Malcolm <dmalcolm@redhat.com>
112 * doc/invoke.texi (Static Analyzer Options): Add
113 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
116 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
119 * config/i386/i386.md (movstrict<mode>): Allow only
120 registers with VALID_INT_MODE_P modes.
122 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
124 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
125 (reduc_insn): Use 'U' and 'B' operand codes.
126 (reduc_<reduc_op>_scal_<mode>): Allow all types.
127 (reduc_<reduc_op>_scal_v64di): Delete.
128 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
129 (*plus_carry_dpp_shr_v64si): Change to ...
130 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
131 (mov_from_lane63_v64di): Change to ...
132 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
133 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
134 Support UNSPEC_MOV_DPP_SHR output formats.
135 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
136 Add "use_extends" reductions.
137 (print_operand_address): Add 'I' and 'U' codes.
138 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
140 2020-03-02 Martin Liska <mliska@suse.cz>
142 * lto-wrapper.c: Fix typo in comment about
143 C++ standard version.
145 2020-03-01 Martin Sebor <msebor@redhat.com>
148 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
150 2020-03-01 Martin Sebor <msebor@redhat.com>
153 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
154 of a pointer in the outermost ADDR_EXPRs.
156 2020-02-28 Jeff Law <law@redhat.com>
158 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
159 * config/v850/v850.c (v850_asm_trampoline_template): Update
162 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
165 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
168 2020-02-28 Martin Liska <mliska@suse.cz>
171 * configure.ac: Improve detection of ld_date by requiring
172 either two dashes or none.
173 * configure: Regenerate.
175 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
177 PR rtl-optimization/93564
178 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
179 do not honor reg alloc order.
181 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
184 * config/aarch64/aarch64.c (aarch64_override_options): Fix
185 misleading warning string.
187 2020-02-27 Martin Sebor <msebor@redhat.com>
189 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
191 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
194 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
195 Split the insn into two parts. This insn only does variable
196 extract from a register.
197 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
198 variable extract from memory.
199 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
200 only does variable extract from a register.
201 (vsx_extract_v4sf_var_load): New insn, do variable extract from
203 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
204 into two parts. This insn only does variable extract from a
206 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
207 do variable extract from memory.
209 2020-02-27 Martin Jambor <mjambor@suse.cz>
210 Feng Xue <fxue@os.amperecomputing.com>
213 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
214 new function calls_same_node_or_its_all_contexts_clone_p.
215 (cgraph_edge_brings_value_p): Use it.
216 (cgraph_edge_brings_value_p): Likewise.
217 (self_recursive_pass_through_p): Return false if caller is a clone.
218 (self_recursive_agg_pass_through_p): Likewise.
220 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
223 * alias.c (ends_tbaa_access_path_p): Break out from ...
224 (component_uses_parent_alias_set_from): ... here.
225 * alias.h (ends_tbaa_access_path_p): Declare.
226 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
227 handle trailing arrays past end of tbaa access path.
228 (aliasing_component_refs_p): ... here; likewise.
229 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
230 path; disambiguate also past end of it.
231 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
234 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
236 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
237 beginning of the file.
238 (vcreate_bf16, vcombine_bf16): New.
239 (vdup_n_bf16, vdupq_n_bf16): New.
240 (vdup_lane_bf16, vdup_laneq_bf16): New.
241 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
242 (vduph_lane_bf16, vduph_laneq_bf16): New.
243 (vset_lane_bf16, vsetq_lane_bf16): New.
244 (vget_lane_bf16, vgetq_lane_bf16): New.
245 (vget_high_bf16, vget_low_bf16): New.
246 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
247 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
248 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
249 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
250 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
251 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
252 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
253 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
254 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
255 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
256 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
257 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
258 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
259 (vreinterpretq_bf16_p128): New.
260 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
261 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
262 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
263 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
264 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
265 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
266 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
267 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
268 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
269 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
270 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
271 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
272 (vreinterpretq_p128_bf16): New.
273 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
275 (V_elem_l): Likewise.
278 (V_DOUBLE): Likewise.
279 (VDQX): Add V4BF and V8BF.
280 (V_two_elem, V_three_elem, V_four_elem): Likewise.
283 (V_double_vector_mode): Likewise.
284 (V_cmp_result): Likewise.
285 (V_uf_sclr): Likewise.
286 (V_sz_elem): Likewise.
287 (Is_d_reg): Likewise.
288 (V_mode_nunits): Likewise.
289 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
291 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
293 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
294 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
295 (<expander><mode>3<exec>): Likewise.
296 (<expander><mode>3): New.
297 (v<expander><mode>3): New.
298 (<expander><mode>3): New.
299 (<expander><mode>3<exec>): Rename to ...
300 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
301 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
303 2020-02-27 Alexandre Oliva <oliva@adacore.com>
305 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
308 2020-02-27 Richard Biener <rguenther@suse.de>
310 PR tree-optimization/93508
311 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
312 non-_CHK variants. Valueize their length arguments.
314 2020-02-27 Richard Biener <rguenther@suse.de>
316 PR tree-optimization/93953
317 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
318 to the hash-map entry.
320 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
322 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
324 2020-02-27 Mark Williams <mwilliams@fb.com>
326 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
327 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
328 -ffile-prefix-map and -fmacro-prefix-map.
329 * lto-streamer-out.c: Include file-prefix-map.h.
330 (lto_output_location): Remap the file part of locations.
332 2020-02-27 Jakub Jelinek <jakub@redhat.com>
335 * gimplify.c (gimplify_init_constructor): Don't promote readonly
336 DECL_REGISTER variables to TREE_STATIC.
338 PR tree-optimization/93582
339 PR tree-optimization/93945
340 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
341 non-zero INTEGER_CST second argument and ref->offset or ref->size
342 not a multiple of BITS_PER_UNIT.
344 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
346 * doc/install.texi (Binaries): Update description of BullFreeware.
348 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
352 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
353 C++ Language Options, Warning Options, and Static Analyzer
354 Options lists. Document negative form of options enabled by
355 default. Move some things around to more accurately sort
356 warnings by category.
357 (C++ Dialect Options, Warning Options, Static Analyzer
358 Options): Document negative form of options when enabled by
359 default. Move some things around to more accurately sort
360 warnings by category. Add some missing index entries.
363 2020-02-26 Carl Love <cel@us.ibm.com>
366 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
367 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
368 for the vector unsigned short arguments. It is also listed as the
369 name of the built-in for arguments vector unsigned short,
370 vector unsigned int and vector unsigned long long built-ins. The
371 name of the builtins for these arguments should be:
372 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
373 __builtin_crypto_vpmsumd respectively.
375 2020-02-26 Richard Biener <rguenther@suse.de>
377 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
378 and load permutation.
380 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
383 * optabs-tree.c (supportable_convert_operation): Reject types with
386 2020-02-26 David Malcolm <dmalcolm@redhat.com>
388 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
390 2020-02-26 Jakub Jelinek <jakub@redhat.com>
392 PR tree-optimization/93820
393 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
394 argument to ALL_INTEGER_CST_P boolean.
395 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
396 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
397 adjacent INTEGER_CST store into merged_store->only_constants like
400 2020-02-25 Jakub Jelinek <jakub@redhat.com>
403 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
405 * cfghooks.c (verify_flow_info): Likewise.
406 * predict.c (combine_predictions_for_bb): Likewise.
407 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
408 sucessor -> successor.
409 (find_traces_1_round): Fix comment typo, destinarion -> destination.
410 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
412 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
413 message typo, sucessors -> successors.
415 2020-02-25 Martin Sebor <msebor@redhat.com>
417 * doc/extend.texi (attribute access): Correct an example.
419 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
421 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
423 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
425 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
426 (VD): Enable for V4BF.
428 (VQ): Enable for V8BF.
431 (VDBL, Vdbl): Add V4BF.
432 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
433 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
434 (bfloat16x8x2_t): Likewise.
435 (bfloat16x4x3_t): Likewise.
436 (bfloat16x8x3_t): Likewise.
437 (bfloat16x4x4_t): Likewise.
438 (bfloat16x8x4_t): Likewise.
439 (vcombine_bf16): New.
440 (vld1_bf16, vld1_bf16_x2): New.
441 (vld1_bf16_x3, vld1_bf16_x4): New.
442 (vld1q_bf16, vld1q_bf16_x2): New.
443 (vld1q_bf16_x3, vld1q_bf16_x4): New.
444 (vld1_lane_bf16): New.
445 (vld1q_lane_bf16): New.
446 (vld1_dup_bf16): New.
447 (vld1q_dup_bf16): New.
450 (vld2_dup_bf16): New.
451 (vld2q_dup_bf16): New.
454 (vld3_dup_bf16): New.
455 (vld3q_dup_bf16): New.
458 (vld4_dup_bf16): New.
459 (vld4q_dup_bf16): New.
460 (vst1_bf16, vst1_bf16_x2): New.
461 (vst1_bf16_x3, vst1_bf16_x4): New.
462 (vst1q_bf16, vst1q_bf16_x2): New.
463 (vst1q_bf16_x3, vst1q_bf16_x4): New.
464 (vst1_lane_bf16): New.
465 (vst1q_lane_bf16): New.
473 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
475 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
476 (VALL_F16): Likewise.
477 (VALLDI_F16): Likewise.
480 (vswap_width_name): Likewise.
481 (VSWAP_WIDTH): Likewise.
485 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
486 (vget_lane_bf16, vgetq_lane_bf16): New.
488 (vdup_n_bf16, vdupq_n_bf16): New.
489 (vdup_lane_bf16, vdup_laneq_bf16): New.
490 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
491 (vduph_lane_bf16, vduph_laneq_bf16): New.
492 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
493 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
494 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
495 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
496 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
497 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
498 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
499 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
500 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
501 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
502 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
503 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
504 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
505 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
506 (vreinterpretq_bf16_p128): New.
507 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
508 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
509 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
510 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
511 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
512 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
513 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
514 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
515 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
516 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
517 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
518 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
519 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
520 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
521 (vreinterpretq_p128_bf16): New.
523 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
525 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
526 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
527 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
528 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
529 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
530 * config/arm/iterators.md (VSF2BF): New attribute.
531 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
532 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
533 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
535 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
537 * config/arm/arm.md (required_for_purecode): New attribute.
538 (enabled): Handle required_for_purecode.
539 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
540 work with -mpure-code.
542 2020-02-25 Jakub Jelinek <jakub@redhat.com>
544 PR rtl-optimization/93908
545 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
548 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
550 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
552 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
554 * doc/install.texi (--enable-checking): Adjust wording.
556 2020-02-25 Richard Biener <rguenther@suse.de>
558 PR tree-optimization/93868
559 * tree-vect-slp.c (slp_copy_subtree): New function.
560 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
561 re-arranging stmts in it.
563 2020-02-25 Jakub Jelinek <jakub@redhat.com>
566 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
567 dummy function and remove it at the end.
570 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
571 paramter -> parameter.
572 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
573 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
575 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
577 * doc/install.texi (--enable-checking): Properly document current
579 (--enable-stage1-checking): Minor clarification about bootstrap.
581 2020-02-24 David Malcolm <dmalcolm@redhat.com>
584 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
585 -fanalyzer-checker=taint is also required.
586 (-fanalyzer-checker=): Note that providing this option enables the
587 given checker, and doing so may be required for checkers that are
590 2020-02-24 David Malcolm <dmalcolm@redhat.com>
592 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
593 significant control flow events; add a "3" which shows all
594 control flow events; the old "3" becomes "4".
596 2020-02-24 Jakub Jelinek <jakub@redhat.com>
598 PR tree-optimization/93582
599 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
600 pd.offset and pd.size to be counted in bits rather than bytes, add
601 support for maxsizei that is not a multiple of BITS_PER_UNIT and
602 handle bitfield stores and loads.
603 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
604 uncomparable quantities - bytes vs. bits. Allow push_partial_def
605 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
606 pd.offset/pd.size to be counted in bits rather than bytes.
607 Formatting fix. Rename shadowed len variable to buflen.
609 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
610 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
613 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
614 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
615 * opts-common.c (parse_options_from_collect_gcc_options): New function.
616 (prepend_xassembler_to_collect_as_options): Likewise.
617 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
618 (prepend_xassembler_to_collect_as_options): Likewise.
619 * lto-opts.c (lto_write_options): Stream assembler options
620 in COLLECT_AS_OPTIONS.
621 * lto-wrapper.c (xassembler_options_error): New static variable.
622 (get_options_from_collect_gcc_options): Move parsing options code to
623 parse_options_from_collect_gcc_options and call it.
624 (merge_and_complain): Validate -Xassembler options.
625 (append_compiler_options): Handle OPT_Xassembler.
626 (run_gcc): Append command line -Xassembler options to
628 * doc/invoke.texi: Add documentation about using Xassembler
631 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
633 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
635 (riscv_rtx_costs): Update cost model for LTGT.
637 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
639 PR rtl-optimization/93564
640 * ira-color.c (struct update_cost_queue_elem): New member start.
641 (queue_update_cost, get_next_update_cost): Add new arg start.
642 (allocnos_conflict_p): New function.
643 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
644 Add checking conflicts with allocnos_conflict_p.
645 (update_costs_from_prefs, restore_costs_from_copies): Adjust
646 update_costs_from_allocno calls.
647 (update_conflict_hard_regno_costs): Add checking conflicts with
648 allocnos_conflict_p. Adjust calls of queue_update_cost and
649 get_next_update_cost.
650 (assign_hard_reg): Adjust calls of queue_update_cost. Add
652 (bucket_allocno_compare_func): Restore previous version.
654 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
656 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
657 double-word size when handling aggregate return values.
658 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
659 that homogeneous SFmode and DFmode aggregates are passed and returned
660 in general registers.
662 2020-02-21 Jakub Jelinek <jakub@redhat.com>
665 * opts.c (print_filtered_help): Translate help before appending
666 messages to it rather than after that.
668 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
670 PR rtl-optimization/PR92989
671 * lra-lives.c (process_bb_lives): Restore the original order
672 of the bb liveness update. Call make_hard_regno_dead for each
673 register clobbered at the start of an EH receiver.
675 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
678 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
679 self-recursively generated.
681 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
684 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
687 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
689 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
690 Document new target supports option.
692 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
694 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
695 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
696 * config/arm/iterators.md (MATMUL): New iterator.
697 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
698 (mmla_sfx): New attribute.
699 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
700 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
701 (UNSPEC_MATMUL_US): New.
703 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
705 * config/arm/arm.md: Prevent scalar shifts from being used when big
708 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
709 Richard Biener <rguenther@suse.de>
711 PR tree-optimization/93586
712 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
713 after mismatched array refs; do not sure type size information to
714 recover from unmatched referneces with !flag_strict_aliasing_p.
716 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
718 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
719 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
720 (scatter_store<mode>): Rename to ...
721 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
722 (scatter<mode>_exec): Delete. Move contents ...
723 (mask_scatter_store<mode>): ... here, and rename that to ...
724 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
725 Remove mode conversion.
726 (mask_gather_load<mode>): Rename to ...
727 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
728 Remove mode conversion.
729 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
731 2020-02-21 Martin Jambor <mjambor@suse.cz>
733 PR tree-optimization/93845
734 * tree-sra.c (verify_sra_access_forest): Only test access size of
737 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
739 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
740 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
741 (addv64di3_exec): Likewise.
742 (subv64di3): Likewise.
743 (subv64di3_exec): Likewise.
744 (addv64di3_zext): Likewise.
745 (addv64di3_zext_exec): Likewise.
746 (addv64di3_zext_dup): Likewise.
747 (addv64di3_zext_dup_exec): Likewise.
748 (addv64di3_zext_dup2): Likewise.
749 (addv64di3_zext_dup2_exec): Likewise.
750 (addv64di3_sext_dup2): Likewise.
751 (addv64di3_sext_dup2_exec): Likewise.
752 (<expander>v64di3): Likewise.
753 (<expander>v64di3_exec): Likewise.
754 (*<reduc_op>_dpp_shr_v64di): Likewise.
755 (*plus_carry_dpp_shr_v64di): Likewise.
756 * config/gcn/gcn.md (adddi3): Likewise.
757 (addptrdi3): Likewise.
758 (<expander>di3): Likewise.
760 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
762 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
764 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
766 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
767 support. Use aarch64_emit_mult instead of emitting multiplication
768 instructions directly.
769 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
770 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
772 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
774 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
775 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
776 instead of emitting multiplication instructions directly.
777 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
778 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
779 (@aarch64_frecps<mode>): New expanders.
781 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
783 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
784 on and produce uint64_ts rather than ints.
785 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
786 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
788 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
790 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
791 an unused xmsk register when handling approximate rsqrt.
793 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
795 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
796 flag_finite_math_only condition.
798 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
801 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
802 to destination operand for shufps alternative.
803 (*vec_extractv2si_1): Ditto.
805 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
808 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
811 2020-02-20 Martin Liska <mliska@suse.cz>
814 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
816 2020-02-20 Martin Liska <mliska@suse.cz>
819 * common/config/avr/avr-common.c: Remote trailing "|".
821 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
823 * collect2.c (maybe_run_lto_and_relink): Fix typo in
826 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
828 PR tree-optimization/93767
829 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
830 access-size bias from the offset calculations for negative strides.
832 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
834 * collect2.c (c_file, o_file): Make const again.
835 (ldout,lderrout, dump_ld_file): Remove.
836 (tool_cleanup): Avoid calling not signal-safe functions.
837 (maybe_run_lto_and_relink): Avoid possible signal handler
838 access to unintialzed memory (lto_o_files).
839 (main): Avoid leaking temp files in $TMPDIR.
840 Initialize c_file/o_file with concat, which avoids exposing
841 uninitialized memory to signal handler, which calls unlink(!).
842 Avoid calling maybe_unlink when the main function returns,
843 since the atexit handler is already doing this.
844 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
846 2020-02-19 Martin Jambor <mjambor@suse.cz>
848 PR tree-optimization/93776
849 * tree-sra.c (create_access): Do not create zero size accesses.
850 (get_access_for_expr): Do not search for zero sized accesses.
852 2020-02-19 Martin Jambor <mjambor@suse.cz>
854 PR tree-optimization/93667
855 * tree-sra.c (scalarizable_type_p): Return false if record fields
856 do not follow wach other.
858 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
860 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
861 rather than fmv.x.s/fmv.s.x.
863 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
865 * config/aarch64/aarch64-simd-builtins.def
866 (intrinsic_vec_smult_lo_): New.
867 (intrinsic_vec_umult_lo_): Likewise.
868 (vec_widen_smult_hi_): Likewise.
869 (vec_widen_umult_hi_): Likewise.
870 * config/aarch64/aarch64-simd.md
871 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
872 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
873 (vmull_high_s16): Likewise.
874 (vmull_high_s32): Likewise.
875 (vmull_high_u8): Likewise.
876 (vmull_high_u16): Likewise.
877 (vmull_high_u32): Likewise.
878 (vmull_s8): Likewise.
879 (vmull_s16): Likewise.
880 (vmull_s32): Likewise.
881 (vmull_u8): Likewise.
882 (vmull_u16): Likewise.
883 (vmull_u32): Likewise.
885 2020-02-18 Martin Liska <mliska@suse.cz>
887 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
888 bootstrap by missing removal of invalid sanity check.
890 2020-02-18 Martin Liska <mliska@suse.cz>
893 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
894 Always compare LHS of gimple_assign.
896 2020-02-18 Martin Liska <mliska@suse.cz>
899 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
900 and return type of functions.
901 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
902 Drop MALLOC attribute for void functions.
903 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
904 malloc_state for a new VOID clone.
906 2020-02-18 Martin Liska <mliska@suse.cz>
909 * common.opt: Add -fprofile-reproducibility.
910 * doc/invoke.texi: Document it.
911 * value-prof.c (dump_histogram_value):
912 Document and support behavior for counters[0]
913 being a negative value.
914 (get_nth_most_common_value): Handle negative
915 counters[0] in respect to flag_profile_reproducible.
917 2020-02-18 Jakub Jelinek <jakub@redhat.com>
920 * cgraph.c (verify_speculative_call): Use speculative_id instead of
921 speculative_uid in messages. Remove trailing whitespace from error
922 message. Use num_speculative_call_targets instead of
923 num_speculative_targets in a message.
924 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
925 edge messages and stmt instead of cal_stmt in reference message.
927 PR tree-optimization/93780
928 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
929 before calling build_vector_type.
930 (execute_update_addresses_taken): Likewise.
933 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
934 typo, functoin -> function.
935 * tree.c (free_lang_data_in_decl): Fix comment typo,
936 functoin -> function.
937 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
939 2020-02-17 David Malcolm <dmalcolm@redhat.com>
941 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
943 (print_option_information): Don't call get_option_url if URLs
946 2020-02-17 Alexandre Oliva <oliva@adacore.com>
948 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
949 handling of register_common-less targets.
951 2020-02-17 Martin Liska <mliska@suse.cz>
954 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
956 2020-02-17 Martin Liska <mliska@suse.cz>
959 * config/rs6000/rs6000.c (rs6000_option_override_internal):
962 2020-02-17 Martin Liska <mliska@suse.cz>
965 * config/rx/elf.opt: Fix typo.
967 2020-02-17 Richard Biener <rguenther@suse.de>
970 * opts-global.c (print_ignored_options): Use inform and
973 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
976 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
978 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
981 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
982 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
984 2020-02-15 Jason Merrill <jason@redhat.com>
986 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
988 2020-02-15 Jakub Jelinek <jakub@redhat.com>
990 PR tree-optimization/93744
991 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
992 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
993 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
994 sure @2 in the first and @1 in the other patterns has no side-effects.
996 2020-02-15 David Malcolm <dmalcolm@redhat.com>
997 Bernd Edlinger <bernd.edlinger@hotmail.de>
1001 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1002 * configure.ac (--with-diagnostics-urls): New configuration
1003 option, based on --with-diagnostics-color.
1004 (DIAGNOSTICS_URLS_DEFAULT): New define.
1005 * config.h: Regenerate.
1006 * configure: Regenerate.
1007 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1008 DIAGNOSTICS_URLS_DEFAULT from configure-time
1009 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1010 and TERM_URLS environment variable.
1011 * diagnostic-url.h (diagnostic_url_format): New enum type.
1012 (diagnostic_urls_enabled_p): rename to...
1013 (determine_url_format): ... this, and change return type.
1014 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1015 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1016 the linux console, and mingw.
1017 (diagnostic_urls_enabled_p): rename to...
1018 (determine_url_format): ... this, and adjust.
1019 * pretty-print.h (pretty_printer::show_urls): rename to...
1020 (pretty_printer::url_format): ... this, and change to enum.
1021 * pretty-print.c (pretty_printer::pretty_printer,
1022 pp_begin_url, pp_end_url, test_urls): Adjust.
1023 * doc/install.texi (--with-diagnostics-urls): Document the new
1024 configuration option.
1025 (--with-diagnostics-color): Document the existing interaction
1026 with GCC_COLORS better.
1027 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1028 vindex reference. Update description of defaults based on the above.
1029 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1030 interacts with GCC_COLORS.
1032 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1035 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1036 conjunction with TARGET_GNU_TLS in early return.
1038 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1040 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1041 the mode is not wider than UNITS_PER_WORD.
1043 2020-02-14 Martin Jambor <mjambor@suse.cz>
1045 PR tree-optimization/93516
1046 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1047 access of the same type as the parent.
1048 (propagate_subaccesses_from_lhs): Likewise.
1050 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1053 * config/i386/avx512vbmi2intrin.h
1054 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1055 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1056 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1057 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1058 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1059 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1060 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1061 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1062 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1063 of lacking a closing parenthesis.
1064 * config/i386/avx512vbmi2vlintrin.h
1065 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1066 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1067 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1068 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1069 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1070 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1071 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1072 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1073 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1074 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1075 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1076 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1077 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1078 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1079 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1080 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1081 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1082 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1084 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1087 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1088 the target function entry.
1090 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1092 * common/config/arc/arc-common.c (arc_option_optimization_table):
1093 Disable if-conversion step when optimized for size.
1095 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1097 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1098 R12-R15 are always in ARCOMPACT16_REGS register class.
1099 * config/arc/arc.opt (mq-class): Deprecate.
1100 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1102 * doc/invoke.texi (mq-class): Update text.
1103 * common/config/arc/arc-common.c (arc_option_optimization_table):
1106 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1108 * config/arc/arc.c (arc_insn_cost): New function.
1109 (TARGET_INSN_COST): Define.
1110 * config/arc/arc.md (cost): New attribute.
1111 (add_n): Use arc_nonmemory_operand.
1112 (ashlsi3_insn): Likewise, also update constraints.
1113 (ashrsi3_insn): Likewise.
1114 (rotrsi3): Likewise.
1115 (add_shift): Likewise.
1116 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1118 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1120 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1122 (umulsidi_600): Likewise.
1124 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1127 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1128 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1129 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1130 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1131 pass __A to the builtin followed by __W instead of __A followed by
1133 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1134 _mm512_mask_popcnt_epi64): Likewise.
1135 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1136 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1137 _mm256_mask_popcnt_epi64): Likewise.
1139 PR tree-optimization/93582
1140 * fold-const.h (shift_bytes_in_array_left,
1141 shift_bytes_in_array_right): Declare.
1142 * fold-const.c (shift_bytes_in_array_left,
1143 shift_bytes_in_array_right): New function, moved from
1144 gimple-ssa-store-merging.c, no longer static.
1145 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1146 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1147 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1148 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1149 shift_bytes_in_array.
1150 (verify_shift_bytes_in_array): Rename to ...
1151 (verify_shift_bytes_in_array_left): ... this. Use
1152 shift_bytes_in_array_left instead of shift_bytes_in_array.
1153 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1154 instead of verify_shift_bytes_in_array.
1155 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1156 / native_interpret_expr where the store covers all needed bits,
1157 punt on PDP-endian, otherwise allow all involved offsets and sizes
1158 not to be byte-aligned.
1161 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1162 use const_0_to_255_operand predicate instead of immediate_operand.
1163 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1164 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1165 vgf2p8affineinvqb_<mode><mask_name>,
1166 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1167 const_0_to_255_operand predicated operands.
1169 2020-02-12 Jeff Law <law@redhat.com>
1171 * config/h8300/h8300.md (comparison shortening peepholes): Use
1172 a mode iterator to merge the HImode and SImode peepholes.
1174 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1177 * real.c (is_even): Make static. Function comment fix.
1178 (is_halfway_below): Make static, don't assert R is not inf/nan,
1179 instead return false for those. Small formatting fixes.
1181 2020-02-12 Martin Sebor <msebor@redhat.com>
1184 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1185 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1186 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1187 (strlen_check_and_optimize_call): Adjust callee name.
1189 2020-02-12 Jeff Law <law@redhat.com>
1191 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1192 (and (xor)) variant. Combine other two into single peephole.
1194 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1196 PR rtl-optimization/93565
1197 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1199 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1201 * config/aarch64/aarch64-simd.md
1202 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1203 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1204 generating separate ADDV and zero_extend patterns.
1205 * config/aarch64/iterators.md (VDQV_E): New iterator.
1207 2020-02-12 Jeff Law <law@redhat.com>
1209 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1210 expanders, splits, etc.
1211 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1212 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1213 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1214 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1215 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1216 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1218 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1220 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1223 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1224 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1225 TARGET_AVX512DQ from condition.
1226 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1227 instead of <mask_mode512bit_condition> in condition. If
1228 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1230 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1233 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1236 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1238 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1240 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1241 where strlen is more legible.
1242 (rs6000_builtin_vectorized_libmass): Ditto.
1243 (rs6000_print_options_internal): Ditto.
1245 2020-02-11 Martin Sebor <msebor@redhat.com>
1247 PR tree-optimization/93683
1248 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1250 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1252 * config/rs6000/predicates.md (cint34_operand): Rename the
1253 -mprefixed-addr option to be -mprefixed.
1254 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1255 the -mprefixed-addr option to be -mprefixed.
1256 (OTHER_FUTURE_MASKS): Likewise.
1257 (POWERPC_MASKS): Likewise.
1258 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1259 the -mprefixed-addr option to be -mprefixed. Change error
1260 messages to refer to -mprefixed.
1261 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1263 (rs6000_legitimate_offset_address_p): Likewise.
1264 (rs6000_mode_dependent_address): Likewise.
1265 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1266 "-mprefixed" for target attributes and pragmas.
1267 (address_to_insn_form): Rename the -mprefixed-addr option to be
1269 (rs6000_adjust_insn_length): Likewise.
1270 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1271 -mprefixed-addr option to be -mprefixed.
1272 (ASM_OUTPUT_OPCODE): Likewise.
1273 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1274 -mprefixed-addr option to be -mprefixed.
1275 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1276 -mprefixed-addr option to be prefixed. Change the option from
1277 being undocumented to being documented.
1278 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1279 -mprefixed option. Update the -mpcrel documentation to mention
1282 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1284 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1285 including FIRST_PSEUDO_REGISTER - 1.
1286 * ira-color.c (print_hard_reg_set): Ditto.
1288 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1290 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1291 (USTERNOP_QUALIFIERS): New define.
1292 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1293 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1294 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1295 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1296 * config/arm/arm_neon.h (vusdot_s32): New.
1297 (vusdot_lane_s32): New.
1298 (vusdotq_lane_s32): New.
1299 (vsudot_lane_s32): New.
1300 (vsudotq_lane_s32): New.
1301 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1302 * config/arm/iterators.md (DOTPROD_I8MM): New.
1303 (sup, opsuffix): Add <us/su>.
1304 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1305 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1307 2020-02-11 Richard Biener <rguenther@suse.de>
1309 PR tree-optimization/93661
1310 PR tree-optimization/93662
1311 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1313 * tree-sra.c (get_access_for_expr): Likewise.
1315 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1318 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1319 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1320 Change condition from TARGET_AVX2 to TARGET_AVX.
1322 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1325 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1326 argument of strncmp.
1328 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1330 Try to generate zero-based comparisons.
1331 * config/cris/cris.c (cris_reduce_compare): New function.
1332 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1333 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1334 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1336 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1339 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1340 in Thumb state and also as a destination in Arm state. Add T16
1343 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1345 * md.texi (Define Subst): Match closing paren in example.
1347 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1351 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1352 arguments of strncmp.
1354 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1357 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1358 but different source value.
1359 (adjust_callers_for_value_intersection): New function.
1360 (gather_edges_for_value): Adjust order of callers to let a
1361 non-self-recursive caller be the first element.
1362 (self_recursive_pass_through_p): Add a new parameter "simple", and
1363 check generalized self-recursive pass-through jump function.
1364 (self_recursive_agg_pass_through_p): Likewise.
1365 (find_more_scalar_values_for_callers_subset): Compute value from
1366 pass-through jump function for self-recursive.
1367 (intersect_with_plats): Cleanup previous implementation code for value
1368 itersection with self-recursive call edge.
1369 (intersect_with_agg_replacements): Likewise.
1370 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1371 function for self-recursive call edge. Cleanup previous implementation
1372 code for value intersection with self-recursive call edge.
1373 (decide_whether_version_node): Remove dead callers and adjust order
1374 to let a non-self-recursive caller be the first element.
1376 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1378 * recog.c: Move pass_split_before_sched2 code in front of
1379 pass_split_before_regstack.
1380 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1381 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1382 (rest_of_handle_split_before_sched2): Remove.
1383 (pass_split_before_sched2::execute): Unconditionally call
1385 (enable_split_before_sched2): New function.
1386 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1387 (pass_split_before_regstack::gate): Ditto.
1388 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1389 Update name check for renamed split4 pass.
1390 * config/sh/sh.c (register_sh_passes): Update pass insertion
1391 point for renamed split4 pass.
1393 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1395 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1396 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1397 copying them around between host and target.
1399 2020-02-08 Andrew Pinski <apinski@marvell.com>
1402 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1403 STRICT_ALIGNMENT also.
1405 2020-02-08 Jim Wilson <jimw@sifive.com>
1408 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1410 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1411 Jakub Jelinek <jakub@redhat.com>
1414 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1415 xmm16-xmm31 call-used even in 64-bit ms-abi.
1417 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1419 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1420 (simd_ummla, simd_usmmla): Likewise.
1421 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1422 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1423 (vusmmlaq_s32): New.
1425 2020-02-07 Richard Biener <rguenther@suse.de>
1428 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1429 skipping unreachable regions.
1430 (optimize_inline_calls): Skip folding stmts when we didn't
1433 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1436 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1437 Don't return aggregates with only SFmode and DFmode in SSE
1439 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1441 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1444 * config/rs6000/rs6000-logue.c
1445 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1446 if it fails, move rs into end_addr and retry. Add
1447 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1448 the insn pattern doesn't describe well what exactly happens to
1452 * config/i386/predicates.md (avx_identity_operand): Remove.
1453 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1454 (avx_<castmode><avxsizesuffix>_<castmode>,
1455 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1456 a VEC_CONCAT of the operand and UNSPEC_CAST.
1457 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1458 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1462 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1463 recog_data.insn if distance_non_agu_define changed it.
1465 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1468 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1469 we only had X-FORM (reg+reg) addressing for vectors. Also before
1470 ISA 3.0, we only had X-FORM addressing for scalars in the
1471 traditional Altivec registers.
1473 2020-02-06 <zhongyunde@huawei.com>
1474 Vladimir Makarov <vmakarov@redhat.com>
1476 PR rtl-optimization/93561
1477 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1478 hard register range.
1480 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1482 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1485 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1487 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1488 where the low and the high 32 bits are equal to each other specially,
1489 with an rldimi instruction.
1491 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1493 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1495 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1497 * config/arm/arm-tables.opt: Regenerate.
1499 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1502 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1503 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1504 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1506 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1508 PR rtl-optimization/87763
1509 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1511 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1513 * config/aarch64/aarch64-simd-builtins.def
1514 (bfmlaq): New built-in function.
1515 (bfmlalb): New built-in function.
1516 (bfmlalt): New built-in function.
1517 (bfmlalb_lane): New built-in function.
1518 (bfmlalt_lane): New built-in function.
1519 * config/aarch64/aarch64-simd.md
1520 (aarch64_bfmmlaqv4sf): New pattern.
1521 (aarch64_bfmlal<bt>v4sf): New pattern.
1522 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1523 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1524 (vbfmlalbq_f32): New intrinsic.
1525 (vbfmlaltq_f32): New intrinsic.
1526 (vbfmlalbq_lane_f32): New intrinsic.
1527 (vbfmlaltq_lane_f32): New intrinsic.
1528 (vbfmlalbq_laneq_f32): New intrinsic.
1529 (vbfmlaltq_laneq_f32): New intrinsic.
1530 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1531 (bt): New int attribute.
1533 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1535 * config/i386/i386.md (*pushtf): Emit "#" instead of
1536 calling gcc_unreachable in insn output.
1539 (*pushsf_rex64): Ditto for alternatives other than 1.
1540 (*pushsf): Ditto for alternatives other than 1.
1542 2020-02-06 Martin Liska <mliska@suse.cz>
1544 PR gcov-profile/91971
1545 PR gcov-profile/93466
1546 * coverage.c (coverage_init): Revert mangling of
1547 path into filename. It can lead to huge filename length.
1548 Creation of subfolders seem more natural.
1550 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1553 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1554 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1555 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1557 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1560 * config/i386/predicates.md (avx_identity_operand): New predicate.
1561 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1562 define_insn_and_split.
1565 * omp-low.c (use_pointer_for_field): For nested constructs, also
1566 look for map clauses on target construct.
1567 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1568 taskreg_nesting_level.
1571 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1572 shared clause, call omp_notice_variable on outer context if any.
1574 2020-02-05 Jason Merrill <jason@redhat.com>
1577 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1578 non-zero address even if weak and not yet defined.
1580 2020-02-05 Martin Sebor <msebor@redhat.com>
1582 PR tree-optimization/92765
1583 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1584 * tree-ssa-strlen.c (compute_string_length): Remove.
1585 (determine_min_objsize): Remove.
1586 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1587 Avoid using type size as the upper bound on string length.
1588 (handle_builtin_string_cmp): Add an argument. Adjust.
1589 (strlen_check_and_optimize_call): Pass additional argument to
1590 handle_builtin_string_cmp.
1592 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1594 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1595 (*pushdi2_rex64 peephole2): Unconditionally split after
1597 (*ashl<mode>3_doubleword): Ditto.
1598 (*<shift_insn><mode>3_doubleword): Ditto.
1600 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1603 * config/rs6000/rs6000.c (get_vector_offset): Fix
1605 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1607 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1609 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1612 (Special Functions for Debugging the Analyzer): Update description
1613 of __analyzer_dump_exploded_nodes.
1615 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1618 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1619 include sets and not clobbers in the vzeroupper pattern.
1620 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1621 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1622 (*avx_vzeroupper_1): New define_insn_and_split.
1625 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1626 don't run when !optimize.
1627 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1630 2020-02-05 Richard Biener <rguenther@suse.de>
1633 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1634 checks before matching calls.
1636 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1638 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1639 function comment typo.
1642 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1643 simd_clone_create failed when i == 0, adjust clone->nargs by
1646 2020-02-05 Martin Liska <mliska@suse.cz>
1649 * doc/invoke.texi: Document that one should
1650 not combine ASLR and -fpch.
1652 2020-02-04 Richard Biener <rguenther@suse.de>
1654 PR tree-optimization/93538
1655 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1657 2020-02-04 Richard Biener <rguenther@suse.de>
1659 PR tree-optimization/91123
1660 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1661 (vn_walk_cb_data::last_vuse): New member.
1662 (vn_walk_cb_data::saved_operands): Likewsie.
1663 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1664 (vn_walk_cb_data::push_partial_def): Use finish.
1665 (vn_reference_lookup_2): Update last_vuse and use finish if
1666 we've saved operands.
1667 (vn_reference_lookup_3): Use finish and update calls to
1668 push_partial_defs everywhere. When translating through
1669 memcpy or aggregate copies save off operands and alias-set.
1670 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1671 operation for redundant store removal.
1673 2020-02-04 Richard Biener <rguenther@suse.de>
1675 PR tree-optimization/92819
1676 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1677 generating more stmts than before.
1679 2020-02-04 Martin Liska <mliska@suse.cz>
1681 * config/arm/arm.c (arm_gen_far_branch): Move the function
1682 outside of selftests.
1684 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1686 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1687 function to adjust PC-relative vector addresses.
1688 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1689 handle vectors with PC-relative addresses.
1691 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1693 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1695 (hard_reg_and_mode_to_addr_mask): Delete.
1696 (rs6000_adjust_vec_address): If the original vector address
1697 was REG+REG or REG+OFFSET and the element is not zero, do the add
1698 of the elements in the original address before adding the offset
1699 for the vector element. Use address_to_insn_form to validate the
1700 address using the register being loaded, rather than guessing
1701 whether the address is a DS-FORM or DQ-FORM address.
1703 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1705 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1706 to calculate the offset in memory from the start of a vector of a
1707 particular element. Add code to keep the element number in
1708 bounds if the element number is variable.
1709 (rs6000_adjust_vec_address): Move calculation of offset of the
1710 vector element to get_vector_offset.
1711 (rs6000_split_vec_extract_var): Do not do the initial AND of
1712 element here, move the code to get_vector_offset.
1714 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1716 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1719 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1721 * config/rs6000/constraints.md: Improve documentation.
1723 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1726 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1727 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1729 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1731 * config.gcc: Remove "carrizo" support.
1732 * config/gcn/gcn-opts.h (processor_type): Likewise.
1733 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1734 * config/gcn/gcn.opt (gpu_type): Likewise.
1735 * config/gcn/t-omp-device: Likewise.
1737 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1740 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1741 * config/arm/arm.c (arm_gen_far_branch): New function
1743 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1745 2020-02-03 Julian Brown <julian@codesourcery.com>
1746 Tobias Burnus <tobias@codesourcery.com>
1748 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1750 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1753 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1754 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1757 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1759 PR rtl-optimization/91333
1760 * ira-color.c (struct allocno_color_data): Add member
1762 (init_allocno_threads): Set the member up.
1763 (bucket_allocno_compare_func): Add compare hard reg
1766 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1768 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1770 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1771 * config.in: Regenerated.
1772 * configure: Regenerated.
1773 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1774 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1775 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1777 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1779 * configure: Regenerate.
1781 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1783 PR rtl-optimization/91333
1784 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1785 reg preferences comparison up.
1787 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1789 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1790 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1791 aarch64-sve-builtins-base.h.
1792 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1793 aarch64-sve-builtins-base.cc.
1794 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1795 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1797 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1798 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1799 (svcvtnt): New functions.
1800 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1801 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1802 (svcvtnt): New functions.
1803 (svcvt): Add a form that converts f32 to bf16.
1804 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1805 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1807 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1808 Treat B as bfloat16_t.
1809 (ternary_bfloat_lane_base): New class.
1810 (ternary_bfloat_def): Likewise.
1811 (ternary_bfloat): New shape.
1812 (ternary_bfloat_lane_def): New class.
1813 (ternary_bfloat_lane): New shape.
1814 (ternary_bfloat_lanex2_def): New class.
1815 (ternary_bfloat_lanex2): New shape.
1816 (ternary_bfloat_opt_n_def): New class.
1817 (ternary_bfloat_opt_n): New shape.
1818 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1819 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1820 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1821 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1822 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1823 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1824 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1825 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1826 the pattern off the narrow mode instead of the wider one.
1827 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1828 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1829 (sve_fp_op): Handle them.
1830 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1831 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1833 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1835 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1836 * config/aarch64/aarch64-modes.def (BF): Move definition before
1837 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1838 (SVE_MODES): Handle BF modes.
1839 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1841 (aarch64_full_sve_mode): Likewise.
1842 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1844 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1845 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1846 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1847 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1849 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1851 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1852 (TYPES_all_data): Add bf16.
1853 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1854 (register_tuple_type): Increase buffer size.
1855 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1856 (bf16): New type suffix.
1857 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1858 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1859 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1860 Change type from all_data to all_arith.
1861 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1864 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1865 Matthew Malcomson <matthew.malcomson@arm.com>
1866 Richard Sandiford <richard.sandiford@arm.com>
1868 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1869 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1870 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1871 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1872 __ARM_FEATURE_MATMUL_FP64.
1873 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1874 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1875 be disabled at the same time.
1876 (f32mm): New extension.
1877 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1878 (AARCH64_FL_F64MM): Bump to the next bit up.
1879 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1880 (TARGET_SVE_F64MM): New macros.
1881 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1882 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1883 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1884 (UNSPEC_ZIP2Q): New unspeccs.
1885 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1886 (optab, sur, perm_insn): Handle the new unspecs.
1887 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1888 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1889 TARGET_SVE_F64MM instead of separate tests.
1890 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1891 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1892 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1893 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1894 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1895 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1896 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1897 (TYPES_s_signed): New macro.
1898 (TYPES_s_integer): Use it.
1899 (TYPES_d_float): New macro.
1900 (TYPES_d_data): Use it.
1901 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1902 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1903 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1904 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1905 (svmmla): New shape.
1906 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1907 template parameters.
1908 (ternary_resize2_lane_base): Likewise.
1909 (ternary_resize2_base): New class.
1910 (ternary_qq_lane_base): Likewise.
1911 (ternary_intq_uintq_lane_def): Likewise.
1912 (ternary_intq_uintq_lane): New shape.
1913 (ternary_intq_uintq_opt_n_def): New class
1914 (ternary_intq_uintq_opt_n): New shape.
1915 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1916 (ternary_uintq_intq_def): New class.
1917 (ternary_uintq_intq): New shape.
1918 (ternary_uintq_intq_lane_def): New class.
1919 (ternary_uintq_intq_lane): New shape.
1920 (ternary_uintq_intq_opt_n_def): New class.
1921 (ternary_uintq_intq_opt_n): New shape.
1922 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1923 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1924 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1925 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1927 (svdotprod_lane_impl): ...this new class.
1928 (svmmla_impl, svusdot_impl): New classes.
1929 (svdot_lane): Update to use svdotprod_lane_impl.
1930 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1931 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1933 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1934 function, with no types defined.
1935 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1936 AARCH64_FL_I8MM functions.
1937 (svmmla): New AARCH64_FL_F32MM function.
1938 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1939 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1940 AARCH64_FL_F64MM function.
1941 (REQUIRED_EXTENSIONS):
1943 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1945 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1948 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1950 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1951 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1952 (*movti_internal): Do not check for
1953 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1954 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1955 just after check for TARGET_AVX.
1956 (*movdf_internal): Ditto.
1957 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1958 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1959 * config/i386/sse.md (mov<mode>_internal): Only check
1960 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1961 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1962 (<sse>_andnot<mode>3<mask_name>): Move check for
1963 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1964 (<code><mode>3<mask_name>): Ditto.
1965 (*andnot<mode>3): Ditto.
1966 (*andnottf3): Ditto.
1967 (*<code><mode>3): Ditto.
1968 (*<code>tf3): Ditto.
1969 (*andnot<VI:mode>3): Remove
1970 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1971 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1972 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1973 (sse4_1_blendv<ssemodesuffix>): Ditto.
1974 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1975 Explain that tune applies to 128bit instructions only.
1977 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1979 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1980 to definition of hsa_kernel_description. Parse assembly to find SGPR
1981 and VGPR count of kernel and store in hsa_kernel_description.
1983 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1985 PR rtl-optimization/91838
1986 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1987 to truncate if allowed or reject combination.
1989 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1991 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1992 (find_inv_vars_cb): Likewise.
1994 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1996 * calls.c (special_function_p): Split out the check for DECL_NAME
1997 being non-NULL and fndecl being extern at file scope into a
1998 new maybe_special_function_p and call it. Drop check for fndecl
1999 being non-NULL that was after a usage of DECL_NAME (fndecl).
2000 * tree.h (maybe_special_function_p): New inline function.
2002 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2004 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2005 (mask_gather_load<mode>): ... here, and zero-initialize the
2007 (maskload<mode>di): Zero-initialize the destination.
2010 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2013 * doc/analyzer.texi (Limitations): Note that constraints on
2014 floating-point values are currently ignored.
2016 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2019 * symtab.c (symtab_node::noninterposable_alias): If localalias
2020 already exists, but is not usable, append numbers after it until
2021 a unique name is found. Formatting fix.
2024 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2027 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2029 * config/gcn/gcn.c (print_operand): Handle LTGT.
2030 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2032 2020-01-30 Richard Biener <rguenther@suse.de>
2034 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2035 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2037 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2039 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2040 without a DECL in .data.rel.ro.local.
2042 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2045 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2049 * config/i386/sse.md
2050 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2051 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2052 any_extend code iterator instead of always zero_extend.
2053 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2054 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2055 Use any_extend code iterator instead of always zero_extend.
2056 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2057 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2058 Use any_extend code iterator instead of always zero_extend.
2059 (*sse2_pmovmskb_ext): New define_insn.
2060 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2063 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2064 (*popcountsi2_zext_falsedep): New define_insn.
2066 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2068 * config.in: Regenerated.
2069 * configure: Regenerated.
2071 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2074 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2075 LLVM's assembler changed the default in version 9.
2077 2020-01-24 Jeff Law <law@redhat.com>
2079 PR tree-optimization/89689
2080 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2082 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2086 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2088 PR rtl-optimization/87763
2089 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2090 simplification to handle subregs as well as bare regs.
2091 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2093 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2096 * ira.c (ira): Revert use of simplified LRA algorithm.
2098 2020-01-29 Martin Jambor <mjambor@suse.cz>
2100 PR tree-optimization/92706
2101 * tree-sra.c (struct access): Fields first_link, last_link,
2102 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2103 next_rhs_queued and grp_rhs_queued respectively, new fields
2104 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2105 (struct assign_link): Field next renamed to next_rhs, new field
2106 next_lhs. Updated comment.
2107 (work_queue_head): Renamed to rhs_work_queue_head.
2108 (lhs_work_queue_head): New variable.
2109 (add_link_to_lhs): New function.
2110 (relink_to_new_repr): Also relink LHS lists.
2111 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2112 (add_access_to_lhs_work_queue): New function.
2113 (pop_access_from_work_queue): Renamed to
2114 pop_access_from_rhs_work_queue.
2115 (pop_access_from_lhs_work_queue): New function.
2116 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2118 (child_would_conflict_in_lacc): Renamed to
2119 child_would_conflict_in_acc. Adjusted parameter names.
2120 (create_artificial_child_access): New parameter set_grp_read, use it.
2121 (subtree_mark_written_and_enqueue): Renamed to
2122 subtree_mark_written_and_rhs_enqueue.
2123 (propagate_subaccesses_across_link): Renamed to
2124 propagate_subaccesses_from_rhs.
2125 (propagate_subaccesses_from_lhs): New function.
2126 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2129 2020-01-29 Martin Jambor <mjambor@suse.cz>
2131 PR tree-optimization/92706
2132 * tree-sra.c (struct access): Adjust comment of
2133 grp_total_scalarization.
2134 (find_access_in_subtree): Look for single children spanning an entire
2136 (scalarizable_type_p): Allow register accesses, adjust callers.
2137 (completely_scalarize): Remove function.
2138 (scalarize_elem): Likewise.
2139 (create_total_scalarization_access): Likewise.
2140 (sort_and_splice_var_accesses): Do not track total scalarization
2142 (analyze_access_subtree): New parameter totally, adjust to new meaning
2143 of grp_total_scalarization.
2144 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2145 (can_totally_scalarize_forest_p): New function.
2146 (create_total_scalarization_access): Likewise.
2147 (create_total_access_and_reshape): Likewise.
2148 (total_should_skip_creating_access): Likewise.
2149 (totally_scalarize_subtree): Likewise.
2150 (analyze_all_variable_accesses): Perform total scalarization after
2151 subaccess propagation using the new functions above.
2152 (initialize_constant_pool_replacements): Output initializers by
2153 traversing the access tree.
2155 2020-01-29 Martin Jambor <mjambor@suse.cz>
2157 * tree-sra.c (verify_sra_access_forest): New function.
2158 (verify_all_sra_access_forests): Likewise.
2159 (create_artificial_child_access): Set parent.
2160 (analyze_all_variable_accesses): Call the verifier.
2162 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2164 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2165 if called on indirect edge.
2166 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2167 speculative call if needed.
2169 2020-01-29 Richard Biener <rguenther@suse.de>
2171 PR tree-optimization/93428
2172 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2173 permutation when the load node is created.
2174 (vect_analyze_slp_instance): Re-use it here.
2176 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2178 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2180 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2182 PR rtl-optimization/93272
2183 * ira-lives.c (process_out_of_region_eh_regs): New function.
2184 (process_bb_node_lives): Call it.
2186 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2188 * coverage.c (read_counts_file): Make error message lowercase.
2190 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2192 * profile-count.c (profile_quality_display_names): Fix ordering.
2194 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2197 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2198 hash only when edge is first within the sequence.
2199 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2200 (symbol_table::create_edge): Do not set target_prob.
2201 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2203 (cgraph_edge::make_speculative): Drop target_prob parameter.
2204 (cgraph_edge::speculative_call_info): Remove.
2205 (cgraph_edge::first_speculative_call_target): New member function.
2206 (update_call_stmt_hash_for_removing_direct_edge): New function.
2207 (cgraph_edge::resolve_speculation): Rewrite to new API.
2208 (cgraph_edge::speculative_call_for_target): New member function.
2209 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2210 multiple speculation targets.
2211 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2213 (verify_speculative_call): Verify that targets form an interval.
2214 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2215 (cgraph_edge::first_speculative_call_target): New member function.
2216 (cgraph_edge::next_speculative_call_target): New member function.
2217 (cgraph_edge::speculative_call_target_ref): New member function.
2218 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2219 (cgraph_edge): Remove target_prob.
2220 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2221 Fix handling of speculative calls.
2222 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2223 * ipa-fnsummary.c (analyze_function_body): Likewise.
2224 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2225 * ipa-profile.c (dump_histogram): Fix formating.
2226 (ipa_profile_generate_summary): Watch for overflows.
2227 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2228 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2229 (update_indirect_edges_after_inlining): Update to new API.
2230 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2232 * profile-count.h: (profile_probability::adjusted): New.
2233 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2234 updating of profile.
2235 * value-prof.c (gimple_ic_transform): Rename to ...
2236 (dump_ic_profile): ... this one; update dumping.
2237 (stream_in_histogram_value): Fix formating.
2238 (gimple_value_profile_transformations): Update.
2240 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2243 * config/i386/i386.md (*movoi_internal_avx): Remove
2244 TARGET_SSE_TYPELESS_STORES check.
2245 (*movti_internal): Prefer TARGET_AVX over
2246 TARGET_SSE_TYPELESS_STORES.
2247 (*movtf_internal): Likewise.
2248 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2249 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2250 from TARGET_SSE_TYPELESS_STORES.
2252 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2254 * diagnostic-core.h (warning_at): Rename overload to...
2255 (warning_meta): ...this.
2256 (emit_diagnostic_valist): Delete decl of overload taking
2257 diagnostic_metadata.
2258 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2259 (warning_at): Rename overload taking diagnostic_metadata to...
2260 (warning_meta): ...this.
2262 2020-01-28 Richard Biener <rguenther@suse.de>
2264 PR tree-optimization/93439
2265 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2266 * tree-cfg.c (move_sese_region_to_fn): ... here.
2267 (verify_types_in_gimple_reference): Verify used cliques are
2270 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2273 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2274 argument of a pointer to struct gcc_options and pass it to
2275 parse_mtune_ctrl_str.
2276 (ix86_function_specific_restore): Pass opts to
2277 set_ix86_tune_features.
2278 (ix86_option_override_internal): Likewise.
2279 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2280 gcc_options and use it for x_ix86_tune_ctrl_string.
2282 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2284 PR rtl-optimization/87763
2285 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2286 simplification to handle subregs as well as bare regs.
2287 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2289 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2291 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2292 for reduction chains that (now) include a call.
2294 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2296 PR tree-optimization/92822
2297 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2298 out the don't-care elements of a vector whose significant elements
2299 are duplicates, make the don't-care elements duplicates too.
2301 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2303 PR tree-optimization/93434
2304 * tree-predcom.c (split_data_refs_to_components): Record which
2305 components have had aliasing loads removed. Prevent store-store
2306 commoning for all such components.
2308 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2311 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2312 -1 or is_vshift is true, use new_vector with number of elts npatterns
2313 rather than new_unary_operation.
2315 PR tree-optimization/93454
2316 * gimple-fold.c (fold_array_ctor_reference): Perform
2317 elt_size.to_uhwi () just once, instead of calling it in every
2318 iteration. Punt if that value is above size of the temporary
2319 buffer. Decrease third native_encode_expr argument when
2320 bufoff + elt_sz is above size of buf.
2322 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2324 * config/mips/mips.c (mips_declare_object_name)
2325 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2327 2020-01-27 Martin Liska <mliska@suse.cz>
2329 PR gcov-profile/93403
2330 * tree-profile.c (gimple_init_gcov_profiler): Generate
2331 both __gcov_indirect_call_profiler_v4 and
2332 __gcov_indirect_call_profiler_v4_atomic.
2334 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2337 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2339 (@aarch64_split_simd_mov<mode>): Use it.
2340 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2341 Leave the vec_extract patterns to handle 2-element vectors.
2342 (aarch64_simd_mov_from_<mode>high): Likewise.
2343 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2344 (vec_extractv2dfv1df): Likewise.
2346 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2348 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2349 jump conditions for *compare_condjump<GPI:mode>.
2351 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2354 * digraph.cc (test_edge::test_edge): Specify template for base
2357 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2359 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2361 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2363 * config/arc/arc-protos.h (gen_mlo): Remove.
2364 (gen_mhi): Likewise.
2365 * config/arc/arc.c (AUX_MULHI): Define.
2366 (arc_must_save_reister): Special handling for r58/59.
2367 (arc_compute_frame_size): Consider mlo/mhi registers.
2368 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2370 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2371 mlo/mhi name selection.
2372 (arc_restore_callee_saves): Don't early restore blink when ISR.
2373 (arc_expand_prologue): Add mlo/mhi saving.
2374 (arc_expand_epilogue): Add mlo/mhi restoring.
2377 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2378 numbering when MUL64 option is used.
2379 (DWARF2_FRAME_REG_OUT): Define.
2380 * config/arc/arc.md (arc600_stall): New pattern.
2381 (VUNSPEC_ARC_ARC600_STALL): Define.
2382 (mulsi64): Use correct mlo/mhi registers.
2383 (mulsi_600): Clean it up.
2384 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2386 (mhi_operand): Likewise.
2388 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2389 Petro Karashchenko <petro.karashchenko@ring.com>
2391 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2392 attributes if needed.
2393 (prepare_move_operands): Generate special unspec instruction for
2395 (arc_isuncached_mem_p): Propagate uncached attribute to each
2397 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2398 (VUNSPEC_ARC_STDI): Likewise.
2399 (ALLI): New mode iterator.
2400 (mALLI): New mode attribute.
2401 (lddi): New instruction pattern.
2403 (stdidi_split): Split instruction for architectures which are not
2404 supporting ll64 option.
2405 (lddidi_split): Likewise.
2407 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2409 PR rtl-optimization/92989
2410 * lra-lives.c (process_bb_lives): Update the live-in set before
2411 processing additional clobbers.
2413 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2415 PR rtl-optimization/93170
2416 * cselib.c (cselib_invalidate_regno_val): New function, split out
2418 (cselib_invalidate_regno): ...here.
2419 (cselib_invalidated_by_call_p): New function.
2420 (cselib_process_insn): Iterate over all the hard-register entries in
2421 REG_VALUES and invalidate any that cross call-clobbered registers.
2423 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2425 * dojump.c (split_comparison): Use HONOR_NANS rather than
2426 HONOR_SNANS when splitting LTGT.
2428 2020-01-27 Martin Liska <mliska@suse.cz>
2431 * opts.c (print_filtered_help): Exclude language-specific
2432 options from --help=common unless enabled in all FEs.
2434 2020-01-27 Martin Liska <mliska@suse.cz>
2436 * opts.c (print_help): Exclude params from
2437 all except --help=param.
2439 2020-01-27 Martin Liska <mliska@suse.cz>
2442 * config/i386/i386-features.c (make_resolver_func):
2443 Align the code with ppc64 target implementation.
2444 Do not generate a unique name for resolver function.
2446 2020-01-27 Richard Biener <rguenther@suse.de>
2448 PR tree-optimization/93397
2449 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2450 converted reduction chain SLP graph adjustment.
2452 2020-01-26 Marek Polacek <polacek@redhat.com>
2455 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2458 2020-01-26 Jason Merrill <jason@redhat.com>
2461 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2464 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2466 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2467 (rx_setmem): Likewise.
2469 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2472 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2473 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2474 drop <di> from constraint of last operand.
2477 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2478 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2479 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2481 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2484 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2487 2020-01-24 Jeff Law <law@redhat.com>
2489 PR tree-optimization/92788
2490 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2493 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2496 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2497 *avx_vperm_broadcast_<mode>,
2498 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2499 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2500 Move before avx2_perm<mode>/avx512f_perm<mode>.
2503 * simplify-rtx.c (simplify_const_unary_operation,
2504 simplify_const_binary_operation): Punt for mode precision above
2505 MAX_BITSIZE_MODE_ANY_INT.
2507 2020-01-24 Andrew Pinski <apinski@marvell.com>
2509 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2512 2020-01-24 Jeff Law <law@redhat.com>
2515 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2516 for REGs. Call output_operand_lossage to get more reasonable
2519 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2521 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2522 gcn_fp_compare_operator.
2523 (vec_cmpu<mode>di): Use gcn_compare_operator.
2524 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2525 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2526 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2527 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2528 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2529 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2530 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2531 gcn_fp_compare_operator.
2532 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2533 gcn_fp_compare_operator.
2534 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2535 gcn_fp_compare_operator.
2536 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2537 gcn_fp_compare_operator.
2539 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2541 * doc/install.texi (Cross-Compiler-Specific Options): Document
2542 `--with-toolexeclibdir' option.
2544 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2546 * target.def (flags_regnum): Also mention effect on delay slot filling.
2547 * doc/tm.texi: Regenerate.
2549 2020-01-23 Jeff Law <law@redhat.com>
2551 PR translation/90162
2552 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2554 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2557 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2560 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2562 PR rtl-optimization/93402
2563 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2566 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2568 * config.in: Regenerated.
2569 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2570 for TARGET_LIBC_GNUSTACK.
2571 * configure: Regenerated.
2572 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2573 found to be 2.31 or greater.
2575 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2577 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2579 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2580 (mips_asm_file_end): New function. Delegate to
2581 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2582 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2584 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2587 * config/i386/i386-modes.def (POImode): New mode.
2588 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2589 * config/i386/i386.md (DPWI): New mode attribute.
2590 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2592 (QPWI): ... this. Use POI instead of OI for TImode.
2593 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2594 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2597 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2600 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2602 (speculation_tracker_rev): New pattern.
2603 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2604 Use speculation_tracker_rev to track the inverse condition.
2606 2020-01-23 Richard Biener <rguenther@suse.de>
2608 PR tree-optimization/93381
2609 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2610 alias-set of the def as argument and record the first one.
2611 (vn_walk_cb_data::first_set): New member.
2612 (vn_reference_lookup_3): Pass the alias-set of the current def
2613 to push_partial_def. Fix alias-set used in the aggregate copy
2615 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2616 * real.c (clear_significand_below): Fix out-of-bound access.
2618 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2621 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2622 New define_insn patterns.
2624 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2626 * doc/sourcebuild.texi (check-function-bodies): Add an
2627 optional target/xfail selector.
2629 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2631 PR rtl-optimization/93124
2632 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2633 bare USE and CLOBBER insns.
2635 2020-01-22 Andrew Pinski <apinski@marvell.com>
2637 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2639 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2642 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2643 diagnostic_manager into "ana" namespace.
2644 * selftest-run-tests.c (selftest::run_tests): Update for move of
2645 selftest::run_analyzer_selftests to
2646 ana::selftest::run_analyzer_selftests.
2648 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2650 * cfgexpand.c (union_stack_vars): Update the size.
2652 2020-01-22 Richard Biener <rguenther@suse.de>
2654 PR tree-optimization/93381
2655 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2656 throughout, handle all conversions the same.
2658 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2661 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2662 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2663 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2664 Call force_reg on high_in2 unconditionally.
2666 2020-01-22 Martin Liska <mliska@suse.cz>
2668 PR tree-optimization/92924
2669 * profile.c (compute_value_histograms): Divide
2672 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2675 * output.h (assemble_name_resolve): Declare.
2676 * varasm.c (assemble_name_resolve): New function.
2677 (assemble_name): Use it.
2678 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2680 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2682 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2683 update_web_docs_git instead of update_web_docs_svn.
2685 2020-01-21 Andrew Pinski <apinski@marvell.com>
2688 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2689 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2690 (*tlsgd_small_<mode>): Likewise.
2691 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2692 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2693 register. Convert that register back to dest using convert_mode.
2695 2020-01-21 Jim Wilson <jimw@sifive.com>
2697 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2700 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2701 Uros Bizjak <ubizjak@gmail.com>
2704 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2706 (legitimize_tls_address): Do GNU2 TLS address computation in
2707 ptr_mode and zero-extend result to Pmode.
2708 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2709 :P with :PTR and Pmode with ptr_mode.
2710 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2711 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2712 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2714 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2717 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2718 the last two operands are CONST_INT_P before using them as such.
2720 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2722 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2723 to get the integer element types.
2725 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2727 * config/aarch64/aarch64-sve-builtins.h
2728 (function_expander::convert_to_pmode): Declare.
2729 * config/aarch64/aarch64-sve-builtins.cc
2730 (function_expander::convert_to_pmode): New function.
2731 (function_expander::get_contiguous_base): Use it.
2732 (function_expander::prepare_gather_address_operands): Likewise.
2733 * config/aarch64/aarch64-sve-builtins-sve2.cc
2734 (svwhilerw_svwhilewr_impl::expand): Likewise.
2736 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2739 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2740 cfun->machine->label_is_assembled.
2741 (aarch64_print_patchable_function_entry): New.
2742 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2743 * config/aarch64/aarch64.h (struct machine_function): New field,
2746 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2749 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2752 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2755 * cgraph.c (cgraph_edge::resolve_speculation,
2756 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2757 call_stmt_site_hash.
2759 2020-01-21 Martin Liska <mliska@suse.cz>
2761 * config/rs6000/rs6000.c (common_mode_defined): Remove
2764 2020-01-21 Richard Biener <rguenther@suse.de>
2766 PR tree-optimization/92328
2767 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2768 type when value-numbering same-sized store by inserting a
2770 (eliminate_dom_walker::eliminate_stmt): When eliminating
2771 a redundant store handle bit-reinterpretation of the same value.
2773 2020-01-21 Andrew Pinski <apinski@marvel.com>
2776 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2778 (prepare_block_for_update): This. Use a worklist instead of
2781 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2783 * gcc/config/arm/arm.c (clear_operation_p):
2784 Initialise last_regno, skip first iteration
2785 based on the first_set value and use ints instead
2786 of the unnecessary HOST_WIDE_INTs.
2788 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2791 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2792 compare_mode other than SFmode or DFmode.
2794 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2797 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2798 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2799 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2801 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2803 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2805 2020-01-20 Andrew Pinski <apinski@marvell.com>
2808 * targhooks.c (default_print_patchable_function_entry): Use
2809 output_asm_insn to emit the nop instruction.
2811 2020-01-20 Fangrui Song <maskray@google.com>
2814 * targhooks.c (default_print_patchable_function_entry): Align to
2817 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2820 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2821 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2822 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2823 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2824 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2825 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2826 Remove the {q} suffix from lea.
2827 (*tls_dynamic_gnu2_call_64): Renamed to ...
2828 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2829 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2830 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2831 Pass Pmode to gen_tls_dynamic_gnu2_64.
2833 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2835 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2837 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2839 * config/aarch64/aarch64-sve-builtins-base.cc
2840 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2842 2020-01-20 Richard Biener <rguenther@suse.de>
2845 * dwarf2out.c (prune_unused_types): Unconditionally mark
2846 called function DIEs.
2848 2020-01-20 Martin Liska <mliska@suse.cz>
2850 PR tree-optimization/93199
2851 * tree-eh.c (struct leh_state): Add
2852 new field outer_non_cleanup.
2853 (cleanup_is_dead_in): Pass leh_state instead
2854 of eh_region. Add a checking that state->outer_non_cleanup
2855 points to outer non-clean up region.
2856 (lower_try_finally): Record outer_non_cleanup
2858 (lower_catch): Likewise.
2859 (lower_eh_filter): Likewise.
2860 (lower_eh_must_not_throw): Likewise.
2861 (lower_cleanup): Likewise.
2863 2020-01-20 Richard Biener <rguenther@suse.de>
2865 PR tree-optimization/93094
2866 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2867 (vect_transform_loop): Likewise.
2868 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2869 loop_vectorized_call to vect_transform_loop.
2870 * tree-vect-loop.c (vect_transform_loop): Pass down
2871 loop_vectorized_call to vect_loop_versioning.
2872 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2873 the earlier discovered loop_vectorized_call.
2875 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2877 * doc/contribute.texi: Update for SVN -> Git transition.
2878 * doc/install.texi: Likewise.
2880 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2883 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2884 speculative targets.
2885 (verify_speculative_call): New function
2886 (cgraph_node::verify_node): Use it.
2887 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2890 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2893 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2894 (cgraph_edge::make_direct): Remove all indirect targets.
2895 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2896 (cgraph_node::verify_node): Verify that only one call_stmt or
2897 lto_stmt_uid is set.
2898 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2900 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2901 (lto_output_ref): Simplify streaming of stmt.
2902 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2904 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2906 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2907 Mark parameter unused.
2909 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2911 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2913 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2915 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2917 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2919 * Makefile.in: Add coroutine-passes.o.
2920 * builtin-types.def (BT_CONST_SIZE): New.
2921 (BT_FN_BOOL_PTR): New.
2922 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2923 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2924 * coroutine-builtins.def: New file.
2925 * coroutine-passes.cc: New file.
2926 * function.h (struct GTY function): Add a bit to indicate that the
2927 function is a coroutine component.
2928 * internal-fn.c (expand_CO_FRAME): New.
2929 (expand_CO_YIELD): New.
2930 (expand_CO_SUSPN): New.
2931 (expand_CO_ACTOR): New.
2932 * internal-fn.def (CO_ACTOR): New.
2936 * passes.def: Add pass_coroutine_lower_builtins,
2937 pass_coroutine_early_expand_ifns.
2938 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2939 (make_pass_coroutine_early_expand_ifns): New.
2940 * doc/invoke.texi: Document the fcoroutines command line
2943 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2945 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2948 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2949 after checking the argument is a REG. Don't use REGNO (reg)
2950 again to set last_regno, reuse regno variable instead.
2952 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2954 * doc/analyzer.texi (Limitations): Add note about NaN.
2956 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2957 Sudakshina Das <sudi.das@arm.com>
2959 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2960 and valid immediate.
2961 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2962 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2963 * config/arm/constraints.md (Pg): New.
2964 * config/arm/predicates.md (long_shift_imm): New.
2965 (arm_reg_or_long_shift_imm): Likewise.
2966 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2967 (thumb2_lsll): Likewise.
2970 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2971 Sudakshina Das <sudi.das@arm.com>
2973 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2974 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2975 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2976 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2977 * config/arm/thumb2.md (thumb2_asrl): New.
2978 (thumb2_lsll): Likewise.
2980 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2982 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2985 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2987 * gdbinit.in (help-gcc-hooks): New command.
2988 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2989 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2992 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2994 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2995 correct target macro.
2997 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2999 * config/aarch64/aarch64-protos.h
3000 (aarch64_sve_ld1ro_operand_p): New.
3001 * config/aarch64/aarch64-sve-builtins-base.cc
3002 (class load_replicate): New.
3003 (class svld1ro_impl): New.
3004 (class svld1rq_impl): Change to inherit from load_replicate.
3005 (svld1ro): New sve intrinsic function base.
3006 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3007 New DEF_SVE_FUNCTION.
3008 * config/aarch64/aarch64-sve-builtins-base.h
3009 (svld1ro): New decl.
3010 * config/aarch64/aarch64-sve-builtins.cc
3011 (function_expander::add_mem_operand): Modify assert to allow
3013 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3015 * config/aarch64/aarch64.c
3016 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3017 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3018 (aarch64_sve_ld1ro_operand_p): New.
3019 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3020 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3021 * config/aarch64/predicates.md
3022 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3024 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3026 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3027 Introduce this ACLE specified predefined macro.
3028 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3029 (fp): Disabling this disables f64mm.
3030 (simd): Disabling this disables f64mm.
3031 (fp16): Disabling this disables f64mm.
3032 (sve): Disabling this disables f64mm.
3033 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3034 (AARCH64_ISA_F64MM): New.
3035 (TARGET_F64MM): New.
3036 * doc/invoke.texi (f64mm): Document new option.
3038 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3040 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3041 (neoversen1_tunings): Likewise.
3043 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3046 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3047 Add assert to ensure prolog has been emitted.
3048 (aarch64_split_atomic_op): Likewise.
3049 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3050 Use epilogue_completed rather than reload_completed.
3051 (aarch64_atomic_exchange<mode>): Likewise.
3052 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3053 (atomic_nand<mode>): Likewise.
3054 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3055 (atomic_fetch_nand<mode>): Likewise.
3056 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3057 (atomic_nand_fetch<mode>): Likewise.
3059 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3062 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3064 (REVERSE_CONDITION): Delete.
3065 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3066 (CCFP_CCFPE): Likewise.
3067 (e): New mode attribute.
3068 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3069 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3070 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3071 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3072 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3073 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3074 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3075 name of generator from gen_ccmpdi to gen_ccmpccdi.
3076 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3077 the previous comparison but aren't able to, use the new ccmp_rev
3080 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3082 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3083 than testing directly for INTEGER_CST.
3084 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3086 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3088 PR tree-optimization/93292
3089 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3090 get_vectype_for_scalar_type returns NULL.
3092 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3094 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3095 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3097 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3099 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3101 * params.opt: (max-predicted-iterations): Set bounds.
3102 * predict.c (real_almost_one, real_br_prob_base,
3103 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3104 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3105 probabilities; do not truncate to reg_br_prob_bases.
3106 (estimate_loops_at_level): Pass max_cyclic_prob.
3107 (estimate_loops): Compute max_cyclic_prob.
3108 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3110 * profile-count.c (profile_probability::to_sreal): New.
3111 * profile-count.h (class sreal): Move up in file.
3112 (profile_probability::to_sreal): Declare.
3114 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3117 (arm_invalid_conversion): New function for target hook.
3118 (arm_invalid_unary_op): New function for target hook.
3119 (arm_invalid_binary_op): New function for target hook.
3121 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3123 * config.gcc: Add arm_bf16.h.
3124 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3125 (arm_simd_builtin_std_type): Add BFmode.
3126 (arm_init_simd_builtin_types): Define element types for vector types.
3127 (arm_init_bf16_types): New function.
3128 (arm_init_builtins): Add arm_init_bf16_types function call.
3129 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3130 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3131 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3132 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3133 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3134 (arm_mangle_type): Add __bf16.
3135 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3136 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3137 arm_bf16_ptr_type_node.
3138 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3139 define_split between ARM registers.
3140 * config/arm/arm_bf16.h: New file.
3141 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3142 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3144 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3145 * config/arm/vfp.md: Add BFmode to movhf patterns.
3147 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3148 Andre Vieira <andre.simoesdiasvieira@arm.com>
3150 * config/arm/arm-cpus.in (mve, mve_float): New features.
3151 (dsp, mve, mve.fp): New options.
3152 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3153 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3154 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3156 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3157 Thomas Preud'homme <thomas.preudhomme@arm.com>
3159 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3161 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3162 error for using -mcmse when targeting Armv8.1-M Mainline.
3164 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3165 Thomas Preud'homme <thomas.preudhomme@arm.com>
3167 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3168 address in r4 when targeting Armv8.1-M Mainline.
3169 (nonsecure_call_value_internal): Likewise.
3170 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3171 a register match_operand again. Emit BLXNS when targeting
3173 (nonsecure_call_value_reg_thumb2): Likewise.
3175 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3176 Thomas Preud'homme <thomas.preudhomme@arm.com>
3178 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3179 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3180 variable as true when floating-point ABI is not hard. Replace
3181 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3182 Generate VLSTM and VLLDM instruction respectively before and
3183 after a function call to cmse_nonsecure_call function.
3184 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3185 (VUNSPEC_VLLDM): Likewise.
3186 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3187 (lazy_load_multiple_insn): Likewise.
3189 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3190 Thomas Preud'homme <thomas.preudhomme@arm.com>
3192 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3193 (arm_emit_vfp_multi_reg_pop): Likewise.
3194 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3195 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3196 restore callee-saved VFP registers.
3198 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3199 Thomas Preud'homme <thomas.preudhomme@arm.com>
3201 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3202 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3203 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3204 callee-saved GPRs as well as clear ip register before doing a nonsecure
3205 call then restore callee-saved GPRs after it when targeting
3207 (arm_reorg): Adapt to function rename.
3209 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3210 Thomas Preud'homme <thomas.preudhomme@arm.com>
3212 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3213 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3214 clear_vfp_multiple pattern based on a new vfp parameter.
3215 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3216 targeting Armv8.1-M Mainline.
3217 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3218 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3219 whether VFP registers are available before looking call_used_regs for a
3221 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3222 of prototype of clear_operation_p.
3223 (clear_vfp_multiple_operation): New predicate.
3224 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3225 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3227 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3228 Thomas Preud'homme <thomas.preudhomme@arm.com>
3230 * config/arm/arm-protos.h (clear_operation_p): Declare.
3231 * config/arm/arm.c (clear_operation_p): New function.
3232 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3233 targeting Armv8.1-M Mainline or successor.
3234 (output_return_instruction): Only output APSR register clearing if
3235 Armv8.1-M Mainline instructions not available.
3236 (thumb_exit): Likewise.
3237 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3238 * config/arm/thumb2.md (clear_apsr): New define_insn.
3239 (clear_multiple): Likewise.
3240 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3242 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3243 Thomas Preud'homme <thomas.preudhomme@arm.com>
3245 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3246 (use_return_insn): Also return false for Armv8.1-M Mainline.
3247 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3248 Mainline instructions are available.
3249 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3250 when targeting Armv8.1-M Mainline Security Extensions.
3251 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3252 Mainline entry function.
3253 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3254 targeting Armv8.1-M Mainline or successor.
3255 (arm_expand_epilogue): Fix indentation of caller-saved register
3256 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3258 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3259 (FP_SYSREGS): Likewise.
3260 (enum vfp_sysregs_encoding): Define enum.
3261 (fp_sysreg_names): Declare.
3262 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3263 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3264 (pop_fpsysreg_insn): Likewise.
3266 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3267 Thomas Preud'homme <thomas.preudhomme@arm.com>
3269 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3270 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3271 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3272 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3273 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3274 (ARMv8_1m_main): New feature group.
3275 (armv8.1-m.main): New architecture.
3276 * config/arm/arm-tables.opt: Regenerate.
3277 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3278 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3279 (arm_options_perform_arch_sanity_checks): Error out when targeting
3280 Armv8.1-M Mainline Security Extensions.
3281 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3283 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3285 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3286 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3287 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3288 aarch64_bfdot_laneq): New.
3289 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3290 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3291 vbfdotq_laneq_f32): New.
3292 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3293 VBFMLA_W, VBF): New.
3294 (isquadop): Add V4BF, V8BF.
3296 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3298 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3299 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3300 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3301 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3302 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3303 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3304 usdot_laneq, sudot_lane,sudot_laneq): New.
3305 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3306 (aarch64_<sur>dot_lane): New.
3307 * config/aarch64/arm_neon.h (vusdot_s32): New.
3309 (vusdot_lane_s32): New.
3310 (vsudot_lane_s32): New.
3311 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3312 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3314 2020-01-16 Martin Liska <mliska@suse.cz>
3316 * value-prof.c (dump_histogram_value): Fix
3317 obvious spacing issue.
3319 2020-01-16 Andrew Pinski <apinski@marvell.com>
3321 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3322 !storage_order_barrier_p.
3324 2020-01-16 Andrew Pinski <apinski@marvell.com>
3326 * sched-int.h (_dep): Add unused bit-field field for the padding.
3327 * sched-deps.c (init_dep_1): Init unused field.
3329 2020-01-16 Andrew Pinski <apinski@marvell.com>
3331 * optabs.h (create_expand_operand): Initialize target field also.
3333 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3335 PR tree-optimization/92429
3336 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3337 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3339 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3342 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3344 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3345 aarch64_sve_int_mode to each mode.
3347 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3349 * doc/analyzer.texi (Overview): Add note about
3350 -fdump-ipa-analyzer.
3352 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3354 PR tree-optimization/93231
3355 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3356 input_type is unsigned. Use tree_to_shwi for shift constant.
3357 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3358 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3359 inputs more efficiently.
3361 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3363 * config/i386/i386.md (*movsf_internal): Do not require
3364 SSE2 ISA for alternatives 14 and 15.
3366 2020-01-15 Richard Biener <rguenther@suse.de>
3369 * tree-eh.c (sink_clobbers): If we already visited the destination
3370 block do not defer insertion.
3371 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3372 the purpose of defered insertion.
3374 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3376 * BASE-VER: Bump to 10.0.1.
3378 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3380 PR tree-optimization/93247
3381 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3382 type of the stmt that we're going to vectorize.
3384 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3386 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3387 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3390 2020-01-15 Martin Liska <mliska@suse.cz>
3392 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3393 2 calls of streamer_read_hwi in a function call.
3395 2020-01-15 Richard Biener <rguenther@suse.de>
3397 * alias.c (record_alias_subset): Avoid redundant work when
3398 subset is already recorded.
3400 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3402 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3403 the analyzer options provide CWE identifiers.
3405 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3407 * tree-diagnostic-path.cc (path_summary::event_range::print):
3408 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3409 using get_pure_location.
3411 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3413 PR tree-optimization/93262
3414 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3415 perform head trimming only if the last argument is constant,
3416 either all ones, or larger or equal to head trim, in the latter
3417 case decrease the last argument by head_trim.
3419 PR tree-optimization/93249
3420 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3421 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3422 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3423 perform head trim unless we can prove there are no '\0' chars
3424 from the source among the first head_trim chars.
3426 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3428 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3430 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3433 * config/i386/sse.md
3434 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3435 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3436 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3437 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3438 just a single alternative instead of two, make operands 1 and 2
3441 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3444 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3447 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3449 * Makefile.in (lang_opt_files): Add analyzer.opt.
3450 (ANALYZER_OBJS): New.
3451 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3452 tristate.o and ANALYZER_OBJS.
3453 (TEXI_GCCINT_FILES): Add analyzer.texi.
3454 * common.opt (-fanalyzer): New driver option.
3455 * config.in: Regenerate.
3456 * configure: Regenerate.
3457 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3458 (gccdepdir): Also create depdir for "analyzer" subdir.
3459 * digraph.cc: New file.
3460 * digraph.h: New file.
3461 * doc/analyzer.texi: New file.
3462 * doc/gccint.texi ("Static Analyzer") New menu item.
3463 (analyzer.texi): Include it.
3464 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3465 ("Warning Options"): Add static analysis warnings to the list.
3466 (-Wno-analyzer-double-fclose): New option.
3467 (-Wno-analyzer-double-free): New option.
3468 (-Wno-analyzer-exposure-through-output-file): New option.
3469 (-Wno-analyzer-file-leak): New option.
3470 (-Wno-analyzer-free-of-non-heap): New option.
3471 (-Wno-analyzer-malloc-leak): New option.
3472 (-Wno-analyzer-possible-null-argument): New option.
3473 (-Wno-analyzer-possible-null-dereference): New option.
3474 (-Wno-analyzer-null-argument): New option.
3475 (-Wno-analyzer-null-dereference): New option.
3476 (-Wno-analyzer-stale-setjmp-buffer): New option.
3477 (-Wno-analyzer-tainted-array-index): New option.
3478 (-Wno-analyzer-use-after-free): New option.
3479 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3480 (-Wno-analyzer-use-of-uninitialized-value): New option.
3481 (-Wanalyzer-too-complex): New option.
3482 (-fanalyzer-call-summaries): New warning.
3483 (-fanalyzer-checker=): New warning.
3484 (-fanalyzer-fine-grained): New warning.
3485 (-fno-analyzer-state-merge): New warning.
3486 (-fno-analyzer-state-purge): New warning.
3487 (-fanalyzer-transitivity): New warning.
3488 (-fanalyzer-verbose-edges): New warning.
3489 (-fanalyzer-verbose-state-changes): New warning.
3490 (-fanalyzer-verbosity=): New warning.
3491 (-fdump-analyzer): New warning.
3492 (-fdump-analyzer-callgraph): New warning.
3493 (-fdump-analyzer-exploded-graph): New warning.
3494 (-fdump-analyzer-exploded-nodes): New warning.
3495 (-fdump-analyzer-exploded-nodes-2): New warning.
3496 (-fdump-analyzer-exploded-nodes-3): New warning.
3497 (-fdump-analyzer-supergraph): New warning.
3498 * doc/sourcebuild.texi (dg-require-dot): New.
3499 (dg-check-dot): New.
3500 * gdbinit.in (break-on-saved-diagnostic): New command.
3501 * graphviz.cc: New file.
3502 * graphviz.h: New file.
3503 * ordered-hash-map-tests.cc: New file.
3504 * ordered-hash-map.h: New file.
3505 * passes.def (pass_analyzer): Add before
3506 pass_ipa_whole_program_visibility.
3507 * selftest-run-tests.c (selftest::run_tests): Call
3508 selftest::ordered_hash_map_tests_cc_tests.
3509 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3511 * shortest-paths.h: New file.
3512 * timevar.def (TV_ANALYZER): New timevar.
3513 (TV_ANALYZER_SUPERGRAPH): Likewise.
3514 (TV_ANALYZER_STATE_PURGE): Likewise.
3515 (TV_ANALYZER_PLAN): Likewise.
3516 (TV_ANALYZER_SCC): Likewise.
3517 (TV_ANALYZER_WORKLIST): Likewise.
3518 (TV_ANALYZER_DUMP): Likewise.
3519 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3520 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3521 * tree-pass.h (make_pass_analyzer): New decl.
3522 * tristate.cc: New file.
3523 * tristate.h: New file.
3525 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3528 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3529 alternatives 9 and 10.
3531 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3533 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3534 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3535 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3536 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3537 (selftest::hash_map_tests_c_tests): Call it.
3538 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3539 New static constant, using the value of = H::empty_zero_p.
3540 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3541 from default_hash_traits <Value>.
3542 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3544 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3545 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3546 calls to mark_empty with !Descriptor::empty_zero_p.
3547 (hash_table::empty_slow): Conditionalize the memset call with a
3548 check that Descriptor::empty_zero_p; otherwise, loop through the
3549 entries calling mark_empty on them.
3550 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3551 (pointer_hash::empty_zero_p): Likewise.
3552 (pair_hash::empty_zero_p): Likewise.
3553 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3555 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3556 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3557 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3558 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3559 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3560 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3561 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3562 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3564 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3567 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3569 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3570 fix typo on return value.
3572 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3575 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3577 (cgraph_edge::make_speculative): Add param for setting speculative_id
3579 (cgraph_edge::speculative_call_info): Update comments and find reference
3580 by speculative_id for multiple indirect targets.
3581 (cgraph_edge::resolve_speculation): Decrease the speculations
3582 for indirect edge, drop it's speculative if not direct target
3583 left. Update comments.
3584 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3585 (cgraph_node::dump): Print num_speculative_call_targets.
3586 (cgraph_node::verify_node): Don't report error if speculative
3587 edge not include statement.
3588 (cgraph_edge::num_speculative_call_targets_p): New function.
3589 * cgraph.h (int common_target_id): Remove.
3590 (int common_target_probability): Remove.
3591 (num_speculative_call_targets): New variable.
3592 (make_speculative): Add param for setting speculative_id.
3593 (cgraph_edge::num_speculative_call_targets_p): New declare.
3594 (target_prob): New variable.
3595 (speculative_id): New variable.
3596 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3597 call summaries for multiple speculative call targets.
3598 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3599 * ipa-profile.c (struct speculative_call_target): New struct.
3600 (class speculative_call_summary): New class.
3601 (class speculative_call_summaries): New class.
3602 (call_sums): New variable.
3603 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3604 (ipa_profile_write_edge_summary): New function.
3605 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3606 (ipa_profile_dump_all_summaries): New function.
3607 (ipa_profile_read_edge_summary): New function.
3608 (ipa_profile_read_summary_section): New function.
3609 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3610 (ipa_profile): Generate num_speculative_call_targets from
3612 * ipa-ref.h (speculative_id): New variable.
3613 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3614 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3615 common_target_probability. Stream out speculative_id and
3616 num_speculative_call_targets.
3617 (input_edge): Likewise.
3618 * predict.c (dump_prediction): Remove edges count assert to be
3620 * symtab.c (symtab_node::create_reference): Init speculative_id.
3621 (symtab_node::clone_references): Clone speculative_id.
3622 (symtab_node::clone_referring): Clone speculative_id.
3623 (symtab_node::clone_reference): Clone speculative_id.
3624 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3625 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3626 if indirect call contains multiple speculative targets.
3627 * value-prof.h (check_ic_target): Remove.
3628 * value-prof.c (gimple_value_profile_transformations):
3629 Use void function gimple_ic_transform.
3630 * value-prof.c (gimple_ic_transform): Handle topn case.
3631 Fix comment typos. Change it to a void function.
3633 2020-01-13 Andrew Pinski <apinski@marvell.com>
3635 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3636 (octeontx2t98): New define.
3637 (octeontx2t96): New define.
3638 (octeontx2t93): New define.
3639 (octeontx2f95): New define.
3640 (octeontx2f95n): New define.
3641 (octeontx2f95mm): New define.
3642 * config/aarch64/aarch64-tune.md: Regenerate.
3643 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3645 2020-01-13 Jason Merrill <jason@redhat.com>
3647 PR c++/33799 - destroy return value if local cleanup throws.
3648 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3650 2020-01-13 Martin Liska <mliska@suse.cz>
3652 * ipa-cp.c (get_max_overall_size): Use newly
3653 renamed param param_ipa_cp_unit_growth.
3654 * params.opt: Remove legacy param name.
3656 2020-01-13 Martin Sebor <msebor@redhat.com>
3658 PR tree-optimization/93213
3659 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3660 stores to be eliminated.
3662 2020-01-13 Martin Liska <mliska@suse.cz>
3664 * opts.c (print_help): Do not print CL_PARAM
3665 and CL_WARNING for CL_OPTIMIZATION.
3667 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3670 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3671 depending on optimization settings.
3673 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3675 PR tree-optimization/90838
3676 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3677 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3678 argument rather than to initialize temporary for targets that
3679 don't use the mode argument at all. Initialize ctzval to avoid
3682 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3684 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3685 * tree-core.h: Document it.
3686 * gimplify.c (gimplify_omp_workshare): Set it.
3687 * omp-low.c (lower_omp_target): Use it.
3688 * tree-pretty-print.c (dump_omp_clause): Print it.
3690 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3691 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3693 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3695 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3696 * common.opt (fdiagnostics-path-format=): New option.
3697 (diagnostic_path_format): New enum.
3698 (fdiagnostics-show-path-depths): New option.
3699 * coretypes.h (diagnostic_event_id_t): New forward decl.
3700 * diagnostic-color.c (color_dict): Add "path".
3701 * diagnostic-event-id.h: New file.
3702 * diagnostic-format-json.cc (json_from_expanded_location): Make
3704 (json_end_diagnostic): Call context->make_json_for_path if it
3705 exists and the diagnostic has a path.
3706 (diagnostic_output_format_init): Clear context->print_path.
3707 * diagnostic-path.h: New file.
3708 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3709 when printing a run of events in a diagnostic_path so that they
3710 all get the same color.
3711 (layout::m_diagnostic_path_p): New field.
3712 (layout::layout): Initialize it.
3713 (layout::print_any_labels): Don't colorize the label text for an
3714 event in a diagnostic_path.
3715 (gcc_rich_location::add_location_if_nearby): Add
3716 "restrict_to_current_line_spans" and "label" params. Pass the
3717 former to layout.maybe_add_location_range; pass the latter
3718 when calling add_range.
3719 * diagnostic.c: Include "diagnostic-path.h".
3720 (diagnostic_initialize): Initialize context->path_format and
3721 context->show_path_depths.
3722 (diagnostic_show_any_path): New function.
3723 (diagnostic_path::interprocedural_p): New function.
3724 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3725 (simple_diagnostic_path::num_events): New function.
3726 (simple_diagnostic_path::get_event): New function.
3727 (simple_diagnostic_path::add_event): New function.
3728 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3729 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3730 (debug): New overload taking a diagnostic_path *.
3731 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3732 * diagnostic.h (enum diagnostic_path_format): New enum.
3733 (json::value): New forward decl.
3734 (diagnostic_context::path_format): New field.
3735 (diagnostic_context::show_path_depths): New field.
3736 (diagnostic_context::print_path): New callback field.
3737 (diagnostic_context::make_json_for_path): New callback field.
3738 (diagnostic_show_any_path): New decl.
3739 (json_from_expanded_location): New decl.
3740 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3741 (-fdiagnostics-show-path-depths): New option.
3742 (-fdiagnostics-color): Add "path" to description of default
3743 GCC_COLORS; describe it.
3744 (-fdiagnostics-format=json): Document how diagnostic paths are
3745 represented in the JSON output format.
3746 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3747 Add optional params "restrict_to_current_line_spans" and "label".
3748 * opts.c (common_handle_option): Handle
3749 OPT_fdiagnostics_path_format_ and
3750 OPT_fdiagnostics_show_path_depths.
3751 * pretty-print.c: Include "diagnostic-event-id.h".
3752 (pp_format): Implement "%@" format code for printing
3753 diagnostic_event_id_t *.
3754 (selftest::test_pp_format): Add tests for "%@".
3755 * selftest-run-tests.c (selftest::run_tests): Call
3756 selftest::tree_diagnostic_path_cc_tests.
3757 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3758 * toplev.c (general_init): Initialize global_dc->path_format and
3759 global_dc->show_path_depths.
3760 * tree-diagnostic-path.cc: New file.
3761 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3762 non-static. Drop "diagnostic" param in favor of storing the
3763 original value of "where" and re-using it.
3764 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3765 maybe_unwind_expanded_macro_loc.
3766 (tree_diagnostics_defaults): Initialize context->print_path and
3767 context->make_json_for_path.
3768 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3770 (default_tree_make_json_for_path): New decl.
3771 (maybe_unwind_expanded_macro_loc): New decl.
3773 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3775 PR tree-optimization/93210
3776 * fold-const.h (native_encode_initializer,
3777 can_native_interpret_type_p): Declare.
3778 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3780 (native_encode_initializer): New function, moved from dwarf2out.c.
3781 Adjust to native_encode_expr compatible arguments, including dry-run
3782 and partial extraction modes. Don't handle STRING_CST.
3783 (can_native_interpret_type_p): No longer static.
3784 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3785 offset / BITS_PER_UNIT fits into int and don't call it if
3786 can_native_interpret_type_p fails. If suboff is NULL and for
3787 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3788 native_encode_initializer.
3789 (fold_const_aggregate_ref_1): Formatting fix.
3790 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3791 (tree_add_const_value_attribute): Adjust caller.
3793 PR tree-optimization/90838
3794 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3795 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3796 CTZ_DEFINED_VALUE_AT_ZERO.
3798 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3801 * lra-constraints.c (match_reload): Permit input operands have the
3802 same mode as output while other input operands have a different
3805 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3807 PR tree-optimization/90838
3808 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3809 (check_ctz_string): Likewise.
3810 (optimize_count_trailing_zeroes): Likewise.
3811 (simplify_count_trailing_zeroes): Likewise.
3812 (pass_forwprop::execute): Try ctz simplification.
3813 * match.pd: Add matching for ctz idioms.
3815 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3817 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3819 (aarch64_invalid_unary_op): New function for target hook.
3820 (aarch64_invalid_binary_op): New function for target hook.
3822 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3824 * config.gcc: Add arm_bf16.h.
3825 * config/aarch64/aarch64-builtins.c
3826 (aarch64_simd_builtin_std_type): Add BFmode.
3827 (aarch64_init_simd_builtin_types): Define element types for vector
3829 (aarch64_init_bf16_types): New function.
3830 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3831 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3833 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3834 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3836 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3837 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3838 * config/aarch64/aarch64.c
3839 (aarch64_classify_vector_mode): Add support for BF types.
3840 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3841 (aarch64_vq_mode): Add support for BF types.
3842 (aarch64_simd_container_mode): Add support for BF types.
3843 (aarch64_mangle_type): Add support for BF scalar type.
3844 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3845 * config/aarch64/arm_bf16.h: New file.
3846 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3847 * config/aarch64/iterators.md: Add BF types to mode attributes.
3848 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3850 2020-01-10 Jason Merrill <jason@redhat.com>
3852 PR c++/93173 - incorrect tree sharing.
3853 * gimplify.c (copy_if_shared): No longer static.
3854 * gimplify.h: Declare it.
3856 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3858 * doc/invoke.texi (-msve-vector-bits=): Document that
3859 -msve-vector-bits=128 now generates VL-specific code for
3860 little-endian targets.
3861 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3862 build_vector_type_for_mode to construct the data vector types.
3863 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3864 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3865 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3866 for 128-bit vectors.
3868 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3870 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3873 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3875 * config/aarch64/aarch64-builtins.c
3876 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3877 rather than checking the number of elements and the element mode.
3879 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3881 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3882 get_related_vectype_for_scalar_type rather than build_vector_type
3883 to create the index type for a conditional reduction.
3885 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3887 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3888 for any type of gather or scatter, including strided accesses.
3890 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3892 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3895 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3897 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3899 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3900 parameter and its use to reset DR_OFFSET's.
3901 (vect_transform_loop): Remove orig_drs_init argument.
3902 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3903 member of dr_vec_info rather than the offset of the associated
3904 data_reference's innermost_loop_behavior.
3905 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3906 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3907 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3908 get_dr_vinfo_offset.
3909 (vectorizable_store): Likewise.
3910 (vectorizable_load): Likewise.
3912 2020-01-10 Richard Biener <rguenther@suse.de>
3914 * gimple-ssa-store-merging
3915 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3917 2020-01-10 Martin Liska <mliska@suse.cz>
3920 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3921 encapsulation that was there before r280040.
3923 2020-01-10 Richard Biener <rguenther@suse.de>
3926 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3927 sequences to avoid walking them again for secondary opportunities.
3928 (pass_lower_eh_dispatch::execute): Instead actually insert
3931 2020-01-10 Richard Biener <rguenther@suse.de>
3934 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3935 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3936 avoid quadraticness.
3938 2020-01-10 Martin Jambor <mjambor@suse.cz>
3940 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3941 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3942 to get param_ipa_sra_max_replacements.
3943 (param_splitting_across_edge): Pass the caller to
3944 pull_accesses_from_callee.
3946 2020-01-10 Martin Jambor <mjambor@suse.cz>
3948 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3949 * ipa-cp.c (max_new_size): Removed.
3950 (orig_overall_size): New variable.
3951 (get_max_overall_size): New function.
3952 (estimate_local_effects): Use it. Adjust dump.
3953 (decide_about_value): Likewise.
3954 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3955 orig_overall_size. Adjust dump.
3956 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3958 2020-01-10 Martin Jambor <mjambor@suse.cz>
3960 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3961 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3962 instead of param_ipa_max_agg_items.
3963 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3964 optimization info for the callee.
3966 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3968 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3969 markers if debug_inline_points is false.
3971 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3973 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3975 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3976 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3977 aarch64-sve-builtins-sve2.h.
3978 (aarch64-sve-builtins-sve2.o): New rule.
3979 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3980 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3981 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3982 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3983 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3984 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3986 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3987 instructions that are handled here.
3988 (@cond_asrd<mode>): Generalize to...
3989 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3990 (*cond_asrd<mode>_2): Generalize to...
3991 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3992 (*cond_asrd<mode>_z): Generalize to...
3993 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3994 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3995 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3996 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3997 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3999 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4000 (@aarch64_scatter_stnt<mode>): Likewise.
4001 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4002 (@aarch64_mul_lane_<mode>): Likewise.
4003 (@aarch64_sve_suqadd<mode>_const): Likewise.
4004 (*<sur>h<addsub><mode>): Generalize to...
4005 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4007 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4008 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4009 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4010 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4011 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4012 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4013 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4014 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4015 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4016 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4017 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4018 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4019 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4020 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4021 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4022 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4023 (@aarch64_sve2_xar<mode>): Likewise.
4024 (@aarch64_sve2_bcax<mode>): Likewise.
4025 (*aarch64_sve2_eor3<mode>): Rename to...
4026 (@aarch64_sve2_eor3<mode>): ...this.
4027 (@aarch64_sve2_bsl<mode>): New expander.
4028 (@aarch64_sve2_nbsl<mode>): Likewise.
4029 (@aarch64_sve2_bsl1n<mode>): Likewise.
4030 (@aarch64_sve2_bsl2n<mode>): Likewise.
4031 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4032 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4033 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4034 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4035 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4036 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4037 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4038 (<su>mull<bt><Vwide>): Generalize to...
4039 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4041 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4042 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4043 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4044 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4045 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4046 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4047 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4048 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4049 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4050 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4051 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4052 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4053 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4054 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4055 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4056 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4057 (<SHRNB:r>shrnb<mode>): Generalize to...
4058 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4060 (<SHRNT:r>shrnt<mode>): Generalize to...
4061 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4063 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4064 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4065 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4066 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4067 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4068 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4069 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4070 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4071 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4072 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4073 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4074 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4075 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4076 (@aarch64_sve2_cvtnt<mode>): Likewise.
4077 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4078 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4079 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4080 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4081 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4082 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4083 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4084 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4085 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4086 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4087 (@aarch64_sve2_pmul<mode>): Likewise.
4088 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4089 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4090 (@aarch64_sve2_tbl2<mode>): Likewise.
4091 (@aarch64_sve2_tbx<mode>): Likewise.
4092 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4093 (@aarch64_sve2_histcnt<mode>): Likewise.
4094 (@aarch64_sve2_histseg<mode>): Likewise.
4095 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4096 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4097 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4098 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4099 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4100 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4101 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4102 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4103 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4104 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4105 (SVE2_PMULL_PAIR_I): New mode iterators.
4106 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4107 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4108 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4109 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4110 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4111 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4112 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4113 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4114 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4115 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4116 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4117 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4118 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4119 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4120 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4121 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4122 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4123 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4124 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4125 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4126 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4127 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4128 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4129 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4130 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4131 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4132 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4133 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4134 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4135 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4136 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4138 (VNARROW, Ventype): New mode attributes.
4139 (Vewtype): Handle VNx2DI. Fix typo in comment.
4140 (VDOUBLE): New mode attribute.
4141 (sve_lane_con): Handle VNx8HI.
4142 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4143 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4144 (sve_int_op, sve_int_op_rev): Handle the above codes.
4145 (sve_pred_int_rhs2_operand): Likewise.
4146 (MULLBT, SHRNB, SHRNT): Delete.
4147 (SVE_INT_SHIFT_IMM): New int iterator.
4148 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4149 and UNSPEC_WHILEHS for TARGET_SVE2.
4150 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4151 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4152 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4153 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4154 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4155 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4156 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4157 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4158 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4159 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4160 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4161 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4162 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4163 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4164 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4165 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4166 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4167 (optab): Handle the new unspecs.
4168 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4170 (lr): Handle the new unspecs.
4172 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4173 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4174 (sve_int_qsub_op): New int attributes.
4175 (sve_fp_op, rot): Handle the new unspecs.
4176 * config/aarch64/aarch64-sve-builtins.h
4177 (function_resolver::require_matching_pointer_type): Declare.
4178 (function_resolver::resolve_unary): Add an optional boolean argument.
4179 (function_resolver::finish_opt_n_resolution): Add an optional
4180 type_suffix_index argument.
4181 (gimple_folder::redirect_call): Declare.
4182 (gimple_expander::prepare_gather_address_operands): Add an optional
4184 * config/aarch64/aarch64-sve-builtins.cc: Include
4185 aarch64-sve-builtins-sve2.h.
4186 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4187 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4188 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4189 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4190 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4191 (TYPES_s_unsigned): Likewise.
4192 (TYPES_s_integer): Use TYPES_s_unsigned.
4193 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4194 (TYPES_sd_integer): Use them.
4195 (TYPES_d_unsigned): New macro.
4196 (TYPES_d_integer): Use it.
4197 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4198 (TYPES_cvt_narrow): Likewise.
4199 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4200 (preds_mx): New variable.
4201 (function_builder::add_overloaded_function): Allow the new feature
4202 set to be more restrictive than the original one.
4203 (function_resolver::infer_pointer_type): Remove qualifiers from
4204 the pointer type before printing it.
4205 (function_resolver::require_matching_pointer_type): New function.
4206 (function_resolver::resolve_sv_displacement): Handle functions
4207 that don't support 32-bit vector indices or svint32_t vector offsets.
4208 (function_resolver::finish_opt_n_resolution): Take the inferred type
4209 as a separate argument.
4210 (function_resolver::resolve_unary): Optionally treat all forms in
4211 the same way as normal merging functions.
4212 (gimple_folder::redirect_call): New function.
4213 (function_expander::prepare_gather_address_operands): Add an argument
4214 that says whether scaled forms are available. If they aren't,
4215 handle scaling of vector indices and don't add the extension and
4217 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4218 fall back to using cond_* instead.
4219 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4220 Split out the member variables into...
4221 (rtx_code_function_base): ...this new base class.
4222 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4223 (unspec_based_function): Split out the member variables into...
4224 (unspec_based_function_base): ...this new base class.
4225 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4226 (unspec_based_function_exact_insn): New class.
4227 (unspec_based_add_function, unspec_based_add_lane_function)
4228 (unspec_based_lane_function, unspec_based_pred_function)
4229 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4230 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4231 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4233 (unspec_based_fused_function): New class.
4234 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4235 (unspec_based_fused_lane_function): New class.
4236 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4238 (CODE_FOR_MODE1): New macro.
4239 (fixed_insn_function): New class.
4240 (while_comparison): Likewise.
4241 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4242 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4243 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4244 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4245 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4246 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4247 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4248 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4249 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4250 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4251 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4252 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4253 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4254 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4255 Also add an initial argument for unary_convert_narrowt, regardless
4256 of the predication type.
4257 (build_32_64): Allow loads and stores to specify MODE_none.
4258 (build_sv_index64, build_sv_uint_offset): New functions.
4259 (long_type_suffix): New function.
4260 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4261 (binary_imm_long_base, load_gather_sv_base): Likewise.
4262 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4263 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4264 (unary_narrowb_base, unary_narrowt_base): Likewise.
4265 (binary_long_lane_def, binary_long_lane): New shape.
4266 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4267 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4268 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4269 (binary_to_uint_def, binary_to_uint): Likewise.
4270 (binary_wide_def, binary_wide): Likewise.
4271 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4272 (compare_def, compare): Likewise.
4273 (compare_ptr_def, compare_ptr): Likewise.
4274 (load_ext_gather_index_restricted_def,
4275 load_ext_gather_index_restricted): Likewise.
4276 (load_ext_gather_offset_restricted_def,
4277 load_ext_gather_offset_restricted): Likewise.
4278 (load_gather_sv_def): Inherit from load_gather_sv_base.
4279 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4280 (shift_left_imm_def, shift_left_imm): Likewise.
4281 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4282 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4283 (store_scatter_index_restricted_def,
4284 store_scatter_index_restricted): Likewise.
4285 (store_scatter_offset_restricted_def,
4286 store_scatter_offset_restricted): Likewise.
4287 (tbl_tuple_def, tbl_tuple): Likewise.
4288 (ternary_long_lane_def, ternary_long_lane): Likewise.
4289 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4290 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4291 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4292 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4293 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4294 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4295 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4296 (ternary_uint_def, ternary_uint): Likewise.
4297 (unary_convert): Fix typo in comment.
4298 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4299 (unary_long_def, unary_long): Likewise.
4300 (unary_narrowb_def, unary_narrowb): Likewise.
4301 (unary_narrowt_def, unary_narrowt): Likewise.
4302 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4303 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4304 (unary_to_int_def, unary_to_int): Likewise.
4305 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4306 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4307 (svasrd_impl): Delete.
4308 (svcadd_impl::expand): Handle integer operations too.
4309 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4310 new functions to derive the unspec numbers.
4311 (svmla_svmls_lane_impl): Replace with...
4312 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4313 integer operations too.
4314 (svwhile_impl): Rename to...
4315 (svwhilelx_impl): ...this and inherit from while_comparison.
4316 (svasrd): Use unspec_based_function.
4317 (svmla_lane): Use svmla_lane_impl.
4318 (svmls_lane): Use svmls_lane_impl.
4319 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4320 (svwhilele, svwhilelt): Use svwhilelx_impl.
4321 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4322 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4323 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4324 * config/aarch64/aarch64-sve-builtins.def: Include
4325 aarch64-sve-builtins-sve2.def.
4327 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4329 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4330 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4331 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4332 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4333 immediates as well as vector ones.
4334 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4335 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4336 (aarch64_sve_qsub_immediate): Update calls accordingly.
4338 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4340 * config/aarch64/aarch64-sve2.md: Add banner comments.
4341 (<su>mulh<r>s<mode>3): Move further up file.
4342 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4343 (*aarch64_sve2_sra<mode>): Move further down file.
4344 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4346 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4348 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4350 (while_optab_cmp): Handle them.
4351 * config/aarch64/aarch64-sve.md
4352 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4353 and add a "@" marker.
4354 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4355 instead of gen_aarch64_sve2_while_ptest.
4356 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4358 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4360 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4361 (UNSPEC_WHILELE): ...this.
4362 (UNSPEC_WHILE_LO): Rename to...
4363 (UNSPEC_WHILELO): ...this.
4364 (UNSPEC_WHILE_LS): Rename to...
4365 (UNSPEC_WHILELS): ...this.
4366 (UNSPEC_WHILE_LT): Rename to...
4367 (UNSPEC_WHILELT): ...this.
4368 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4369 (cmp_op, while_optab_cmp): Likewise.
4370 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4371 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4372 (svwhilelt): Likewise.
4374 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4376 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4377 (unary_to_uint): Define.
4378 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4379 (unary_count): Rename to...
4380 (unary_to_uint_def, unary_to_uint): ...this.
4381 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4383 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4385 * config/aarch64/aarch64-sve-builtins-functions.h
4386 (code_for_mode_function): New class.
4387 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4388 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4389 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4390 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4391 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4393 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4395 * config/aarch64/iterators.md (addsub): New code attribute.
4396 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4398 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4399 in the asm string and attributes. Fix indentation.
4400 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4402 (@aarch64_sve_<optab><mode>): ...this.
4403 * config/aarch64/aarch64-sve-builtins.h
4404 (function_expander::expand_signed_unpred_op): Delete.
4405 * config/aarch64/aarch64-sve-builtins.cc
4406 (function_expander::expand_signed_unpred_op): Likewise.
4407 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4408 try using code_for_aarch64_sve instead.
4409 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4410 (svqsub_impl): Likewise.
4411 (svqadd, svqsub): Use rtx_code_function instead.
4413 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4415 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4416 (HADDSUB, sur, addsub): Remove them.
4418 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4420 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4421 functions the same way, rather than singling out those that
4422 aren't mapped directly to optabs.
4424 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4426 * target.def (compatible_vector_types_p): New target hook.
4427 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4428 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4429 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4430 * doc/tm.texi: Regenerate.
4431 * gimple-expr.c: Include target.h.
4432 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4433 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4435 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4436 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4437 Use the original predicate if it already has a suitable type.
4439 2020-01-09 Martin Jambor <mjambor@suse.cz>
4441 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4442 resolve_speculation and redirect_call_stmt_to_callee static. Change
4443 return type of set_call_stmt to cgraph_edge *.
4444 * auto-profile.c (afdo_indirect_call): Adjust call to
4445 redirect_call_stmt_to_callee.
4446 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4447 make the this pointer explicit, adjust self-recursive calls and the
4448 call top make_direct. Return the resulting edge.
4449 (cgraph_edge::remove): Make this pointer explicit.
4450 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4451 (cgraph_edge::make_direct): Likewise, adjust call to
4452 resolve_speculation.
4453 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4454 call to set_call_stmt.
4455 (cgraph_update_edges_for_call_stmt_node): Update call to
4456 set_call_stmt and remove.
4457 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4458 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4459 (cgraph_node::create_edge_including_clones): Moved "first" definition
4460 of edge to the block where it was used. Adjusted calls to
4462 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4463 cgraph_edge::remove.
4464 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4465 make_direct and redirect_call_stmt_to_callee.
4466 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4467 resolve_speculation and make_direct.
4468 * ipa-inline-transform.c (inline_transform): Adjust call to
4469 redirect_call_stmt_to_callee.
4470 (check_speculations_1):: Adjust call to resolve_speculation.
4471 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4472 resolve-speculation.
4473 (inline_small_functions): Adjust call to resolve_speculation.
4474 (ipa_inline): Likewise.
4475 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4477 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4478 safe with regards to edge removal, adjust calls to
4479 redirect_call_stmt_to_callee.
4480 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4481 and redirect_call_stmt_to_callee.
4482 * multiple_target.c (create_dispatcher_calls): Adjust call to
4483 redirect_call_stmt_to_callee
4484 (redirect_to_specific_clone): Likewise.
4485 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4486 Adjust calls to cgraph_edge::remove.
4487 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4488 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4489 (expand_call_inline): Adjust call to cgraph_edge::remove.
4491 2020-01-09 Martin Liska <mliska@suse.cz>
4493 * params.opt: Set Optimization for
4494 param_max_speculative_devirt_maydefs.
4496 2020-01-09 Martin Sebor <msebor@redhat.com>
4500 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4502 2020-01-09 Martin Liska <mliska@suse.cz>
4504 * auto-profile.c (auto_profile): Use opt_for_fn
4506 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4507 (propagate_vals_across_arith_jfunc): Likewise.
4508 (hint_time_bonus): Likewise.
4509 (incorporate_penalties): Likewise.
4510 (good_cloning_opportunity_p): Likewise.
4511 (perform_estimation_of_a_value): Likewise.
4512 (estimate_local_effects): Likewise.
4513 (ipcp_propagate_stage): Likewise.
4514 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4515 (set_switch_stmt_execution_predicate): Likewise.
4516 (analyze_function_body): Likewise.
4517 * ipa-inline-analysis.c (offline_size): Likewise.
4518 * ipa-inline.c (early_inliner): Likewise.
4519 * ipa-prop.c (ipa_analyze_node): Likewise.
4520 (ipcp_transform_function): Likewise.
4521 * ipa-sra.c (process_scan_results): Likewise.
4522 (ipa_sra_summarize_function): Likewise.
4523 * params.opt: Rename ipcp-unit-growth to
4524 ipa-cp-unit-growth. Add Optimization for various
4525 IPA-related parameters.
4527 2020-01-09 Richard Biener <rguenther@suse.de>
4530 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4532 2020-01-09 Richard Biener <rguenther@suse.de>
4534 PR tree-optimization/93040
4535 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4537 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4539 * common/config/avr/avr-common.c (avr_option_optimization_table)
4540 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4542 2020-01-09 Martin Liska <mliska@suse.cz>
4544 * cgraphclones.c (symbol_table::materialize_all_clones):
4545 Use cgraph_node::dump_name.
4547 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4550 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4551 output_operand_lossage instead of gcc_unreachable.
4552 * doc/md.texi (riscv f constraint): Fix typo.
4555 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4556 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4557 CONST_SCALAR_INT_P instead of CONST_INT_P.
4558 (*subv<mode>4_1): Rename to ...
4559 (subv<mode>4_1): ... this.
4560 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4561 define_insn_and_split patterns.
4562 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4565 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4567 * vec.c (class selftest::count_dtor): New class.
4568 (selftest::test_auto_delete_vec): New test.
4569 (selftest::vec_c_tests): Call it.
4570 * vec.h (class auto_delete_vec): New class template.
4571 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4573 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4575 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4577 2020-01-08 Jim Wilson <jimw@sifive.com>
4579 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4580 use of TLS_MODEL_LOCAL_EXEC when not pic.
4582 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4584 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4587 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4590 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4591 *stack_protect_set_3 peephole2): Also check that the second
4592 insns source is general_operand.
4595 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4596 predicate for output operand instead of register_operand.
4597 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4598 memory destination and non-memory operands[2].
4600 2020-01-08 Martin Liska <mliska@suse.cz>
4602 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4603 ::dump_asm_name instead of (::name or ::asm_name).
4604 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4605 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4606 (analyze_functions): Likewise.
4607 (expand_all_functions): Likewise.
4608 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4609 (propagate_bits_across_jump_function): Likewise.
4610 (dump_profile_updates): Likewise.
4611 (ipcp_store_bits_results): Likewise.
4612 (ipcp_store_vr_results): Likewise.
4613 * ipa-devirt.c (dump_targets): Likewise.
4614 * ipa-fnsummary.c (analyze_function_body): Likewise.
4615 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4616 (process_hsa_functions): Likewise.
4617 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4618 (set_alias_uids): Likewise.
4619 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4620 * ipa-inline.c (recursive_inlining): Likewise.
4621 (inline_to_all_callers_1): Likewise.
4622 (ipa_inline): Likewise.
4623 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4624 (ipa_propagate_frequency): Likewise.
4625 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4626 (remove_described_reference): Likewise.
4627 * ipa-pure-const.c (worse_state): Likewise.
4628 (check_retval_uses): Likewise.
4629 (analyze_function): Likewise.
4630 (propagate_pure_const): Likewise.
4631 (propagate_nothrow): Likewise.
4632 (dump_malloc_lattice): Likewise.
4633 (propagate_malloc): Likewise.
4634 (pass_local_pure_const::execute): Likewise.
4635 * ipa-visibility.c (optimize_weakref): Likewise.
4636 (function_and_variable_visibility): Likewise.
4637 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4638 (ipa_discover_variable_flags): Likewise.
4639 * lto-streamer-out.c (output_function): Likewise.
4640 (output_constructor): Likewise.
4641 * tree-inline.c (copy_bb): Likewise.
4642 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4643 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4645 2020-01-08 Richard Biener <rguenther@suse.de>
4648 * tree-eh.c (sink_clobbers): Update virtual operands for
4649 the first and last stmt only. Add a dry-run capability.
4650 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4651 after CFG manipulations and in RPO order to catch all
4652 secondary opportunities reliably.
4654 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4657 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4659 2019-01-08 Richard Biener <rguenther@suse.de>
4662 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4663 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4664 virtual operand, also updating SSA use.
4665 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4666 Update stmt after resetting virtual operand.
4667 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4668 * gimple-iterator.c (gsi_remove): When not removing the stmt
4669 permanently do not delink immediate uses or mark the stmt modified.
4671 2020-01-08 Martin Liska <mliska@suse.cz>
4673 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4674 (ipa_call_context::estimate_size_and_time): Likewise.
4675 (inline_analyze_function): Likewise.
4677 2020-01-08 Martin Liska <mliska@suse.cz>
4679 * cgraph.c (cgraph_node::dump): Use systematically
4682 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4684 Add -nodevicespecs option for avr.
4687 * config/avr/avr.opt (-nodevicespecs): New driver option.
4688 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4689 "-specs=device-specs/..." if that option is not set.
4690 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4692 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4694 Implement 64-bit double functions for avr.
4697 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4698 --with-double-comparison.
4699 * doc/install.texi: Document them.
4700 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4701 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4702 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4703 * doc/invoke.texi (AVR Built-in Macros): Document them.
4704 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4705 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4706 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4708 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4711 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4712 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4713 when only building rm-profile multilibs.
4715 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4718 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4719 lattice for a value to check.
4720 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4721 finite propagation in self-recursive scc.
4723 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4725 * ipa-inline.c (caller_growth_limits): Restore the AND.
4727 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4729 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4730 (VEC_ALLREG_ALT): New iterator.
4731 (VEC_ALLREG_INT_MODE): New iterator.
4732 (VCMP_MODE): New iterator.
4733 (VCMP_MODE_INT): New iterator.
4734 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4735 (vec_cmp<u>v64qidi): New define_expand.
4736 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4737 (vec_cmpu<mode>di_exec): New define_expand.
4738 (vec_cmp<u>v64qidi_exec): New define_expand.
4739 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4740 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4741 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4742 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4743 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4744 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4745 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4746 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4747 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4748 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4750 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4751 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4753 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4755 * config/gcn/constraints.md (DA): Update description and match.
4757 (Db): New constraint.
4758 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4760 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4761 Implement 'Db' mixed immediate type.
4762 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4763 (addcv64si3_dup<exec_vcc>): Delete.
4764 (subcv64si3<exec_vcc>): Rework constraints.
4765 (addv64di3): Rework constraints.
4766 (addv64di3_exec): Rework constraints.
4767 (subv64di3): Rework constraints.
4768 (addv64di3_dup): Delete.
4769 (addv64di3_dup_exec): Delete.
4770 (addv64di3_zext): Rework constraints.
4771 (addv64di3_zext_exec): Rework constraints.
4772 (addv64di3_zext_dup): Rework constraints.
4773 (addv64di3_zext_dup_exec): Rework constraints.
4774 (addv64di3_zext_dup2): Rework constraints.
4775 (addv64di3_zext_dup2_exec): Rework constraints.
4776 (addv64di3_sext_dup2): Rework constraints.
4777 (addv64di3_sext_dup2_exec): Rework constraints.
4779 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4781 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4782 existing target checks.
4784 2020-01-07 Richard Biener <rguenther@suse.de>
4786 * doc/install.texi: Bump minimal supported MPC version.
4788 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4790 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4791 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4792 * langhooks.c: Include stor-layout.h.
4793 (lhd_simulate_enum_decl): New function.
4794 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4795 handle_arm_sve_h for the LTO frontend.
4796 (register_vector_type): Cope with null returns from pushdecl.
4798 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4800 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4801 (aarch64_sve::nvectors_if_data_type): Replace with...
4802 (aarch64_sve::builtin_type_p): ...this.
4803 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4804 (find_vector_type): Delete.
4805 (add_sve_type_attribute): New function.
4806 (lookup_sve_type_attribute): Likewise.
4807 (register_builtin_types): Add an "SVE type" attribute to each type.
4808 (register_tuple_type): Likewise.
4809 (svbool_type_p, nvectors_if_data_type): Delete.
4810 (mangle_builtin_type): Use lookup_sve_type_attribute.
4811 (builtin_type_p): Likewise. Add an overload that returns the
4812 number of constituent vector and predicate registers.
4813 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4814 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4815 instead of aarch64_sve_argument_p.
4816 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4817 (aarch64_pass_by_reference): Likewise.
4818 (aarch64_function_value_1): Likewise.
4819 (aarch64_return_in_memory): Likewise.
4820 (aarch64_layout_arg): Likewise.
4822 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4824 PR tree-optimization/93156
4825 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4826 least significant bit is always clear.
4828 PR tree-optimization/93118
4829 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4830 simplifier with two intermediate conversions.
4832 2020-01-07 Martin Liska <mliska@suse.cz>
4834 * params.opt: Add Optimization for various parameters.
4836 2020-01-07 Martin Liska <mliska@suse.cz>
4839 * doc/extend.texi: Explain cloning for target_clone
4842 2020-01-07 Martin Liska <mliska@suse.cz>
4844 PR tree-optimization/92860
4845 * common.opt: Make in Optimization option
4846 as it is affected by -O0, which is an Optimization
4848 * tree-inline.c (tree_inlinable_function_p):
4849 Use opt_for_fn for warn_inline.
4850 (expand_call_inline): Likewise.
4852 2020-01-07 Martin Liska <mliska@suse.cz>
4854 PR tree-optimization/92860
4855 * common.opt: Make flag_ree as optimization
4858 2020-01-07 Martin Liska <mliska@suse.cz>
4860 PR optimization/92860
4861 * params.opt: Mark param_min_crossjump_insns with Optimization
4864 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4866 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4867 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4869 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4871 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4872 helper function to return the valid addressing formats for a given
4873 hard register and mode.
4874 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4876 * config/rs6000/constraints.md (Q constraint): Update
4878 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4881 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4882 Use 'Q' for doing vector extract from memory.
4883 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4885 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4886 doing vector extract from memory.
4887 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4888 extract from memory.
4890 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4891 for the offset being 34-bits when -mcpu=future is used.
4893 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4895 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4896 instead of cmpib_comparison_operator in cmpib patterns.
4897 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4898 of cmpib_comparison_operator. Revise comment.
4900 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4902 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4903 in an IFN_DIV_POW2 node to be equal.
4905 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4907 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4908 (vect_check_scalar_mask): ...this.
4909 (vectorizable_store, vectorizable_load): Update call accordingly.
4910 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4911 argument in calls to conditional internal functions.
4913 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4915 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4916 '0' matching inputs.
4917 (subv64di3_exec): Likewise.
4919 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4921 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4922 * doc/md.texi (movstr): Likewise.
4924 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4926 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4929 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4931 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4933 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4934 to a temporary file and use move-if-change to update the real
4935 file where necessary.
4937 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4939 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4940 rather than Upa for CPY /M.
4942 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4944 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4947 2020-01-06 Martin Liska <mliska@suse.cz>
4949 PR tree-optimization/92860
4950 * params.opt: Mark param_max_combine_insns with Optimization
4953 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4956 * config/i386/i386.md (SWIDWI): New mode iterator.
4957 (DWI, dwi): Add TImode variants.
4958 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4959 <general_hilo_operand> instead of <general_operand>. Use
4960 CONST_SCALAR_INT_P instead of CONST_INT_P.
4961 (*addv<mode>4_1): Rename to ...
4962 (addv<mode>4_1): ... this.
4963 (QWI): New mode attribute.
4964 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4965 define_insn_and_split patterns.
4966 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4968 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4969 <general_hilo_operand> instead of <general_operand>.
4970 (*addcarry<mode>_1): New define_insn.
4971 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4973 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4975 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4976 Use "call" instead of "set".
4978 2020-01-03 Martin Jambor <mjambor@suse.cz>
4981 * ipa-cp.c (print_all_lattices): Skip functions without info.
4983 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4986 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4987 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4988 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4989 for 'e' simd clones.
4992 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4994 (mprefer-vector-width=): Add Save.
4995 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4996 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4997 (ix86_debug_options, ix86_function_specific_print): Adjust
4998 ix86_target_string callers.
4999 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5000 (ix86_valid_target_attribute_tree): Likewise.
5001 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5002 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5003 ix86_target_string caller.
5006 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5007 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5008 instead of gen_int_shift_amount + convert_modes.
5010 PR rtl-optimization/93088
5011 * loop-iv.c (find_single_def_src): Punt after looking through
5012 128 reg copies for regs with single definitions. Move definitions
5015 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5017 * config/arm/arm-c.c (arm_cpu_builtins): Define
5018 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5019 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5020 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5021 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5022 * config/arm/arm-tables.opt: Regenerated.
5023 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5024 arm_arch_i8mm and arm_arch_bf16 when enabled.
5025 * config/arm/arm.h (TARGET_I8MM): New macro.
5026 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5027 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5028 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5029 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5030 (v8_6_a_simd_variants): New.
5031 (v8_*_a_simd_variants): Add i8mm and bf16.
5032 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5034 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5037 * predict.c (compute_function_frequency): Don't call
5038 warn_function_cold on functions that already have cold attribute.
5040 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5043 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5044 COMDAT group function labels in .data.rel.ro.local section.
5045 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5048 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5049 comparison_operator in B and S integer comparisons. Likewise, use
5050 ordered_comparison_operator instead of cmpib_comparison_operator in
5052 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5054 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5056 Update copyright years.
5058 * gcc.c (process_command): Update copyright notice dates.
5059 * gcov-dump.c (print_version): Ditto.
5060 * gcov.c (print_version): Ditto.
5061 * gcov-tool.c (print_version): Ditto.
5062 * gengtype.c (create_file): Ditto.
5063 * doc/cpp.texi: Bump @copying's copyright year.
5064 * doc/cppinternals.texi: Ditto.
5065 * doc/gcc.texi: Ditto.
5066 * doc/gccint.texi: Ditto.
5067 * doc/gcov.texi: Ditto.
5068 * doc/install.texi: Ditto.
5069 * doc/invoke.texi: Ditto.
5071 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5073 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5076 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5078 PR tree-optimization/93098
5079 * match.pd (popcount): For shift amounts, use integer_onep
5080 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5081 tests. Make sure that precision is power of two larger than or equal
5082 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5083 instead of ULL suffixed constants. Formatting fixes.
5085 Copyright (C) 2020 Free Software Foundation, Inc.
5087 Copying and distribution of this file, with or without modification,
5088 are permitted in any medium without royalty provided the copyright
5089 notice and this notice are preserved.